Corporate Finance

Description
Corporate finance is the area of finance dealing with the sources of funding and the capital structure of corporations and the actions that managers take to increase the value of the firm to the shareholders, as well as the tools and analysis used to allocate financial resources.

ABSTRACT

Title of dissertation:

EMPIRICAL ESSAYS IN CORPORATE FINANCE Kristina Leigh Minnick, Doctor of Philosophy, 2005

Dissertation directed by:

Professor Lemma Senbet and Professor Nagpurnanand Prabhala Department of Finance

Over the past twenty years, write-o?s have grown in popularity. With the increased usage of write-o?s, it is becoming more important to understand the mechanisms behind why companies take write-o?s and how write-o?s a?ect company performance. In this paper, I examine the cross-sectional determinants of the decision to take write-o?s. I use a hand-collected dataset on write-o?s that is much more comprehensive than existing write-o? datasets. Contrary to much hype and scandals surrounding a few write-o?s, I ?nd that quality of governance is positively related to write-o? decisions in the cross-section. My results also suggest that poor governance companies wait to take write-o?s until it becomes inevitable, while well-monitored companies take write-o?s sooner. As a result, the charge is substantially larger than the average write-o? charge. When these poor governance companies announce write-o?s, the announcement generates negative abnormal returns. However, when good corporate governance companies announce write-o?s, the charge is substantially smaller than the average charge. These well-monitored companies take write-o?s immediately following a problem. Following the write-o? announcements of these types of companies, average announcement day e?ects exceed a positive six percent. These results suggest that companies with quality monitoring mechanisms use write-o?s in a manner that is consistent with enhancing shareholder value. In my second essay I examine the e?ect of write-o? announcements on the stock market liquidity of ?rms taking write-o?s from 1980 to 2000. I ?nd that there are substantial improvements in stock market liquidity following corporate write-o?s. Spreads decrease and turnover volume increases after write-o? announcements, which indicates an improvement in liquidity. The liquidity

improvement is greater for better governed companies. I decompose bid-ask spreads and show that adverse selection costs decrease substantially as market participants respond to the write-o? announcement. The evidence suggests a liquidity bene?t of write-o?s that must be weighed against any other perceived cost of write-o?s. Such a liquidity bene?t may validate that write-o?s convey favorable information about the ?rm.

EMPIRICAL ESSAYS IN CORPORATE FINANCE

by Kristina Leigh Minnick

Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial ful?llment of the requirements for the degree of Doctor of Philosophy 2005

Advisory Committee: Professor Professor Professor Professor Professor Lemma Senbet, Chair/Advisor Nagpurnanand Prabhala, Chair/Advisor Gordon Phillips Oliver KIm Rachel Kranton

c Copyright by Kristina Leigh Minnick 2005

This dissertation is dedicated to my parents Pam and Tom Minnick.

ii

ACKNOWLEDGMENTS

I owe my gratitude to all the people who have made this dissertation possible. Because of you my graduate experience has been one that I will cherish forever. First and foremost, I would like to thank my advisors, Professors Nagpurnanand Prabhala and Lemma Senbet for giving me an invaluable opportunity to work on challenging and extremely interesting projects over the past ?ve years. They have always made themselves available for help and advice. It has been a pleasure to work with and learn from such extraordinary individuals. Thanks are due to Professor Gordon Phillips for agreeing to serve on my thesis committee and for sparing his invaluable time reviewing the manuscript. In addition, I am very thankful for the feedback that Professor Oliver Kim, Professor Ginger Jin, and Professor Rachel Kranton have provided along the way. I owe my deepest thanks to my family - my mother and father who have always stood by me and guided me through my career, and have supported me with every endeavor I take on. Words cannot express the gratitude I owe them. It is impossible to remember all, and I apologize to those I have inadvertently left out. Lastly, thank you all and thank God!

iii

TABLE OF CONTENTS

1

Write-o?s and Corporate Governance 1.1 1.2 1.3 1.4 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characteristics of Write-o? Companies . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 1.4.2 1.4.3 1.4.4 1.4.5 1.4.6 1.5 1.6 1.7 Corporate Cleanup Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . Executive Compensation and the Write-o? Decision . . . . . . . . . . . . . Monitoring Mechanism Hypothesis . . . . . . . . . . . . . . . . . . . . . . . Multivariate Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Weak Shareholder Protection and Write-o?s . . . . . . . . . . . . . . . . . . Governance and Size of Write-o?s . . . . . . . . . . . . . . . . . . . . . . .

1 1 2 5 14 14 18 20 23 25 26 28 30 32 43 44 45 46 47 48 49 51 52 53 65

Market Reaction and Write-o? Announcements . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Write-o?s and Liquidity 2.1 2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sample and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 2.3 2.4 2.5 2.6 2.7 2.8 Liquidity Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Liquidity E?ects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Governance and Liquidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adverse Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Bibliography

iv

Chapter 1 Write-o?s and Corporate Governance 1.1 Introduction
Write-o?s have become increasingly common in the past two decades. The consumer-manufacturing sector alone had write-o? charges totaling over $2 billion in 2000 compared to $800 million in 1980, an increase of 250 percent.

There are three potential explanations for why companies take write-o?s. First, write-o?s are a consequence of poor managerial decision-making. Write-o?s become inevitable actions for companies that are su?ering from a chain of management errors. Second, write-o?s can be a response by management proactively responding to negative shocks to the company. Even quality managers that take calculated risks can have problems ex-post. Management consciously decides to amend these problems by taking a write-o? that also provides private information to the market concerning the quality of the ?rm. Third, in companies with CEO turnover, write-o?s can act as a tool that allows the new CEO to get rid of bad accounts left by the previous CEO.

In this paper, I examine why companies take write-o?s and how the market reacts to writeo? announcements; for this analysis, I use a carefully collected dataset of consumer manufacturing companies, focusing on asset and lay-o? based write-o?s. I characterize what de?nes good and bad write-o?s, and analyze the characteristics of companies that take di?erent types of write-o?s. Finally, I examine the shareholder wealth e?ects of write-o?s, and whether ?rm speci?c factors in?uence the market’s reaction to the write-o? announcement.

I ?nd that the write-o? decision is linked to industry shocks. Governance mechanisms also

1

a?ect the write-o? decision. Companies with high pay-performance sensitivity, desirable board composition, strong shareholder protection measures, and CEO turnover resulting in an external replacement are all signi?cantly correlated to a tendency to take write-o?s. I ?nd a negative relationship between governance quality and the size of write-o?s, which suggests that poorly monitored companies wait to take write-o?s and continue to accumulate problems. Eventually the problems become so large that a write-o? is inevitable. Conversely, well-monitored companies take write-o?s sooner. Since these companies act quickly, there is comparatively less that they can write-o?, so the charges of well-monitored companies are less than the charges of poorly monitored companies. I also ?nd that these well-monