Description
Turnarounds Modeling The Probability Of A Turnaround
TURNAROUNDS
-MODELING THE PROBABILITY OF A
TURNAROUND-
Master Thesis
Spring 2011
Supervisor:
Göran Andersson
Authors:
Eduard Ciorogariu
Andreas Goumas
2 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
ABSTRACT
The objective of this paper is to examine the possibility of predicting the recovery of a
distressed firm into a turnaround based on its current financial situation and a set of variables
that are considered of having a significant impact on the turnaround probability. To assess this
problem 150 firms are used that were distressed at some point during the period 1991 to 2001.
These firms were all listed in one of the major US stock exchanges and were all randomly
chosen, with 86 failing to recover from distress and 64 making a successful turnaround. In
order to establish a forecast model, two different quantitative econometrical methods are
applied; Linear Discriminant Analysis and Logistic Regression. The model predicting the
outcome of the 150 distressed firms with the highest accuracy is tested for its prediction
power on a holdout sample that consisted of 3140 distressed firms. These 3140 firms were all
listed at one of the major US stock exchanges and are distressed at some point during the
period 2002 to 2008. The prediction accuracy of the best model amounted to 92.7 % in the in-
sample and 89% in the holdout sample. The decisive variables that were selected by this
model are firm size, severity of distress and total debt to total assets.
Finally, we compare the returns yielded by a portfolio consisting of the turnarounds that were
predicted by the model out of the holdout sample to the returns generated by the S&P 500.
The annual returns for the seven years back-testing period, 2004-2010, for our portfolio
amounted to 18%, while the annual return for the S&P 500 was 4%.
Keywords: Financial distress, Turnaround, Turnaround prediction, Altman Z-score,
Discriminant Analysis, Logistic model.
3 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table of Contents
Introduction ............................................................................................................................................. 5
1.1 Background .................................................................................................................................... 5
1.2 Discussion of the problem ............................................................................................................. 8
1.3 Purpose .......................................................................................................................................... 9
1.4 Limitations ..................................................................................................................................... 9
1.5 Outline of the thesis .................................................................................................................... 10
Theoretical Framework ......................................................................................................................... 12
2.1 Prior research on turnarounds and its determinants.................................................................. 12
2.1.1 Causes of performance decline ............................................................................................ 13
2.1.2 Strategies to reverse performance decline .......................................................................... 18
Data and Methodology .......................................................................................................................... 22
3.1 Sources of information ................................................................................................................ 22
3.2 Criticism of sources ..................................................................................................................... 22
3.3 Definitions ................................................................................................................................... 23
3.4 Data ............................................................................................................................................. 25
3.5 Variables and hypothesis development ...................................................................................... 29
3.5.1 Size (X1) ................................................................................................................................ 29
3.5.2 Severity of distress (X2) ........................................................................................................ 30
3.5.3 Capital structure (X3, X4) ...................................................................................................... 30
3.5.4 Long-term financial health (X5, X6) ...................................................................................... 31
3.5.5 Short-term financial health / Liquidity (X7, X8, X9) .............................................................. 32
3.5.6 Profitability / Efficiency (X10, X11) ....................................................................................... 33
3.5.7 Investments / Divestments (X12, X13, X14, X15) ................................................................. 34
3.5.8 Management Expertise (X16) ............................................................................................... 36
3.5.9 Overview of examined variables .......................................................................................... 37
3.6 Methodology ............................................................................................................................... 38
3.6.1 Linear Discriminant analysis (LDA) ....................................................................................... 39
3.6.2 Binary Logistic regression model (Binary LOGIT) ................................................................. 44
Empirical Findings .................................................................................................................................. 48
4.1 Initial situation ............................................................................................................................. 48
4.2 Results of LDA .............................................................................................................................. 49
4.2.1 Means and correlation procedure (Model I) ........................................................................ 49
4.2.2 Stepdisc procedures (Model 2) ............................................................................................ 51
4 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.3 Results of LOGIT .......................................................................................................................... 52
4.3.1 Forward Conditional Logistic Regression (Model 3) ............................................................. 53
4.3.2 Backward Conditional Logistic Regression (Model 4) .......................................................... 54
4.3.3 Model comparison, selection and interpretation of results ................................................ 55
4.4 Model interpretation ................................................................................................................... 58
4.4.1 Size (X
1
) ................................................................................................................................. 58
4.4.2 Severity of distress (X
2
) ......................................................................................................... 58
4.4.3 Total debt to total assets (X
3
) ............................................................................................... 59
4.5 Stock returns of turnarounds from the holdout sample ............................................................. 64
4.6 Variable testing ............................................................................................................................ 64
4.6.1 Normality test ....................................................................................................................... 64
4.6.2 Heteroskedasticity test ......................................................................................................... 65
4.6.3 Multicollinearity test ............................................................................................................ 65
Conclusion ............................................................................................................................................. 67
References ............................................................................................................................................. 69
Published References ........................................................................................................................ 69
Internet References ........................................................................................................................... 72
Manuals ............................................................................................................................................. 72
Database ............................................................................................................................................ 72
Appendix ................................................................................................................................................ 73
Appendix 1: Overview of prior studies on turnaround determinants .............................................. 73
Appendix 2: Firms included in the in-sample .................................................................................... 74
Appendix 3: Scores model 1-4, in-sample ......................................................................................... 77
5 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 1
Introduction
1.1 Background
When plotting the movement of the S&P 500 over the last thirteen years, the index depicts
four extreme turning points. Driven by the elation of the emergence of a New Economy, the
S&P 500 surged at the end of the century and started to decline with the burst of the dot.com
bubble at the beginning of 2001, hitting rock bottom two years later. In 2003, the index
headed off to regain strength, reaching a new peak four years later. The outbreak of the
financial crisis triggered a new downswing of the index at the end of 2007. After a year, the
S&P 500 had to register an annual return of -38.5%
1
, its lowest result in sixty years of history.
During the first quarter of 2009 the index bounced back again and approaches its all-time high
of 1565.15
2
. The share index seems to display a reverting pattern, enabling market
participants to realise high capital gains, by selling when markets are at peak and buying
when they bottom out.
While the strategy “Buy low – Sell high” adds up for indices, it might fail for a single share,
because the risk of lasting underperformance or at worst bankruptcy cannot be diversified
away. However, to generate high returns investors don’t have to look out for the next global
crisis that will cause indices to plummet before they rally again. There are plenty of company-
specific financial crises occurring every year out of which a high-yield portfolio can be
constructed. Recalling that the share price reflects the investors’ expectations about the
company’s future performance, the stock of a firm sliding into financial distress is likely to
slump, regardless of whether the distressed state is expected to be temporary or long lasting.
A possible explanation for this behaviour is the market’s inability to capture the economic
fundamentals of distressed shares. Compliant with behavioural finance theory the
convergence between irrationality and barriers to arbitrage impede a separation between
transient and ongoing distressed stocks.
3
Moreover, since the variables of distressed
companies outweigh the variables of non-distressed companies in number, complexity and
1
http://www.forecast-chart.com/historical-sp-500.html
2
Twin A. (2009-03-09), “For Dow another 12-year low”, CNN Money,http://money.cnn.com/2009/03/09/markets/markets_newyork/index.htm
3
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, pp. 388
6 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
degree of uncertainty, deriving the intrinsic value of a distressed firm is a delicate endeavour,
enticing many intrinsic investors to refrain from their initial intention to invest.
4
Hence,
intrinsic investors fail to engage in a price correction and the stocks continue to fall. This
implies an undervaluation of some of the firms that are in financial distress, allowing market
participants to buy stocks with ample upside potential at marked-down prices. In fact, the
degree of upside potential can be expressed as a function of markdown. To be a value-
creating investment the share price needs to stop its downfall and start increasing again. It is
expected that the downward trend of a financial distressed company is reversed by the
implementation of a successful turnaround. Despite succeeded turnaround, market participants
will place low expectations in the future performance of recent distressed companies,
facilitating the management’s task to exceed shareholder expectations, which winds up in an
increase of the share price.
5
There exist plenty of studies, intending to identify the decisive factors in the turnaround
process. Some studies centre on quantitative variables, while other studies involve a
combination of quantitative and qualitative variables, taking into account that management
expertise and stakeholder support are crucial for conducting a successful turnaround.
Researchers distinguish between an efficiency-oriented and an entrepreneurial-oriented
strategy, firms can embark on during the turnaround process.
6
While several researchers like
Zeni et. al (2010)
7
develop turnaround prediction models, only few of them test their model
with respect to the stock returns yielded by the predicted turnarounds.
While investing in distressed companies is a popular research area, the bulk of research
focuses on investing in defaulted debt securities. Edward Altman, the inventor of the Z-score
that is widely used for determining distressed firms, has undertaken extensive research in this
field. He concentrates on the “risk and return performance of defaulted debt”.
8
In 2003 Altman
and Pompeii laid out an analysis of the historical performance of investments in defaulted
4
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the thoughtful investor,
Harper Business, pp. 189.
5
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, pp. 46.
66
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
7
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence from
Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
8
Altman, Edward I., (1998). Market Dynamics and Investment Performance of Distressed and Defaulted Debt
Securities, New York University, Center for Law and Business, Working Paper No. 98-022, pp. 2.
7 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
bonds and bank loans, covering the time period 1987 - 2001.
9
Practitioners like fund manager
Joel Greenblatt, recommend against investing in distressed stocks, because in case of
bankruptcy shareholder interests are the last to serve and equity holders might come away
empty-handed. Thus, his focus lies also on investing in distressed bonds, bank loans and trade
claims. However, this area is dominated by vulture investors, who can cope with the
complexity of such investments, which results from the legal and financial issues brought
about by different classes of creditors with different priorities and claims.
10
Nevertheless, researchers and practitioners see a potential benefit from investing in companies
that emerged from financial distress. Greenblatt states that recently emerged shares are
available at substantial discount, partially because they suffer from low analyst coverage and
partially because market participants still attach a high risk profile to the stock.
11
Another
reason is that a big stake of creditors’ bankruptcy claims rather tends to be converted into
equity claims than paid out in cash. Former debt holders such as banks, bondholders and trade
creditors have no incentive to engage in a long-term commitment in the emerged company
and aim at cashing out by selling the new share (we assume that the participation of former
shareholders was either bought up for a liquidating dividend or cancelled).
12
This creates a
negotiating range, which permits interested investors to purchase the share at a discount.
Altman et. al (1998)
13
investigated the stock performance of firms emerging from Chapter 11.
The authors observed significant positive excess returns over the long-term (200 trading days
from emergence date from Chapter 11) and ascribed it to the market’s inefficiency, which
causes a paucity of information that in turn leads to a stock’s mispricing. Besides, the study
points towards the existence of a positive relationship between the nature of securities
accepted by creditors and the appearance of excess stock returns. According to this, stocks of
emerged firms for which debt holders approved a complete equity-for-debt exchange
demonstrate strong positive long-term abnormal returns.
14
The identified linkage between the
type of arrangement the emerged firm and its debt holders agreed upon and its stock returns
let us infer that the creditors dispose of information, which is not captured by the market,
allowing them to compute the firm’s intrinsic value. The conclusion is reasonable, as creditors
9
Altman, Edward I. and Jha, Shubin, (2003), “Market size and investment performance of defaulted bonds and
bank loans: 1987-2001”, Economic Notes, Vol. 32, Iss. 2, pp. 147-176.
10
Greenblatt, Joel, (1999), You can be a stock market genius, Fireside, pp. 166 – 168.
11
Ibid, pp. 175.
12
Ibid, pp. 169 – 170.
13
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of Firms
Emerging from Bankruptcy, Journal of Finance, Vol. 54, Iss. 5, pp. 1855-1868.
14
Ibid, pp. 1865-1867.
8 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
like banks and other financial institutions e.g. life insurance, pension funds etc. are classified
as informed investors and do not confront information asymmetry issues.
15
However,
according to Kahl (2002)
16
debt holders cannot eliminate the degree of uncertainty in respect
of the viability of distressed firms completely, which drives them to engender a decision
model comprising three options: recovery, controlled liquidation and immediate liquidation.
Consequent, if creditors agree to swap their entire debt position for equity, they have strong
beliefs in the recovery potential and growth opportunities of the firm and choose the first
option.
1.2 Discussion of the problem
Altman et. al (1998)
17
attested that stocks of financially distressed firms, which were likely to
succeed in the turnaround process, yielded abnormal returns over a time frame of 200 trading
days, outstripping market indices by roughly 20%. Indro et. al (1999)
18
established a model
consisting of five variables, which can be applied to distinguish between successful and failed
restructurings. They demonstrated that for a portfolio comprising distressed stocks and having
an accumulated turnaround probability of more than 50%, excess compounded returns amount
to 42% over a one-year period.
While most studies focus on firms that have submitted an official bankruptcy petition, such as
e.g. filing under Chapter 11, our empirical research is not restricted to this formal procedure.
We refrained from constraining our study on the stock performance of firms emerging from
Chapter 11. In this manner, we took into consideration the findings of Hotchkiss (1995)
19
,
who challenges the accuracy of the Chapter 11 process in separating economically inefficient
from economically efficient companies.
Instead, we consider the Altman Z-score to be more precise in distinguishing potential
turnarounds from non-turnarounds, because it encompasses company-specific financial data,
15
Ogden, Joseph P., Jen, Frank C. and O’Connor, Philip F.,(2003), Advanced Corporate Finance. Policies and
Strategies, Prentice Hall.
16
Kahl, Matthias, (2002), Economic Distress, Financial Distress and Dynamic Liquidation, The Journal of Finance,
Vol. 57, pp. 135-168.
17
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of Firms
Emerging from Bankruptcy, Journal of Finance, Vol. 54, Iss. 5, pp. 1855-1868.
18
Indro, D. C., Leach, R.T. and Lee, W. Y., (1999), Sources of gains to shareholders from bankruptcy resolution,
Journal of Banking & Finance, Vol. 23, Issue 1, pp. 21-47.
19
Hotchkiss, Edith S., (1995), Postbankruptcy Performance and Management Turnover, Journal of Finance,
Vol. 50, Issue 1, pp. 3-21.
9 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
which makes it possible to pinpoint operational performance changes over time. Additionally,
we investigate the impact of other quantitative variables on the likelihood of turnaround,
aiming at determining the drivers that lie behind a successful restructuring.
1.3 Purpose
The underlying thesis follows the purpose of deriving the decisive variables that allow for
distinguishing financially distressed firms with turnaround potential from those without.
Based on a sample of 150 companies, it intends to conceive the main drivers in the turnaround
process and their relation to the turnaround potential of a firm. Further, the identified drivers
are used to establish a prediction model, which can be adopted for classifying financially
distressed companies into turnarounds and non-turnarounds. To some extent, the thesis aims
at analyzing whether recovery is achieved by focusing on efficiency-oriented or on
entrepreneurial-oriented strategies. In addition, the thesis touches upon the opportunity that
arises from investing in financially distressed firms with high turnaround potential, by
exhibiting generated returns of the turnarounds detected in the holdout sample. In this manner,
the thesis tends to pave the way for future, profound research in distress investing.
1.4 Limitations
The states Distress and Recovery are defined by a company’s Z-score falling short of a given
threshold and subsequently exceeding this threshold and are not dependent on an official
filing for bankruptcy proceeding and a pursuant announcement of recovery. Hence, the
sample might consist of some firms that did not file under the US Bankruptcy Code and might
omit some firms that did so.
The sample employed to determine the decisive factors in the turnaround process contains
companies listed at one of the three major US stock exchanges, New York Stock Exchange
(NYSE), American Stock Exchange (NYSE Amex) and NASDAQ Exchange and embraces
the time period 1991 to 2003. The main reason for deciding to collect data from the specified
exchanges and over the specified period is to ensure the availability of a representative sample
in terms of size and industry coverage. On top of that, the sample comprises only non-
financial companies. The exclusion of financial firms from the study group is motivated by
the belief that inclusion would lead to biased results. This expectation can be motivated by
10 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
two points. Firstly, financial companies are characterized by an extremely high leverage ratio,
so that their involvement would most likely distort the impact of debt on the turnaround
process. Secondly, due to their importance for the overall economic stability of a country the
probability of governmental interventions in case of distress is much higher than for non-
financial companies. However, this study does not make a point of developing a way to
identify companies, which will be bailed out with the utmost probability in the event of
distress. Neither does it advise market participants to invest in such companies. The variables
analysed by the study are of quantitative nature, as collecting reliable and satisfying
qualitative data would require access to resources that were not approachable for us and
would go beyond the time period scheduled for this work.
The data employed for back-testing the prediction power of the established model is also
raised from the three major US stock exchanges.
1.5 Outline of the thesis
Chapter 1 – Introduction
The first chapter explains why we became interested in examining the drivers behind a
successful turnaround. It states the opportunities arising for investors to make money by
investing in financially distressed firms with a high turnaround potential. Studies of other
researchers are named and the main conclusions are summarised. Some of the difficulties a
financially distressed firm has to deal with in the process of restructuring are mentioned,
providing an idea about the variables that play a decisive role in the turnaround process. The
chapter ends with a listing of the study’s limitations.
Chapter 2 – Theoretical foundation
The second chapter delves into the subject of identifying the determinants of a successful
turnaround, so as to be able to distinguish firms with turnaround potential from firms that are
likely to remain in distress, denoted as non-turnarounds. Prior research on turnarounds is
covered and the key results of these studies are discussed.
The entire chapter provides the theoretical background on which the underlying thesis is
based.
11 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 3 – Data and Methodology
The third chapter gives a description of the process of data collection and brings in the
variables that were considered in the empirical study. Hereby, it develops the hypotheses that
are to be examined empirically. Last but not least, the two econometrical models that are
applied to come up with the turnaround prediction model, Multiple Discriminant Analysis
(MDA) and Logistic Regression (Logit Model), are introduced.
Chapter 4 – Empirical Results
The fourth chapter presents and further discusses the determinants that were found to be
critical for a successful turnaround by each of the applied models. In this way, it comes up
with four prediction models, which are assessed based on their in-sample accuracy. It also
includes the back-testing of the two best prediction models. The excess returns of the holdout
sample turnarounds, selected by the best prediction model, are computed. The input data of
the final chosen model is subject to some statistical tests.
Chapter 5 – Conclusion
The fifth chapter aims at wrapping up the key findings of the implemented study. Besides, a
recommendation about what issues future research in this area could cover is administered.
12 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 2
Theoretical Framework
Plenty of research has been undertaken on turnarounds with the purpose of identifying the
determinants of success. An overview of prior studies in this field is outlined in the appendix.
While empirical studies use financial metrics and quantitative ratios to specify a successful
turnaround, Balgobin et al. (2001)
20
provide a more qualitative definition:
“A corporate turnaround may be defined simply as a recovery of a firm’s economic
performance following an existence-threatening decline.”
2.1 Prior research on turnarounds and its determinants
Empirical studies concentrate on the reasons that drive companies into such performance
decline and the corresponding strategies the management team employs to reach
rehabilitation. Schendel et al. (1976)
21
22
and Hofer et al. (1978, 1980)
23
24
attributed
deteriorating firm performance to operational or strategic issues and emphasized the
importance of correctly recognizing the source of decline, which enables the firm to adopt the
adequate measures to reverse decreasing performance. They point out that failure to locate the
catalyst of decline leads to the implementation of wrong measures and can hinder the firm to
achieve a successful turnaround.
In the following sections several research studies on turnarounds are reviewed, outlining the
causes of performance decline and the strategies and measures firms take to return to a stable
state. This review provides the basis for the selection of the potential turnaround determinants
analyzed in the underlying thesis.
20
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
21
Schendel Dan E. and Patton, G.R., (1976), Corporate stagnation and turnaround, Journal of Economics and
Business, Vol. 28, Iss. 3, pp. 236-242.
22
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), Corporate turnaround strategies: A study of profit
decline and recovery, Journal of General Management, Vol. 3, Iss.. 3, pp. 3-12.
23
Hofer, Charles W. and Schendel, Dan, (1978), Strategy Formulation: Analytical Concepts, West Publishing.
24
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp. 19-31.
13 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
2.1.1 Causes of performance decline
Balgobin et al. (2001)
25
summarize the main performance decline triggers that were identified
by six independent research studies. A separation is made between internal and external
causes of performance decline, where internal causes refer to company-specific issues and
external causes are affiliated to weak economic conditions or industry-specific issues.
External causes
External causes include downturn in demand, increase in competition and increase in input
costs. The causes show a certain degree of interdependency, as fierce competition and
increased input costs both are likely to affect demand negatively.
Fall in demand can be traced back to several other reasons, such as the contraction of the
industry, an overall economic recession that weakens purchasing power or the failure of the
company to meet customer expectations. While the first two affect all competitors within an
industry, the last one applies to single companies. In regard of customer expectations, Lepak
et al. (2007)
26
define the task of a firm in creating value for its customers. They call this the
use value. In return, the customers are willing to provide the exchange value, which is
measured in monetary units. If a company cannot constantly provide the value demanded by
its customers at the expected conditions (e.g. quality, sales price), or if other companies have
the resources to provide equal or higher value, or to provide the same value at improved
conditions a firm-related cutback in demand should be expected.
The intensity of competition is determined by the characteristics of the industry. The concept
of industrial organization competition summarized by Barney (1986)
27
considers the barriers
established in an industry as one determinant of competitive pressure. These barriers involve
barriers to entry, barriers to competition, barriers to imitation and barriers to exit.
28
Other
decisive factors are the amount and size of rivals, the nature of products (customized vs. mass
product) and the demand elasticity.
29
In addition, substitute products pose a threat, as firms
might not recognize immediately to whom they lose market share, aggravating the necessity
25
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
26
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture: A Multilevel
Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
27
Barney, Jay B., (1986), Types of competition and the theory of strategy: Toward an integrative framework,
Academy of Management, Vol. 11, No. 4 , pp. 791-800.
28
Ibid.
29
Ibid.
14 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
to retaliate.
30
Lepak et. al (2007) mention the problem of value slippage
31
, which leads to an
erosion of the firm’s competitive advantage and can be avoided by the establishment of
isolating mechanisms
32
(e.g. patents, trademarks, special knowledge). Fierce competition will
undermine the financial position of a company, as it has to engage in costly retaliation
campaigns, such as price cuts or expensive marketing and promotion campaigns to guarantee
competitiveness. Adopting the wrong retaliation tactics will lead to a further weakening of the
firm’s market position and absorption of financial means.
Price increases in input costs bring forth a rise in costs of goods sold, putting pressure on the
gross profit margin. Given that companies are able to pass on the increase in costs fully to the
customers, margins will not be suppressed. While this might be imaginable in a monopolistic
market, it is unlikely to hold for a situation of perfect competition. Price pressure will
originate from some companies that enjoy advantages on the input market e.g. strong
bargaining power, bulk purchase etc. and squeeze margins of firms that do not dispose of
these advantages by luring away customers.
I nternal causes
Internal causes of performance decline involve poor management, inadequate financial
control/policy and high cost structure. As for the external causes a kind of interdependency
can be observed, as inadequate financial control/policy and high cost structure both can
originate from poor management.
The management team needs to be endued with the capabilities to steer the company through
times of prosperity and decline. There exists no commonly accepted definition of poor
management performance, but some conclusions about its meaning can be drawn from the
examined literature. According to Hedberg et al. (1976)
33
firm decline bears upon the
omission of the management team to align the strategy of the company to its evolving
environment. This problem tends to exacerbate for companies with a long track record, as the
management team is prone to hubris and overconfidence and has strong beliefs in the
30
Koller et. al (2010), Valuation: Measuring and managing the value of companies, pp. 79-98
31
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture: A Multilevel
Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
32
Ibid.
33
Hedberg, Bo L. T., Nystrom, Paul C. and Starbuck, William H., (1976), Camping on Seesaws: Prescriptions for
a Self-Designing Organization, Administrative Science Quarterly, Vol. 21, No. 1, pp. 41-65.
15 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
underlying strategy, rendering it immune to necessary strategic change.
34
Harker (1996)
35
stresses the importance for a company to understand its industry, markets and customers, to
know its position and future potential with respect to its industry and markets and to know its
competitors and their position and future potential. Only if the management team is able to
capture and process these variables, can it come up with an adequate strategy. Another
important task of the management team is the delegation of responsibilities to lower
hierarchical levels. Concentrating too much decision power at the top level can lead to inertia
and delayed responsiveness to changes in customer preferences, as it was the case for IBM
UK in the 90s.
36
Regarding inadequate financial control/policy the focus lies on the firm’s capital structure and
the sources of financing.
37
Owing to the fact that interest expenses lower the taxable income
and in this way increase the free cash flow to firm, the market value of the firm would be
maximized by employing a gearing of 100%. However, with an increase in leverage the
probability of future financial distress and the cost of financial distress also raise, which
results in a depression of the firm’s market value. This relationship is depicted by the
following formula:
V
L
= V
U
* ?
c
* D – PV[E(CFFD)]
V
L
= Market value of the levered firm
V
U
= Market value of the unlevered firm
?
c
= Corporate tax rate
D = Face value of debt
PV[E(CFFD)] = Present value of expected cost of future financial distress
In addition, having an extremely high gearing might force a company to postpone or
completely abandon some value creating investments, causing an underinvestment problem.
34
Barker, Vincent L. and Barr, Pamela S., (2000), Linking top manager attributions to strategic reorientation in
declining firms attempting turnarounds, Journal of Business Research, Vol. 55, Iss. 12, pp. 963-979.
35
Harker, Michael, (1996), Managing company turnarounds: how to develop “destiny”, Marketing Intelligence
& Plannings, Vol. 14, Iss. 3 pp. 5-10.
36
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
37
Ibid, pp. 303.
16 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
An example is provided to illustrate the underinvestment problem. Assumptions are based on
Myers (1977)
38
:
It is assumed that the firm’s assets in place V
A
are equal to 0. The firm takes on risky debt P
(P is defined as risky debt, as the firms has no assets in place and thus does not dispose of
collaterals to secure the debt), which together with the contribution I (equity investments) of
shareholders account for the required cash outlay to realize V
G
, the real option. The debt is to
be repaid after the expiration of the real option V
G
and the owners know the value of the real
option in the event of exertion, which is depicted V(s).
In such a case, shareholders will exercise the real option only on condition of:
V(s) > I + P
If the cash outlay (I + P) exceeds V(s), shareholders will refuse to exercise the real option, as
their equity investment I will be higher than the market value of their shares.
39
Thus, even if
the real option is value creating (V(s) > I), it won’t be realized, because of the effect of the
debt burden. This is also referred to as the debt overhang problem. Abandoning value creating
projects can accelerate a firm’s performance decline, as its competitive position might be
undermined, triggering a decrease in operational performance. Generating fewer turnovers
will raise the financial pressure, amplifying the debt overhang problem and prompting further
sacrifice of value enhancing investments. However, the abandonment of promising projects
must not always be the actuator of a drop in operational performance. Sales can also be
depressed by the various external factors described in the previous section, like e.g. industry
contraction. Although there exists no optimal debt-to-equity ratio, when deciding on leverage
a firm should allow for a reasonable financial buffer to be able to absorb unexpected
economic or industry-specific declines and leave the door open to undertake value enhancing
investments when they appear.
With reference to the sources of financing, failure to abide by the maturity matching principle
can bring a company into financial distress. In general, the maturity on an interest-bearing
liability should coincide with the life expectancy of the asset or project, for which the credit is
raised. Ignoring this guideline gives rise to either a refinancing risk or a debt overhang
problem. In the case that the maturity of interest-bearing debt falls short of the life expectancy
38
Myers, Stewart C., (1977), Determinants of Corporate Borrowing, Journal of Financial Economics, Vol. 5,
No. 2, pp. 147-175.
39
Ibid, pp. 153
17 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
of the asset/project, interest and principal payments might be due before the asset/project
managed to generate any cash-flow. Hence, the company will try to roll over the outstanding
debt. If the debt holder refuses refinancing, the firm will be forced to repay the credit
immediately, although the asset/project missed to yield a return so far. This will lead to a
financial bottleneck that can culminate in defaulting on debt. Alternatively, if the maturity of
interest-bearing debt exceeds the life expectancy of the asset/project, the firm will have to
bear an ongoing debt burden, which is related to an asset/project that does not generate cash-
flows anymore. Thus, it would carry on the default risk and the outstanding debt would
interfere with the realization of valuable real options (debt overhang problem).
Last but not least Balgobin et. al (2001) cite a high cost structure as a reason for performance
decline. In case of IBM UK an overstated cost base was build up, because the top
management expected revenues to grow at historical rates. When realized revenues did not
comply with expected revenues, the company had to face a cost burden that outstripped the
one of competitors significantly.
40
This example also elucidates the interdependency of
performance decline causes, as the high cost base was a result of management’s inability to
correctly anticipate future market demand.
All in all, deteriorating performance can be traced back to several internal and external
causes, which are closely intertwined and act jointly, making it impossible to relate
performance decline to one single source. The table below provides an overview of the
different research studies that focused on the same external and internal causes of decline.
Researcher Schendler et al. (1976) Bibeault (1982) Slatter (1984) Thain et. al (1989) Grinyer et. Al (1990) Gopal (1991)
External causes
Decrease in demand x x x x x x
Increase in competition x x x x x x
Increase in input costs x x x x n.a. x
Internal causes
Poor management x x x x x x
Inadequate financial
control/policy n.a. x x x x x
High cost structure x n.a. x x x n.a.
Table 1: The Causes of Declining Performance
Source: Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
X = clearly referred to; n.a. = not referred to
40
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Issue 3, pp. 308.
18 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
2.1.2 Strategies to reverse performance decline
This section outlines the various strategies that companies adopt, in order to cope with an
existence-threatening decline and to achieve a successful turnaround.
According to Schendel et al. (1976)
41
a company can either apply an efficiency-oriented or an
entrepreneurial-oriented strategy. Which strategy is chosen depends on the cause of the
downturn. Efficiency-oriented restructurings imply the enforcement of retrenchment, which
incorporates cost-cutting measures, downsizing and asset reduction, while entrepreneurial-
oriented restructurings aim at aligning the underlying strategy to the prevalent market
conditions.
42
The viewpoint of Schendel et. al (1976) is supported by Barker et. al (1997)
43
,
who distinguish between two sources of decline: industry-specific and firm-specific decline.
Cameron et. al (1988)
44
explain firm-specific decline as the inability of a company to perform
at eye level with its competitors, suffering from a competitive disadvantage. Thus, if a firm
acts in a growing industry, but faces deteriorating performance, the adoption of an
entrepreneurial-oriented strategy is compulsory. It can be concluded that companies, which
suffer from performance decline due to a contraction of the industry, should put more weight
on efficiency-oriented strategies.
As opposed to this, Robbins et. al (1992)
45
hold that independent of the cause of decline the
implementation of efficiency-oriented strategies is crucial for succeeding with the turnaround.
Other researchers opt for separating the turnaround process into two subsequent stages:
Stabilization and Recovery. The purpose of the first stage is to prevent a continuation in
performance decline and to build the foundations for the implementation of recovery
strategies. This process involves convincing stakeholders to support the turnaround intention,
stop the financial drainage and ensure a constructive internal climate. In the second stage the
recovery strategies are introduced, according to the trigger of decline. The necessity of
41
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), Corporate turnaround strategies: A study of profit
decline and recovery, Journal of General Management, Vol. 3, Issue 3, pp. 3-12.
42
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Issue 3, pp. 305.
43
Barker, Vincent and Duhaime, Irene M., (1997), Strategic change in the turnaround process: Theory and
empirical evidence, Strategic Management Journal, Vol. 18, Issue 1, pp. 13-38.
44
Cameron, Kim S., Sutton, Robert I. and Whetten, David A., (1988), “Readings in Organizational decline:
Frameworks”, Ballinger Publishing.
45
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
19 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
administering a two-stage approach is determined by the severity of distress, the firm size and
the availability of free assets.
46
Maintaining stakeholder promotion is critical for facilitating the continuation of the
operational business. Companies that slide into financial distress are likely to experience a
large number of resignations from key employees, resulting in a brain drain that aggravates
the competitive situation. Suppliers and customers must be persuaded to uphold business
relations with the firm and debt holders must be prepared to compromise on contractual terms.
In order to restore the stakeholders’ trust in the firm’s survival potential and accomplish its
support, quick actions that yield immediate results are applied at the beginning of the
turnaround process, aiming at improving efficiency.
47
These actions narrow down to cutbacks,
which are concentrated on downsizing, reduction of inventory levels, cost of goods sold and
selling, general and administrative expenses.
48
Cost cuttings and efficiency enhancements will
free up resources that can be reallocated.
49
However, the assertion of cutbacks might backfire
and even cause a further drop in firm efficiency. This can be expected when managers decide
to cut costs on the wrong positions. For example, switching to cheaper suppliers might also
have a degrading effect on product quality, causing more customers to discontinue business
relations. Furthermore, the management’s decision to lay off employees and undertake salary
cuts and cancellations of one-time bonus payments can create a working climate that is
characterized by insecurity about the workplace and frustration, releasing a loss of motivation
and associated increase in absenteeism, more production of scrap, decreased product quality,
extended processing time and delayed deliveries.
50
Arogyaswamy et. al (1997)
51
conclude that
turnarounds and non-turnarounds have a strong tendency to engage in cutbacks. However,
non-turnarounds apply this measure more excessively than turnarounds. Also, turnarounds are
more successful in translating cutbacks to efficiency improvements than non-turnarounds.
From this it can be concluded that managers of turnarounds pick the write spots for cost-
46
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Issue 3, pp. 304-320, p. 305.
47
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11, p. 4.
48
Ibid, p. 3.
49
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 306.
50
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11, p. 3.
51
Ibid.
20 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
cutting and prove able to convince remaining employees from the necessity of the undertaken
retrenchment measures. A more extensive form of cutbacks is operational asset reduction,
which is carried out to lower the firm’s capacity to the current production level. In so doing,
manufacturing facilities are employed more efficiently and cash-inflows are generated
52
through the sales of assets, which can be used for lowering the debt burden or to make
necessary capital expenditures, like maintenance investments in PP&E.
As has been represented at length, stopping performance decline presupposes the execution of
retrenchment measures to improve efficiency. Hofer et al. (1980)
53
note that a financially
distressed company will restructure its operations and thereby repel the threat of bankruptcy
before it starts to analyze its strategic position in the market.
In the second stage, the distressed company will either continue to implement more profound,
long-lasting operational changes or strive for a strategic reorientation, depending on the cause
of decline. Yet, according to Grinyer et. al (1988)
54
a sustainable operational improvement is
achieved, when turnarounds put emphasis on strategic reorientation, redefining their product
and market portfolio. Strategic reorientation implies divesting in unrelated areas and investing
in related areas, thereby strengthening the focus of the company on its core-capabilities.
Basically, a company has to separate its products and markets following the criterion of value
creation. Value-destroying business units are sold and value-destroying markets are
abandoned. That way, the company obtains a cash-inflow in terms of the sales price and
reduces cash-outflow, which was attributed to the maintenance of the sold business units and
exited markets. The cash-inflow obtained from the divestments can be used partially to lower
the debt burden and partially to invest into value-creating business units and markets.
Nevertheless, depending on the severity of distress and the support from stakeholders,
especially the willingness of debt holders to grant further moratorium or even provide
additional financial funds, a company might be forced to sell off profitable business units to
generate sufficient cash.
55
Schlingemann et. al (2002)
56
reinforce this assumption,
52
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 185.
53
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp. 19-31.
54
Grinyer, Peter H., Mayes, David and McKiernan, Peter, (1988), “Sharpbenders: The secrets of unleashing
corporate potential”, Blackwell Publishers, Oxford.
55
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 186.
56
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the Liquidity of the
Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1, pp. 117-144.
21 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
demonstrating that firms rather divest the most profitable segment over the least liquid
segment and the most liquid segment over the least profitable segment.
In terms of asset investment, Hambrick et. al (1983)
57
argue that internal capital expenditures
are geared towards obtaining efficiency improvements, by means of e.g. improved monitoring
and steering of the process flow. Arogyaswamy et. al (1997)
58
demonstrate the importance of
capital expenditure on PP&E in their study. However, they point out that the amount invested
in PP&E is almost equal between turnarounds and non-turnarounds. Nevertheless, a
significant difference exists regarding investments in R&D, which are clearly higher for
turnarounds. They take the view that investing in new technology is vital for manufacturers to
adapt to a changing environment
59
and meet market expectations. External investments in the
sense of acquisitions can be conducted, as part of the strategic reorientation process to
strengthen the product and market portfolio and accelerate revenue growth, given that the
company disposes of sufficient financial slack.
60
The table below summarizes the main measures implemented under the corresponding
strategy.
Efficiency-oriented Entrepreneurial-oriented
Table 2: Turnaround strategies
Strategy alignment to changing
environment through:
Investments in R&D
strategic asset divestment
strategic asset investment
Retrenchment of the firm:
reduction of operational cost
Downsizing
Operational asset reduction
Internal capital expenditures
57
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
58
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11.
59
Ibid, p. 3.
60
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies and their
implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
22 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 3
Data and Methodology
This chapter describes the steps undertaken in our empirical research on determining the
decisive factors for a successful turnaround. It provides information on applied sources, data
collection and data processing and outlines the methodology employed to perform the
research.
3.1 Sources of information
As our empirical research is to the greatest extent based on financial data, we perceived it as a
vital prerequisite to obtain our input data from a single and reliable source. Taking into
account that each database has established its own technique of providing financial
information, such as balance sheet items, financial ratios, share prices etc., falling back on
several databases could lead to distorted results and conclusions. Thus, all financial data used
in our study is collected from the Standard & Poor’s database.
In terms of literature employed for presenting the theoretical background of our study, we
went back to course literature, course material and scientific articles, which were extracted
from LibHub.
3.2 Criticism of sources
The financial data used in our empirical study was not generated at first hand, as we collected
it from a database provider. It has to be taken into account that due to the extent of our sample
and its encompassed time span, it would have been inefficient to extract the data separately
from the balance sheet statements and the income statements of each company. In order to
examine the reliability of the database, we randomly selected firms from our sample and
cross-checked the provided data with the data reported in the respective SEC filings.
With respect to the literature applied for the theoretical background, we made reference to
scientific articles published in distinct journals and to literature that covers the topic of
restructuring financially distressed companies. With a view to critically scrutinize the
readings, we abandoned demonstrating only the viewpoint and results of one researcher, but
23 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
were constantly anxious to provide corroborating or confuting ideas of other research
colleagues.
3.3 Definitions
Our sample consists of companies that are classified as either distressed firms or turnarounds.
We applied the Altman Z-score to obtain a sample of distressed firms and to separate
companies, which remained distressed from companies which achieved a turnaround.
Altman (1968)
61
introduced the Z-score as a measure of predicting the firm’s probability of
going bankrupt. It is defined as follows:
Where, X
1
: Working capital/Total assets
X
2
: Retained Earnings/Total Assets
X
3
: Earnings before Interest and Taxes/Total Assets
X
4
: Market value equity/Book value of total liabilities
X
5
: Sales/Total assets
According to Altman, a Z-score of greater than 2.99 classifies the firm into the “non-
bankrupt” zone. If the Z-score is below 1.8, the firm falls into the “bankrupt” zone and firms
lying in-between 1.8 and 2.99 are assigned to the “zone of ignorance” or “grey area”.
62
We defined a firm as financially distressed, if it exhibited a Z-score below 1.8 for two
consecutive years. In the event that the Z-score of the company increased above 1.8 in the
third year and above 2.99 in the fourth year, or was above 2.99 for two successive years after
being classified as financially distressed, it was perceived as a successful turnaround.
Companies, whose Z-score remained below 1.8 for two further years were categorized as
failed turnarounds.
61
Altman, Edward I., (2000). “Predicting Financial Distress of Companies: Revisiting the Z-score and Zeta
Models“, New York University, Center for Law and Business.
62
Ibid.
24 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 1: Illustration of z-score development for 2 companies
The above graph depicts three companies (A, B & C), which would have been eligible for our
sample based on the development of their Z-scores over a total period of four years. All three
firms had a Z-score below 1.8 for two consecutive years. However, in year 3 and 4 the Z-
score of company B surpassed the critical threshold of 2.99, while the Z-score of company A
stayed below the critical threshold of 1.8. The Z-score of company C went beyond the critical
threshold of 1.8 in year 3 and exceeded the critical threshold of 2.99 in year 4. In this
instance, we would have categorized firms B and C as successful turnarounds (1) and firm A
as a failed turnaround (0). The two tables below summarize the main definitions, on which the
entire study is based.
Distressed firm: two consecutive years of Z-score below 1.8
Successful turnaround: two consecutive years of Z-score below 1.8
followed by either two consecutive years of
Z-score above 2.99 or a Z-score above 1.8
in year 3 and a Z-score above 2.99 in year 4.
Failed turnaround: two consecutive years of Z-score below 1.8
followed by two consecutive years of Z-score
below 1.8
Table 3: Summary of definitions
Status Year 1 Year 2 Year 3 Year 4
0 < 1.8 < 1.8 < 1.8 < 1.8
1 < 1.8 < 1.8
> 1.8
>2.99
> 2.99
Table 4: Status based on Z-score
0 = failed turnaround 1 = successful turnaround
0
1
2
3
4
5
0 1 2 3 4 5
Z
-
s
c
o
r
e
Years
Company A
Company B
Company C
25 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.4 Data
The firms and the appertaining financial figures of our sample were gathered from the
Standard and Poor’s database. In order to obtain a reasonable sample size, we focused on a
sample period ranging from 1991 to 2003, within which data was collected. This period was
divided into ten sub-periods, with each sub-period covering four years.
1990 2004
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
1991 - 1994
Sub-period 1
1992 - 1995
Sub-period 2
Graph 2: Illustration of sample period and corresponding sub-periods
Graph 2 depicts the process of data collection over the sample period and emphasizes that
data was gathered from each of the ten sub-periods.
Sub-period Turnaround Distress
1991-94 0 3
1992-95 1 1
1993-96 3 1
1994-97 6 5
1995-98 5 5
1996-99 2 6
1997-00 7 7
1998-01 11 16
1999-02 7 18
2000-03 22 24
Total 64 86
Table 5: No. of firms per sub-period
The precedent table displays the amount of firms collected from each sub-period,
differentiating between turnarounds and distressed companies. The sample comprised a total
of 150 companies, out of which 64 were classified as successful turnarounds and 86 were
categorized as failed turnarounds.
26 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table 6 provides an overview of the industries that were covered by the data.
Industry Turnaround Distress
Automobiles & Components 0 2
Biotechnology 6 3
Capital Goods 8 9
Commercial & Professional Services 2 6
Consumer Services 1 8
Energy 5 12
Food, Beverage and Tobacco 3 2
Gold 5 0
Healthcare 9 7
Household & Personal Products 2 1
Materials 1 9
Media 0 8
Oil, Gas & Coal Exploration & Production 4 1
Retailing 1 3
Software and Services 2 3
Technology Hardware and Equipment 6 4
Transportation 0 8
Others 9 0
Total 64 86
Table 6: No. of firms per covered industry
Involving firms from different industries was important, because the study aims at deriving
universal implications about decisive factors in the turnaround process and does not narrow
down its research to a particular industry. Nevertheless, we excluded financial institutions
from our sample, due to their highly levered capital structure and because some financial
institutions enjoy governmental bankruptcy protection through bail-out guarantees.
All firms used in the study were publicly traded and listed at one of the following U.S. stock
exchanges during the period 1991 to 2003:
i. (NYSE) ? New York Stock Exchange
ii. (AMEX) ? American Stock Exchange
iii. (NasdaqGM) ? Nasdaq Global Market
iv. (NasdaqCM) ? Nasdaq Capital Market
v. (NasdaqGS) ? Nasdaq Global Select
27 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
We have chosen these exchanges, because when accumulated they rank first place in two
categories over the whole length of our study period:
1. Amount of listed firms
2. Trading volume
A high amount of listed firms provides broad industry coverage, which is necessary
concerning that we do not restrict our analysis to a specific industry.
The variable trading volume is a crucial determinant of stock returns. The Wall Street believes
in a relationship between trading volume and stock returns, stating that “It takes volume to
make the prices move”
63
. Ying (1966)
64
demonstrated that small trading volumes are related
to negative returns (price fall) and large trading volumes are related to positive returns (price
rise). Several other researchers substantiated a positive correlation between trading volume
and stock returns. Below an excerpt of a list of studies on this issue is provided.
Researcher Year Sample Data Sample Period Interval
Positive
Correlation
found?
Comiskey et. al 1984 211 common stocks 1976-79 yearly Yes
Morgan 1976 44 common stocks 1926-68 monthly Yes
Richardson et. al 1987 106 common stocks 1973-82 weekly Yes
Jain et. al 1986 Stocks market aggregates 1979-83 hourly Yes
Table 7: Prior studies on correlation between returns and trading volume
Source: Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A survey”, The
Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126, p. 112.
A high trading volume is of interest for our study, since we also point to significant
differences in stock returns of distressed firms and turnarounds.
The amount of listed firms and trading volume was accumulated for the considered US stock
exchanges and compared to the London SE and the Tokyo SE over the sample period. The
63
Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A survey”, The
Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126, p. 112.
64
Ying, Charles C., (1966), “Stock Market Prices and Volumes of Sales”, Econometrica, Vol. 34, No. 3, pp. 676-
685, p. 676.
28 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
London SE and the Tokyo SE were chosen as benchmark stock exchanges, because they were
ranked among the top five largest stock exchanges throughout the whole study period.
65
Graph 3: Listed firms by stock exchange
Graph 4: Trading volume in USD millions by stock exchange
Graph 3 and 4 demonstrate that the involved US stock exchanges outperformed the London
SE and the Tokyo SE in number of listed firms and trading volume throughout the whole
sample period.
65
World Federation of Exchanges,http://www.world-exchanges.org/statistics/time-series/market-
capitalization
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
1990 1993 1996 1999 2002
US accumulated
London
Tokyo
0.0
5,000,000.0
10,000,000.0
15,000,000.0
20,000,000.0
25,000,000.0
30,000,000.0
35,000,000.0
40,000,000.0
45,000,000.0
1990 1993 1996 1999 2002
Tokyo
London
US accumulated
29 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.5 Variables and hypothesis development
This section presents the variables we have taken into account in our empirical research. We
assume that these variables have an impact on a company’s probability to recover from
financial distress. Therefore they act as discriminating predictors, enabling a separation of
distressed companies into turnarounds and failed turnarounds (firms that remain in financial
distress). A hypothesis was formulated for each variable.
3.5.1 Size (X1)
We measure size by means of total tangible assets. Some studies use total sales as an indicator
of size. However, we follow Smith et. al (2005)
66
, who relate the size of a company to its
borrowing capacity. According to the Collateral Hypothesis, a firm’s debt capacity is
restricted to its collateralizable assets, which are represented by its total tangible assets. Thus,
a firm with a higher debt capacity will have easier access to the credit market, to raise funds
necessary for the restructuring. White (1989)
67
highlights the positive impact of the track
record of large companies in raising external funds on their ability to obtain additional
financial support. Besides, strong stakeholder support is expected for large firms, as their
stakeholders have more to lose in the event of bankruptcy.
68
On top of this, large companies
dispose of more assets that can be sold and more business units that can be divested,
triggering a release of internally generated financial means, which contribute to the
restructuring process by reducing the leverage or enabling the realization of value-creating
investments. A contrary opinion is provided by Paint (1991)
69
, who reveals a negative
relationship between size and turnaround potential. According to him, smaller companies can
adapt more readily to altering conditions of their environment.
Although small firms might be characterized by a flat hierarchy and exhibit few layers of
management, allowing them to react fast to market changes, there access to external capital
markets might be restricted, forcing them to resort to internal financial means. As we perceive
66
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 310.
67
White, M. (1989), “Bankruptcy, liquidation and reorganization”, in: D.E. Logue (ed.), Handbook of Modern
Finance, Warren, Gorham & Lamont, New York.
68
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 306.
69
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in corporate
turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
30 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
the availability of capital as a vital factor for a successful turnaround, we do not share Paint’s
viewpoint.
Hypothesis 1: Firm size and turnaround potential are positively correlated.
3.5.2 Severity of distress (X2)
Robbins et al. (1992)
70
investigate a positive relationship between the severity of distress and
the degree of cutbacks and asset divestments. As stated by Slatter (1984)
71
, the
implementation of retrenchment measures might face organizational resistance, which results
in a decrease of operational efficiency, thereby aggravating the distressed state. Sudarsanam
et. al (2001)
72
indicate that the severity of distress negatively influences the required time for
the restructuring and can inhibit the completion of certain restructuring measures. These
measures primarily embrace actions aiming at reshaping the firm’s strategy and call for
capital expenditures. Debt holders might oppose the implementation of such actions, because
they consume capital funds that can be used to settle part of their claims. The severity of
distress is measured by the Z-score. Companies displaying a Z-score, which is below 1.8 or
even negative, are considered to be severely distressed.
Hypothesis 2: Severity of distress and turnaround potential are negatively correlated.
3.5.3 Capital structure (X3, X4)
Klarman (1991)
73
points out that financial distress in the majority of cases can be traced back
to excessive leverage. For a distressed firm reorganizing its capital structure and reducing the
debt burden might be an essential step towards a successful turnaround. With respect to
Gilson (1990)
74
an alleviation of the indebtedness can be reached by renegotiating existing
debt contracts, in such a way that the creditor offers either a composition (reduction of interest
or principal) or an extension or even an exchange of debt for equity or a combination of all
70
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
71
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies and their
implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
72
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
73
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the thoughtful investor,
Harper Business.
74
Gilson, Stuart C., (1990), “Bankruptcy, boards, banks and bondholders – Evidence on changes in corporate
ownership and control when firms default”, Journal of Financial Economics, Vol. 27, Issue 2, pp. 355-387.
31 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
three. Brown et. al (1993)
75
underscore the signaling effect of exchanges by stating that
positive information is conveyed to the market, if a firm achieves an exchange with its banks.
However, pushing through an exchange with bondholders transmits negative information, as
contrasted with banks bondholders do not have the capabilities to assess a firm’s recovery
potential. In order to get a more precise picture of the indebtedness of a firm, we focused on
the ratio total debt to total assets, which states how much of the total assets are externally
financed. A decrease in the ratio can either stem from the implementation of debt reducing
measures or from an internally financed increase of total assets. While it is obvious that
deleveraging will promote recovery from financial distress, expanding the asset base will also
foster the completion of a successful turnaround, given that the additional assets are value
creating. Also, an increase in the equity position is expected to promote a successful
turnaround, making it possible to interpret that existing shareholders believe in the turnaround
potential of the firm.
Hypothesis 3: Change in total debt to total assets and turnaround potential are negatively
correlated.
Hypothesis 4: Change in total equity and turnaround potential are positively correlated.
3.5.4 Long-term financial health (X5, X6)
Free cash-flow to total liabilities (X5)
Free cash-flow is the part of the cash-flow generated by a firm’s operational business, which
is left over after subtracting the capital expenditures that were reinvested back into the
operations. It excludes the impact of financial and non-operating items and is available to debt
holders and shareholders.
76
The ratio free cash-flow to total liabilities measures to what extent
a firm is able to cover its liabilities by means of financial funds yielded from its operations.
Sudarsanam et. al (2001)
77
employed PBITD (profit before interest taxes and depreciation) as
a cash-flow proxy. We refrained from adopting the same cash-flow proxy and relied on free
cash-flow. Unlike PBITD free cash-flow takes into consideration the cash-outflows resulting
from tax payment and reinvestment in the operational business. Therefore, we contemplated
75
Brown, David T., James, Christopher M. and Mooradian, Robert M., (1993), “The information content of
distressed restructurings involving public and private debt claims”, Journal of Financial Economics, Vol. 33,
No. 1, pp. 93-118.
76
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, p. 135.
77
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
32 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
that if a firm is in a distressed state and it strives for reorganization, expenses related to
operations will still arise and give cause for reinvestments. We measure the change of the
ratio free cash-flow over total liabilities. An increase of the ratio can either be affiliated to a
reduction of total liabilities or to an increase of free cash-flow.
Hypothesis 5: Change of free cash-flow over total liabilities and turnaround potential are
positively correlated.
Solvency ratio (X6)
We included the solvency ratio, which is defined as follows:
Non-cash expenses were added back to net income, to reflect the entire funds available for the
redemption of a firm’s liabilities. In contrast to times interest earned, the solvency ratio
incorporates also cash that stems from non-operational actions, such as asset sales. As the
occurrence of bankruptcy is often contingent on a firm’s insolvency, we perceive an increase
of the solvency ratio as a sign of financial recovery.
Hypothesis 6: Change of solvency ratio and turnaround potential are positively correlated.
3.5.5 Short-term financial health / Liquidity (X7, X8, X9)
Times interest earned (X7)
In accordance with Zeni et. al (2010)
78
we include the interest coverage ratio in our analysis.
It measures if the financial funds generated by a firm’s operations (EBIT) are sufficient to
comply with its interest charges. An increase in the ratio can originate either from a rise in
EBIT or from a decrease in interest charges. The latter can be motivated by a reduction of the
debt burden through a composition, an exchange or a debt retirement.
Hypothesis 7: Change of times interest earned and turnaround potential are positively
correlated.
78
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence
from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
33 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Quick ratio (X8, X9)
By including the quick ratio in our analysis, we bear in mind the importance of financial
liquidity for a company. The quick ratio is computed thusly:
Remaining liquid is a prerequisite for being able to continue operations, since an illiquid firm
might be confronted with the termination of its business relations with suppliers, as it cannot
make timely payments. We examine the relation between liquidity, change in liquidity and
turnaround potential. We included also change in liquidity by arguing that an increase of
liquidity will invigorate stakeholder support, as suppliers can feel more certain about the firm
meeting their financial obligations. Loyal suppliers might cause customers losing their fear
regarding the firm’s ability to carry out orders, making them to refrain from switching to
competitors.
Hypothesis 8: Quick ratio and turnaround potential are positively correlated.
Hypothesis 9: Change of quick ratio and turnaround potential are positively correlated.
3.5.6 Profitability / Efficiency (X10, X11)
Free cash-flow to sales (X10)
We apply this ratio as a profitability measure, instead of using the profit margin. In
comparison to earnings, free cash-flow provides a more undistorted measure of a company’s
profit creation, as it is not subject to estimation and judgment of the top management team.
Earnings can be forged by earnings management and accrual manipulation, with the objective
of artificially improving the firm’s profit creation.
79
Free cash-flow to sales states the amount of cash generated by a firm’s revenues after
subtracting capital expenditures. It gives an indication of a company’s proficiency to control
its cost structure. However, a low ratio cannot always be attributed to a high cost structure,
but might result from undertaken investments in e.g. new technology, which would suppress
free-cash flow downward. If the investments are value creating, cash-flows will be generated
and the ratio will be revised upwards next year, under the assumption that new investments
79
Dechow, Patricia M. and Schrand, Catherine M., (2004), “Earnings Quality”, The Research Foundation of CFA
Institute.
34 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
will start generating cash-flows one year after their implementation. We measure the change
in free-cash flow to sales, as we believe that an increase of the ratio can be effectuated either
by an improvement of the cost structure or by a cash-flow rise due to implemented value
creating investments. An increase in profitability will promote the turnaround process, as the
firm will be able to use free cash-flow for alleviating the indebtedness or for pursuing further
value creating investments.
Hypothesis 10: Change of free cash-flow over sales and turnaround potential are positively
correlated.
Operating profit margin (X11)
Hambrick et. al (1983)
80
and Robbins et. al (1992)
81
are in agreement about the necessity of
enforcing retrenchment actions to pave the way for a turnaround. We adopt operating profit
margin as a ratio of efficiency improvement through the consummation of cost-cutting
measures. It is computed as follows:
An increase in operating profit margin does most likely stem from a decrease in variable
costs, like wages and raw material prices etc. Another possible source would be an
acquisition, which would cause sales to rise at a faster pace than variable costs, due to the
realization of synergies.
Hypothesis 11: Change of operating profit margin and turnaround potential are positively
correlated.
3.5.7 I nvestments / Divestments (X12, X13, X14, X15)
Free assets (X12)
This variable is directly related to a firm’s debt capacity. Smith et. al (2005)
82
define it as
follows:
80
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
81
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
82
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
35 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The ratio indicates a firm’s amount of unutilized collateralizable assets. Thus, the higher the
ratio the more leeway the firm enjoys in taking on additional debt, as new loans could be
endowed with collaterals, permitting the creditors to seize ownership of the assets in the event
of default. This facilitates the access to external capital for the distressed firm. Moreover,
while covenants in the debt contract inhibit the sale of assets pledged as collateral, free assets
can be sold, therefore representing a possible source of cash-inflow.
Hypothesis 12: Free assets and turnaround potential are positively correlated.
Degree of Downsizing (X13)
Robbins et. al (1992)
83
perceive the enforcement of retrenchment measures as the first step of
a successful turnaround. Such measures are not constraint to headcount reduction, but involve
cost-cutting efforts and asset divestments with a view to improving efficiency and generating
cash flows. Based on Smith et. al (2005)
84
we define downsizing in the following way:
Hypothesis 13: Degree of downsizing and turnaround potential are positively correlated.
Goodwill (X14)
The position goodwill in the balance sheet statement contains the premiums paid for the
acquisitions a company has undertaken.
85
We refer to change in goodwill as a measure of
strategic asset investment/divestment. Goodwill impairments and amortizations were added
back to avoid making wrong inferences about asset divestments that did not occur. Thus,
changes in the goodwill position reflect acquisitions and divestments, respectively. Since
several researchers suggest that a firm should consider strategic asset investments in the
83
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
84
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
85
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, p. 141.
36 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
turnaround process, in order to adjust its asset portfolio to the evolving environment
86
, we
investigate the relationship between goodwill increases and turnaround potential.
Hypothesis 14: Change in goodwill and turnaround potential are positively correlated.
R&D expenses (X15)
In regard to Hambrick et. al (1983)
87
, we include R&D expenses as a measure of a firm’s
concentration on the development of new products, which shall promote the company in the
process of strategic reorientation. However, we argue that constantly interpreting investments
in R&D as a strategic measure would be wrong. A firm might expend R&D effort in
developing new technologies or in designing innovative ways to organize and manage the
process cycles, both of which would rather be related to an efficiency improvement than to a
strategic reorientation. In addition, we decided not to focus on the ratio of R&D to sales, as
has been done by Hambrick et. al (1983). Instead, we measure the change in R&D expenses
over a one-year period. That way, we prevent the emergence of wrong conclusions about a
firm’s R&D policy, as a decline in the ratio could be explained by a cutback in R&D expenses
or by an increase in sales, or by both.
Hypothesis 15: Change in R&D and turnaround potential are positively correlated.
3.5.8 Management Expertise (X16)
ROE (X16)
For a successful turnaround it is crucial that stakeholders and shareholders believe in the
incumbent management team’s ability to steer the company out of distress. Otherwise, they
will refrain from promoting the turnaround attempt, putting at risk the firm’s recovery.
Zeni et. al (2010)
88
included ROE (Return on equity) as a measure of top management
expertise in their Z-score, which they developed for the Malaysian market.
We apply ROE as an indicator of the top management team’s capability to initiate a process
of recovery from distress and thereby ensure stakeholder and shareholder support.
86
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 186.
87
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
88
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence
from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
37 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Hypothesis 16: ROE and turnaround potential are positively correlated.
3.5.9 Overview of examined variables
The table below provides an overview of the variables taken into account in our empirical
study. It also shows the encoding used in the statistical programs SAS and SPSS for each
predictor. Besides, it states the scale unit employed to each variable and points out which
empirical research motivated the inclusion of the variable.
I. Size Encoding Empirical support
Total tangible assets X1
White (1989)
Smith et. al (2005)
II. Severity of distress
Z-score X2 Sudarsanam et. al (2001)
III. Capital structure
? total debt/total assets X3
Gilson (1990)
Klarman (1991)
? total equity X4 Klarman (1991)
IV. Long-term financial health
? FCF/total liabilities X5
Sudarsanam et. al (2001)
Own intuition
? Solvency ratio X6 Own intuition
V. Financial health/ Liquidity
? Times interest earned X7 Zeni et. al (2010)
Quick ratio X8 Own intuition
? Quick ratio X9 Own intuition
VI. Profitability/Efficiency
? FCF/Sales X10 Goumas et. al (2011)
? EBIT/Sales X11
Hambrick et. al (1983)
Robbins et. al (1992)
Chowdhury et. al (1996)
VII. Investments/Divestments
Free assets X12 Smith et. al (2005)
Degree of downsizing X13
Robbins et.al (1992)
Smith et. al (2005)
? Goodwill X14
Hofer (1980)
Grinyer et. al (1988)
Sudarsanam et. al (2001)
? R&D expenses X15
Hambrick et. al (1983)
Goumas et. al (2011)
VIII. Management Expertise
ROE X16
Abdullah et. al (2008)
Zeni et. al (2010)
Table 8: Overview of examined variabels
? = change no ? = no change (the variable was taken into account, instead of the change in the variable) yoy = year on
year
38 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table 9 summarizes the examined hypotheses with respect to each variable.
I. Size Encoding Correlation with turnaround potential
Total tangible assets X1 +
II. Severity of distress
Z-score X2 -
III. Capital structure
? total debt/total assets X3 -
? total equity X4 +
IV. Long-term financial health
? FCF/total liabilities X5 +
? Solvency ratio X6 +
V. Financial health/ Liquidity
? Times interest earned X7 +
Quick ratio X8 +
? Quick ratio X9 +
VI. Profitability/Efficiency
? FCF/Sales X10 +
? EBIT/Sales X11 +
VII. Investments/Divestments
Free assets X12 +
Degree of downsizing X13 +
? Goodwill X14 +
? R&D expenses X15 +
VIII. Management performance
ROE X16 +
Table 9: Summary of hypothesis formulation
3.6 Methodology
In order to investigate which of the outlined independent variables are most qualified to
separate our sample into turnarounds and non-turnarounds, we run a linear discriminant
analysis (LDA) and a logistic regression (LOGIT) on our sample data. Both methods will
enable us to test the developed hypotheses and create a discriminant function (DF) and
logistic function. They act as a prediction model, facilitating the categorization of financially
distressed firms into turnarounds and non-turnarounds, based on a cut-off point. The DF
obtained from the LDA and the logistic function given by LOGIT will be applied to a holdout
sample for reasons of back-testing their prediction accuracy.
The statistical programs employed on the sample data were SAS (with respect to LDA) and
SPSS (with respect to LOGIT).
39 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.6.1 Linear Discriminant analysis (LDA)
LDA aims at ascribing an unknown subject (e.g. financially distressed firm) to one of two
groups (e.g. turnarounds; non-turnarounds)
89
, with the aid of discriminating variables
(explanatory variables).
The DF is expressed by the following equation:
Where, Z = discriminant score
? = constant term
? = discriminant coefficients
X = discriminating variables
Discriminating
Variables
I
m
p
a
c
t
Financially
Distressed
Firms
Non-
turnarounds
(0)
Turnarounds
(1)
Graph 5: The discriminating process in the linear discriminant analysis
The discriminating process is described in the graph above. The dependent variable (e.g.
turnaround outcome of financially distressed firms) is categorical and the two groups must be
definite distinguishable from each other i.e. they need to be mutually exclusive.
89
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No. 1, pp. 69-85.
40 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The discriminating variables, which are considered to be suitable for differentiating the two
groups from each other, are chosen according to their ability to maximize the distance
between the means of the probability distributions of the two groups and included in the DF.
90
Graph 6: Probability distributions of turnarounds and non-turnarounds and Mahalanobis-Distance
?
0
= Mean of non-turnaround distribution based on DF ?
1
= Mean of turnaround distribution based on DF
C = Cut-off point
The probability distributions of the two groups are depicted in graph 6. LDA assumes that the
explanatory variables follow a normal distribution, having equal variances and covariances.
91
The distance between the means of the two groups is called Mahalanobis-Distance and
calculated thusly:
S
2
= pooled sample variance
With an increasing D
2
the overlapping area of the normal distributions becomes smaller,
enabling an almost unambiguous differentiation between the two groups.
90
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No. 1, pp. 69-85.
91
Cox, D. R. and Snell, E.J., (1989), “The analysis of binary data”, 2
nd
Edition, Chapman and Hall.
?
0
?
1
C
Probability distribution of
non-turnarounds
Probability distribution of
turnarounds
Mahalanobis-Distance
41 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 7: Overlapping of the probability distributions of two populations I and II
Graph 7 depicts an example of a small Mahalanobis-Distance between two population means,
giving rise to a large overlapping area and increasing the probability of misclassification. In
most cases, a perfect seperation of two populations cannot be achieved, so that the tails of the
distributions will cross, leading to the emergence of type I and type II errors (misclassification
errors).
92
In our particular study, a type I error corresponds to the classification of a non-
turnaround as a turnaround and a type II error is equivalent to the classification of a
turnaround as a non-turnaround.
So as to be able to categorize an observation in one of the two groups, a cut-off point needs to
be determined. Graph 6 shows that the seperation line is located where the tails of the two
groups’ probability distributions cross. Thus, assuming a normal distribution, the cut-off point
is given by the formula:
For a new observation X the discriminant score must be calculated on the basis of the DF. If
the discriminant score lies below the cut-off point, the observation is classified into group I
and vice versa.
92
Lachenbruch, P. A., (1968), “On expected probabilities of misclassification in discriminant analysis, necessary
sample size, and a relation with the multiple correlation coefficient”, Biometrics, Vol. 24, No. 4.
I +I I
I I I
42 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 8: Classification of a new observation X
Graph 8 visualizes the classification of a new observation X in one of the two groups. The red
circle depicts the discriminant score of the observation X. As the score is smaller than the cut-
off point C, the observation is categorized into the group of non-turnarounds. Another
approach is to argue that the discriminant score of X is closer to the mean of the probability
distribution of non-turnarounds than to the mean of the probability distribution of
turnarounds, leading to a classification into the first group.
As already stated, the independent variables involved in the DF are the most qualified for
discriminating the two groups from each other. The graph below shows an example for two
independent variables, applied to classify financially distressed firms into turnarounds and
non-turnarounds.
S
e
v
e
r
i
t
y
o
f
D
i
s
t
r
e
s
s
Size
Graph 9: Classification of financially distressed firms into turnarounds and non-turnarounds based on size and severity of
distress
Probability distribution of
non-turnarounds
Probability distribution of
turnarounds
C ?
1
?
0
X
43 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
For this example the turnarounds are depicted by the black triangles and the non-turnarounds
correspond to the red circles. It is apparent that the non-turnarounds are located in the
northwestern part of the scatterplot, while the turnarounds lie southeasterly. The blue line,
seperating the observations of the two groups, is called the linear discriminant boundary and
is described by the following formula:
93
This formula complies with the DF formula. An observation that is situated on the
discriminant boundary has a discriminant score equal to zero. As a consequence, observations
located on one side of the boundary will be distinguished from observations located on the
other side by having an opposite sign in the discriminant score. This is ensured by including
the constant term ? in the formula. The value of the discriminant score indicates the distance
of the observation from the discriminant boundary.
94
We used SAS to conduct the LDA. The program offers different approaches to create a subset
of explanatory variables (predictors) out of an initial set of variables. The most important
approaches are described briefly.
Means and correlation procedure:
SAS provides an overview of the means and the standard deviations for each variable that is
entailed in both groups (0, 1). Taking e.g. the variable Size, the mean ?(Size
0
) and the standard
deviation ?(Size
0
) are compared with the mean ?(Size
1
) and the standard deviation ?(Size
1
).
The higher the difference in the two means and the lower the intra-group standard deviation,
the better the variable Size discriminates between the two groups. Given close to each other
located means and high intra-group standard deviation, increases the chance of overlapping
probability distributions, producing large misclassification errors.
Stepdisc Forward Variables Selection
At the beginning, no variable is included in the model. Then, the variable exhibiting the
highest discriminatory power is selected. In the following steps, the variables that paired with
the initial variable lead to the highest increase of the model’s discriminatory power are
93
Cooper, Ron A. and Weekes, Tony J., (1983), “Data, Models and statistical analysis”, Philip Allan Publishers
Limited, New Jersey, USA, p. 280-285.
94
Ibid.
44 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
included. The selection process stops as soon as no further significant increase in the model’s
discriminatory power is achieved.
95
Stepdisc Backward Variables Selection
This selection procedure is equal to an elimination process. Initially, all variables are included
in the model. Then, the variables having the smallest impact on the model’s discriminatory
power are excluded. In this manner, the process ensures that only the best predictors are kept
in the DF.
96
3.6.2 Binary Logistic regression model (Binary LOGI T)
Binary LOGIT is an alternative to LDA when the dependent variable is dichotomous. In
contrast to LDA it does not require the explanatory variables to be normally distributed and
have equal variance and covariance, making it a more flexible and robust model.
97
As
financial data does not follow a normal but rather a leptokurtic distribution, binary LOGIT
appears to be more appropriate for the underlying study.
98
In practice, binary LOGIT is applied in many different fields, to determine the explanatory
variables that cause a separation of two groups from each other. One example is the
application of binary LOGIT in medical science to determine the factors for predicting the
emergence of heart diseases. The dependent variable is binary and comprises the two
outcomes i. heart disease and ii. no heart disease, which are equivalent to the two groups.
The explanatory variables (predictors) that allow for a classification of a patient in one of the
two groups would be e.g. age, weight, blood pressure, smoking habits etc.
99
The explanatory variables can be of quantitative or binary character, or a mixture of both. The
binary variable, whether dependent or explanatory, is encoded by the use of 0 and 1, where 0
denotes the absence of a situation and 1 denotes the presence of a situation respectively.
Thus, the binary logistic regression aims at explaining differences between two groups on the
basis of a common set of variables. It identifies the explanatory variables, which are most
95
SAS Institute Inc. 2010. SAS/STAT® 9.22 User’s Guide. Cary, NC: SAS Institute Inc.
96
Ibid.
97
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition, John Wiley &
Sons, Inc.
98
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press.
99
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
45 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
qualified to discriminate between two groups, as well as their direction and intensity of impact
on the respective group.
100
To give an example, based on our study the binary logistic regression extracts the relevant
variables out of a set of fifteen variables, which best separate our sample into turnarounds and
non-turnarounds. This allows us to make inferences about the decisive drivers in the
turnaround process.
The logistic function is given by:
101
Where, Y = dependent binary variable
? = Probability of the outcome of category I e.g. turnaround
1-? = Probability of the outcome of category II e.g. non-turnaround
X = explanatory variable
? = intercept
? = regression coefficient
u = random disturbance term
The above equation shows that binary LOGIT predicts “the probability that a case will be
classified into one as opposed to the other of the two categories of the dependent variable”.
102
This is known as the odds ratio, which can be expressed as follows:
Where, P (Y = 1) = Probability of Y = 1
1 – P (Y = 1) = Probability of Y ? 1
The probability of turnaround is defined by:
This equation is known as the logistic regression equation.
103
100
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
101
Peng, Chao-Ying J., Lee, Kuk L. and Ingersoll, Gary M., (2002), “An Introduction to Logistic Regression
Analysis and Reporting”, The Journal of Educational Research, Vol. 96, No. 1, pp. 3-14.
102
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications, Inc.
46 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Taking the natural logarithm of the odds ratio answers the purpose of constraining the
estimated probability within the boundaries of 0 and 1.
104
In this way, values for the
dependent variable will lie between 0 and 1, as shown in the graph below.
Graph 10: Logistic curve model for a dichotomous dependent variable
105
For the logistic regression, SPSS sets the cut-off point automatically at 0.5. With respect to
our study, we encoded turnarounds as ones and non-turnarounds as zeros. Thus, turnarounds
should yield a score above 0.5 and non-turnarounds below 0.5 respectively, on condition that
they are correctly classified. If graph 10 was the classification result of the binary LOGIT of
financially distressed firms based on an explanatory variable e.g. size, according to the
beforehand mentioned encoding, the four dots above 0.5 would match firms classified as
turnarounds and the three dots below 0.5 would correspond to firms classified as non-
turnarounds.
There exist two stepwise procedures to extract the most qualified predictors for discriminating
between the two categories, out of a set of explanatory variables.
Forward Conditional Logistic Regression:
In the beginning block (step 0) no explanatory variable is included in the model, but only the
intercept. Then, explanatory variables are entered in a stepwise procedure. First, the
explanatory variable with the highest discriminatory power is included into the model, which
103
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
104
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications, Inc.
105
47 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
is consistent with the variable that has the highest statistically significant chi-square.
106
This
variable causes “the greatest change in the log-likelihood relative to a model not containing
the variable”
107
. The quality of fit of the model is indicated by the deviance (-2 Log
likelihood). A decreasing deviance indicates that the model fits the data well.
108
This process
is repeated until no further improvement in the model can be obtained, as no more statistically
significant chi-squares are computed.
109
As including additional predictors to the model
causes an upward bias of the goodness of fit measure, the model is penalized by an increase in
degrees of freedom.
110
Backward Conditional Logistic Regression:
The backward conditional logistic regression includes all of the variables in the beginning
block (step 0) and stepwise removes variables, which are estimated as statistically
insignificant. These are the explanatory variables, which demonstrate the largest p-value in
terms of the likelihood ratio chi-square test.
111
It stops removing variables when all of the
remaining predictors show a statistically significant contribution to the model.
112
106
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
107
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition, John Wiley &
Sons, Inc.
108
Ibid.
109
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
110
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press.
111
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
112
SPSS Regression 17.0
48 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 4
Empirical Findings
In this chapter the empirical findings of our study are presented and discussed. The results
obtained by the two approaches, LDA and LOGIT, are summarized and inferences about the
factors playing a decisive role in the turnaround process are made. In addition, the findings
provide answers regarding the rejection or non-rejection of the hypotheses developed in the
previous chapter. The performance of the models is assessed based on their forecasting
accuracy with respect to the in-sample data. The two models showing the highest in-sample
prediction accuracy are evaluated based on their prediction performance on a holdout sample.
Furthermore, the predictors proposed by the model presenting the best forecasting accuracy
are tested for normality and heteroskedasticity, thereby making allowances for possible
violations of the assumptions underlying LDA and LOGIT.
4.1 Initial situation
The time frame for investigating the impact of the explanatory variables on the turnaround
outcome embraced year two and three of the four year window, which was exhibited in graph
2. We also took into account the period spanning year one and two, the two years for which
all of the companies in our sample displayed a Z-score below 1.8. However, for this time
period the degree of discrimination between the two groups was very low, leading to large
misclassification errors. Below the time period of analysis is depicted graphically.
1991 1995
1992 1993 1994
1992 - 1993
LDA & LOGIT on year 2 & 3
Graph 11: Time period of analysis regarding the impact of the explanatory variables on the turnaround outcome
49 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.2 Results of LDA
The LDA was performed by application of the statistical program SAS. The first step
consisted of extracting the most qualified predictors out of our set of explanatory variables,
followed by the derivation of the DF and computation of the cut-off point. SAS offers four
approaches for determining the variables that discriminate best between two groups:
i. Means and correlation procedure
ii. Stepdisc Forward Variables Selection
iii. Stepdisc Backward Variables Selection
iv. Stepdisc Stepwise Variables Selection
All four procedures were taken into account. The results provided by each procedure are
presented hereafter.
4.2.1 Means and correlation procedure (Model I )
The table below was generated by SAS and includes means and standard deviations for both
groups and for each variable taken into consideration in our study.
Variable Mean Stand. Dev. Mean Stand. Dev.
X1 7.647,47000 24.791,22000 479,80522 1.100,58000
X2 0,07189 2,24010 4,00548 2,80354
X3 0,00905 0,10599 -0,09743 0,16494
X4 0,02040 0,49043 0,67246 1,08285
X5 -0,00299 0,16569 0,02719 0,27306
X6 -0,00204 0,44957 0,28519 0,63444
X7 -0,08362 1,00394 0,07012 1,14753
X8 1,00256 1,08011 1,53116 1,19166
X9 -0,04107 0,32322 0,11107 0,39723
X10 -0,04533 0,22052 0,01140 0,24605
X11 0,05215 0,28629 0,05551 0,43024
X12 -0,03622 0,14050 -0,00097 0,10077
X13 0,03945 0,26643 0,15411 0,45012
X14 122,95094 573,49013 28,72262 175,23403
X15 -0,24562 18,23560 -0,99882 9,24052
X16 -0,05992 0,41028 0,02149 0,56350
Status 0 Status 1
Table 10: Means procedure
Based on this table and on intuition a mean procedure prediction model was established. The
variable X2 in table 10, equivalent to the Z-score, has a large difference in the means between
50 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
the two groups. The intra-group standard deviation is almost equal for both categories. Thus,
we included X2 as a predictor that allows for discrimination between the two groups.
However, also X1 was included in our mean procedure prediction model, although the
variable shows an extremely high intra-group standard deviation for both groups.
Nevertheless, based on our intuition and the theoretical foundation presented in chapter 2, we
believe that it is a decisive factor in the turnaround process, enabling a separation between
turnarounds and non-turnarounds. We also examined the correlation matrix generated by SAS
without finding significant correlations among the included variables. Several mean procedure
prediction models were tested, covering various combinations of the explanatory variables. In
the following, the mean procedure prediction model that revealed the highest in-sample
forecasting accuracy is demonstrated. The DF of this model is given by:
X
1
= total tangible assets
X
2
= Z-score
X
3
= change in total debt/total assets
The relationship between the predictors suggested by the means and correlation procedure and
the categories turnarounds and non-turnarounds is displayed in table 11.
Variable Coefficients (0) Impact direction Coefficients (1) Impact direction
X1 0,0000216 + -5,00E-06 -
X2 -0,0006751 - 0,6262 +
X3 0,41596 + -4,38467 -
Table 11: Relation between predictors and categories for X1 X2 X3
The table corroborates the discriminating power of the selected predictors, as their attached
coefficients have an opposite sign for each of the two categories. According to the model,
there exists a positive relation between size and non-turnarounds, while size and turnarounds
are negatively related. Thus, the model suggests that smaller firms have a higher probability
to be successful in the turnaround process. Even though the coefficient size is close to zero, it
has a significant impact on the Z
CEGA
, considering that the variable size includes large
numerical values (firm’s total assets). Moreover, firms for which the state of distress is less
severe, as measured by the Altman Z-score (X2), are more likely to achieve a turnaround. The
last variable considered to be decisive in the turnaround process is the change in the ratio total
debt to total assets (X3). According to the model, a decrease in the ratio is associated with a
successful recovery from financial distress. This empirical result was expected intuitively, as
51 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
financial distress is traced back to excessive leverage, in most of the cases. The magnitude of
the coefficient underpins the importance of the variable in the turnaround process, stating that
a reduction of X3 will trigger a substantial increase in the discriminant score Z
CEGA
.
We computed Z
CEGA
scores for the firms in our in-sample and calculated the cut-off point,
which accounted for – 0.00016632. The prediction accuracy of the model was estimated based
on our in-sample, which involved 150 firms. The classification matrix is provided below.
Status 0 1 Total
0
83
96,51%
3
3,49%
86
100%
1
8
12,50%
56
87,50%
64
100%
Total 91 59 150
Table 12: Classification Matrix in-sample X1 X2 X3
The blue shaded areas in the classification matrix display the correctly categorized firms per
outcome. The prediction accuracy on the in-sample amounted to 92.7%, which is computed
by taking the sum of the correctly classified firms divided through the total number of firms.
Misclassifications were restricted to 3 type I errors and 8 type II errors.
4.2.2 Stepdisc procedures (Model 2)
All of the three stepdisc procedures selected the same variables for inclusion in the final DF.
Thus, the DF on the basis of the stepdisc procedures is given by:
X
1
= total tangible assets
X
2
= Z-score
X
3
= change in total debt/total assets
X
6
= change in solvency ratio
Table 13 provides an overview of the coefficients belonging to the probability distributions of
turnarounds and non-turnarounds. They indicate direction and size of the impact of the
respective explanatory variables on the turnaround process.
52 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Coefficients (0) Impact direction Coefficients (1) Impact direction
X1 0,0000216 + -4,77E-06 -
X2 -0,0007326 - 0,62249 +
X3 0,42103 + -4,05877 -
X6 0,01242 + 0,79903 +
Table 13: Relation between predictors and categories for X1 X2 X3 X6
The first three variables are equivalent to the variables chosen by the means and correlation
procedure and their attached coefficients coincide in sign and are of similar magnitude.
However, the stepdisc procedures selected X6 as an additional variable, which corresponds to
the change in the solvency ratio. Although the coefficients of the predictor X6 are positive for
both groups, the coefficient belonging to turnarounds is of larger numerical value, leading to
the conclusion that companies with an increase in the solvency ratio are more likely to belong
to the group of turnarounds.
The cut-off point for the DF of the second model was computed based on the Z
CEGA
discriminant scores of the in-sample firms and amounted to -0.000165657. Below, the
classification matrix for in-sample data is displayed.
Status 0 1 Total
0
83
96,51%
3
3,49%
86
100%
1
10
15,63%
54
84,37%
64
100%
Total 93 57 150
Table 14: Classification Matrix in-sample X1 X2 X3 X6
The in-sample forecasting accuracy of model 2 adds up to 91.33%, falling short of the
prediction performance of model 1 by only 1.37%.
4.3 Results of LOGIT
The LOGIT was conducted by use of SPSS 17.0. Two procedures were applied to obtain the
independent variables acting as predictors, so that two LOGIT functions were computed. The
models were assessed regarding to their forecasting accuracy relative to the in-sample data.
The procedures were performed at different confidence intervals, which varied from 95% to
85%, taking into account that additional variables might be viewed as significant for
differentiating into the two groups, given lower confidence intervals. However, no additional
53 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
variables were included in the LOGIT function at lower confidence intervals. Hence, the
presented results were obtained at a confidence interval of 95%.
We excluded the variable severity of distress (X
2
) from the set of potential discriminating
variables. The LOGIT is based on a different algorithm than the LDA. It computes the
probabilities of a company, being classified in one of two groups. As our definition of a
turnaround was connected to a firm’s Z-score, inserting the variable severity of distress
(measured by Z-score) as a potential discriminator between turnarounds and non-turnarounds
into the logistic algorithm of SPSS, yielded a logistic function consisting solely of the Z-score
and displaying a prediction accuracy of 100%. Since this result is flawed, because of
conformity between the defining variable and a potentially predictive variable, we conducted
the LOGIT without inclusion of the Z-score.
4.3.1 Forward Conditional Logistic Regression (Model 3)
This procedure created a subset of three variables, which it regarded as eligible for
categorizing financially distressed firms into turnarounds and non-turnarounds.
The LOGIT function is given by:
X
3
= change in total debt/total assets
X
4
= change in total equity
X
6
= change in solvency ratio
When interpreting the regression coefficients, it is important to bear in mind that the above
LOGIT function complies with the log-odds of turnarounds.
Referring to the ratio total debt to total assets (X
3
), LOGIT estimates a negative correlation
between the predictor and the outcome turnaround. Hence, an increase in the ratio will lower
the chances for turnaround, while a decrease will promote recovery from financial distress. As
opposed to this, the model suggests that change in solvency ratio (X
6
) is positively related
with the probability of turnaround. The predictors X
3
and X
6
are represented in at least one of
the two LDA-based models. Change in total equity (X
4
), not taken into account by LDA, is
expected to be positively connected with turnaround. Raising new equity will trigger an
increase of the turnaround likelihood, as it reshapes the capital structure and allocates
financial funds, which are disposable for deleveraging or investments in value-creating
projects.
54 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The cut-off point for the LOGIT was set automatically by SPSS at 0.5. For each firm the
probability of turnaround is computed by means of the values of the predictors. A company is
classified as a turnaround, if the computed probability exceeds 0.5. The in-sample
classification matrix is provided below.
Status 0 1 Total
0
77
89.5%
9
10.5%
86
100%
1
26
40.6%
38
59.4%
64
100%
Total 103 47 150
Table 15:Classification Matrix in-sample X3 X4 X6
The in-sample forecasting accuracy of the model suggested by LOGIT forward amounts to
76.67%, being around 16% lower than the prediction accuracy given by LDA-model 1.
4.3.2 Backward Conditional Logistic Regression (Model 4)
The backward procedure included four variables in the LOGIT function. In addition to the
three variables considered by the forward procedure, it also took into account total tangible
assets (X
1
).
X
1
= total tangible assets
X
3
= change in total debt/total assets
X
4
= change in total equity
X
6
= change in solvency ratio
According to the model, total tangible assets negatively influence the turnaround likelihood,
allowing us to infer that larger firms are less likely to recover from financial distress than
smaller firms. The same conclusion was drawn from the LDA models. In respect of the other
three variables, the relationship assumed by model 4 coincides with the relationship that
model 3 predicted. To grasp the prediction accuracy of model 4, an in-sample classification
matrix was generated.
55 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Status 0 1 Total
0
72
84%
14
16%
86
100%
1
20
31%
44
69%
64
100%
Total 92 58 150
Table 16:Classification Matrix in-sample X1 X3 X4 X6
The model’s forecasting accuracy accounts for 77.33%. Thus, the inclusion of total tangible
assets as an additional predictor led to a marginal improvement of the prediction power by 66
basis points. However, also the second LOGIT model performs significantly poorer than the
LDA models in categorizing financially distressed companies into turnarounds and non-
turnarounds.
4.3.3 Model comparison, selection and interpretation of results
The table below provides an overview of the in-sample prediction accuracy and the number of
misclassifications of all four models.
Model 1 Model 2 Model 3 Model 4
Forecasting accuracy 92,70% 91,33% 76,67% 77,33%
Type I errors 3 3 9 14
Type II errors 8 10 26 20
LDA LOGIT
Table 17:Comparison of in-sample forecasting accuracy among models
The first LDA model shows the highest forecasting accuracy and makes the fewest
misclassification errors, followed by the second LDA model. The LOGIT models are inferior
to the LDA models, making three to five times more type I errors and two to three times more
type 2 errors, which leads to a much lower forecasting accuracy.
As the performance of the two LDA models is almost equally good, we examine their ability
to make correct forecasts by means of a holdout sample, which comprises 3140 companies.
The classification matrices for both LDA models are displayed below.
56 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Status 0 1 Total
0
2606
88,85%
327
11,15%
2933
100%
1
17
8,21%
190
91,79%
207
100%
Total 2623 517 3140
Table 18: Classification Matrix Holdout Sample X1 X2 X3
Status 0 1 Total
0
2481
84,59%
452
15,41%
2933
100%
1
49
23,67%
158
76,33%
207
100%
Total 2530 610 3140
Table 19: Classification Matrix holdout sample X1 X2 X3 X6
LDA model 1 has a prediction accuracy of 89% in the holdout sample. Type I and type II
errors increased, as did the sample size. The considerable rise in type I errors is explained by
the boost of non-turnarounds from 86 in the in-sample to 2933 in the holdout sample.
The second LDA model classifies only 84% of the cases correctly. Hence, model 2
misclassifies significantly more firms than model 1, especially in terms of type II errors,
which occur almost three times as often as in the first model. On these grounds, LDA model 1
was chosen as the best model for discriminating between turnarounds and non-turnarounds.
57 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The two graphs below visualize the model’s forecasting accuracy for the in-sample and the
holdout sample. Considering that the cut-off point of the LDA-model 1 was close to zero, the
two graphs display very well the amount of committed type 1 and type 2 errors.
Graph 12: Z
CEGA
scores for turnarounds and non-turnarounds
For the in-sample, all Z
CEGA
scores of the type 1 errors were below 0.5, not so far away from
the cut-off point. The most external type 2 errors were lying between -2 and -1. Allowing for
a margin of safety, firms displaying Z
CEGA
scores larger than 1, can be perceived as real
turnaround candidates, with respect to the in-sample.
Graph 13: Z
CEGA
scores for turnarounds and non-turnarounds
For the holdout sample the most external type 2 errors were again laying between -2 and -1.
However, the type 1 errors were more dispersed with some exhibiting a Z
CEGA
score larger
0%
5%
10%
15%
20%
25%
30%
35%
40%
Failed Turnaround
Successful turnaround
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Failed Turnaround
Successful turnaround
58 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
than 2. Thus, although the probability of selecting a non-turnaround decreases with an
increasing Z
CEGA
score, the risk cannot be ruled out completely.
4.4 Model interpretation
Since this model contained only the three variables total tangible assets (X
1
), the Z-score as a
measure of distress severity (X
2
) and the ratio total debt to total assets (X
3
), no direct
relationship could be identified between the dependent variable and the remaining thirteen
explanatory variables outlined in table 8. Hence, the model assumes that they are statistically
insignificant in determining the turnaround potential of a financially distressed company,
leading to a rejection of the developed hypotheses for these particular variables. Table 20
displays the proposed impact of the three explanatory variables, which entered the final DF of
model 1 and compares it to the impact revealed by the model.
Variable Expected impact Revealed impact Developed hypothesis
X1 + - rejected
X2 - - not rejected
X3 - - not rejected
Table 20: Comparison of expected and revealed impact of predictors
4.4.1 Size (X
1
)
With respect to firm size, measured by total tangible assets, we developed the hypothesis that
the larger the firms the higher their turnaround potential. We motivated our hypothesis by
stating that larger firms can easier access capital markets to raise external funds. Moreover
they dispose of the possibility to sell assets, such as e.g. unrelated business units, to generate
internal funds. However, our hypothesis was rejected, as the model reveals that smaller
companies are more likely to succeed in the turnaround process. A reason for that can be their
swiftness in implementing strategic changes, as was argued by Paint (1991)
113
. Another
possible argument would be that small companies are accustomed to share a closer
relationship with its stakeholders, because there are less hierarchical levels, which obstruct the
flow of communication and information. As a consequence, it might be easier for smaller
firms to assure stakeholder support compared to larger firms.
4.4.2 Severity of distress (X
2
)
Measuring the severity of distress by the Z-score is reasonable, as it indicates the firm’s
probability of experiencing bankruptcy. Less severe distressed firms face fewer hurdles when
113
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in corporate
turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
59 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
approaching capital markets and can easier convince stakeholders to support them in the
restructuring process as firms in a more severe state of distress. The model did not reject our
developed hypothesis, so that severity of distress is negatively related with turnaround
potential.
In order to understand the concept of severity of distress, one has to keep in mind that it is
measured by the Altman Z-score. For a company displaying a high Z-score, the state of
distress is less severe than for a company with a low Z-score. This implies a negative
correlation between Z-score and severity of distress. The coefficients in table 11 for X
2
refer
to the Z-score. According to them, turnarounds (1) and Z-score are positively correlated,
while non-turnarounds (0) and Z-score are negatively correlated, leading to a negative
relationship between turnaround potential and severity of distress. The model suggests:
i. The higher the Z-score is, the more likely the turnaround outcome. As a high Z-
score implies low severity of distress, the model suggests a negative relation
between turnaround potential and severity of distress.
ii. The lower a firm’s Z-score is, the more likely the non-turnaround outcome. As a
low Z-score implies high severity of distress, the model suggests a negative
relation between turnaround potential and severity of distress.
4.4.3 Total debt to total assets (X
3
)
The model estimated a negative relation between change in the ratio total debt to total assets
and turnaround potential, corroborating our hypothesis. Both, deleveraging and expansion of
the asset base are assumed to support a firm in the process of recovery. Debt reduction will
send out a positive signal, strengthening stakeholder support. Expanding the asset base by
undertaking value-creating investments is an important step in the restructuring process of a
firm striving for strategic reorientation and will generate future cash-flows, available for
further investments or deleveraging.
Hence, we believe that changes in this ratio can reveal the implementation of efficiency-
oriented or entrepreneurial-oriented strategies, or a combination of both. Therefore we
scrutinized the in-sample firms to gauge what drives changes in total debt to total assets.
60 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Turnarounds Non-turnarounds Total
+?total assets 42 48 90
-?total assets 22 38 60
Total 64 86 150
+?total debt 17 43 60
-?total debt 47 43 90
Total 64 86 150
+?sales 50 54 104
-?sales 14 32 46
Total 64 86 150
Table 21: ? Total assets ? Total debt ? total sales
The above table depicts the changes in total assets, total debt and total sales for all firms in the
in-sample, separated by turnarounds and non-turnarounds. We also focus on change in total
sales, because we suppose that an increase in the asset base will trigger a rise in total
revenues, given that value-creating investments have been undertaken. The rows highlighted
in red depict significant differences between turnarounds and non-turnarounds. Considering
that the in-sample consists of 43% turnarounds and 57% non-turnarounds, a difference of a
factor larger than 1.5 between the two groups with respect to a variable was regarded as
significant. It appears that non-turnarounds are more prone to reducing the asset base than
turnarounds. Regarding total debt, turnarounds rather refrain from approaching external
financer and centre on deleveraging. For non-turnarounds the situation is different, with half
of them increasing the debt level and the other half decreasing it.
Aiming at obtaining a more concrete picture about the strategies employed by the firms in our
sample, we examined the direction of change in the three variables stated in table 21 for each
single company and formulated eight strategies, which were rated as efficiency-oriented,
entrepreneurial-oriented or a combination of both and are displayed in graph 14. However,
none of the strategies could be identified as being only efficiency-oriented.
61 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 14: Overview of developed strategies
Table 22 presents a summary of the classification of each in-sample company to a strategy
group.
Strategies Turnaround Non-turnaround Total
1 5 12 17
2 2 5 7
3 2 5 7
4 10 23 33
5 25 10 35
6 3 10 13
7 4 5 9
8 13 16 29
Total 64 86 150
Table 22: Employed strategies by financially distressed firms
• Decrease in total assets, total
sales, total debt
• combination of both
Strategy 1
• Decrease total assets, total
sales and increase total debt
• entrepreneurial-oriented
Strategy 2
• decrease total assets and
increase total sales, total debt
• combination of both
Strategy 3
• increase total assets, total sales,
total debt
• entrepreneurial-oriented
Strategy 4
• increase total assets, total sales
and decrease total debt
• entrepreneurial-oriented
Strategy 5
• increase total assets, total debt
and decrease total sales
• entrepreneurial-oriented
Strategy 6
• increase total assets and
decrease total sales, total debt
• entrepreneurial-oriented
Strategy 7
• decrease total assets, total debt
and increase total sales
• combination of both
Strategy 8
62 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Four of the eight strategies were highlighted due to their frequency in appearance among both
groups and the fact that they were applied more often by one group than the other, rendering
possible to make conclusions about differences in the strategies pursued by turnarounds and
non-turnarounds.
Strategy 1:
Compliant with our sample this strategy was implemented by 12 non-turnarounds and 5
turnarounds. It is a combination between efficiency-oriented and entrepreneurial-oriented
strategies, as asset divestments can be operational or strategic or both. Hence, turnarounds and
non-turnarounds engaged in operational and/or strategic asset divestments, freeing up
financial funds that can be used for deleveraging. Simultaneously, divestments led to
reductions in turnover, due to decrease of capacitance and/or exit of business segments.
For non-turnarounds following this strategy, recovery might fail because the distress state is
too severe, so that they are forced to sell the most profitable segment, as was proposed by
Schlingemann et. al (2002)
114
. Even though this will cause a unique cash-inflow, the sale of
the most profitable segment will deteriorate the firm’s ability to generate turnover in the
future.
Firms succeeding with this strategy, rather clean up their portfolio by divesting unrelated
businesses and focusing on core capabilities.
Strategy 4:
This strategy was adopted twice as often by non-turnarounds than by turnarounds with respect
to our sample. It is denoted as an entrepreneurial-strategy, where strategic asset investments
are undertaken to re-orientate in the market. The strategic reorientation is financed with
outside capital and leads to sales increases, as value-creating investments are assumed to be
realized.
As contrasted with turnarounds, non-turnarounds sticking to this strategy might fail in the
event that the rise in sales provided by new investments cannot absorb the aggravated gearing,
resulting in an exacerbation of the distress state.
114
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the Liquidity of the
Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1, pp. 117-144.
63 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Strategy 5: More than twice as many turnarounds followed this strategy, compared to non-
turnarounds. It is termed to be an entrepreneurial-oriented strategy, where firms undertake
strategic investments through use of internal funds, which are obtained for example by raising
new equity. The additionally generated sales increase the cash-flows, which in return are
employed to pay down debt. Non-turnarounds might come unstuck with this strategy, if the
undertaken investments do not generate sufficient cash-flows to reach a relief of the
oppressive debt burden.
Strategy 6: This variable combination is predominant in the sub-sample of non-turnarounds. It
is thought as an entrepreneurial-oriented strategy, which misses the point. An increase in total
assets, whether it is related to capacity expansion or business diversification, should aim at
triggering a rise in total sales, as exemplified by strategy 4. The appearance of the opposite
can be motivated through value-destroying investments, for which costs exceed generated
cash-flows. Such a situation paired with rising debt levels will worsen the severity of distress.
For the three turnarounds displaying this variable combination, the change was marginal and
might be induced by temporary demand fluctuations distorting forecasted capital budgeting.
An example would be a company overestimating demand, leading to temporary inflated
inventory levels.
Although, there seems to be a difference in strategies applied by turnarounds and non-
turnarounds in the restructuring process, no specific strategy could be identified by our
sample that was solely implemented by one of the two groups. Ultimately, all above outlined
strategies, with exception of the flawed strategy 6, can result in a successful turnaround when
implemented correctly. To choose the adequate strategy is task of the management team. For
example, adopting an entrepreneurial-oriented strategy can bring about financial recovery, if
the cause of distress lies in a misfit between strategy of the company and operating markets.
Nevertheless, given that disadvantages in efficiency provoke financial distress, a strategic
reorientation is unlikely to bring the company back on track. An important finding of this
detailed analysis about what lies behind the changes in the ratio total debt to total assets is that
there exists evidence of turnarounds making use of efficiency-oriented and entrepreneurial-
oriented strategies.
64 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.5 Stock returns of turnarounds from the holdout sample
We computed the excess returns of the holdout sample firms that were identified as
turnarounds by the LDA-model 1. The S&P 500 was used as a benchmark index. The results
are presented in table 23.
Our model S&P 500 Excess Return Our model S&P 500
2004 30.9% 12.0% 18.9% 1.31 1.12
2005 38.3% 4.9% 33.4% 1.81 1.17
2006 28.8% 15.8% 13.0% 2.33 1.36
2007 -9.3% 5.6% -14.9% 2.12 1.44
2008 -47.1% -37.0% -10.1% 1.12 0.91
2009 111.7% 26.5% 85.2% 2.37 1.15
2010 34.9% 15.1% 19.8% 3.20 1.32
Total Return 220% 32%
Cumulative Yearly
Table 23: Excess returns of identified holdout sample turnarounds
2004-2010
The yearly returns of the identified turnarounds substantially exceeded the returns yielded by
the S&P 500. Only for 2007 and 2008, the time of the manifestation of the financial crisis, the
returns realized by our model fall short of the returns provided by the S&P 500. A possible
reason for this is that in times of crisis investors may be overly pessimistic and put not much
faith in the prospects of a recently financial distressed company. Nevertheless, an average
annual excess return of 14% over the average annual return yielded by the S&P 500 is a good
reason for considering investing in turnarounds identified by our model.
4.6 Variable testing
As the chosen model is generated by LDA, we test whether the residuals of the included
variables violate the assumptions that are imposed by discriminant analysis. The tests are
restricted to the in-sample data.
4.6.1 Normality test
The Bera-Jarque test is applied to examine whether the residuals of the included variables are
normally distributed. A normal distribution is characterized by its bell-shaped, symmetric
form, having a kurtosis of 3. The null-hypothesis of the test assumes normally distributed
residuals.
115
Table 24 summarizes the results of the Bera-Jarque test conducted in EViews.
115
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press, p.
181-182.
65 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Skewness Kurtosis p-Value
X1 9,79 109,38 0,00
X2 0,45 8,45 0,00
X3 -0,26 7,91 0,00
Table 24: Results of Bera-J arque test
The distributions of the residuals of the predictors were skewed and had excess kurtosis.
Besides, the null-hypothesis of the Bera-Jarque test was rejected for all three variables,
concluding that they are not normally distributed. We did not expect something different, as
financial data uses to display a leptokurtic distribution.
116
4.6.2 Heteroskedasticity test
LDA assumes constant variance in the residual terms of the model, which is denoted as
homoskedasticity. The opposite is known as heteroskedasticity and describes the situation
where the residual terms do not display a constant variance.
117
We apply the Breusch-Pagan-
Godfrey test, in order to investigate whether our established LDA model shows signs of
heteroskedasticity. The null-hypothesis of the Breusch-Pagan-Godfrey test assumes
homoskedasticity of the residual terms. The test is performed in EViews and the test statistic
is presented below.
Heteroskedasticity Test: Breusch-Pagan-Godfrey
F-statistic 1.777394 Prob. F(3,146) 0.1541
Obs*R-squared 5.285244 Prob. Chi-Square(3) 0.1521
Scaled explained SS 1.543019 Prob. Chi-Square(3) 0.6724
Graph 15: Breusch-Pagan-Godfrey test statistic output for LDA-model 1
The results of the Breusch-Pagan-Godfrey test provide no indication for heteroskedasticity.
All three versions of the test statistic have p-values exceeding the critical threshold of 0.05,
concluding that the null-hypothesis of homoskedasticity cannot be rejected. Thus, the chosen
LDA model did not violate the assumption of constant variance.
4.6.3 Multicollinearity test
Last but not least, we address the problem of multicollinearity, which occurs when the
predictors are highly correlated with each other. We define high correlation by a correlation
116
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press,
p. 179-180
117
Ibid, p. 147
66 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
coefficient greater or equal to 0.7. The following correlation matrix was generated by SAS
and belongs to the first LDA model, which showed the highest prediction accuracy.
Variables X1 X2 X3
X1 1,000 -0,04786 0,09415
X2 -0,04786 1,000 -0,28953
X3 0,09415 -0,28953 1,000
Table 25: Correlation Matrix X1 X2 X3 LDA model 1
As can be seen, the correlation among the predictors was very low and no correlation came
close to the critical value of 0.7. Accordingly, our model does not face a multicollinearity
issue.
67 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 5
Conclusion
The underlying thesis aimed at deriving the decisive factors in the turnaround process and
establishing a prediction model, thereby making possible to discriminate between future
turnarounds and non-turnarounds. The employed models were Linear Discriminant Analysis
and Logistic Regression and the model with the highest prediction accuracy was derived by
the LDA-approach and amounted to 92.7%. Taking into account that testing the model’s
prediction accuracy on the same sample, which was used to establish the model, will trigger
an upward bias in the results, the model was further evaluated based on its forecasting ability
on a holdout sample consisting of 3140 financially distressed firms. A marginal decrease in
the forecasting accuracy to 89% was recorded.
The selected variables failed to discriminate between turnarounds and non-turnarounds for the
first two years, where all companies in our sample displayed a Z-score below 1.8. A
significant discrimination between the two groups was obtained when focusing on year 2 and
3, where some companies realized an alleviation of the severity of distress, while for other
firms the distress state remained unchanged or even became worse.
With respect to the included predictors, the chosen model restricted itself to 3 explanatory
variables, which were firm size, severity of distress and total debt to total assets. While the
developed hypotheses for severity of distress and total debt to total assets were corroborated
by the empirical results of the chosen LDA-model, the hypothesis that firm size and
turnaround potential are positively related was rejected. Hence, for our sample smaller firms
are more likely to succeed in the turnaround process than larger firms. A possible explanation
is that smaller firms can implement entrepreneurial-oriented strategies faster and meet with
less resistance from high-level, old-established managers, who might interpret strategic
changes as a critique of their decision-making ability. Smaller firms tend to pursue a corporate
policy that is less dominated by complex hierarchical structures, which impede quick
decision-making. As severity of distress is measured by the Altman Z-score, this variable
substantiates the importance of the implementation of efficiency-oriented strategies to
improve financial performance, because the score is compounded of five variables that fall
into the category of efficiency measures. Changes in the ratio total debt to total assets can be
driven either by efficiency-oriented or by entrepreneurial-oriented strategies, or by a mix of
both. Which strategy is adopted and to what extent is determined by the cause of distress.
68 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Having at hand a turnaround prediction model with a high forecasting accuracy opens new
possibilities in yielding excessive returns from investing in distressed companies that will
transform to turnarounds. We briefly touched upon these opportunities by computing the
excess returns of the holdout sample turnarounds over the S&P 500, which amounted to 14%
on an annual basis.
Last but not least, we suggest that further research should aim at modeling the impact of
qualitative variables on the turnaround potential, such as internal firm climate and CEO
turnover. Moreover, other quantitative variables that are less firm-specific and more industry-
specific should be considered, like e.g. the growth of the industry a company is classified to.
69 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
References
Published References
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd
edition, Chapman and Hall.
Altman, Edward I., (1998). “Market Dynamics and Investment Performance of Distressed and
Defaulted Debt Securities“, New York University, Center for Law and Business, Working Paper
No. 98-022.
Altman, Edward I., (2000). “Predicting Financial Distress of Companies: Revisiting the Z-score
and Zeta Models“, New York University, Center for Law and Business.
Altman, Edward I. and Jha, Shubin, (2003), “Market size and investment performance of
defaulted bonds and bank loans: 1987-2001”, Economic Notes, Vol. 32, Issue 2, pp. 147-176.
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding
the Role of Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE
Transactions on Engineering Management, Vol. 44, No. 1, pp. 3-11.
Balgobin, R. and Pandit, N., (2001), “Stages in the Turnaround Process: The Case of IBM UK”,
European Management Journal, Vol. 19, Issue 3, pp. 301-316.
Barker, Vincent L. and Barr, Pamela S., (2000), “Linking top manager attributions to strategic
reorientation in declining firms attempting turnarounds”, Journal of Business Research, Vol. 55,
Issue 12, pp. 963-979.
Barker, Vincent and Duhaime, Irene M., (1997), Strategic change in the turnaround process:
Theory and empirical evidence, Strategic Management Journal, Vol. 18, Issue 1, pp. 13-38.
Barney, Jay B., (1986), “Types of competition and the theory of strategy: Toward an integrative
framework”, Academy of Management, Vol. 11, No. 4 , pp. 791-800.
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge
University Press.
Brown, David T., James, Christopher M. and Mooradian, Robert M., (1993), “The information
content of distressed restructurings involving public and private debt claims”, Journal of
Financial Economics, Vol. 33, No. 1, pp. 93-118.
Cameron, Kim S., Sutton, Robert I. and Whetten, David A., (1988), “Readings in Organizational
decline: Frameworks”, Ballinger Publishing.
Cooper, Ron A. and Weekes, Tony J., (1983), “Data, Models and statistical analysis”, Philip
Allan Publishers Limited, New Jersey, USA.
Cox, D. R. and Snell, E.J., (1989), “The analysis of binary data”, 2
nd
Edition, Chapman and Hall.
70 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Dechow, Patricia M. and Schrand, Catherine M., (2004), “Earnings Quality”, The Research
Foundation of CFA Institute.
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of
Firms Emerging from Bankruptcy, Journal of Finance, Vol. 54, Issue 5, pp. 1855-1868.
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
Gilson, Stuart C., (1990), “Bankruptcy, boards, banks and bondholders – Evidence on changes in
corporate ownership and control when firms default”, Journal of Financial Economics, Vol. 27,
Issue 2, pp. 355-387.
Greenblatt, Joel, (1999), “You can be a stock market genius”, Fireside.
Grinyer, Peter H., Mayes, David and McKiernan, Peter, (1988), “Sharpbenders: The secrets of
unleashing corporate potential”, Blackwell Publishers, Oxford.
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature
Industrial-Product Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-
248.
Harker, Michael, (1996), Managing company turnarounds: how to develop “destiny”, Marketing
Intelligence & Plannings, Vol. 14, Issue 3 pp. 5-10.
Hedberg, Bo L. T., Nystrom, Paul C. and Starbuck, William H., (1976), Camping on Seesaws:
Prescriptions for a Self-Designing Organization, Administrative Science Quarterly, Vol. 21, No.
1, pp. 41-65.
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp.
19-31.
Hofer, Charles W. and Schendel, Dan, (1978), Strategy Formulation: Analytical Concepts, West
Publishing.
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition,
John Wiley & Sons, Inc.
Hotchkiss, Edith S., (1995), Postbankruptcy Performance and Management Turnover, Journal of
Finance, Vol. 50, Iss. 1, pp. 3-21.
Indro, D. C., Leach, R.T. and Lee, W. Y., (1999), Sources of gains to shareholders from
bankruptcy resolution, Journal of Banking & Finance, Vol. 23, Iss. 1, pp. 21-47.
Kahl, Matthias, (2002), Economic Distress, Financial Distress and Dynamic Liquidation, The
Journal of Finance, Vol. 57, pp. 135-168.
Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A
survey”, The Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126.
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the
thoughtful investor, Harper Business.
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the
value of companies, John Wiley & Sons.
71 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Lachenbruch, P. A., (1968), “On expected probabilities of misclassification in discriminant
analysis, necessary sample size, and a relation with the multiple correlation coefficient”,
Biometrics, Vol. 24, No. 4.
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No.
1, pp. 69-85.
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture:
A Multilevel Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications,
Inc.
Myers, Stewart C., (1977), Determinants of Corporate Borrowing, Journal of Financial
Economics, Vol. 5, No. 2, pp. 147-175.
Ogden, Joseph P., Jen, Frank C. and O’Connor, Philip F.,(2003), Advanced Corporate Finance.
Policies and Strategies, Prentice Hall.
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in
corporate turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
Peng, Chao-Ying J., Lee, Kuk L. and Ingersoll, Gary M., (2002), “An Introduction to Logistic
Regression Analysis and Reporting”, The Journal of Educational Research, Vol. 96, No. 1, pp. 3-
14.
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”,
Strategic Management Journal, Vol. 13, Issue 4, pp. 287-309.
Schendel Dan E. and Patton, G.R., (1976), “Corporate stagnation and turnaround”, Journal of
Economics and Business, Vol. 28, Iss. 3, pp. 236-242.
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), “Corporate turnaround strategies: A study
of profit decline and recovery”, Journal of General Management, Vol. 3, Iss.. 3, pp. 3-12.
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the
Liquidity of the Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1,
pp. 117-144.
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies
and their implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress,
Managerial Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies:
An empirical analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
White, M. (1989), “Bankruptcy, liquidation and reorganization”, in: D.E. Logue (ed.), Handbook
of Modern Finance, Warren, Gorham & Lamont, New York
Ying, Charles C., (1966), “Stock Market Prices and Volumes of Sales”, Econometrica, Vol. 34,
No. 3, pp. 676-685.
72 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies:
evidence from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-
159.
Internet References
Buffet, Warren E., (2008), “Buy American, I Am”, The New York Times,http://www.nytimes.com/2008/10/17/opinion/17buffett.html
Forbes (2010):http://www.forbes.com/lists/2010/10/billionaires-2010_Warren-Buffett_C0R3.html
Twin A. (2009-03-09), “For Dow another 12-year low”, CNN Money,http://money.cnn.com/2009/03/09/markets/markets_newyork/index.htm
World Federation of Exchanges:http://www.world-exchanges.org/statistics/time-series/market-capitalization
Historical Chart of S&P 500:http://www.forecast-chart.com/historical-sp-500.html
VassarStats, Logistic:
Manuals
SAS Institute Inc. 2010. SAS/STAT® 9.22 User’s Guide. Cary, NC: SAS Institute Inc.
SPSS Regression 17.0
Database
Statndard and Poor’s Database
(http://www.standardandpoors.com)
73 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix
Appendix 1: Overview of prior studies on turnaround determinants
Researchers
Types of turnaround and
nonturnaround firms
Measures of
management actions
Actions associated with
turnaround or improved
financial performance
Schendel and Patton (1976)
Manufacturing firms
matched by SIC code Financial ratios
Decreased costs/sales
Increased sales
Increased investment
Hambrick and Schecter (1983) Mature, industrial product SBUs
Financial ratios and some
perceptual measures
Decreased R&D expenditures/sales
Decreased marketing expenditures/sales
Decreased receivables/sales
Decreased inventory/sales
Increased employee productivity
Increased plant & equipment newness
Increased market share
Ramanujam (1984)
Undiversified manufacturing
firms Financial ratios
Decreased cost of goods sold/sales
Decreased inventory/sales
Decreased receivables/sales
Increased sales
Thietart (1988)
SBUs across varying industry
environments
Financial ratios and some
perceptual measures
Combination of actions that cut costs and
increase productivity
Robbins and Pearce (1992) Textile firms 1976 - 85
Financial statement changes,
some perceptual measures
Asset reduction
Cost reduction
Arogyaswamy (1992) Manufacturing firms
Financial ratios, financial
statement changes
Decreasing at least three of the following:
Employees/Sales
Receivables/Sales
Inventory/Sales
Cost of goods sold/sales
SGA Expenses/sales
Combining above efficiency posture with
increased R&D or plant expenditures
Increasing R&D expenditures
Not decreasing at least three of the
following:
Employees
Receivables
Inventory
Cost of goods sold
SGA expenses
74 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix 2: Firms included in the in-sample
I n-sample companies by turnaround outcome
Non-turnarounds Turnarounds
AeroCentury Corp. (1998-99)
Abaxis Inc. (1997-98)
Air Methods Corp. (1996-97)
AmSurg Corp. (1998-99)
AK Steel Holding Corporation (1998-99)
Ariad Pharmaceuticals Inc. (1997-98)
Akamai Technologies Inc. (2000-01)
ASM International NV (1997-98)
AMR Corporation (1999-00)
Atwood Oceanics, Inc. (1993-94)
Apache Corp. (1995-96)
Aurizon Mines Ltd. (2000-01)
Appliance Recycling Centers of America Inc.
(1996-97)
BioDelivery Sciences International Inc. (2000-
01)
Arabian American Development Company (1999-
00)
Biogen Idec Inc. (1993-94)
Avis Budget Group, Inc. (1999-00)
Biolase Technology, Inc. (2000-01)
Bally Technologies, Inc. (1996-97)
Bio-Reference Laboratories Inc. (1998-99)
Belo Corp. (1997-98)
Birner Dental Management Services Inc.
(2000-01)
Belo Corp. (1998-99)
Blue Dolphin Energy Company (1994-95)
Boyd Gaming Corp. (1999-00)
China Yuchai International Limited (2000-01)
Breeze-Eastern Corporation (2000-01)
Comfort Systems USA Inc. (2000-01)
Cabot Oil & Gas Corporation (1994-95)
Concurrent Computer Corporation (1995-96)
Cabot Oil & Gas Corporation (1996-97)
CPI Aerostructures Inc. (2000-01)
Carriage Services Inc. (1999-00)
Cray Inc. (2000-01)
Chyron Corporation (2000-01)
Eldorado Gold Corp. (2000-01)
CNH Global NV (1998-99)
Energy Conversion Devices, Inc. (1998-99)
Comstock Resources Inc. (1998-99)
EOG Resources, Inc. (1998-99)
Corrections Corporation of America (1999-00)
Fieldpoint Petroleum Corp. (1996-97)
Craft Brewers Alliance, Inc. (2000-01)
FLIR Systems, Inc. (1999-00)
Crown Holdings Inc. (1998-99)
Food Technology Service Inc. (1995-96)
CSX Corp. (2000-01)
Fuel-Tech, Inc. (1999-00)
Earthstone Energy, Inc. (1998-99)
Gardner Denver Inc. (1995-96)
El Paso Corp. (2000-01)
Golden Star Resources, Ltd. (1998-99)
Female Health Co. (1997-98)
H&R Block, Inc. (1999-00)
Fonar Corp. (2000-01)
Hallador Energy Company (1996-97)
Ford Motor Co. (1992-93)
Headwaters Inc. (1998-99)
Forest Oil Corp. (1999-00)
Hollywood Media Corp. (1997-98)
Furmanite Corporation (1995-96)
Hurco Companies Inc. (1994-95)
GATX Corp. (1991-92)
ImmuCell Corp. (1994-95)
Good Times Restaurants Inc. (2000-01)
Imperial Sugar Co. (1999-00)
GP Strategies Corp. (2000-01)
Insignia Systems Inc. (1997-98)
Gray Television Inc. (2000-01)
InterDigital, Inc. (1992-93)
75 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Non-turnarounds Turnarounds
Hallador Energy Company (1998-99)
Inventure Foods, Inc. (1998-99)
HearUSA Inc. (2000-01)
Itron, Inc. (1999-00)
Hecla Mining Co. (1996-97)
Ivanhoe Mines Ltd. (2000-01)
Heska Corp. (2000-01)
Kinross Gold Corporation (2000-01)
Hexcel Corp. (2000-01)
Laboratory Corp. of America Holdings (1998-
99)
Hill-Rom Holdings, Inc. (1999-00)
Magnetek Inc. (1998-99)
Hill-Rom Holdings, Inc. (2000-01)
Medifast Inc. (1999-00)
HKN, Inc. (1999-00)
Medifast Inc. (2000-01)
icad Inc. (1998-99)
MEMC Electronic Materials Inc. (2000-01)
InsWeb Corp. (2000-01)
New Dragon Asia Corp. (2000-01)
International Shipholding Corp. (1994-95)
NovaGold Resources Inc. (2000-01)
Isle of Capri Casinos Inc. (1998-99)
Onyx Pharmaceuticals, Inc. (1997-98)
Iteris, Inc. (2000-01)
Orbit International Corp. (1998-99)
Joe's Jeans Inc. (2000-01)
Palatin Technologies Inc. (1997-98)
LodgeNet Interactive Corporation (1997-98)
Quest Diagnostics Inc. (1998-99)
MEDTOX Scientific Inc. (1999-00)
Retractable Technologies Inc. (2000-01)
Mercer International Inc. (1999-00)
RTI International Metals, Inc. (1994-95)
Norfolk Southern Corp. (1997-98)
Schawk Inc. (1995-96)
Norfolk Southern Corp. (1999-00)
Simulations Plus Inc. (2000-01)
NTN Buzztime Inc. (1998-99)
Stericycle, Inc. (1994-95)
Occidental Petroleum Corporation (1998-99)
Stericycle, Inc. (1999-00)
Orbit International Corp. (2000-01)
Tri-Valley Corp. (1995-96)
Parker Drilling Co. (2000-01)
Tri-Valley Corp. (2000-01)
Perma-Fix Environmental Services Inc. (1994-95)
Unisys Corporation (1993-94)
Pride International Inc. (1997-98)
Unit Corp. (1993-94)
PRIMEDIA Inc. (1999-00)
Universal Security Instruments Inc. (2000-01)
Ramtron International Corp. (1998-99)
Valero Energy Corp. (1994-95)
Reynolds American Inc. (2000-01)
Verint Systems Inc. (2000-01)
Ryder System, Inc. (1994-95)
Western Digital Corp. (1998-99)
Service Corp. International (1994-95)
Service Corp. International (1995-96)
Sinclair Broadcast Group Inc. (1998-99)
Six Flags Entertainment Corp. (2000-01)
Southwall Technologies Inc. (1998-99)
StemCells Inc. (1997-98)
Streamline Health Solutions, Inc. (1999-00)
Swift Energy Co. (1991-92)
76 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Non-turnarounds
Temple-Inland Inc. (1991-92)
Tenneco Inc. (1999-00)
The Hallwood Group Incorporated (1998-99)
The Interpublic Group of Companies, Inc.
(2000-01)
Titan International Inc. (2000-01)
Titanium Metals Corporation (1999-00)
Union Pacific Corporation (1997-98)
Unisys Corporation (1996-97)
Valhi, Inc. (1995-96)
Viad Corp (1995-96)
Waste Connections Inc. (2000-01)
Waste Management, Inc. (1999-00)
Willis Lease Finance Corp. (2000-01)
Xerox Corp. (1999-00)
77 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix 3: Scores model 1-4, in-sample
Model 1 Model 2 Model 3 Model 4
-1,41 -1,49 0,34 0,48
-1,06 -1,18 0,31 0,45
-0,72 -0,77 0,32 0,12
-3,18 -2,99 0,02 0,08
-2,27 -2,61 0,15 0,00
-0,54 -0,63 0,32 0,17
-3,42 -3,61 0,03 0,09
-1,55 -1,57 0,30 0,43
-2,27 -2,41 0,38 0,00
-2,53 -2,71 0,02 0,04
-0,79 -0,84 0,31 0,17
-0,64 -0,74 0,33 0,18
-0,68 -0,80 0,30 0,30
-1,46 -1,85 0,08 0,13
-1,54 -1,76 0,16 0,26
-0,66 -0,80 0,35 0,42
-1,07 -0,77 0,45 0,51
-3,43 -2,27 0,54 0,67
-1,52 -1,64 0,36 0,00
0,16 0,19 0,54 0,61
-1,33 -0,91 0,49 0,44
-0,73 -0,81 0,31 0,44
-1,20 -1,34 0,21 0,01
-1,25 -1,32 0,32 0,00
-5,25 -4,71 0,71 0,77
-2,61 -2,83 0,17 0,00
-8,76 -8,45 0,19 0,32
-0,79 -2,20 0,18 0,28
-6,66 -6,68 0,34 0,00
-0,34 -0,42 0,35 0,34
-0,61 -0,69 0,32 0,43
-0,72 -0,78 0,36 0,23
-0,46 -0,43 0,33 0,47
-0,58 -0,81 0,33 0,44
-0,08 -0,27 0,86 0,85
-1,46 -1,68 0,34 0,46
-1,72 -1,54 0,60 0,68
-1,65 -2,16 0,15 0,26
LDA-models LOGIT-models
Z(CEGA) scores and predicted probabilities,
models 1-4, in-sample
78 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
-5,61 -4,60 0,36 0,49
-1,50 -1,11 0,41 0,48
-0,37 -0,39 0,36 0,06
-0,52 -0,58 0,42 0,17
-5,32 -5,63 0,12 0,21
-3,43 -3,40 0,28 0,43
-4,79 -4,76 0,29 0,42
-0,74 -0,85 0,29 0,37
-0,46 -0,52 0,66 0,69
-1,97 -2,20 0,27 0,39
0,03 0,13 0,64 0,70
-1,28 -1,34 0,10 0,15
-1,01 -0,72 0,39 0,55
-0,66 -0,87 0,30 0,40
-1,30 -1,44 0,29 0,00
-1,15 -1,22 0,32 0,00
-3,81 -4,50 0,37 0,53
-0,75 -0,72 0,56 0,00
-0,71 -0,67 0,39 0,52
-1,36 -1,58 0,16 0,22
-0,58 0,18 0,72 0,79
-1,07 -1,23 0,27 0,24
-2,68 -2,96 0,10 0,08
-2,44 -3,83 0,32 0,46
-1,39 -1,85 0,16 0,00
-0,57 -0,70 0,30 0,09
-0,85 -0,93 0,33 0,04
-0,77 -0,85 0,36 0,02
-1,01 -1,13 0,31 0,19
-1,28 -1,41 0,28 0,14
-1,41 -1,58 0,24 0,36
-5,21 -5,16 0,15 0,25
-0,91 -0,99 0,39 0,52
-0,35 -0,42 0,43 0,54
-0,85 -0,95 0,31 0,01
-1,38 -1,51 0,16 0,12
0,01 -0,26 0,38 0,48
-1,00 -0,99 0,39 0,02
-0,71 -0,80 0,24 0,33
-0,34 -1,42 0,16 0,23
-1,50 -1,48 0,35 0,00
79 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
-1,15 -1,54 0,21 0,04
-0,46 -0,55 0,38 0,34
-0,57 -0,66 0,30 0,35
-0,72 -0,76 0,38 0,00
-1,02 -0,99 0,43 0,25
-1,18 -1,26 0,30 0,40
-1,38 -1,46 0,33 0,00
2,08 3,30 0,61 0,70
-0,76 -0,91 0,20 0,33
0,18 0,15 0,67 0,76
2,11 2,01 0,79 0,81
0,89 0,82 0,44 0,54
3,99 3,92 0,70 0,78
2,08 1,93 0,88 0,88
6,06 6,73 0,57 0,66
0,34 1,05 0,96 0,98
-0,21 -0,03 0,39 0,53
0,43 0,67 0,45 0,57
-0,22 -0,20 0,25 0,40
0,41 1,08 0,60 0,36
1,01 0,95 0,45 0,50
0,70 0,70 0,42 0,57
1,74 2,79 0,97 0,98
0,19 0,16 0,66 0,76
5,47 6,57 0,87 0,89
1,54 1,54 0,86 0,93
1,39 1,15 0,44 0,31
-1,08 -1,38 0,07 0,16
5,38 6,19 0,99 0,99
2,64 1,56 0,98 0,97
2,02 1,34 0,21 0,33
0,65 0,75 0,49 0,60
1,78 2,88 0,98 0,98
0,14 0,20 0,32 0,17
2,81 2,76 0,51 0,62
4,55 5,58 0,99 0,99
10,02 9,85 0,81 0,82
0,91 0,90 0,77 0,83
0,26 0,07 0,27 0,41
0,27 -0,82 0,08 0,13
2,04 2,07 0,14 0,27
80 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
0,54 0,74 0,57 0,61
-0,69 -1,45 0,14 0,32
1,51 1,65 0,72 0,78
-0,22 -0,21 0,28 0,38
0,53 0,71 0,46 0,59
7,57 7,50 0,89 0,94
1,41 1,40 0,24 0,34
0,37 1,20 0,60 0,69
-0,51 -0,43 0,05 0,11
0,41 0,39 0,45 0,33
1,57 0,35 0,60 0,66
4,55 4,41 0,98 0,98
1,35 1,64 0,60 0,67
4,74 4,68 0,76 0,85
0,72 0,59 0,54 0,64
0,83 1,06 0,87 0,90
0,15 0,15 0,26 0,39
1,31 1,30 0,79 0,89
0,59 1,12 0,64 0,26
-0,25 -0,34 0,32 0,45
1,48 2,40 0,85 0,86
1,03 1,02 0,57 0,48
1,01 1,12 0,62 0,69
1,14 1,04 0,60 0,64
doc_661398817.pdf
Turnarounds Modeling The Probability Of A Turnaround
TURNAROUNDS
-MODELING THE PROBABILITY OF A
TURNAROUND-
Master Thesis
Spring 2011
Supervisor:
Göran Andersson
Authors:
Eduard Ciorogariu
Andreas Goumas
2 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
ABSTRACT
The objective of this paper is to examine the possibility of predicting the recovery of a
distressed firm into a turnaround based on its current financial situation and a set of variables
that are considered of having a significant impact on the turnaround probability. To assess this
problem 150 firms are used that were distressed at some point during the period 1991 to 2001.
These firms were all listed in one of the major US stock exchanges and were all randomly
chosen, with 86 failing to recover from distress and 64 making a successful turnaround. In
order to establish a forecast model, two different quantitative econometrical methods are
applied; Linear Discriminant Analysis and Logistic Regression. The model predicting the
outcome of the 150 distressed firms with the highest accuracy is tested for its prediction
power on a holdout sample that consisted of 3140 distressed firms. These 3140 firms were all
listed at one of the major US stock exchanges and are distressed at some point during the
period 2002 to 2008. The prediction accuracy of the best model amounted to 92.7 % in the in-
sample and 89% in the holdout sample. The decisive variables that were selected by this
model are firm size, severity of distress and total debt to total assets.
Finally, we compare the returns yielded by a portfolio consisting of the turnarounds that were
predicted by the model out of the holdout sample to the returns generated by the S&P 500.
The annual returns for the seven years back-testing period, 2004-2010, for our portfolio
amounted to 18%, while the annual return for the S&P 500 was 4%.
Keywords: Financial distress, Turnaround, Turnaround prediction, Altman Z-score,
Discriminant Analysis, Logistic model.
3 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table of Contents
Introduction ............................................................................................................................................. 5
1.1 Background .................................................................................................................................... 5
1.2 Discussion of the problem ............................................................................................................. 8
1.3 Purpose .......................................................................................................................................... 9
1.4 Limitations ..................................................................................................................................... 9
1.5 Outline of the thesis .................................................................................................................... 10
Theoretical Framework ......................................................................................................................... 12
2.1 Prior research on turnarounds and its determinants.................................................................. 12
2.1.1 Causes of performance decline ............................................................................................ 13
2.1.2 Strategies to reverse performance decline .......................................................................... 18
Data and Methodology .......................................................................................................................... 22
3.1 Sources of information ................................................................................................................ 22
3.2 Criticism of sources ..................................................................................................................... 22
3.3 Definitions ................................................................................................................................... 23
3.4 Data ............................................................................................................................................. 25
3.5 Variables and hypothesis development ...................................................................................... 29
3.5.1 Size (X1) ................................................................................................................................ 29
3.5.2 Severity of distress (X2) ........................................................................................................ 30
3.5.3 Capital structure (X3, X4) ...................................................................................................... 30
3.5.4 Long-term financial health (X5, X6) ...................................................................................... 31
3.5.5 Short-term financial health / Liquidity (X7, X8, X9) .............................................................. 32
3.5.6 Profitability / Efficiency (X10, X11) ....................................................................................... 33
3.5.7 Investments / Divestments (X12, X13, X14, X15) ................................................................. 34
3.5.8 Management Expertise (X16) ............................................................................................... 36
3.5.9 Overview of examined variables .......................................................................................... 37
3.6 Methodology ............................................................................................................................... 38
3.6.1 Linear Discriminant analysis (LDA) ....................................................................................... 39
3.6.2 Binary Logistic regression model (Binary LOGIT) ................................................................. 44
Empirical Findings .................................................................................................................................. 48
4.1 Initial situation ............................................................................................................................. 48
4.2 Results of LDA .............................................................................................................................. 49
4.2.1 Means and correlation procedure (Model I) ........................................................................ 49
4.2.2 Stepdisc procedures (Model 2) ............................................................................................ 51
4 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.3 Results of LOGIT .......................................................................................................................... 52
4.3.1 Forward Conditional Logistic Regression (Model 3) ............................................................. 53
4.3.2 Backward Conditional Logistic Regression (Model 4) .......................................................... 54
4.3.3 Model comparison, selection and interpretation of results ................................................ 55
4.4 Model interpretation ................................................................................................................... 58
4.4.1 Size (X
1
) ................................................................................................................................. 58
4.4.2 Severity of distress (X
2
) ......................................................................................................... 58
4.4.3 Total debt to total assets (X
3
) ............................................................................................... 59
4.5 Stock returns of turnarounds from the holdout sample ............................................................. 64
4.6 Variable testing ............................................................................................................................ 64
4.6.1 Normality test ....................................................................................................................... 64
4.6.2 Heteroskedasticity test ......................................................................................................... 65
4.6.3 Multicollinearity test ............................................................................................................ 65
Conclusion ............................................................................................................................................. 67
References ............................................................................................................................................. 69
Published References ........................................................................................................................ 69
Internet References ........................................................................................................................... 72
Manuals ............................................................................................................................................. 72
Database ............................................................................................................................................ 72
Appendix ................................................................................................................................................ 73
Appendix 1: Overview of prior studies on turnaround determinants .............................................. 73
Appendix 2: Firms included in the in-sample .................................................................................... 74
Appendix 3: Scores model 1-4, in-sample ......................................................................................... 77
5 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 1
Introduction
1.1 Background
When plotting the movement of the S&P 500 over the last thirteen years, the index depicts
four extreme turning points. Driven by the elation of the emergence of a New Economy, the
S&P 500 surged at the end of the century and started to decline with the burst of the dot.com
bubble at the beginning of 2001, hitting rock bottom two years later. In 2003, the index
headed off to regain strength, reaching a new peak four years later. The outbreak of the
financial crisis triggered a new downswing of the index at the end of 2007. After a year, the
S&P 500 had to register an annual return of -38.5%
1
, its lowest result in sixty years of history.
During the first quarter of 2009 the index bounced back again and approaches its all-time high
of 1565.15
2
. The share index seems to display a reverting pattern, enabling market
participants to realise high capital gains, by selling when markets are at peak and buying
when they bottom out.
While the strategy “Buy low – Sell high” adds up for indices, it might fail for a single share,
because the risk of lasting underperformance or at worst bankruptcy cannot be diversified
away. However, to generate high returns investors don’t have to look out for the next global
crisis that will cause indices to plummet before they rally again. There are plenty of company-
specific financial crises occurring every year out of which a high-yield portfolio can be
constructed. Recalling that the share price reflects the investors’ expectations about the
company’s future performance, the stock of a firm sliding into financial distress is likely to
slump, regardless of whether the distressed state is expected to be temporary or long lasting.
A possible explanation for this behaviour is the market’s inability to capture the economic
fundamentals of distressed shares. Compliant with behavioural finance theory the
convergence between irrationality and barriers to arbitrage impede a separation between
transient and ongoing distressed stocks.
3
Moreover, since the variables of distressed
companies outweigh the variables of non-distressed companies in number, complexity and
1
http://www.forecast-chart.com/historical-sp-500.html
2
Twin A. (2009-03-09), “For Dow another 12-year low”, CNN Money,http://money.cnn.com/2009/03/09/markets/markets_newyork/index.htm
3
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, pp. 388
6 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
degree of uncertainty, deriving the intrinsic value of a distressed firm is a delicate endeavour,
enticing many intrinsic investors to refrain from their initial intention to invest.
4
Hence,
intrinsic investors fail to engage in a price correction and the stocks continue to fall. This
implies an undervaluation of some of the firms that are in financial distress, allowing market
participants to buy stocks with ample upside potential at marked-down prices. In fact, the
degree of upside potential can be expressed as a function of markdown. To be a value-
creating investment the share price needs to stop its downfall and start increasing again. It is
expected that the downward trend of a financial distressed company is reversed by the
implementation of a successful turnaround. Despite succeeded turnaround, market participants
will place low expectations in the future performance of recent distressed companies,
facilitating the management’s task to exceed shareholder expectations, which winds up in an
increase of the share price.
5
There exist plenty of studies, intending to identify the decisive factors in the turnaround
process. Some studies centre on quantitative variables, while other studies involve a
combination of quantitative and qualitative variables, taking into account that management
expertise and stakeholder support are crucial for conducting a successful turnaround.
Researchers distinguish between an efficiency-oriented and an entrepreneurial-oriented
strategy, firms can embark on during the turnaround process.
6
While several researchers like
Zeni et. al (2010)
7
develop turnaround prediction models, only few of them test their model
with respect to the stock returns yielded by the predicted turnarounds.
While investing in distressed companies is a popular research area, the bulk of research
focuses on investing in defaulted debt securities. Edward Altman, the inventor of the Z-score
that is widely used for determining distressed firms, has undertaken extensive research in this
field. He concentrates on the “risk and return performance of defaulted debt”.
8
In 2003 Altman
and Pompeii laid out an analysis of the historical performance of investments in defaulted
4
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the thoughtful investor,
Harper Business, pp. 189.
5
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, pp. 46.
66
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
7
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence from
Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
8
Altman, Edward I., (1998). Market Dynamics and Investment Performance of Distressed and Defaulted Debt
Securities, New York University, Center for Law and Business, Working Paper No. 98-022, pp. 2.
7 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
bonds and bank loans, covering the time period 1987 - 2001.
9
Practitioners like fund manager
Joel Greenblatt, recommend against investing in distressed stocks, because in case of
bankruptcy shareholder interests are the last to serve and equity holders might come away
empty-handed. Thus, his focus lies also on investing in distressed bonds, bank loans and trade
claims. However, this area is dominated by vulture investors, who can cope with the
complexity of such investments, which results from the legal and financial issues brought
about by different classes of creditors with different priorities and claims.
10
Nevertheless, researchers and practitioners see a potential benefit from investing in companies
that emerged from financial distress. Greenblatt states that recently emerged shares are
available at substantial discount, partially because they suffer from low analyst coverage and
partially because market participants still attach a high risk profile to the stock.
11
Another
reason is that a big stake of creditors’ bankruptcy claims rather tends to be converted into
equity claims than paid out in cash. Former debt holders such as banks, bondholders and trade
creditors have no incentive to engage in a long-term commitment in the emerged company
and aim at cashing out by selling the new share (we assume that the participation of former
shareholders was either bought up for a liquidating dividend or cancelled).
12
This creates a
negotiating range, which permits interested investors to purchase the share at a discount.
Altman et. al (1998)
13
investigated the stock performance of firms emerging from Chapter 11.
The authors observed significant positive excess returns over the long-term (200 trading days
from emergence date from Chapter 11) and ascribed it to the market’s inefficiency, which
causes a paucity of information that in turn leads to a stock’s mispricing. Besides, the study
points towards the existence of a positive relationship between the nature of securities
accepted by creditors and the appearance of excess stock returns. According to this, stocks of
emerged firms for which debt holders approved a complete equity-for-debt exchange
demonstrate strong positive long-term abnormal returns.
14
The identified linkage between the
type of arrangement the emerged firm and its debt holders agreed upon and its stock returns
let us infer that the creditors dispose of information, which is not captured by the market,
allowing them to compute the firm’s intrinsic value. The conclusion is reasonable, as creditors
9
Altman, Edward I. and Jha, Shubin, (2003), “Market size and investment performance of defaulted bonds and
bank loans: 1987-2001”, Economic Notes, Vol. 32, Iss. 2, pp. 147-176.
10
Greenblatt, Joel, (1999), You can be a stock market genius, Fireside, pp. 166 – 168.
11
Ibid, pp. 175.
12
Ibid, pp. 169 – 170.
13
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of Firms
Emerging from Bankruptcy, Journal of Finance, Vol. 54, Iss. 5, pp. 1855-1868.
14
Ibid, pp. 1865-1867.
8 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
like banks and other financial institutions e.g. life insurance, pension funds etc. are classified
as informed investors and do not confront information asymmetry issues.
15
However,
according to Kahl (2002)
16
debt holders cannot eliminate the degree of uncertainty in respect
of the viability of distressed firms completely, which drives them to engender a decision
model comprising three options: recovery, controlled liquidation and immediate liquidation.
Consequent, if creditors agree to swap their entire debt position for equity, they have strong
beliefs in the recovery potential and growth opportunities of the firm and choose the first
option.
1.2 Discussion of the problem
Altman et. al (1998)
17
attested that stocks of financially distressed firms, which were likely to
succeed in the turnaround process, yielded abnormal returns over a time frame of 200 trading
days, outstripping market indices by roughly 20%. Indro et. al (1999)
18
established a model
consisting of five variables, which can be applied to distinguish between successful and failed
restructurings. They demonstrated that for a portfolio comprising distressed stocks and having
an accumulated turnaround probability of more than 50%, excess compounded returns amount
to 42% over a one-year period.
While most studies focus on firms that have submitted an official bankruptcy petition, such as
e.g. filing under Chapter 11, our empirical research is not restricted to this formal procedure.
We refrained from constraining our study on the stock performance of firms emerging from
Chapter 11. In this manner, we took into consideration the findings of Hotchkiss (1995)
19
,
who challenges the accuracy of the Chapter 11 process in separating economically inefficient
from economically efficient companies.
Instead, we consider the Altman Z-score to be more precise in distinguishing potential
turnarounds from non-turnarounds, because it encompasses company-specific financial data,
15
Ogden, Joseph P., Jen, Frank C. and O’Connor, Philip F.,(2003), Advanced Corporate Finance. Policies and
Strategies, Prentice Hall.
16
Kahl, Matthias, (2002), Economic Distress, Financial Distress and Dynamic Liquidation, The Journal of Finance,
Vol. 57, pp. 135-168.
17
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of Firms
Emerging from Bankruptcy, Journal of Finance, Vol. 54, Iss. 5, pp. 1855-1868.
18
Indro, D. C., Leach, R.T. and Lee, W. Y., (1999), Sources of gains to shareholders from bankruptcy resolution,
Journal of Banking & Finance, Vol. 23, Issue 1, pp. 21-47.
19
Hotchkiss, Edith S., (1995), Postbankruptcy Performance and Management Turnover, Journal of Finance,
Vol. 50, Issue 1, pp. 3-21.
9 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
which makes it possible to pinpoint operational performance changes over time. Additionally,
we investigate the impact of other quantitative variables on the likelihood of turnaround,
aiming at determining the drivers that lie behind a successful restructuring.
1.3 Purpose
The underlying thesis follows the purpose of deriving the decisive variables that allow for
distinguishing financially distressed firms with turnaround potential from those without.
Based on a sample of 150 companies, it intends to conceive the main drivers in the turnaround
process and their relation to the turnaround potential of a firm. Further, the identified drivers
are used to establish a prediction model, which can be adopted for classifying financially
distressed companies into turnarounds and non-turnarounds. To some extent, the thesis aims
at analyzing whether recovery is achieved by focusing on efficiency-oriented or on
entrepreneurial-oriented strategies. In addition, the thesis touches upon the opportunity that
arises from investing in financially distressed firms with high turnaround potential, by
exhibiting generated returns of the turnarounds detected in the holdout sample. In this manner,
the thesis tends to pave the way for future, profound research in distress investing.
1.4 Limitations
The states Distress and Recovery are defined by a company’s Z-score falling short of a given
threshold and subsequently exceeding this threshold and are not dependent on an official
filing for bankruptcy proceeding and a pursuant announcement of recovery. Hence, the
sample might consist of some firms that did not file under the US Bankruptcy Code and might
omit some firms that did so.
The sample employed to determine the decisive factors in the turnaround process contains
companies listed at one of the three major US stock exchanges, New York Stock Exchange
(NYSE), American Stock Exchange (NYSE Amex) and NASDAQ Exchange and embraces
the time period 1991 to 2003. The main reason for deciding to collect data from the specified
exchanges and over the specified period is to ensure the availability of a representative sample
in terms of size and industry coverage. On top of that, the sample comprises only non-
financial companies. The exclusion of financial firms from the study group is motivated by
the belief that inclusion would lead to biased results. This expectation can be motivated by
10 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
two points. Firstly, financial companies are characterized by an extremely high leverage ratio,
so that their involvement would most likely distort the impact of debt on the turnaround
process. Secondly, due to their importance for the overall economic stability of a country the
probability of governmental interventions in case of distress is much higher than for non-
financial companies. However, this study does not make a point of developing a way to
identify companies, which will be bailed out with the utmost probability in the event of
distress. Neither does it advise market participants to invest in such companies. The variables
analysed by the study are of quantitative nature, as collecting reliable and satisfying
qualitative data would require access to resources that were not approachable for us and
would go beyond the time period scheduled for this work.
The data employed for back-testing the prediction power of the established model is also
raised from the three major US stock exchanges.
1.5 Outline of the thesis
Chapter 1 – Introduction
The first chapter explains why we became interested in examining the drivers behind a
successful turnaround. It states the opportunities arising for investors to make money by
investing in financially distressed firms with a high turnaround potential. Studies of other
researchers are named and the main conclusions are summarised. Some of the difficulties a
financially distressed firm has to deal with in the process of restructuring are mentioned,
providing an idea about the variables that play a decisive role in the turnaround process. The
chapter ends with a listing of the study’s limitations.
Chapter 2 – Theoretical foundation
The second chapter delves into the subject of identifying the determinants of a successful
turnaround, so as to be able to distinguish firms with turnaround potential from firms that are
likely to remain in distress, denoted as non-turnarounds. Prior research on turnarounds is
covered and the key results of these studies are discussed.
The entire chapter provides the theoretical background on which the underlying thesis is
based.
11 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 3 – Data and Methodology
The third chapter gives a description of the process of data collection and brings in the
variables that were considered in the empirical study. Hereby, it develops the hypotheses that
are to be examined empirically. Last but not least, the two econometrical models that are
applied to come up with the turnaround prediction model, Multiple Discriminant Analysis
(MDA) and Logistic Regression (Logit Model), are introduced.
Chapter 4 – Empirical Results
The fourth chapter presents and further discusses the determinants that were found to be
critical for a successful turnaround by each of the applied models. In this way, it comes up
with four prediction models, which are assessed based on their in-sample accuracy. It also
includes the back-testing of the two best prediction models. The excess returns of the holdout
sample turnarounds, selected by the best prediction model, are computed. The input data of
the final chosen model is subject to some statistical tests.
Chapter 5 – Conclusion
The fifth chapter aims at wrapping up the key findings of the implemented study. Besides, a
recommendation about what issues future research in this area could cover is administered.
12 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 2
Theoretical Framework
Plenty of research has been undertaken on turnarounds with the purpose of identifying the
determinants of success. An overview of prior studies in this field is outlined in the appendix.
While empirical studies use financial metrics and quantitative ratios to specify a successful
turnaround, Balgobin et al. (2001)
20
provide a more qualitative definition:
“A corporate turnaround may be defined simply as a recovery of a firm’s economic
performance following an existence-threatening decline.”
2.1 Prior research on turnarounds and its determinants
Empirical studies concentrate on the reasons that drive companies into such performance
decline and the corresponding strategies the management team employs to reach
rehabilitation. Schendel et al. (1976)
21
22
and Hofer et al. (1978, 1980)
23
24
attributed
deteriorating firm performance to operational or strategic issues and emphasized the
importance of correctly recognizing the source of decline, which enables the firm to adopt the
adequate measures to reverse decreasing performance. They point out that failure to locate the
catalyst of decline leads to the implementation of wrong measures and can hinder the firm to
achieve a successful turnaround.
In the following sections several research studies on turnarounds are reviewed, outlining the
causes of performance decline and the strategies and measures firms take to return to a stable
state. This review provides the basis for the selection of the potential turnaround determinants
analyzed in the underlying thesis.
20
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
21
Schendel Dan E. and Patton, G.R., (1976), Corporate stagnation and turnaround, Journal of Economics and
Business, Vol. 28, Iss. 3, pp. 236-242.
22
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), Corporate turnaround strategies: A study of profit
decline and recovery, Journal of General Management, Vol. 3, Iss.. 3, pp. 3-12.
23
Hofer, Charles W. and Schendel, Dan, (1978), Strategy Formulation: Analytical Concepts, West Publishing.
24
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp. 19-31.
13 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
2.1.1 Causes of performance decline
Balgobin et al. (2001)
25
summarize the main performance decline triggers that were identified
by six independent research studies. A separation is made between internal and external
causes of performance decline, where internal causes refer to company-specific issues and
external causes are affiliated to weak economic conditions or industry-specific issues.
External causes
External causes include downturn in demand, increase in competition and increase in input
costs. The causes show a certain degree of interdependency, as fierce competition and
increased input costs both are likely to affect demand negatively.
Fall in demand can be traced back to several other reasons, such as the contraction of the
industry, an overall economic recession that weakens purchasing power or the failure of the
company to meet customer expectations. While the first two affect all competitors within an
industry, the last one applies to single companies. In regard of customer expectations, Lepak
et al. (2007)
26
define the task of a firm in creating value for its customers. They call this the
use value. In return, the customers are willing to provide the exchange value, which is
measured in monetary units. If a company cannot constantly provide the value demanded by
its customers at the expected conditions (e.g. quality, sales price), or if other companies have
the resources to provide equal or higher value, or to provide the same value at improved
conditions a firm-related cutback in demand should be expected.
The intensity of competition is determined by the characteristics of the industry. The concept
of industrial organization competition summarized by Barney (1986)
27
considers the barriers
established in an industry as one determinant of competitive pressure. These barriers involve
barriers to entry, barriers to competition, barriers to imitation and barriers to exit.
28
Other
decisive factors are the amount and size of rivals, the nature of products (customized vs. mass
product) and the demand elasticity.
29
In addition, substitute products pose a threat, as firms
might not recognize immediately to whom they lose market share, aggravating the necessity
25
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
26
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture: A Multilevel
Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
27
Barney, Jay B., (1986), Types of competition and the theory of strategy: Toward an integrative framework,
Academy of Management, Vol. 11, No. 4 , pp. 791-800.
28
Ibid.
29
Ibid.
14 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
to retaliate.
30
Lepak et. al (2007) mention the problem of value slippage
31
, which leads to an
erosion of the firm’s competitive advantage and can be avoided by the establishment of
isolating mechanisms
32
(e.g. patents, trademarks, special knowledge). Fierce competition will
undermine the financial position of a company, as it has to engage in costly retaliation
campaigns, such as price cuts or expensive marketing and promotion campaigns to guarantee
competitiveness. Adopting the wrong retaliation tactics will lead to a further weakening of the
firm’s market position and absorption of financial means.
Price increases in input costs bring forth a rise in costs of goods sold, putting pressure on the
gross profit margin. Given that companies are able to pass on the increase in costs fully to the
customers, margins will not be suppressed. While this might be imaginable in a monopolistic
market, it is unlikely to hold for a situation of perfect competition. Price pressure will
originate from some companies that enjoy advantages on the input market e.g. strong
bargaining power, bulk purchase etc. and squeeze margins of firms that do not dispose of
these advantages by luring away customers.
I nternal causes
Internal causes of performance decline involve poor management, inadequate financial
control/policy and high cost structure. As for the external causes a kind of interdependency
can be observed, as inadequate financial control/policy and high cost structure both can
originate from poor management.
The management team needs to be endued with the capabilities to steer the company through
times of prosperity and decline. There exists no commonly accepted definition of poor
management performance, but some conclusions about its meaning can be drawn from the
examined literature. According to Hedberg et al. (1976)
33
firm decline bears upon the
omission of the management team to align the strategy of the company to its evolving
environment. This problem tends to exacerbate for companies with a long track record, as the
management team is prone to hubris and overconfidence and has strong beliefs in the
30
Koller et. al (2010), Valuation: Measuring and managing the value of companies, pp. 79-98
31
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture: A Multilevel
Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
32
Ibid.
33
Hedberg, Bo L. T., Nystrom, Paul C. and Starbuck, William H., (1976), Camping on Seesaws: Prescriptions for
a Self-Designing Organization, Administrative Science Quarterly, Vol. 21, No. 1, pp. 41-65.
15 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
underlying strategy, rendering it immune to necessary strategic change.
34
Harker (1996)
35
stresses the importance for a company to understand its industry, markets and customers, to
know its position and future potential with respect to its industry and markets and to know its
competitors and their position and future potential. Only if the management team is able to
capture and process these variables, can it come up with an adequate strategy. Another
important task of the management team is the delegation of responsibilities to lower
hierarchical levels. Concentrating too much decision power at the top level can lead to inertia
and delayed responsiveness to changes in customer preferences, as it was the case for IBM
UK in the 90s.
36
Regarding inadequate financial control/policy the focus lies on the firm’s capital structure and
the sources of financing.
37
Owing to the fact that interest expenses lower the taxable income
and in this way increase the free cash flow to firm, the market value of the firm would be
maximized by employing a gearing of 100%. However, with an increase in leverage the
probability of future financial distress and the cost of financial distress also raise, which
results in a depression of the firm’s market value. This relationship is depicted by the
following formula:
V
L
= V
U
* ?
c
* D – PV[E(CFFD)]
V
L
= Market value of the levered firm
V
U
= Market value of the unlevered firm
?
c
= Corporate tax rate
D = Face value of debt
PV[E(CFFD)] = Present value of expected cost of future financial distress
In addition, having an extremely high gearing might force a company to postpone or
completely abandon some value creating investments, causing an underinvestment problem.
34
Barker, Vincent L. and Barr, Pamela S., (2000), Linking top manager attributions to strategic reorientation in
declining firms attempting turnarounds, Journal of Business Research, Vol. 55, Iss. 12, pp. 963-979.
35
Harker, Michael, (1996), Managing company turnarounds: how to develop “destiny”, Marketing Intelligence
& Plannings, Vol. 14, Iss. 3 pp. 5-10.
36
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
37
Ibid, pp. 303.
16 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
An example is provided to illustrate the underinvestment problem. Assumptions are based on
Myers (1977)
38
:
It is assumed that the firm’s assets in place V
A
are equal to 0. The firm takes on risky debt P
(P is defined as risky debt, as the firms has no assets in place and thus does not dispose of
collaterals to secure the debt), which together with the contribution I (equity investments) of
shareholders account for the required cash outlay to realize V
G
, the real option. The debt is to
be repaid after the expiration of the real option V
G
and the owners know the value of the real
option in the event of exertion, which is depicted V(s).
In such a case, shareholders will exercise the real option only on condition of:
V(s) > I + P
If the cash outlay (I + P) exceeds V(s), shareholders will refuse to exercise the real option, as
their equity investment I will be higher than the market value of their shares.
39
Thus, even if
the real option is value creating (V(s) > I), it won’t be realized, because of the effect of the
debt burden. This is also referred to as the debt overhang problem. Abandoning value creating
projects can accelerate a firm’s performance decline, as its competitive position might be
undermined, triggering a decrease in operational performance. Generating fewer turnovers
will raise the financial pressure, amplifying the debt overhang problem and prompting further
sacrifice of value enhancing investments. However, the abandonment of promising projects
must not always be the actuator of a drop in operational performance. Sales can also be
depressed by the various external factors described in the previous section, like e.g. industry
contraction. Although there exists no optimal debt-to-equity ratio, when deciding on leverage
a firm should allow for a reasonable financial buffer to be able to absorb unexpected
economic or industry-specific declines and leave the door open to undertake value enhancing
investments when they appear.
With reference to the sources of financing, failure to abide by the maturity matching principle
can bring a company into financial distress. In general, the maturity on an interest-bearing
liability should coincide with the life expectancy of the asset or project, for which the credit is
raised. Ignoring this guideline gives rise to either a refinancing risk or a debt overhang
problem. In the case that the maturity of interest-bearing debt falls short of the life expectancy
38
Myers, Stewart C., (1977), Determinants of Corporate Borrowing, Journal of Financial Economics, Vol. 5,
No. 2, pp. 147-175.
39
Ibid, pp. 153
17 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
of the asset/project, interest and principal payments might be due before the asset/project
managed to generate any cash-flow. Hence, the company will try to roll over the outstanding
debt. If the debt holder refuses refinancing, the firm will be forced to repay the credit
immediately, although the asset/project missed to yield a return so far. This will lead to a
financial bottleneck that can culminate in defaulting on debt. Alternatively, if the maturity of
interest-bearing debt exceeds the life expectancy of the asset/project, the firm will have to
bear an ongoing debt burden, which is related to an asset/project that does not generate cash-
flows anymore. Thus, it would carry on the default risk and the outstanding debt would
interfere with the realization of valuable real options (debt overhang problem).
Last but not least Balgobin et. al (2001) cite a high cost structure as a reason for performance
decline. In case of IBM UK an overstated cost base was build up, because the top
management expected revenues to grow at historical rates. When realized revenues did not
comply with expected revenues, the company had to face a cost burden that outstripped the
one of competitors significantly.
40
This example also elucidates the interdependency of
performance decline causes, as the high cost base was a result of management’s inability to
correctly anticipate future market demand.
All in all, deteriorating performance can be traced back to several internal and external
causes, which are closely intertwined and act jointly, making it impossible to relate
performance decline to one single source. The table below provides an overview of the
different research studies that focused on the same external and internal causes of decline.
Researcher Schendler et al. (1976) Bibeault (1982) Slatter (1984) Thain et. al (1989) Grinyer et. Al (1990) Gopal (1991)
External causes
Decrease in demand x x x x x x
Increase in competition x x x x x x
Increase in input costs x x x x n.a. x
Internal causes
Poor management x x x x x x
Inadequate financial
control/policy n.a. x x x x x
High cost structure x n.a. x x x n.a.
Table 1: The Causes of Declining Performance
Source: Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Iss. 3, pp. 301-316.
X = clearly referred to; n.a. = not referred to
40
Balgobin, R. and Pandit, N., (2001), Stages in the Turnaround Process: The Case of IBM UK, European
Management Journal, Vol. 19, Issue 3, pp. 308.
18 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
2.1.2 Strategies to reverse performance decline
This section outlines the various strategies that companies adopt, in order to cope with an
existence-threatening decline and to achieve a successful turnaround.
According to Schendel et al. (1976)
41
a company can either apply an efficiency-oriented or an
entrepreneurial-oriented strategy. Which strategy is chosen depends on the cause of the
downturn. Efficiency-oriented restructurings imply the enforcement of retrenchment, which
incorporates cost-cutting measures, downsizing and asset reduction, while entrepreneurial-
oriented restructurings aim at aligning the underlying strategy to the prevalent market
conditions.
42
The viewpoint of Schendel et. al (1976) is supported by Barker et. al (1997)
43
,
who distinguish between two sources of decline: industry-specific and firm-specific decline.
Cameron et. al (1988)
44
explain firm-specific decline as the inability of a company to perform
at eye level with its competitors, suffering from a competitive disadvantage. Thus, if a firm
acts in a growing industry, but faces deteriorating performance, the adoption of an
entrepreneurial-oriented strategy is compulsory. It can be concluded that companies, which
suffer from performance decline due to a contraction of the industry, should put more weight
on efficiency-oriented strategies.
As opposed to this, Robbins et. al (1992)
45
hold that independent of the cause of decline the
implementation of efficiency-oriented strategies is crucial for succeeding with the turnaround.
Other researchers opt for separating the turnaround process into two subsequent stages:
Stabilization and Recovery. The purpose of the first stage is to prevent a continuation in
performance decline and to build the foundations for the implementation of recovery
strategies. This process involves convincing stakeholders to support the turnaround intention,
stop the financial drainage and ensure a constructive internal climate. In the second stage the
recovery strategies are introduced, according to the trigger of decline. The necessity of
41
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), Corporate turnaround strategies: A study of profit
decline and recovery, Journal of General Management, Vol. 3, Issue 3, pp. 3-12.
42
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Issue 3, pp. 305.
43
Barker, Vincent and Duhaime, Irene M., (1997), Strategic change in the turnaround process: Theory and
empirical evidence, Strategic Management Journal, Vol. 18, Issue 1, pp. 13-38.
44
Cameron, Kim S., Sutton, Robert I. and Whetten, David A., (1988), “Readings in Organizational decline:
Frameworks”, Ballinger Publishing.
45
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
19 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
administering a two-stage approach is determined by the severity of distress, the firm size and
the availability of free assets.
46
Maintaining stakeholder promotion is critical for facilitating the continuation of the
operational business. Companies that slide into financial distress are likely to experience a
large number of resignations from key employees, resulting in a brain drain that aggravates
the competitive situation. Suppliers and customers must be persuaded to uphold business
relations with the firm and debt holders must be prepared to compromise on contractual terms.
In order to restore the stakeholders’ trust in the firm’s survival potential and accomplish its
support, quick actions that yield immediate results are applied at the beginning of the
turnaround process, aiming at improving efficiency.
47
These actions narrow down to cutbacks,
which are concentrated on downsizing, reduction of inventory levels, cost of goods sold and
selling, general and administrative expenses.
48
Cost cuttings and efficiency enhancements will
free up resources that can be reallocated.
49
However, the assertion of cutbacks might backfire
and even cause a further drop in firm efficiency. This can be expected when managers decide
to cut costs on the wrong positions. For example, switching to cheaper suppliers might also
have a degrading effect on product quality, causing more customers to discontinue business
relations. Furthermore, the management’s decision to lay off employees and undertake salary
cuts and cancellations of one-time bonus payments can create a working climate that is
characterized by insecurity about the workplace and frustration, releasing a loss of motivation
and associated increase in absenteeism, more production of scrap, decreased product quality,
extended processing time and delayed deliveries.
50
Arogyaswamy et. al (1997)
51
conclude that
turnarounds and non-turnarounds have a strong tendency to engage in cutbacks. However,
non-turnarounds apply this measure more excessively than turnarounds. Also, turnarounds are
more successful in translating cutbacks to efficiency improvements than non-turnarounds.
From this it can be concluded that managers of turnarounds pick the write spots for cost-
46
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Issue 3, pp. 304-320, p. 305.
47
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11, p. 4.
48
Ibid, p. 3.
49
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 306.
50
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11, p. 3.
51
Ibid.
20 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
cutting and prove able to convince remaining employees from the necessity of the undertaken
retrenchment measures. A more extensive form of cutbacks is operational asset reduction,
which is carried out to lower the firm’s capacity to the current production level. In so doing,
manufacturing facilities are employed more efficiently and cash-inflows are generated
52
through the sales of assets, which can be used for lowering the debt burden or to make
necessary capital expenditures, like maintenance investments in PP&E.
As has been represented at length, stopping performance decline presupposes the execution of
retrenchment measures to improve efficiency. Hofer et al. (1980)
53
note that a financially
distressed company will restructure its operations and thereby repel the threat of bankruptcy
before it starts to analyze its strategic position in the market.
In the second stage, the distressed company will either continue to implement more profound,
long-lasting operational changes or strive for a strategic reorientation, depending on the cause
of decline. Yet, according to Grinyer et. al (1988)
54
a sustainable operational improvement is
achieved, when turnarounds put emphasis on strategic reorientation, redefining their product
and market portfolio. Strategic reorientation implies divesting in unrelated areas and investing
in related areas, thereby strengthening the focus of the company on its core-capabilities.
Basically, a company has to separate its products and markets following the criterion of value
creation. Value-destroying business units are sold and value-destroying markets are
abandoned. That way, the company obtains a cash-inflow in terms of the sales price and
reduces cash-outflow, which was attributed to the maintenance of the sold business units and
exited markets. The cash-inflow obtained from the divestments can be used partially to lower
the debt burden and partially to invest into value-creating business units and markets.
Nevertheless, depending on the severity of distress and the support from stakeholders,
especially the willingness of debt holders to grant further moratorium or even provide
additional financial funds, a company might be forced to sell off profitable business units to
generate sufficient cash.
55
Schlingemann et. al (2002)
56
reinforce this assumption,
52
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 185.
53
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp. 19-31.
54
Grinyer, Peter H., Mayes, David and McKiernan, Peter, (1988), “Sharpbenders: The secrets of unleashing
corporate potential”, Blackwell Publishers, Oxford.
55
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 186.
56
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the Liquidity of the
Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1, pp. 117-144.
21 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
demonstrating that firms rather divest the most profitable segment over the least liquid
segment and the most liquid segment over the least profitable segment.
In terms of asset investment, Hambrick et. al (1983)
57
argue that internal capital expenditures
are geared towards obtaining efficiency improvements, by means of e.g. improved monitoring
and steering of the process flow. Arogyaswamy et. al (1997)
58
demonstrate the importance of
capital expenditure on PP&E in their study. However, they point out that the amount invested
in PP&E is almost equal between turnarounds and non-turnarounds. Nevertheless, a
significant difference exists regarding investments in R&D, which are clearly higher for
turnarounds. They take the view that investing in new technology is vital for manufacturers to
adapt to a changing environment
59
and meet market expectations. External investments in the
sense of acquisitions can be conducted, as part of the strategic reorientation process to
strengthen the product and market portfolio and accelerate revenue growth, given that the
company disposes of sufficient financial slack.
60
The table below summarizes the main measures implemented under the corresponding
strategy.
Efficiency-oriented Entrepreneurial-oriented
Table 2: Turnaround strategies
Strategy alignment to changing
environment through:
Investments in R&D
strategic asset divestment
strategic asset investment
Retrenchment of the firm:
reduction of operational cost
Downsizing
Operational asset reduction
Internal capital expenditures
57
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
58
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding the Role of
Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE Transactions on Engineering
Management, Vol. 44, No. 1, pp. 3-11.
59
Ibid, p. 3.
60
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies and their
implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
22 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 3
Data and Methodology
This chapter describes the steps undertaken in our empirical research on determining the
decisive factors for a successful turnaround. It provides information on applied sources, data
collection and data processing and outlines the methodology employed to perform the
research.
3.1 Sources of information
As our empirical research is to the greatest extent based on financial data, we perceived it as a
vital prerequisite to obtain our input data from a single and reliable source. Taking into
account that each database has established its own technique of providing financial
information, such as balance sheet items, financial ratios, share prices etc., falling back on
several databases could lead to distorted results and conclusions. Thus, all financial data used
in our study is collected from the Standard & Poor’s database.
In terms of literature employed for presenting the theoretical background of our study, we
went back to course literature, course material and scientific articles, which were extracted
from LibHub.
3.2 Criticism of sources
The financial data used in our empirical study was not generated at first hand, as we collected
it from a database provider. It has to be taken into account that due to the extent of our sample
and its encompassed time span, it would have been inefficient to extract the data separately
from the balance sheet statements and the income statements of each company. In order to
examine the reliability of the database, we randomly selected firms from our sample and
cross-checked the provided data with the data reported in the respective SEC filings.
With respect to the literature applied for the theoretical background, we made reference to
scientific articles published in distinct journals and to literature that covers the topic of
restructuring financially distressed companies. With a view to critically scrutinize the
readings, we abandoned demonstrating only the viewpoint and results of one researcher, but
23 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
were constantly anxious to provide corroborating or confuting ideas of other research
colleagues.
3.3 Definitions
Our sample consists of companies that are classified as either distressed firms or turnarounds.
We applied the Altman Z-score to obtain a sample of distressed firms and to separate
companies, which remained distressed from companies which achieved a turnaround.
Altman (1968)
61
introduced the Z-score as a measure of predicting the firm’s probability of
going bankrupt. It is defined as follows:
Where, X
1
: Working capital/Total assets
X
2
: Retained Earnings/Total Assets
X
3
: Earnings before Interest and Taxes/Total Assets
X
4
: Market value equity/Book value of total liabilities
X
5
: Sales/Total assets
According to Altman, a Z-score of greater than 2.99 classifies the firm into the “non-
bankrupt” zone. If the Z-score is below 1.8, the firm falls into the “bankrupt” zone and firms
lying in-between 1.8 and 2.99 are assigned to the “zone of ignorance” or “grey area”.
62
We defined a firm as financially distressed, if it exhibited a Z-score below 1.8 for two
consecutive years. In the event that the Z-score of the company increased above 1.8 in the
third year and above 2.99 in the fourth year, or was above 2.99 for two successive years after
being classified as financially distressed, it was perceived as a successful turnaround.
Companies, whose Z-score remained below 1.8 for two further years were categorized as
failed turnarounds.
61
Altman, Edward I., (2000). “Predicting Financial Distress of Companies: Revisiting the Z-score and Zeta
Models“, New York University, Center for Law and Business.
62
Ibid.
24 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 1: Illustration of z-score development for 2 companies
The above graph depicts three companies (A, B & C), which would have been eligible for our
sample based on the development of their Z-scores over a total period of four years. All three
firms had a Z-score below 1.8 for two consecutive years. However, in year 3 and 4 the Z-
score of company B surpassed the critical threshold of 2.99, while the Z-score of company A
stayed below the critical threshold of 1.8. The Z-score of company C went beyond the critical
threshold of 1.8 in year 3 and exceeded the critical threshold of 2.99 in year 4. In this
instance, we would have categorized firms B and C as successful turnarounds (1) and firm A
as a failed turnaround (0). The two tables below summarize the main definitions, on which the
entire study is based.
Distressed firm: two consecutive years of Z-score below 1.8
Successful turnaround: two consecutive years of Z-score below 1.8
followed by either two consecutive years of
Z-score above 2.99 or a Z-score above 1.8
in year 3 and a Z-score above 2.99 in year 4.
Failed turnaround: two consecutive years of Z-score below 1.8
followed by two consecutive years of Z-score
below 1.8
Table 3: Summary of definitions
Status Year 1 Year 2 Year 3 Year 4
0 < 1.8 < 1.8 < 1.8 < 1.8
1 < 1.8 < 1.8
> 1.8
>2.99
> 2.99
Table 4: Status based on Z-score
0 = failed turnaround 1 = successful turnaround
0
1
2
3
4
5
0 1 2 3 4 5
Z
-
s
c
o
r
e
Years
Company A
Company B
Company C
25 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.4 Data
The firms and the appertaining financial figures of our sample were gathered from the
Standard and Poor’s database. In order to obtain a reasonable sample size, we focused on a
sample period ranging from 1991 to 2003, within which data was collected. This period was
divided into ten sub-periods, with each sub-period covering four years.
1990 2004
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
1991 - 1994
Sub-period 1
1992 - 1995
Sub-period 2
Graph 2: Illustration of sample period and corresponding sub-periods
Graph 2 depicts the process of data collection over the sample period and emphasizes that
data was gathered from each of the ten sub-periods.
Sub-period Turnaround Distress
1991-94 0 3
1992-95 1 1
1993-96 3 1
1994-97 6 5
1995-98 5 5
1996-99 2 6
1997-00 7 7
1998-01 11 16
1999-02 7 18
2000-03 22 24
Total 64 86
Table 5: No. of firms per sub-period
The precedent table displays the amount of firms collected from each sub-period,
differentiating between turnarounds and distressed companies. The sample comprised a total
of 150 companies, out of which 64 were classified as successful turnarounds and 86 were
categorized as failed turnarounds.
26 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table 6 provides an overview of the industries that were covered by the data.
Industry Turnaround Distress
Automobiles & Components 0 2
Biotechnology 6 3
Capital Goods 8 9
Commercial & Professional Services 2 6
Consumer Services 1 8
Energy 5 12
Food, Beverage and Tobacco 3 2
Gold 5 0
Healthcare 9 7
Household & Personal Products 2 1
Materials 1 9
Media 0 8
Oil, Gas & Coal Exploration & Production 4 1
Retailing 1 3
Software and Services 2 3
Technology Hardware and Equipment 6 4
Transportation 0 8
Others 9 0
Total 64 86
Table 6: No. of firms per covered industry
Involving firms from different industries was important, because the study aims at deriving
universal implications about decisive factors in the turnaround process and does not narrow
down its research to a particular industry. Nevertheless, we excluded financial institutions
from our sample, due to their highly levered capital structure and because some financial
institutions enjoy governmental bankruptcy protection through bail-out guarantees.
All firms used in the study were publicly traded and listed at one of the following U.S. stock
exchanges during the period 1991 to 2003:
i. (NYSE) ? New York Stock Exchange
ii. (AMEX) ? American Stock Exchange
iii. (NasdaqGM) ? Nasdaq Global Market
iv. (NasdaqCM) ? Nasdaq Capital Market
v. (NasdaqGS) ? Nasdaq Global Select
27 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
We have chosen these exchanges, because when accumulated they rank first place in two
categories over the whole length of our study period:
1. Amount of listed firms
2. Trading volume
A high amount of listed firms provides broad industry coverage, which is necessary
concerning that we do not restrict our analysis to a specific industry.
The variable trading volume is a crucial determinant of stock returns. The Wall Street believes
in a relationship between trading volume and stock returns, stating that “It takes volume to
make the prices move”
63
. Ying (1966)
64
demonstrated that small trading volumes are related
to negative returns (price fall) and large trading volumes are related to positive returns (price
rise). Several other researchers substantiated a positive correlation between trading volume
and stock returns. Below an excerpt of a list of studies on this issue is provided.
Researcher Year Sample Data Sample Period Interval
Positive
Correlation
found?
Comiskey et. al 1984 211 common stocks 1976-79 yearly Yes
Morgan 1976 44 common stocks 1926-68 monthly Yes
Richardson et. al 1987 106 common stocks 1973-82 weekly Yes
Jain et. al 1986 Stocks market aggregates 1979-83 hourly Yes
Table 7: Prior studies on correlation between returns and trading volume
Source: Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A survey”, The
Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126, p. 112.
A high trading volume is of interest for our study, since we also point to significant
differences in stock returns of distressed firms and turnarounds.
The amount of listed firms and trading volume was accumulated for the considered US stock
exchanges and compared to the London SE and the Tokyo SE over the sample period. The
63
Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A survey”, The
Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126, p. 112.
64
Ying, Charles C., (1966), “Stock Market Prices and Volumes of Sales”, Econometrica, Vol. 34, No. 3, pp. 676-
685, p. 676.
28 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
London SE and the Tokyo SE were chosen as benchmark stock exchanges, because they were
ranked among the top five largest stock exchanges throughout the whole study period.
65
Graph 3: Listed firms by stock exchange
Graph 4: Trading volume in USD millions by stock exchange
Graph 3 and 4 demonstrate that the involved US stock exchanges outperformed the London
SE and the Tokyo SE in number of listed firms and trading volume throughout the whole
sample period.
65
World Federation of Exchanges,http://www.world-exchanges.org/statistics/time-series/market-
capitalization
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
1990 1993 1996 1999 2002
US accumulated
London
Tokyo
0.0
5,000,000.0
10,000,000.0
15,000,000.0
20,000,000.0
25,000,000.0
30,000,000.0
35,000,000.0
40,000,000.0
45,000,000.0
1990 1993 1996 1999 2002
Tokyo
London
US accumulated
29 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.5 Variables and hypothesis development
This section presents the variables we have taken into account in our empirical research. We
assume that these variables have an impact on a company’s probability to recover from
financial distress. Therefore they act as discriminating predictors, enabling a separation of
distressed companies into turnarounds and failed turnarounds (firms that remain in financial
distress). A hypothesis was formulated for each variable.
3.5.1 Size (X1)
We measure size by means of total tangible assets. Some studies use total sales as an indicator
of size. However, we follow Smith et. al (2005)
66
, who relate the size of a company to its
borrowing capacity. According to the Collateral Hypothesis, a firm’s debt capacity is
restricted to its collateralizable assets, which are represented by its total tangible assets. Thus,
a firm with a higher debt capacity will have easier access to the credit market, to raise funds
necessary for the restructuring. White (1989)
67
highlights the positive impact of the track
record of large companies in raising external funds on their ability to obtain additional
financial support. Besides, strong stakeholder support is expected for large firms, as their
stakeholders have more to lose in the event of bankruptcy.
68
On top of this, large companies
dispose of more assets that can be sold and more business units that can be divested,
triggering a release of internally generated financial means, which contribute to the
restructuring process by reducing the leverage or enabling the realization of value-creating
investments. A contrary opinion is provided by Paint (1991)
69
, who reveals a negative
relationship between size and turnaround potential. According to him, smaller companies can
adapt more readily to altering conditions of their environment.
Although small firms might be characterized by a flat hierarchy and exhibit few layers of
management, allowing them to react fast to market changes, there access to external capital
markets might be restricted, forcing them to resort to internal financial means. As we perceive
66
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 310.
67
White, M. (1989), “Bankruptcy, liquidation and reorganization”, in: D.E. Logue (ed.), Handbook of Modern
Finance, Warren, Gorham & Lamont, New York.
68
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320, p. 306.
69
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in corporate
turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
30 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
the availability of capital as a vital factor for a successful turnaround, we do not share Paint’s
viewpoint.
Hypothesis 1: Firm size and turnaround potential are positively correlated.
3.5.2 Severity of distress (X2)
Robbins et al. (1992)
70
investigate a positive relationship between the severity of distress and
the degree of cutbacks and asset divestments. As stated by Slatter (1984)
71
, the
implementation of retrenchment measures might face organizational resistance, which results
in a decrease of operational efficiency, thereby aggravating the distressed state. Sudarsanam
et. al (2001)
72
indicate that the severity of distress negatively influences the required time for
the restructuring and can inhibit the completion of certain restructuring measures. These
measures primarily embrace actions aiming at reshaping the firm’s strategy and call for
capital expenditures. Debt holders might oppose the implementation of such actions, because
they consume capital funds that can be used to settle part of their claims. The severity of
distress is measured by the Z-score. Companies displaying a Z-score, which is below 1.8 or
even negative, are considered to be severely distressed.
Hypothesis 2: Severity of distress and turnaround potential are negatively correlated.
3.5.3 Capital structure (X3, X4)
Klarman (1991)
73
points out that financial distress in the majority of cases can be traced back
to excessive leverage. For a distressed firm reorganizing its capital structure and reducing the
debt burden might be an essential step towards a successful turnaround. With respect to
Gilson (1990)
74
an alleviation of the indebtedness can be reached by renegotiating existing
debt contracts, in such a way that the creditor offers either a composition (reduction of interest
or principal) or an extension or even an exchange of debt for equity or a combination of all
70
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
71
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies and their
implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
72
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
73
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the thoughtful investor,
Harper Business.
74
Gilson, Stuart C., (1990), “Bankruptcy, boards, banks and bondholders – Evidence on changes in corporate
ownership and control when firms default”, Journal of Financial Economics, Vol. 27, Issue 2, pp. 355-387.
31 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
three. Brown et. al (1993)
75
underscore the signaling effect of exchanges by stating that
positive information is conveyed to the market, if a firm achieves an exchange with its banks.
However, pushing through an exchange with bondholders transmits negative information, as
contrasted with banks bondholders do not have the capabilities to assess a firm’s recovery
potential. In order to get a more precise picture of the indebtedness of a firm, we focused on
the ratio total debt to total assets, which states how much of the total assets are externally
financed. A decrease in the ratio can either stem from the implementation of debt reducing
measures or from an internally financed increase of total assets. While it is obvious that
deleveraging will promote recovery from financial distress, expanding the asset base will also
foster the completion of a successful turnaround, given that the additional assets are value
creating. Also, an increase in the equity position is expected to promote a successful
turnaround, making it possible to interpret that existing shareholders believe in the turnaround
potential of the firm.
Hypothesis 3: Change in total debt to total assets and turnaround potential are negatively
correlated.
Hypothesis 4: Change in total equity and turnaround potential are positively correlated.
3.5.4 Long-term financial health (X5, X6)
Free cash-flow to total liabilities (X5)
Free cash-flow is the part of the cash-flow generated by a firm’s operational business, which
is left over after subtracting the capital expenditures that were reinvested back into the
operations. It excludes the impact of financial and non-operating items and is available to debt
holders and shareholders.
76
The ratio free cash-flow to total liabilities measures to what extent
a firm is able to cover its liabilities by means of financial funds yielded from its operations.
Sudarsanam et. al (2001)
77
employed PBITD (profit before interest taxes and depreciation) as
a cash-flow proxy. We refrained from adopting the same cash-flow proxy and relied on free
cash-flow. Unlike PBITD free cash-flow takes into consideration the cash-outflows resulting
from tax payment and reinvestment in the operational business. Therefore, we contemplated
75
Brown, David T., James, Christopher M. and Mooradian, Robert M., (1993), “The information content of
distressed restructurings involving public and private debt claims”, Journal of Financial Economics, Vol. 33,
No. 1, pp. 93-118.
76
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, p. 135.
77
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
32 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
that if a firm is in a distressed state and it strives for reorganization, expenses related to
operations will still arise and give cause for reinvestments. We measure the change of the
ratio free cash-flow over total liabilities. An increase of the ratio can either be affiliated to a
reduction of total liabilities or to an increase of free cash-flow.
Hypothesis 5: Change of free cash-flow over total liabilities and turnaround potential are
positively correlated.
Solvency ratio (X6)
We included the solvency ratio, which is defined as follows:
Non-cash expenses were added back to net income, to reflect the entire funds available for the
redemption of a firm’s liabilities. In contrast to times interest earned, the solvency ratio
incorporates also cash that stems from non-operational actions, such as asset sales. As the
occurrence of bankruptcy is often contingent on a firm’s insolvency, we perceive an increase
of the solvency ratio as a sign of financial recovery.
Hypothesis 6: Change of solvency ratio and turnaround potential are positively correlated.
3.5.5 Short-term financial health / Liquidity (X7, X8, X9)
Times interest earned (X7)
In accordance with Zeni et. al (2010)
78
we include the interest coverage ratio in our analysis.
It measures if the financial funds generated by a firm’s operations (EBIT) are sufficient to
comply with its interest charges. An increase in the ratio can originate either from a rise in
EBIT or from a decrease in interest charges. The latter can be motivated by a reduction of the
debt burden through a composition, an exchange or a debt retirement.
Hypothesis 7: Change of times interest earned and turnaround potential are positively
correlated.
78
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence
from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
33 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Quick ratio (X8, X9)
By including the quick ratio in our analysis, we bear in mind the importance of financial
liquidity for a company. The quick ratio is computed thusly:
Remaining liquid is a prerequisite for being able to continue operations, since an illiquid firm
might be confronted with the termination of its business relations with suppliers, as it cannot
make timely payments. We examine the relation between liquidity, change in liquidity and
turnaround potential. We included also change in liquidity by arguing that an increase of
liquidity will invigorate stakeholder support, as suppliers can feel more certain about the firm
meeting their financial obligations. Loyal suppliers might cause customers losing their fear
regarding the firm’s ability to carry out orders, making them to refrain from switching to
competitors.
Hypothesis 8: Quick ratio and turnaround potential are positively correlated.
Hypothesis 9: Change of quick ratio and turnaround potential are positively correlated.
3.5.6 Profitability / Efficiency (X10, X11)
Free cash-flow to sales (X10)
We apply this ratio as a profitability measure, instead of using the profit margin. In
comparison to earnings, free cash-flow provides a more undistorted measure of a company’s
profit creation, as it is not subject to estimation and judgment of the top management team.
Earnings can be forged by earnings management and accrual manipulation, with the objective
of artificially improving the firm’s profit creation.
79
Free cash-flow to sales states the amount of cash generated by a firm’s revenues after
subtracting capital expenditures. It gives an indication of a company’s proficiency to control
its cost structure. However, a low ratio cannot always be attributed to a high cost structure,
but might result from undertaken investments in e.g. new technology, which would suppress
free-cash flow downward. If the investments are value creating, cash-flows will be generated
and the ratio will be revised upwards next year, under the assumption that new investments
79
Dechow, Patricia M. and Schrand, Catherine M., (2004), “Earnings Quality”, The Research Foundation of CFA
Institute.
34 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
will start generating cash-flows one year after their implementation. We measure the change
in free-cash flow to sales, as we believe that an increase of the ratio can be effectuated either
by an improvement of the cost structure or by a cash-flow rise due to implemented value
creating investments. An increase in profitability will promote the turnaround process, as the
firm will be able to use free cash-flow for alleviating the indebtedness or for pursuing further
value creating investments.
Hypothesis 10: Change of free cash-flow over sales and turnaround potential are positively
correlated.
Operating profit margin (X11)
Hambrick et. al (1983)
80
and Robbins et. al (1992)
81
are in agreement about the necessity of
enforcing retrenchment actions to pave the way for a turnaround. We adopt operating profit
margin as a ratio of efficiency improvement through the consummation of cost-cutting
measures. It is computed as follows:
An increase in operating profit margin does most likely stem from a decrease in variable
costs, like wages and raw material prices etc. Another possible source would be an
acquisition, which would cause sales to rise at a faster pace than variable costs, due to the
realization of synergies.
Hypothesis 11: Change of operating profit margin and turnaround potential are positively
correlated.
3.5.7 I nvestments / Divestments (X12, X13, X14, X15)
Free assets (X12)
This variable is directly related to a firm’s debt capacity. Smith et. al (2005)
82
define it as
follows:
80
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
81
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
82
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
35 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The ratio indicates a firm’s amount of unutilized collateralizable assets. Thus, the higher the
ratio the more leeway the firm enjoys in taking on additional debt, as new loans could be
endowed with collaterals, permitting the creditors to seize ownership of the assets in the event
of default. This facilitates the access to external capital for the distressed firm. Moreover,
while covenants in the debt contract inhibit the sale of assets pledged as collateral, free assets
can be sold, therefore representing a possible source of cash-inflow.
Hypothesis 12: Free assets and turnaround potential are positively correlated.
Degree of Downsizing (X13)
Robbins et. al (1992)
83
perceive the enforcement of retrenchment measures as the first step of
a successful turnaround. Such measures are not constraint to headcount reduction, but involve
cost-cutting efforts and asset divestments with a view to improving efficiency and generating
cash flows. Based on Smith et. al (2005)
84
we define downsizing in the following way:
Hypothesis 13: Degree of downsizing and turnaround potential are positively correlated.
Goodwill (X14)
The position goodwill in the balance sheet statement contains the premiums paid for the
acquisitions a company has undertaken.
85
We refer to change in goodwill as a measure of
strategic asset investment/divestment. Goodwill impairments and amortizations were added
back to avoid making wrong inferences about asset divestments that did not occur. Thus,
changes in the goodwill position reflect acquisitions and divestments, respectively. Since
several researchers suggest that a firm should consider strategic asset investments in the
83
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”, Strategic
Management Journal, Vol. 13, Issue 4, pp. 287-309.
84
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress, Managerial
Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
85
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the value of
companies, John Wiley & Sons, p. 141.
36 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
turnaround process, in order to adjust its asset portfolio to the evolving environment
86
, we
investigate the relationship between goodwill increases and turnaround potential.
Hypothesis 14: Change in goodwill and turnaround potential are positively correlated.
R&D expenses (X15)
In regard to Hambrick et. al (1983)
87
, we include R&D expenses as a measure of a firm’s
concentration on the development of new products, which shall promote the company in the
process of strategic reorientation. However, we argue that constantly interpreting investments
in R&D as a strategic measure would be wrong. A firm might expend R&D effort in
developing new technologies or in designing innovative ways to organize and manage the
process cycles, both of which would rather be related to an efficiency improvement than to a
strategic reorientation. In addition, we decided not to focus on the ratio of R&D to sales, as
has been done by Hambrick et. al (1983). Instead, we measure the change in R&D expenses
over a one-year period. That way, we prevent the emergence of wrong conclusions about a
firm’s R&D policy, as a decline in the ratio could be explained by a cutback in R&D expenses
or by an increase in sales, or by both.
Hypothesis 15: Change in R&D and turnaround potential are positively correlated.
3.5.8 Management Expertise (X16)
ROE (X16)
For a successful turnaround it is crucial that stakeholders and shareholders believe in the
incumbent management team’s ability to steer the company out of distress. Otherwise, they
will refrain from promoting the turnaround attempt, putting at risk the firm’s recovery.
Zeni et. al (2010)
88
included ROE (Return on equity) as a measure of top management
expertise in their Z-score, which they developed for the Malaysian market.
We apply ROE as an indicator of the top management team’s capability to initiate a process
of recovery from distress and thereby ensure stakeholder and shareholder support.
86
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies: An empirical
analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199, p. 186.
87
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature Industrial-Product
Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-248.
88
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies: evidence
from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-159.
37 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Hypothesis 16: ROE and turnaround potential are positively correlated.
3.5.9 Overview of examined variables
The table below provides an overview of the variables taken into account in our empirical
study. It also shows the encoding used in the statistical programs SAS and SPSS for each
predictor. Besides, it states the scale unit employed to each variable and points out which
empirical research motivated the inclusion of the variable.
I. Size Encoding Empirical support
Total tangible assets X1
White (1989)
Smith et. al (2005)
II. Severity of distress
Z-score X2 Sudarsanam et. al (2001)
III. Capital structure
? total debt/total assets X3
Gilson (1990)
Klarman (1991)
? total equity X4 Klarman (1991)
IV. Long-term financial health
? FCF/total liabilities X5
Sudarsanam et. al (2001)
Own intuition
? Solvency ratio X6 Own intuition
V. Financial health/ Liquidity
? Times interest earned X7 Zeni et. al (2010)
Quick ratio X8 Own intuition
? Quick ratio X9 Own intuition
VI. Profitability/Efficiency
? FCF/Sales X10 Goumas et. al (2011)
? EBIT/Sales X11
Hambrick et. al (1983)
Robbins et. al (1992)
Chowdhury et. al (1996)
VII. Investments/Divestments
Free assets X12 Smith et. al (2005)
Degree of downsizing X13
Robbins et.al (1992)
Smith et. al (2005)
? Goodwill X14
Hofer (1980)
Grinyer et. al (1988)
Sudarsanam et. al (2001)
? R&D expenses X15
Hambrick et. al (1983)
Goumas et. al (2011)
VIII. Management Expertise
ROE X16
Abdullah et. al (2008)
Zeni et. al (2010)
Table 8: Overview of examined variabels
? = change no ? = no change (the variable was taken into account, instead of the change in the variable) yoy = year on
year
38 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Table 9 summarizes the examined hypotheses with respect to each variable.
I. Size Encoding Correlation with turnaround potential
Total tangible assets X1 +
II. Severity of distress
Z-score X2 -
III. Capital structure
? total debt/total assets X3 -
? total equity X4 +
IV. Long-term financial health
? FCF/total liabilities X5 +
? Solvency ratio X6 +
V. Financial health/ Liquidity
? Times interest earned X7 +
Quick ratio X8 +
? Quick ratio X9 +
VI. Profitability/Efficiency
? FCF/Sales X10 +
? EBIT/Sales X11 +
VII. Investments/Divestments
Free assets X12 +
Degree of downsizing X13 +
? Goodwill X14 +
? R&D expenses X15 +
VIII. Management performance
ROE X16 +
Table 9: Summary of hypothesis formulation
3.6 Methodology
In order to investigate which of the outlined independent variables are most qualified to
separate our sample into turnarounds and non-turnarounds, we run a linear discriminant
analysis (LDA) and a logistic regression (LOGIT) on our sample data. Both methods will
enable us to test the developed hypotheses and create a discriminant function (DF) and
logistic function. They act as a prediction model, facilitating the categorization of financially
distressed firms into turnarounds and non-turnarounds, based on a cut-off point. The DF
obtained from the LDA and the logistic function given by LOGIT will be applied to a holdout
sample for reasons of back-testing their prediction accuracy.
The statistical programs employed on the sample data were SAS (with respect to LDA) and
SPSS (with respect to LOGIT).
39 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
3.6.1 Linear Discriminant analysis (LDA)
LDA aims at ascribing an unknown subject (e.g. financially distressed firm) to one of two
groups (e.g. turnarounds; non-turnarounds)
89
, with the aid of discriminating variables
(explanatory variables).
The DF is expressed by the following equation:
Where, Z = discriminant score
? = constant term
? = discriminant coefficients
X = discriminating variables
Discriminating
Variables
I
m
p
a
c
t
Financially
Distressed
Firms
Non-
turnarounds
(0)
Turnarounds
(1)
Graph 5: The discriminating process in the linear discriminant analysis
The discriminating process is described in the graph above. The dependent variable (e.g.
turnaround outcome of financially distressed firms) is categorical and the two groups must be
definite distinguishable from each other i.e. they need to be mutually exclusive.
89
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No. 1, pp. 69-85.
40 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The discriminating variables, which are considered to be suitable for differentiating the two
groups from each other, are chosen according to their ability to maximize the distance
between the means of the probability distributions of the two groups and included in the DF.
90
Graph 6: Probability distributions of turnarounds and non-turnarounds and Mahalanobis-Distance
?
0
= Mean of non-turnaround distribution based on DF ?
1
= Mean of turnaround distribution based on DF
C = Cut-off point
The probability distributions of the two groups are depicted in graph 6. LDA assumes that the
explanatory variables follow a normal distribution, having equal variances and covariances.
91
The distance between the means of the two groups is called Mahalanobis-Distance and
calculated thusly:
S
2
= pooled sample variance
With an increasing D
2
the overlapping area of the normal distributions becomes smaller,
enabling an almost unambiguous differentiation between the two groups.
90
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No. 1, pp. 69-85.
91
Cox, D. R. and Snell, E.J., (1989), “The analysis of binary data”, 2
nd
Edition, Chapman and Hall.
?
0
?
1
C
Probability distribution of
non-turnarounds
Probability distribution of
turnarounds
Mahalanobis-Distance
41 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 7: Overlapping of the probability distributions of two populations I and II
Graph 7 depicts an example of a small Mahalanobis-Distance between two population means,
giving rise to a large overlapping area and increasing the probability of misclassification. In
most cases, a perfect seperation of two populations cannot be achieved, so that the tails of the
distributions will cross, leading to the emergence of type I and type II errors (misclassification
errors).
92
In our particular study, a type I error corresponds to the classification of a non-
turnaround as a turnaround and a type II error is equivalent to the classification of a
turnaround as a non-turnaround.
So as to be able to categorize an observation in one of the two groups, a cut-off point needs to
be determined. Graph 6 shows that the seperation line is located where the tails of the two
groups’ probability distributions cross. Thus, assuming a normal distribution, the cut-off point
is given by the formula:
For a new observation X the discriminant score must be calculated on the basis of the DF. If
the discriminant score lies below the cut-off point, the observation is classified into group I
and vice versa.
92
Lachenbruch, P. A., (1968), “On expected probabilities of misclassification in discriminant analysis, necessary
sample size, and a relation with the multiple correlation coefficient”, Biometrics, Vol. 24, No. 4.
I +I I
I I I
42 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 8: Classification of a new observation X
Graph 8 visualizes the classification of a new observation X in one of the two groups. The red
circle depicts the discriminant score of the observation X. As the score is smaller than the cut-
off point C, the observation is categorized into the group of non-turnarounds. Another
approach is to argue that the discriminant score of X is closer to the mean of the probability
distribution of non-turnarounds than to the mean of the probability distribution of
turnarounds, leading to a classification into the first group.
As already stated, the independent variables involved in the DF are the most qualified for
discriminating the two groups from each other. The graph below shows an example for two
independent variables, applied to classify financially distressed firms into turnarounds and
non-turnarounds.
S
e
v
e
r
i
t
y
o
f
D
i
s
t
r
e
s
s
Size
Graph 9: Classification of financially distressed firms into turnarounds and non-turnarounds based on size and severity of
distress
Probability distribution of
non-turnarounds
Probability distribution of
turnarounds
C ?
1
?
0
X
43 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
For this example the turnarounds are depicted by the black triangles and the non-turnarounds
correspond to the red circles. It is apparent that the non-turnarounds are located in the
northwestern part of the scatterplot, while the turnarounds lie southeasterly. The blue line,
seperating the observations of the two groups, is called the linear discriminant boundary and
is described by the following formula:
93
This formula complies with the DF formula. An observation that is situated on the
discriminant boundary has a discriminant score equal to zero. As a consequence, observations
located on one side of the boundary will be distinguished from observations located on the
other side by having an opposite sign in the discriminant score. This is ensured by including
the constant term ? in the formula. The value of the discriminant score indicates the distance
of the observation from the discriminant boundary.
94
We used SAS to conduct the LDA. The program offers different approaches to create a subset
of explanatory variables (predictors) out of an initial set of variables. The most important
approaches are described briefly.
Means and correlation procedure:
SAS provides an overview of the means and the standard deviations for each variable that is
entailed in both groups (0, 1). Taking e.g. the variable Size, the mean ?(Size
0
) and the standard
deviation ?(Size
0
) are compared with the mean ?(Size
1
) and the standard deviation ?(Size
1
).
The higher the difference in the two means and the lower the intra-group standard deviation,
the better the variable Size discriminates between the two groups. Given close to each other
located means and high intra-group standard deviation, increases the chance of overlapping
probability distributions, producing large misclassification errors.
Stepdisc Forward Variables Selection
At the beginning, no variable is included in the model. Then, the variable exhibiting the
highest discriminatory power is selected. In the following steps, the variables that paired with
the initial variable lead to the highest increase of the model’s discriminatory power are
93
Cooper, Ron A. and Weekes, Tony J., (1983), “Data, Models and statistical analysis”, Philip Allan Publishers
Limited, New Jersey, USA, p. 280-285.
94
Ibid.
44 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
included. The selection process stops as soon as no further significant increase in the model’s
discriminatory power is achieved.
95
Stepdisc Backward Variables Selection
This selection procedure is equal to an elimination process. Initially, all variables are included
in the model. Then, the variables having the smallest impact on the model’s discriminatory
power are excluded. In this manner, the process ensures that only the best predictors are kept
in the DF.
96
3.6.2 Binary Logistic regression model (Binary LOGI T)
Binary LOGIT is an alternative to LDA when the dependent variable is dichotomous. In
contrast to LDA it does not require the explanatory variables to be normally distributed and
have equal variance and covariance, making it a more flexible and robust model.
97
As
financial data does not follow a normal but rather a leptokurtic distribution, binary LOGIT
appears to be more appropriate for the underlying study.
98
In practice, binary LOGIT is applied in many different fields, to determine the explanatory
variables that cause a separation of two groups from each other. One example is the
application of binary LOGIT in medical science to determine the factors for predicting the
emergence of heart diseases. The dependent variable is binary and comprises the two
outcomes i. heart disease and ii. no heart disease, which are equivalent to the two groups.
The explanatory variables (predictors) that allow for a classification of a patient in one of the
two groups would be e.g. age, weight, blood pressure, smoking habits etc.
99
The explanatory variables can be of quantitative or binary character, or a mixture of both. The
binary variable, whether dependent or explanatory, is encoded by the use of 0 and 1, where 0
denotes the absence of a situation and 1 denotes the presence of a situation respectively.
Thus, the binary logistic regression aims at explaining differences between two groups on the
basis of a common set of variables. It identifies the explanatory variables, which are most
95
SAS Institute Inc. 2010. SAS/STAT® 9.22 User’s Guide. Cary, NC: SAS Institute Inc.
96
Ibid.
97
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition, John Wiley &
Sons, Inc.
98
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press.
99
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
45 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
qualified to discriminate between two groups, as well as their direction and intensity of impact
on the respective group.
100
To give an example, based on our study the binary logistic regression extracts the relevant
variables out of a set of fifteen variables, which best separate our sample into turnarounds and
non-turnarounds. This allows us to make inferences about the decisive drivers in the
turnaround process.
The logistic function is given by:
101
Where, Y = dependent binary variable
? = Probability of the outcome of category I e.g. turnaround
1-? = Probability of the outcome of category II e.g. non-turnaround
X = explanatory variable
? = intercept
? = regression coefficient
u = random disturbance term
The above equation shows that binary LOGIT predicts “the probability that a case will be
classified into one as opposed to the other of the two categories of the dependent variable”.
102
This is known as the odds ratio, which can be expressed as follows:
Where, P (Y = 1) = Probability of Y = 1
1 – P (Y = 1) = Probability of Y ? 1
The probability of turnaround is defined by:
This equation is known as the logistic regression equation.
103
100
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
101
Peng, Chao-Ying J., Lee, Kuk L. and Ingersoll, Gary M., (2002), “An Introduction to Logistic Regression
Analysis and Reporting”, The Journal of Educational Research, Vol. 96, No. 1, pp. 3-14.
102
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications, Inc.
46 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Taking the natural logarithm of the odds ratio answers the purpose of constraining the
estimated probability within the boundaries of 0 and 1.
104
In this way, values for the
dependent variable will lie between 0 and 1, as shown in the graph below.
Graph 10: Logistic curve model for a dichotomous dependent variable
105
For the logistic regression, SPSS sets the cut-off point automatically at 0.5. With respect to
our study, we encoded turnarounds as ones and non-turnarounds as zeros. Thus, turnarounds
should yield a score above 0.5 and non-turnarounds below 0.5 respectively, on condition that
they are correctly classified. If graph 10 was the classification result of the binary LOGIT of
financially distressed firms based on an explanatory variable e.g. size, according to the
beforehand mentioned encoding, the four dots above 0.5 would match firms classified as
turnarounds and the three dots below 0.5 would correspond to firms classified as non-
turnarounds.
There exist two stepwise procedures to extract the most qualified predictors for discriminating
between the two categories, out of a set of explanatory variables.
Forward Conditional Logistic Regression:
In the beginning block (step 0) no explanatory variable is included in the model, but only the
intercept. Then, explanatory variables are entered in a stepwise procedure. First, the
explanatory variable with the highest discriminatory power is included into the model, which
103
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
104
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications, Inc.
105
47 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
is consistent with the variable that has the highest statistically significant chi-square.
106
This
variable causes “the greatest change in the log-likelihood relative to a model not containing
the variable”
107
. The quality of fit of the model is indicated by the deviance (-2 Log
likelihood). A decreasing deviance indicates that the model fits the data well.
108
This process
is repeated until no further improvement in the model can be obtained, as no more statistically
significant chi-squares are computed.
109
As including additional predictors to the model
causes an upward bias of the goodness of fit measure, the model is penalized by an increase in
degrees of freedom.
110
Backward Conditional Logistic Regression:
The backward conditional logistic regression includes all of the variables in the beginning
block (step 0) and stepwise removes variables, which are estimated as statistically
insignificant. These are the explanatory variables, which demonstrate the largest p-value in
terms of the likelihood ratio chi-square test.
111
It stops removing variables when all of the
remaining predictors show a statistically significant contribution to the model.
112
106
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
107
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition, John Wiley &
Sons, Inc.
108
Ibid.
109
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
110
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press.
111
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd edition,
Chapman and Hall.
112
SPSS Regression 17.0
48 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 4
Empirical Findings
In this chapter the empirical findings of our study are presented and discussed. The results
obtained by the two approaches, LDA and LOGIT, are summarized and inferences about the
factors playing a decisive role in the turnaround process are made. In addition, the findings
provide answers regarding the rejection or non-rejection of the hypotheses developed in the
previous chapter. The performance of the models is assessed based on their forecasting
accuracy with respect to the in-sample data. The two models showing the highest in-sample
prediction accuracy are evaluated based on their prediction performance on a holdout sample.
Furthermore, the predictors proposed by the model presenting the best forecasting accuracy
are tested for normality and heteroskedasticity, thereby making allowances for possible
violations of the assumptions underlying LDA and LOGIT.
4.1 Initial situation
The time frame for investigating the impact of the explanatory variables on the turnaround
outcome embraced year two and three of the four year window, which was exhibited in graph
2. We also took into account the period spanning year one and two, the two years for which
all of the companies in our sample displayed a Z-score below 1.8. However, for this time
period the degree of discrimination between the two groups was very low, leading to large
misclassification errors. Below the time period of analysis is depicted graphically.
1991 1995
1992 1993 1994
1992 - 1993
LDA & LOGIT on year 2 & 3
Graph 11: Time period of analysis regarding the impact of the explanatory variables on the turnaround outcome
49 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.2 Results of LDA
The LDA was performed by application of the statistical program SAS. The first step
consisted of extracting the most qualified predictors out of our set of explanatory variables,
followed by the derivation of the DF and computation of the cut-off point. SAS offers four
approaches for determining the variables that discriminate best between two groups:
i. Means and correlation procedure
ii. Stepdisc Forward Variables Selection
iii. Stepdisc Backward Variables Selection
iv. Stepdisc Stepwise Variables Selection
All four procedures were taken into account. The results provided by each procedure are
presented hereafter.
4.2.1 Means and correlation procedure (Model I )
The table below was generated by SAS and includes means and standard deviations for both
groups and for each variable taken into consideration in our study.
Variable Mean Stand. Dev. Mean Stand. Dev.
X1 7.647,47000 24.791,22000 479,80522 1.100,58000
X2 0,07189 2,24010 4,00548 2,80354
X3 0,00905 0,10599 -0,09743 0,16494
X4 0,02040 0,49043 0,67246 1,08285
X5 -0,00299 0,16569 0,02719 0,27306
X6 -0,00204 0,44957 0,28519 0,63444
X7 -0,08362 1,00394 0,07012 1,14753
X8 1,00256 1,08011 1,53116 1,19166
X9 -0,04107 0,32322 0,11107 0,39723
X10 -0,04533 0,22052 0,01140 0,24605
X11 0,05215 0,28629 0,05551 0,43024
X12 -0,03622 0,14050 -0,00097 0,10077
X13 0,03945 0,26643 0,15411 0,45012
X14 122,95094 573,49013 28,72262 175,23403
X15 -0,24562 18,23560 -0,99882 9,24052
X16 -0,05992 0,41028 0,02149 0,56350
Status 0 Status 1
Table 10: Means procedure
Based on this table and on intuition a mean procedure prediction model was established. The
variable X2 in table 10, equivalent to the Z-score, has a large difference in the means between
50 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
the two groups. The intra-group standard deviation is almost equal for both categories. Thus,
we included X2 as a predictor that allows for discrimination between the two groups.
However, also X1 was included in our mean procedure prediction model, although the
variable shows an extremely high intra-group standard deviation for both groups.
Nevertheless, based on our intuition and the theoretical foundation presented in chapter 2, we
believe that it is a decisive factor in the turnaround process, enabling a separation between
turnarounds and non-turnarounds. We also examined the correlation matrix generated by SAS
without finding significant correlations among the included variables. Several mean procedure
prediction models were tested, covering various combinations of the explanatory variables. In
the following, the mean procedure prediction model that revealed the highest in-sample
forecasting accuracy is demonstrated. The DF of this model is given by:
X
1
= total tangible assets
X
2
= Z-score
X
3
= change in total debt/total assets
The relationship between the predictors suggested by the means and correlation procedure and
the categories turnarounds and non-turnarounds is displayed in table 11.
Variable Coefficients (0) Impact direction Coefficients (1) Impact direction
X1 0,0000216 + -5,00E-06 -
X2 -0,0006751 - 0,6262 +
X3 0,41596 + -4,38467 -
Table 11: Relation between predictors and categories for X1 X2 X3
The table corroborates the discriminating power of the selected predictors, as their attached
coefficients have an opposite sign for each of the two categories. According to the model,
there exists a positive relation between size and non-turnarounds, while size and turnarounds
are negatively related. Thus, the model suggests that smaller firms have a higher probability
to be successful in the turnaround process. Even though the coefficient size is close to zero, it
has a significant impact on the Z
CEGA
, considering that the variable size includes large
numerical values (firm’s total assets). Moreover, firms for which the state of distress is less
severe, as measured by the Altman Z-score (X2), are more likely to achieve a turnaround. The
last variable considered to be decisive in the turnaround process is the change in the ratio total
debt to total assets (X3). According to the model, a decrease in the ratio is associated with a
successful recovery from financial distress. This empirical result was expected intuitively, as
51 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
financial distress is traced back to excessive leverage, in most of the cases. The magnitude of
the coefficient underpins the importance of the variable in the turnaround process, stating that
a reduction of X3 will trigger a substantial increase in the discriminant score Z
CEGA
.
We computed Z
CEGA
scores for the firms in our in-sample and calculated the cut-off point,
which accounted for – 0.00016632. The prediction accuracy of the model was estimated based
on our in-sample, which involved 150 firms. The classification matrix is provided below.
Status 0 1 Total
0
83
96,51%
3
3,49%
86
100%
1
8
12,50%
56
87,50%
64
100%
Total 91 59 150
Table 12: Classification Matrix in-sample X1 X2 X3
The blue shaded areas in the classification matrix display the correctly categorized firms per
outcome. The prediction accuracy on the in-sample amounted to 92.7%, which is computed
by taking the sum of the correctly classified firms divided through the total number of firms.
Misclassifications were restricted to 3 type I errors and 8 type II errors.
4.2.2 Stepdisc procedures (Model 2)
All of the three stepdisc procedures selected the same variables for inclusion in the final DF.
Thus, the DF on the basis of the stepdisc procedures is given by:
X
1
= total tangible assets
X
2
= Z-score
X
3
= change in total debt/total assets
X
6
= change in solvency ratio
Table 13 provides an overview of the coefficients belonging to the probability distributions of
turnarounds and non-turnarounds. They indicate direction and size of the impact of the
respective explanatory variables on the turnaround process.
52 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Coefficients (0) Impact direction Coefficients (1) Impact direction
X1 0,0000216 + -4,77E-06 -
X2 -0,0007326 - 0,62249 +
X3 0,42103 + -4,05877 -
X6 0,01242 + 0,79903 +
Table 13: Relation between predictors and categories for X1 X2 X3 X6
The first three variables are equivalent to the variables chosen by the means and correlation
procedure and their attached coefficients coincide in sign and are of similar magnitude.
However, the stepdisc procedures selected X6 as an additional variable, which corresponds to
the change in the solvency ratio. Although the coefficients of the predictor X6 are positive for
both groups, the coefficient belonging to turnarounds is of larger numerical value, leading to
the conclusion that companies with an increase in the solvency ratio are more likely to belong
to the group of turnarounds.
The cut-off point for the DF of the second model was computed based on the Z
CEGA
discriminant scores of the in-sample firms and amounted to -0.000165657. Below, the
classification matrix for in-sample data is displayed.
Status 0 1 Total
0
83
96,51%
3
3,49%
86
100%
1
10
15,63%
54
84,37%
64
100%
Total 93 57 150
Table 14: Classification Matrix in-sample X1 X2 X3 X6
The in-sample forecasting accuracy of model 2 adds up to 91.33%, falling short of the
prediction performance of model 1 by only 1.37%.
4.3 Results of LOGIT
The LOGIT was conducted by use of SPSS 17.0. Two procedures were applied to obtain the
independent variables acting as predictors, so that two LOGIT functions were computed. The
models were assessed regarding to their forecasting accuracy relative to the in-sample data.
The procedures were performed at different confidence intervals, which varied from 95% to
85%, taking into account that additional variables might be viewed as significant for
differentiating into the two groups, given lower confidence intervals. However, no additional
53 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
variables were included in the LOGIT function at lower confidence intervals. Hence, the
presented results were obtained at a confidence interval of 95%.
We excluded the variable severity of distress (X
2
) from the set of potential discriminating
variables. The LOGIT is based on a different algorithm than the LDA. It computes the
probabilities of a company, being classified in one of two groups. As our definition of a
turnaround was connected to a firm’s Z-score, inserting the variable severity of distress
(measured by Z-score) as a potential discriminator between turnarounds and non-turnarounds
into the logistic algorithm of SPSS, yielded a logistic function consisting solely of the Z-score
and displaying a prediction accuracy of 100%. Since this result is flawed, because of
conformity between the defining variable and a potentially predictive variable, we conducted
the LOGIT without inclusion of the Z-score.
4.3.1 Forward Conditional Logistic Regression (Model 3)
This procedure created a subset of three variables, which it regarded as eligible for
categorizing financially distressed firms into turnarounds and non-turnarounds.
The LOGIT function is given by:
X
3
= change in total debt/total assets
X
4
= change in total equity
X
6
= change in solvency ratio
When interpreting the regression coefficients, it is important to bear in mind that the above
LOGIT function complies with the log-odds of turnarounds.
Referring to the ratio total debt to total assets (X
3
), LOGIT estimates a negative correlation
between the predictor and the outcome turnaround. Hence, an increase in the ratio will lower
the chances for turnaround, while a decrease will promote recovery from financial distress. As
opposed to this, the model suggests that change in solvency ratio (X
6
) is positively related
with the probability of turnaround. The predictors X
3
and X
6
are represented in at least one of
the two LDA-based models. Change in total equity (X
4
), not taken into account by LDA, is
expected to be positively connected with turnaround. Raising new equity will trigger an
increase of the turnaround likelihood, as it reshapes the capital structure and allocates
financial funds, which are disposable for deleveraging or investments in value-creating
projects.
54 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The cut-off point for the LOGIT was set automatically by SPSS at 0.5. For each firm the
probability of turnaround is computed by means of the values of the predictors. A company is
classified as a turnaround, if the computed probability exceeds 0.5. The in-sample
classification matrix is provided below.
Status 0 1 Total
0
77
89.5%
9
10.5%
86
100%
1
26
40.6%
38
59.4%
64
100%
Total 103 47 150
Table 15:Classification Matrix in-sample X3 X4 X6
The in-sample forecasting accuracy of the model suggested by LOGIT forward amounts to
76.67%, being around 16% lower than the prediction accuracy given by LDA-model 1.
4.3.2 Backward Conditional Logistic Regression (Model 4)
The backward procedure included four variables in the LOGIT function. In addition to the
three variables considered by the forward procedure, it also took into account total tangible
assets (X
1
).
X
1
= total tangible assets
X
3
= change in total debt/total assets
X
4
= change in total equity
X
6
= change in solvency ratio
According to the model, total tangible assets negatively influence the turnaround likelihood,
allowing us to infer that larger firms are less likely to recover from financial distress than
smaller firms. The same conclusion was drawn from the LDA models. In respect of the other
three variables, the relationship assumed by model 4 coincides with the relationship that
model 3 predicted. To grasp the prediction accuracy of model 4, an in-sample classification
matrix was generated.
55 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Status 0 1 Total
0
72
84%
14
16%
86
100%
1
20
31%
44
69%
64
100%
Total 92 58 150
Table 16:Classification Matrix in-sample X1 X3 X4 X6
The model’s forecasting accuracy accounts for 77.33%. Thus, the inclusion of total tangible
assets as an additional predictor led to a marginal improvement of the prediction power by 66
basis points. However, also the second LOGIT model performs significantly poorer than the
LDA models in categorizing financially distressed companies into turnarounds and non-
turnarounds.
4.3.3 Model comparison, selection and interpretation of results
The table below provides an overview of the in-sample prediction accuracy and the number of
misclassifications of all four models.
Model 1 Model 2 Model 3 Model 4
Forecasting accuracy 92,70% 91,33% 76,67% 77,33%
Type I errors 3 3 9 14
Type II errors 8 10 26 20
LDA LOGIT
Table 17:Comparison of in-sample forecasting accuracy among models
The first LDA model shows the highest forecasting accuracy and makes the fewest
misclassification errors, followed by the second LDA model. The LOGIT models are inferior
to the LDA models, making three to five times more type I errors and two to three times more
type 2 errors, which leads to a much lower forecasting accuracy.
As the performance of the two LDA models is almost equally good, we examine their ability
to make correct forecasts by means of a holdout sample, which comprises 3140 companies.
The classification matrices for both LDA models are displayed below.
56 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Status 0 1 Total
0
2606
88,85%
327
11,15%
2933
100%
1
17
8,21%
190
91,79%
207
100%
Total 2623 517 3140
Table 18: Classification Matrix Holdout Sample X1 X2 X3
Status 0 1 Total
0
2481
84,59%
452
15,41%
2933
100%
1
49
23,67%
158
76,33%
207
100%
Total 2530 610 3140
Table 19: Classification Matrix holdout sample X1 X2 X3 X6
LDA model 1 has a prediction accuracy of 89% in the holdout sample. Type I and type II
errors increased, as did the sample size. The considerable rise in type I errors is explained by
the boost of non-turnarounds from 86 in the in-sample to 2933 in the holdout sample.
The second LDA model classifies only 84% of the cases correctly. Hence, model 2
misclassifies significantly more firms than model 1, especially in terms of type II errors,
which occur almost three times as often as in the first model. On these grounds, LDA model 1
was chosen as the best model for discriminating between turnarounds and non-turnarounds.
57 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
The two graphs below visualize the model’s forecasting accuracy for the in-sample and the
holdout sample. Considering that the cut-off point of the LDA-model 1 was close to zero, the
two graphs display very well the amount of committed type 1 and type 2 errors.
Graph 12: Z
CEGA
scores for turnarounds and non-turnarounds
For the in-sample, all Z
CEGA
scores of the type 1 errors were below 0.5, not so far away from
the cut-off point. The most external type 2 errors were lying between -2 and -1. Allowing for
a margin of safety, firms displaying Z
CEGA
scores larger than 1, can be perceived as real
turnaround candidates, with respect to the in-sample.
Graph 13: Z
CEGA
scores for turnarounds and non-turnarounds
For the holdout sample the most external type 2 errors were again laying between -2 and -1.
However, the type 1 errors were more dispersed with some exhibiting a Z
CEGA
score larger
0%
5%
10%
15%
20%
25%
30%
35%
40%
Failed Turnaround
Successful turnaround
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Failed Turnaround
Successful turnaround
58 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
than 2. Thus, although the probability of selecting a non-turnaround decreases with an
increasing Z
CEGA
score, the risk cannot be ruled out completely.
4.4 Model interpretation
Since this model contained only the three variables total tangible assets (X
1
), the Z-score as a
measure of distress severity (X
2
) and the ratio total debt to total assets (X
3
), no direct
relationship could be identified between the dependent variable and the remaining thirteen
explanatory variables outlined in table 8. Hence, the model assumes that they are statistically
insignificant in determining the turnaround potential of a financially distressed company,
leading to a rejection of the developed hypotheses for these particular variables. Table 20
displays the proposed impact of the three explanatory variables, which entered the final DF of
model 1 and compares it to the impact revealed by the model.
Variable Expected impact Revealed impact Developed hypothesis
X1 + - rejected
X2 - - not rejected
X3 - - not rejected
Table 20: Comparison of expected and revealed impact of predictors
4.4.1 Size (X
1
)
With respect to firm size, measured by total tangible assets, we developed the hypothesis that
the larger the firms the higher their turnaround potential. We motivated our hypothesis by
stating that larger firms can easier access capital markets to raise external funds. Moreover
they dispose of the possibility to sell assets, such as e.g. unrelated business units, to generate
internal funds. However, our hypothesis was rejected, as the model reveals that smaller
companies are more likely to succeed in the turnaround process. A reason for that can be their
swiftness in implementing strategic changes, as was argued by Paint (1991)
113
. Another
possible argument would be that small companies are accustomed to share a closer
relationship with its stakeholders, because there are less hierarchical levels, which obstruct the
flow of communication and information. As a consequence, it might be easier for smaller
firms to assure stakeholder support compared to larger firms.
4.4.2 Severity of distress (X
2
)
Measuring the severity of distress by the Z-score is reasonable, as it indicates the firm’s
probability of experiencing bankruptcy. Less severe distressed firms face fewer hurdles when
113
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in corporate
turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
59 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
approaching capital markets and can easier convince stakeholders to support them in the
restructuring process as firms in a more severe state of distress. The model did not reject our
developed hypothesis, so that severity of distress is negatively related with turnaround
potential.
In order to understand the concept of severity of distress, one has to keep in mind that it is
measured by the Altman Z-score. For a company displaying a high Z-score, the state of
distress is less severe than for a company with a low Z-score. This implies a negative
correlation between Z-score and severity of distress. The coefficients in table 11 for X
2
refer
to the Z-score. According to them, turnarounds (1) and Z-score are positively correlated,
while non-turnarounds (0) and Z-score are negatively correlated, leading to a negative
relationship between turnaround potential and severity of distress. The model suggests:
i. The higher the Z-score is, the more likely the turnaround outcome. As a high Z-
score implies low severity of distress, the model suggests a negative relation
between turnaround potential and severity of distress.
ii. The lower a firm’s Z-score is, the more likely the non-turnaround outcome. As a
low Z-score implies high severity of distress, the model suggests a negative
relation between turnaround potential and severity of distress.
4.4.3 Total debt to total assets (X
3
)
The model estimated a negative relation between change in the ratio total debt to total assets
and turnaround potential, corroborating our hypothesis. Both, deleveraging and expansion of
the asset base are assumed to support a firm in the process of recovery. Debt reduction will
send out a positive signal, strengthening stakeholder support. Expanding the asset base by
undertaking value-creating investments is an important step in the restructuring process of a
firm striving for strategic reorientation and will generate future cash-flows, available for
further investments or deleveraging.
Hence, we believe that changes in this ratio can reveal the implementation of efficiency-
oriented or entrepreneurial-oriented strategies, or a combination of both. Therefore we
scrutinized the in-sample firms to gauge what drives changes in total debt to total assets.
60 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Turnarounds Non-turnarounds Total
+?total assets 42 48 90
-?total assets 22 38 60
Total 64 86 150
+?total debt 17 43 60
-?total debt 47 43 90
Total 64 86 150
+?sales 50 54 104
-?sales 14 32 46
Total 64 86 150
Table 21: ? Total assets ? Total debt ? total sales
The above table depicts the changes in total assets, total debt and total sales for all firms in the
in-sample, separated by turnarounds and non-turnarounds. We also focus on change in total
sales, because we suppose that an increase in the asset base will trigger a rise in total
revenues, given that value-creating investments have been undertaken. The rows highlighted
in red depict significant differences between turnarounds and non-turnarounds. Considering
that the in-sample consists of 43% turnarounds and 57% non-turnarounds, a difference of a
factor larger than 1.5 between the two groups with respect to a variable was regarded as
significant. It appears that non-turnarounds are more prone to reducing the asset base than
turnarounds. Regarding total debt, turnarounds rather refrain from approaching external
financer and centre on deleveraging. For non-turnarounds the situation is different, with half
of them increasing the debt level and the other half decreasing it.
Aiming at obtaining a more concrete picture about the strategies employed by the firms in our
sample, we examined the direction of change in the three variables stated in table 21 for each
single company and formulated eight strategies, which were rated as efficiency-oriented,
entrepreneurial-oriented or a combination of both and are displayed in graph 14. However,
none of the strategies could be identified as being only efficiency-oriented.
61 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Graph 14: Overview of developed strategies
Table 22 presents a summary of the classification of each in-sample company to a strategy
group.
Strategies Turnaround Non-turnaround Total
1 5 12 17
2 2 5 7
3 2 5 7
4 10 23 33
5 25 10 35
6 3 10 13
7 4 5 9
8 13 16 29
Total 64 86 150
Table 22: Employed strategies by financially distressed firms
• Decrease in total assets, total
sales, total debt
• combination of both
Strategy 1
• Decrease total assets, total
sales and increase total debt
• entrepreneurial-oriented
Strategy 2
• decrease total assets and
increase total sales, total debt
• combination of both
Strategy 3
• increase total assets, total sales,
total debt
• entrepreneurial-oriented
Strategy 4
• increase total assets, total sales
and decrease total debt
• entrepreneurial-oriented
Strategy 5
• increase total assets, total debt
and decrease total sales
• entrepreneurial-oriented
Strategy 6
• increase total assets and
decrease total sales, total debt
• entrepreneurial-oriented
Strategy 7
• decrease total assets, total debt
and increase total sales
• combination of both
Strategy 8
62 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Four of the eight strategies were highlighted due to their frequency in appearance among both
groups and the fact that they were applied more often by one group than the other, rendering
possible to make conclusions about differences in the strategies pursued by turnarounds and
non-turnarounds.
Strategy 1:
Compliant with our sample this strategy was implemented by 12 non-turnarounds and 5
turnarounds. It is a combination between efficiency-oriented and entrepreneurial-oriented
strategies, as asset divestments can be operational or strategic or both. Hence, turnarounds and
non-turnarounds engaged in operational and/or strategic asset divestments, freeing up
financial funds that can be used for deleveraging. Simultaneously, divestments led to
reductions in turnover, due to decrease of capacitance and/or exit of business segments.
For non-turnarounds following this strategy, recovery might fail because the distress state is
too severe, so that they are forced to sell the most profitable segment, as was proposed by
Schlingemann et. al (2002)
114
. Even though this will cause a unique cash-inflow, the sale of
the most profitable segment will deteriorate the firm’s ability to generate turnover in the
future.
Firms succeeding with this strategy, rather clean up their portfolio by divesting unrelated
businesses and focusing on core capabilities.
Strategy 4:
This strategy was adopted twice as often by non-turnarounds than by turnarounds with respect
to our sample. It is denoted as an entrepreneurial-strategy, where strategic asset investments
are undertaken to re-orientate in the market. The strategic reorientation is financed with
outside capital and leads to sales increases, as value-creating investments are assumed to be
realized.
As contrasted with turnarounds, non-turnarounds sticking to this strategy might fail in the
event that the rise in sales provided by new investments cannot absorb the aggravated gearing,
resulting in an exacerbation of the distress state.
114
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the Liquidity of the
Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1, pp. 117-144.
63 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Strategy 5: More than twice as many turnarounds followed this strategy, compared to non-
turnarounds. It is termed to be an entrepreneurial-oriented strategy, where firms undertake
strategic investments through use of internal funds, which are obtained for example by raising
new equity. The additionally generated sales increase the cash-flows, which in return are
employed to pay down debt. Non-turnarounds might come unstuck with this strategy, if the
undertaken investments do not generate sufficient cash-flows to reach a relief of the
oppressive debt burden.
Strategy 6: This variable combination is predominant in the sub-sample of non-turnarounds. It
is thought as an entrepreneurial-oriented strategy, which misses the point. An increase in total
assets, whether it is related to capacity expansion or business diversification, should aim at
triggering a rise in total sales, as exemplified by strategy 4. The appearance of the opposite
can be motivated through value-destroying investments, for which costs exceed generated
cash-flows. Such a situation paired with rising debt levels will worsen the severity of distress.
For the three turnarounds displaying this variable combination, the change was marginal and
might be induced by temporary demand fluctuations distorting forecasted capital budgeting.
An example would be a company overestimating demand, leading to temporary inflated
inventory levels.
Although, there seems to be a difference in strategies applied by turnarounds and non-
turnarounds in the restructuring process, no specific strategy could be identified by our
sample that was solely implemented by one of the two groups. Ultimately, all above outlined
strategies, with exception of the flawed strategy 6, can result in a successful turnaround when
implemented correctly. To choose the adequate strategy is task of the management team. For
example, adopting an entrepreneurial-oriented strategy can bring about financial recovery, if
the cause of distress lies in a misfit between strategy of the company and operating markets.
Nevertheless, given that disadvantages in efficiency provoke financial distress, a strategic
reorientation is unlikely to bring the company back on track. An important finding of this
detailed analysis about what lies behind the changes in the ratio total debt to total assets is that
there exists evidence of turnarounds making use of efficiency-oriented and entrepreneurial-
oriented strategies.
64 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
4.5 Stock returns of turnarounds from the holdout sample
We computed the excess returns of the holdout sample firms that were identified as
turnarounds by the LDA-model 1. The S&P 500 was used as a benchmark index. The results
are presented in table 23.
Our model S&P 500 Excess Return Our model S&P 500
2004 30.9% 12.0% 18.9% 1.31 1.12
2005 38.3% 4.9% 33.4% 1.81 1.17
2006 28.8% 15.8% 13.0% 2.33 1.36
2007 -9.3% 5.6% -14.9% 2.12 1.44
2008 -47.1% -37.0% -10.1% 1.12 0.91
2009 111.7% 26.5% 85.2% 2.37 1.15
2010 34.9% 15.1% 19.8% 3.20 1.32
Total Return 220% 32%
Cumulative Yearly
Table 23: Excess returns of identified holdout sample turnarounds
2004-2010
The yearly returns of the identified turnarounds substantially exceeded the returns yielded by
the S&P 500. Only for 2007 and 2008, the time of the manifestation of the financial crisis, the
returns realized by our model fall short of the returns provided by the S&P 500. A possible
reason for this is that in times of crisis investors may be overly pessimistic and put not much
faith in the prospects of a recently financial distressed company. Nevertheless, an average
annual excess return of 14% over the average annual return yielded by the S&P 500 is a good
reason for considering investing in turnarounds identified by our model.
4.6 Variable testing
As the chosen model is generated by LDA, we test whether the residuals of the included
variables violate the assumptions that are imposed by discriminant analysis. The tests are
restricted to the in-sample data.
4.6.1 Normality test
The Bera-Jarque test is applied to examine whether the residuals of the included variables are
normally distributed. A normal distribution is characterized by its bell-shaped, symmetric
form, having a kurtosis of 3. The null-hypothesis of the test assumes normally distributed
residuals.
115
Table 24 summarizes the results of the Bera-Jarque test conducted in EViews.
115
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press, p.
181-182.
65 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Variable Skewness Kurtosis p-Value
X1 9,79 109,38 0,00
X2 0,45 8,45 0,00
X3 -0,26 7,91 0,00
Table 24: Results of Bera-J arque test
The distributions of the residuals of the predictors were skewed and had excess kurtosis.
Besides, the null-hypothesis of the Bera-Jarque test was rejected for all three variables,
concluding that they are not normally distributed. We did not expect something different, as
financial data uses to display a leptokurtic distribution.
116
4.6.2 Heteroskedasticity test
LDA assumes constant variance in the residual terms of the model, which is denoted as
homoskedasticity. The opposite is known as heteroskedasticity and describes the situation
where the residual terms do not display a constant variance.
117
We apply the Breusch-Pagan-
Godfrey test, in order to investigate whether our established LDA model shows signs of
heteroskedasticity. The null-hypothesis of the Breusch-Pagan-Godfrey test assumes
homoskedasticity of the residual terms. The test is performed in EViews and the test statistic
is presented below.
Heteroskedasticity Test: Breusch-Pagan-Godfrey
F-statistic 1.777394 Prob. F(3,146) 0.1541
Obs*R-squared 5.285244 Prob. Chi-Square(3) 0.1521
Scaled explained SS 1.543019 Prob. Chi-Square(3) 0.6724
Graph 15: Breusch-Pagan-Godfrey test statistic output for LDA-model 1
The results of the Breusch-Pagan-Godfrey test provide no indication for heteroskedasticity.
All three versions of the test statistic have p-values exceeding the critical threshold of 0.05,
concluding that the null-hypothesis of homoskedasticity cannot be rejected. Thus, the chosen
LDA model did not violate the assumption of constant variance.
4.6.3 Multicollinearity test
Last but not least, we address the problem of multicollinearity, which occurs when the
predictors are highly correlated with each other. We define high correlation by a correlation
116
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge University Press,
p. 179-180
117
Ibid, p. 147
66 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
coefficient greater or equal to 0.7. The following correlation matrix was generated by SAS
and belongs to the first LDA model, which showed the highest prediction accuracy.
Variables X1 X2 X3
X1 1,000 -0,04786 0,09415
X2 -0,04786 1,000 -0,28953
X3 0,09415 -0,28953 1,000
Table 25: Correlation Matrix X1 X2 X3 LDA model 1
As can be seen, the correlation among the predictors was very low and no correlation came
close to the critical value of 0.7. Accordingly, our model does not face a multicollinearity
issue.
67 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Chapter 5
Conclusion
The underlying thesis aimed at deriving the decisive factors in the turnaround process and
establishing a prediction model, thereby making possible to discriminate between future
turnarounds and non-turnarounds. The employed models were Linear Discriminant Analysis
and Logistic Regression and the model with the highest prediction accuracy was derived by
the LDA-approach and amounted to 92.7%. Taking into account that testing the model’s
prediction accuracy on the same sample, which was used to establish the model, will trigger
an upward bias in the results, the model was further evaluated based on its forecasting ability
on a holdout sample consisting of 3140 financially distressed firms. A marginal decrease in
the forecasting accuracy to 89% was recorded.
The selected variables failed to discriminate between turnarounds and non-turnarounds for the
first two years, where all companies in our sample displayed a Z-score below 1.8. A
significant discrimination between the two groups was obtained when focusing on year 2 and
3, where some companies realized an alleviation of the severity of distress, while for other
firms the distress state remained unchanged or even became worse.
With respect to the included predictors, the chosen model restricted itself to 3 explanatory
variables, which were firm size, severity of distress and total debt to total assets. While the
developed hypotheses for severity of distress and total debt to total assets were corroborated
by the empirical results of the chosen LDA-model, the hypothesis that firm size and
turnaround potential are positively related was rejected. Hence, for our sample smaller firms
are more likely to succeed in the turnaround process than larger firms. A possible explanation
is that smaller firms can implement entrepreneurial-oriented strategies faster and meet with
less resistance from high-level, old-established managers, who might interpret strategic
changes as a critique of their decision-making ability. Smaller firms tend to pursue a corporate
policy that is less dominated by complex hierarchical structures, which impede quick
decision-making. As severity of distress is measured by the Altman Z-score, this variable
substantiates the importance of the implementation of efficiency-oriented strategies to
improve financial performance, because the score is compounded of five variables that fall
into the category of efficiency measures. Changes in the ratio total debt to total assets can be
driven either by efficiency-oriented or by entrepreneurial-oriented strategies, or by a mix of
both. Which strategy is adopted and to what extent is determined by the cause of distress.
68 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Having at hand a turnaround prediction model with a high forecasting accuracy opens new
possibilities in yielding excessive returns from investing in distressed companies that will
transform to turnarounds. We briefly touched upon these opportunities by computing the
excess returns of the holdout sample turnarounds over the S&P 500, which amounted to 14%
on an annual basis.
Last but not least, we suggest that further research should aim at modeling the impact of
qualitative variables on the turnaround potential, such as internal firm climate and CEO
turnover. Moreover, other quantitative variables that are less firm-specific and more industry-
specific should be considered, like e.g. the growth of the industry a company is classified to.
69 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
References
Published References
Afifi, Abdelmonem and Clark, Virginia A., (1996), “Computer-aided multivariate analysis”, 3rd
edition, Chapman and Hall.
Altman, Edward I., (1998). “Market Dynamics and Investment Performance of Distressed and
Defaulted Debt Securities“, New York University, Center for Law and Business, Working Paper
No. 98-022.
Altman, Edward I., (2000). “Predicting Financial Distress of Companies: Revisiting the Z-score
and Zeta Models“, New York University, Center for Law and Business.
Altman, Edward I. and Jha, Shubin, (2003), “Market size and investment performance of
defaulted bonds and bank loans: 1987-2001”, Economic Notes, Vol. 32, Issue 2, pp. 147-176.
Arogyaswamy, K. and Yasai-Ardekani, M., (1997), “Organizational Turnaround: Understanding
the Role of Cutbacks, Efficiency Improvements, and Investment in Technology”, IEEE
Transactions on Engineering Management, Vol. 44, No. 1, pp. 3-11.
Balgobin, R. and Pandit, N., (2001), “Stages in the Turnaround Process: The Case of IBM UK”,
European Management Journal, Vol. 19, Issue 3, pp. 301-316.
Barker, Vincent L. and Barr, Pamela S., (2000), “Linking top manager attributions to strategic
reorientation in declining firms attempting turnarounds”, Journal of Business Research, Vol. 55,
Issue 12, pp. 963-979.
Barker, Vincent and Duhaime, Irene M., (1997), Strategic change in the turnaround process:
Theory and empirical evidence, Strategic Management Journal, Vol. 18, Issue 1, pp. 13-38.
Barney, Jay B., (1986), “Types of competition and the theory of strategy: Toward an integrative
framework”, Academy of Management, Vol. 11, No. 4 , pp. 791-800.
Brooks, Chris, (2002), “Introductory Econometrics for Finance”, 1
st
Edition, Cambridge
University Press.
Brown, David T., James, Christopher M. and Mooradian, Robert M., (1993), “The information
content of distressed restructurings involving public and private debt claims”, Journal of
Financial Economics, Vol. 33, No. 1, pp. 93-118.
Cameron, Kim S., Sutton, Robert I. and Whetten, David A., (1988), “Readings in Organizational
decline: Frameworks”, Ballinger Publishing.
Cooper, Ron A. and Weekes, Tony J., (1983), “Data, Models and statistical analysis”, Philip
Allan Publishers Limited, New Jersey, USA.
Cox, D. R. and Snell, E.J., (1989), “The analysis of binary data”, 2
nd
Edition, Chapman and Hall.
70 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Dechow, Patricia M. and Schrand, Catherine M., (2004), “Earnings Quality”, The Research
Foundation of CFA Institute.
Eberhart, Allan C., Altman, Edward I. and Aggarwal, Reena, (1998), The Equity Performance of
Firms Emerging from Bankruptcy, Journal of Finance, Vol. 54, Issue 5, pp. 1855-1868.
Fromm, Sabine, (2005), “Binäre logistische Regressionsanalyse”, Universität Bamberg.
Gilson, Stuart C., (1990), “Bankruptcy, boards, banks and bondholders – Evidence on changes in
corporate ownership and control when firms default”, Journal of Financial Economics, Vol. 27,
Issue 2, pp. 355-387.
Greenblatt, Joel, (1999), “You can be a stock market genius”, Fireside.
Grinyer, Peter H., Mayes, David and McKiernan, Peter, (1988), “Sharpbenders: The secrets of
unleashing corporate potential”, Blackwell Publishers, Oxford.
Hambrick, Donald C. and Schecter, Steven M., (1983), “Turnaround Strategies for Mature
Industrial-Product Business Units”, Academy of Management Journal, Vol. 26, No. 2, pp 231-
248.
Harker, Michael, (1996), Managing company turnarounds: how to develop “destiny”, Marketing
Intelligence & Plannings, Vol. 14, Issue 3 pp. 5-10.
Hedberg, Bo L. T., Nystrom, Paul C. and Starbuck, William H., (1976), Camping on Seesaws:
Prescriptions for a Self-Designing Organization, Administrative Science Quarterly, Vol. 21, No.
1, pp. 41-65.
Hofer, Charles W., (1980), Turnaround Strategies, Journal of Business Strategy, Vol. 1, Iss.1, pp.
19-31.
Hofer, Charles W. and Schendel, Dan, (1978), Strategy Formulation: Analytical Concepts, West
Publishing.
Hosmer, David W. and Lemeshow, Stanley, (2000), “Applied Logistic Regression”, 2nd Edition,
John Wiley & Sons, Inc.
Hotchkiss, Edith S., (1995), Postbankruptcy Performance and Management Turnover, Journal of
Finance, Vol. 50, Iss. 1, pp. 3-21.
Indro, D. C., Leach, R.T. and Lee, W. Y., (1999), Sources of gains to shareholders from
bankruptcy resolution, Journal of Banking & Finance, Vol. 23, Iss. 1, pp. 21-47.
Kahl, Matthias, (2002), Economic Distress, Financial Distress and Dynamic Liquidation, The
Journal of Finance, Vol. 57, pp. 135-168.
Karpoff, Jonathan M., (1987), “The relation between price changes and trading volume: A
survey”, The Journal of Financial and Quantitive Analysis, Vol. 22, No. 1, pp. 109-126.
Klarman, Seth. A.,(1991),Margin of safety: Risk-averse value investing strategies for the
thoughtful investor, Harper Business.
Koller, Tim; Goedhart, Marc and Wessels, David, (2010)Valuation: Measuring and managing the
value of companies, John Wiley & Sons.
71 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Lachenbruch, P. A., (1968), “On expected probabilities of misclassification in discriminant
analysis, necessary sample size, and a relation with the multiple correlation coefficient”,
Biometrics, Vol. 24, No. 4.
Lachenbruch, P. A. and Goldstein, M., (1979), “Discriminant Analysis”, Biometrics, Vol. 35, No.
1, pp. 69-85.
Lepak, David P., Smith, Ken G. and Taylor, M. Susan, (2007), Value Creation and Value Capture:
A Multilevel Perspective, Academy of Management Review, Vol. 32, No. 1, pp. 180-194.
Menard, Scott (1995), “Applied Logistic Regression Analysis”, 2
nd
Edition, Sage Publications,
Inc.
Myers, Stewart C., (1977), Determinants of Corporate Borrowing, Journal of Financial
Economics, Vol. 5, No. 2, pp. 147-175.
Ogden, Joseph P., Jen, Frank C. and O’Connor, Philip F.,(2003), Advanced Corporate Finance.
Policies and Strategies, Prentice Hall.
Paint, Laurie W., (1991), “An investigation of industry and firm structural characteristics in
corporate turnarounds”, Journal of Management Studies, Vol. 28, Issue 6, pp. 623-643.
Peng, Chao-Ying J., Lee, Kuk L. and Ingersoll, Gary M., (2002), “An Introduction to Logistic
Regression Analysis and Reporting”, The Journal of Educational Research, Vol. 96, No. 1, pp. 3-
14.
Robbins, D. Keith, and Pearce, John A., (1992), “Turnaround: retrenchment and recovery”,
Strategic Management Journal, Vol. 13, Issue 4, pp. 287-309.
Schendel Dan E. and Patton, G.R., (1976), “Corporate stagnation and turnaround”, Journal of
Economics and Business, Vol. 28, Iss. 3, pp. 236-242.
Schendel, Dan, Patton, G.R. and Riggs, James, (1976), “Corporate turnaround strategies: A study
of profit decline and recovery”, Journal of General Management, Vol. 3, Iss.. 3, pp. 3-12.
Schlingemann, Frederik P., Stulz, René M. and Walkling, Ralph A. (2002), “Divestures and the
Liquidity of the Market for Corporate Assets”, Journal of Financial Economics, Vol. 64, Issue 1,
pp. 117-144.
Slatter, Stuart and Johnson, Gerry, (1984), “Corporate Recovery: Successful turnaround strategies
and their implementation”, Strategic Management Journal, Vol. 7, Issue 1, pp. 99-100.
Smith, Malcom and Graves, Christopher, (2005), Corporate turnaround and financial distress,
Managerial Auditing Journal, Vol. 20, Iss. 3, pp. 304-320.
Sudarsanam, Sudi and Lai, Jim (2001), “Corporate Financial Distress and Turnaround Strategies:
An empirical analysis”, British Journal of Management, Vol. 12, Issue 3, pp. 183-199.
White, M. (1989), “Bankruptcy, liquidation and reorganization”, in: D.E. Logue (ed.), Handbook
of Modern Finance, Warren, Gorham & Lamont, New York
Ying, Charles C., (1966), “Stock Market Prices and Volumes of Sales”, Econometrica, Vol. 34,
No. 3, pp. 676-685.
72 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Zeni, Syahida Binti and Ameer, Rashid, (2010), Turnaround prediction of distressed companies:
evidence from Malaysia, Journal of Financial Reporting and Accounting, Vol. 8, Issue 2, pp 143-
159.
Internet References
Buffet, Warren E., (2008), “Buy American, I Am”, The New York Times,http://www.nytimes.com/2008/10/17/opinion/17buffett.html
Forbes (2010):http://www.forbes.com/lists/2010/10/billionaires-2010_Warren-Buffett_C0R3.html
Twin A. (2009-03-09), “For Dow another 12-year low”, CNN Money,http://money.cnn.com/2009/03/09/markets/markets_newyork/index.htm
World Federation of Exchanges:http://www.world-exchanges.org/statistics/time-series/market-capitalization
Historical Chart of S&P 500:http://www.forecast-chart.com/historical-sp-500.html
VassarStats, Logistic:
Manuals
SAS Institute Inc. 2010. SAS/STAT® 9.22 User’s Guide. Cary, NC: SAS Institute Inc.
SPSS Regression 17.0
Database
Statndard and Poor’s Database
(http://www.standardandpoors.com)
73 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix
Appendix 1: Overview of prior studies on turnaround determinants
Researchers
Types of turnaround and
nonturnaround firms
Measures of
management actions
Actions associated with
turnaround or improved
financial performance
Schendel and Patton (1976)
Manufacturing firms
matched by SIC code Financial ratios
Decreased costs/sales
Increased sales
Increased investment
Hambrick and Schecter (1983) Mature, industrial product SBUs
Financial ratios and some
perceptual measures
Decreased R&D expenditures/sales
Decreased marketing expenditures/sales
Decreased receivables/sales
Decreased inventory/sales
Increased employee productivity
Increased plant & equipment newness
Increased market share
Ramanujam (1984)
Undiversified manufacturing
firms Financial ratios
Decreased cost of goods sold/sales
Decreased inventory/sales
Decreased receivables/sales
Increased sales
Thietart (1988)
SBUs across varying industry
environments
Financial ratios and some
perceptual measures
Combination of actions that cut costs and
increase productivity
Robbins and Pearce (1992) Textile firms 1976 - 85
Financial statement changes,
some perceptual measures
Asset reduction
Cost reduction
Arogyaswamy (1992) Manufacturing firms
Financial ratios, financial
statement changes
Decreasing at least three of the following:
Employees/Sales
Receivables/Sales
Inventory/Sales
Cost of goods sold/sales
SGA Expenses/sales
Combining above efficiency posture with
increased R&D or plant expenditures
Increasing R&D expenditures
Not decreasing at least three of the
following:
Employees
Receivables
Inventory
Cost of goods sold
SGA expenses
74 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix 2: Firms included in the in-sample
I n-sample companies by turnaround outcome
Non-turnarounds Turnarounds
AeroCentury Corp. (1998-99)
Abaxis Inc. (1997-98)
Air Methods Corp. (1996-97)
AmSurg Corp. (1998-99)
AK Steel Holding Corporation (1998-99)
Ariad Pharmaceuticals Inc. (1997-98)
Akamai Technologies Inc. (2000-01)
ASM International NV (1997-98)
AMR Corporation (1999-00)
Atwood Oceanics, Inc. (1993-94)
Apache Corp. (1995-96)
Aurizon Mines Ltd. (2000-01)
Appliance Recycling Centers of America Inc.
(1996-97)
BioDelivery Sciences International Inc. (2000-
01)
Arabian American Development Company (1999-
00)
Biogen Idec Inc. (1993-94)
Avis Budget Group, Inc. (1999-00)
Biolase Technology, Inc. (2000-01)
Bally Technologies, Inc. (1996-97)
Bio-Reference Laboratories Inc. (1998-99)
Belo Corp. (1997-98)
Birner Dental Management Services Inc.
(2000-01)
Belo Corp. (1998-99)
Blue Dolphin Energy Company (1994-95)
Boyd Gaming Corp. (1999-00)
China Yuchai International Limited (2000-01)
Breeze-Eastern Corporation (2000-01)
Comfort Systems USA Inc. (2000-01)
Cabot Oil & Gas Corporation (1994-95)
Concurrent Computer Corporation (1995-96)
Cabot Oil & Gas Corporation (1996-97)
CPI Aerostructures Inc. (2000-01)
Carriage Services Inc. (1999-00)
Cray Inc. (2000-01)
Chyron Corporation (2000-01)
Eldorado Gold Corp. (2000-01)
CNH Global NV (1998-99)
Energy Conversion Devices, Inc. (1998-99)
Comstock Resources Inc. (1998-99)
EOG Resources, Inc. (1998-99)
Corrections Corporation of America (1999-00)
Fieldpoint Petroleum Corp. (1996-97)
Craft Brewers Alliance, Inc. (2000-01)
FLIR Systems, Inc. (1999-00)
Crown Holdings Inc. (1998-99)
Food Technology Service Inc. (1995-96)
CSX Corp. (2000-01)
Fuel-Tech, Inc. (1999-00)
Earthstone Energy, Inc. (1998-99)
Gardner Denver Inc. (1995-96)
El Paso Corp. (2000-01)
Golden Star Resources, Ltd. (1998-99)
Female Health Co. (1997-98)
H&R Block, Inc. (1999-00)
Fonar Corp. (2000-01)
Hallador Energy Company (1996-97)
Ford Motor Co. (1992-93)
Headwaters Inc. (1998-99)
Forest Oil Corp. (1999-00)
Hollywood Media Corp. (1997-98)
Furmanite Corporation (1995-96)
Hurco Companies Inc. (1994-95)
GATX Corp. (1991-92)
ImmuCell Corp. (1994-95)
Good Times Restaurants Inc. (2000-01)
Imperial Sugar Co. (1999-00)
GP Strategies Corp. (2000-01)
Insignia Systems Inc. (1997-98)
Gray Television Inc. (2000-01)
InterDigital, Inc. (1992-93)
75 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Non-turnarounds Turnarounds
Hallador Energy Company (1998-99)
Inventure Foods, Inc. (1998-99)
HearUSA Inc. (2000-01)
Itron, Inc. (1999-00)
Hecla Mining Co. (1996-97)
Ivanhoe Mines Ltd. (2000-01)
Heska Corp. (2000-01)
Kinross Gold Corporation (2000-01)
Hexcel Corp. (2000-01)
Laboratory Corp. of America Holdings (1998-
99)
Hill-Rom Holdings, Inc. (1999-00)
Magnetek Inc. (1998-99)
Hill-Rom Holdings, Inc. (2000-01)
Medifast Inc. (1999-00)
HKN, Inc. (1999-00)
Medifast Inc. (2000-01)
icad Inc. (1998-99)
MEMC Electronic Materials Inc. (2000-01)
InsWeb Corp. (2000-01)
New Dragon Asia Corp. (2000-01)
International Shipholding Corp. (1994-95)
NovaGold Resources Inc. (2000-01)
Isle of Capri Casinos Inc. (1998-99)
Onyx Pharmaceuticals, Inc. (1997-98)
Iteris, Inc. (2000-01)
Orbit International Corp. (1998-99)
Joe's Jeans Inc. (2000-01)
Palatin Technologies Inc. (1997-98)
LodgeNet Interactive Corporation (1997-98)
Quest Diagnostics Inc. (1998-99)
MEDTOX Scientific Inc. (1999-00)
Retractable Technologies Inc. (2000-01)
Mercer International Inc. (1999-00)
RTI International Metals, Inc. (1994-95)
Norfolk Southern Corp. (1997-98)
Schawk Inc. (1995-96)
Norfolk Southern Corp. (1999-00)
Simulations Plus Inc. (2000-01)
NTN Buzztime Inc. (1998-99)
Stericycle, Inc. (1994-95)
Occidental Petroleum Corporation (1998-99)
Stericycle, Inc. (1999-00)
Orbit International Corp. (2000-01)
Tri-Valley Corp. (1995-96)
Parker Drilling Co. (2000-01)
Tri-Valley Corp. (2000-01)
Perma-Fix Environmental Services Inc. (1994-95)
Unisys Corporation (1993-94)
Pride International Inc. (1997-98)
Unit Corp. (1993-94)
PRIMEDIA Inc. (1999-00)
Universal Security Instruments Inc. (2000-01)
Ramtron International Corp. (1998-99)
Valero Energy Corp. (1994-95)
Reynolds American Inc. (2000-01)
Verint Systems Inc. (2000-01)
Ryder System, Inc. (1994-95)
Western Digital Corp. (1998-99)
Service Corp. International (1994-95)
Service Corp. International (1995-96)
Sinclair Broadcast Group Inc. (1998-99)
Six Flags Entertainment Corp. (2000-01)
Southwall Technologies Inc. (1998-99)
StemCells Inc. (1997-98)
Streamline Health Solutions, Inc. (1999-00)
Swift Energy Co. (1991-92)
76 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Non-turnarounds
Temple-Inland Inc. (1991-92)
Tenneco Inc. (1999-00)
The Hallwood Group Incorporated (1998-99)
The Interpublic Group of Companies, Inc.
(2000-01)
Titan International Inc. (2000-01)
Titanium Metals Corporation (1999-00)
Union Pacific Corporation (1997-98)
Unisys Corporation (1996-97)
Valhi, Inc. (1995-96)
Viad Corp (1995-96)
Waste Connections Inc. (2000-01)
Waste Management, Inc. (1999-00)
Willis Lease Finance Corp. (2000-01)
Xerox Corp. (1999-00)
77 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Appendix 3: Scores model 1-4, in-sample
Model 1 Model 2 Model 3 Model 4
-1,41 -1,49 0,34 0,48
-1,06 -1,18 0,31 0,45
-0,72 -0,77 0,32 0,12
-3,18 -2,99 0,02 0,08
-2,27 -2,61 0,15 0,00
-0,54 -0,63 0,32 0,17
-3,42 -3,61 0,03 0,09
-1,55 -1,57 0,30 0,43
-2,27 -2,41 0,38 0,00
-2,53 -2,71 0,02 0,04
-0,79 -0,84 0,31 0,17
-0,64 -0,74 0,33 0,18
-0,68 -0,80 0,30 0,30
-1,46 -1,85 0,08 0,13
-1,54 -1,76 0,16 0,26
-0,66 -0,80 0,35 0,42
-1,07 -0,77 0,45 0,51
-3,43 -2,27 0,54 0,67
-1,52 -1,64 0,36 0,00
0,16 0,19 0,54 0,61
-1,33 -0,91 0,49 0,44
-0,73 -0,81 0,31 0,44
-1,20 -1,34 0,21 0,01
-1,25 -1,32 0,32 0,00
-5,25 -4,71 0,71 0,77
-2,61 -2,83 0,17 0,00
-8,76 -8,45 0,19 0,32
-0,79 -2,20 0,18 0,28
-6,66 -6,68 0,34 0,00
-0,34 -0,42 0,35 0,34
-0,61 -0,69 0,32 0,43
-0,72 -0,78 0,36 0,23
-0,46 -0,43 0,33 0,47
-0,58 -0,81 0,33 0,44
-0,08 -0,27 0,86 0,85
-1,46 -1,68 0,34 0,46
-1,72 -1,54 0,60 0,68
-1,65 -2,16 0,15 0,26
LDA-models LOGIT-models
Z(CEGA) scores and predicted probabilities,
models 1-4, in-sample
78 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
-5,61 -4,60 0,36 0,49
-1,50 -1,11 0,41 0,48
-0,37 -0,39 0,36 0,06
-0,52 -0,58 0,42 0,17
-5,32 -5,63 0,12 0,21
-3,43 -3,40 0,28 0,43
-4,79 -4,76 0,29 0,42
-0,74 -0,85 0,29 0,37
-0,46 -0,52 0,66 0,69
-1,97 -2,20 0,27 0,39
0,03 0,13 0,64 0,70
-1,28 -1,34 0,10 0,15
-1,01 -0,72 0,39 0,55
-0,66 -0,87 0,30 0,40
-1,30 -1,44 0,29 0,00
-1,15 -1,22 0,32 0,00
-3,81 -4,50 0,37 0,53
-0,75 -0,72 0,56 0,00
-0,71 -0,67 0,39 0,52
-1,36 -1,58 0,16 0,22
-0,58 0,18 0,72 0,79
-1,07 -1,23 0,27 0,24
-2,68 -2,96 0,10 0,08
-2,44 -3,83 0,32 0,46
-1,39 -1,85 0,16 0,00
-0,57 -0,70 0,30 0,09
-0,85 -0,93 0,33 0,04
-0,77 -0,85 0,36 0,02
-1,01 -1,13 0,31 0,19
-1,28 -1,41 0,28 0,14
-1,41 -1,58 0,24 0,36
-5,21 -5,16 0,15 0,25
-0,91 -0,99 0,39 0,52
-0,35 -0,42 0,43 0,54
-0,85 -0,95 0,31 0,01
-1,38 -1,51 0,16 0,12
0,01 -0,26 0,38 0,48
-1,00 -0,99 0,39 0,02
-0,71 -0,80 0,24 0,33
-0,34 -1,42 0,16 0,23
-1,50 -1,48 0,35 0,00
79 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
-1,15 -1,54 0,21 0,04
-0,46 -0,55 0,38 0,34
-0,57 -0,66 0,30 0,35
-0,72 -0,76 0,38 0,00
-1,02 -0,99 0,43 0,25
-1,18 -1,26 0,30 0,40
-1,38 -1,46 0,33 0,00
2,08 3,30 0,61 0,70
-0,76 -0,91 0,20 0,33
0,18 0,15 0,67 0,76
2,11 2,01 0,79 0,81
0,89 0,82 0,44 0,54
3,99 3,92 0,70 0,78
2,08 1,93 0,88 0,88
6,06 6,73 0,57 0,66
0,34 1,05 0,96 0,98
-0,21 -0,03 0,39 0,53
0,43 0,67 0,45 0,57
-0,22 -0,20 0,25 0,40
0,41 1,08 0,60 0,36
1,01 0,95 0,45 0,50
0,70 0,70 0,42 0,57
1,74 2,79 0,97 0,98
0,19 0,16 0,66 0,76
5,47 6,57 0,87 0,89
1,54 1,54 0,86 0,93
1,39 1,15 0,44 0,31
-1,08 -1,38 0,07 0,16
5,38 6,19 0,99 0,99
2,64 1,56 0,98 0,97
2,02 1,34 0,21 0,33
0,65 0,75 0,49 0,60
1,78 2,88 0,98 0,98
0,14 0,20 0,32 0,17
2,81 2,76 0,51 0,62
4,55 5,58 0,99 0,99
10,02 9,85 0,81 0,82
0,91 0,90 0,77 0,83
0,26 0,07 0,27 0,41
0,27 -0,82 0,08 0,13
2,04 2,07 0,14 0,27
80 Turnarounds – Modeling the probability of a turnaround Master Thesis Spring 2011
Model 1 Model 2 Model 3 Model 4
0,54 0,74 0,57 0,61
-0,69 -1,45 0,14 0,32
1,51 1,65 0,72 0,78
-0,22 -0,21 0,28 0,38
0,53 0,71 0,46 0,59
7,57 7,50 0,89 0,94
1,41 1,40 0,24 0,34
0,37 1,20 0,60 0,69
-0,51 -0,43 0,05 0,11
0,41 0,39 0,45 0,33
1,57 0,35 0,60 0,66
4,55 4,41 0,98 0,98
1,35 1,64 0,60 0,67
4,74 4,68 0,76 0,85
0,72 0,59 0,54 0,64
0,83 1,06 0,87 0,90
0,15 0,15 0,26 0,39
1,31 1,30 0,79 0,89
0,59 1,12 0,64 0,26
-0,25 -0,34 0,32 0,45
1,48 2,40 0,85 0,86
1,03 1,02 0,57 0,48
1,01 1,12 0,62 0,69
1,14 1,04 0,60 0,64
doc_661398817.pdf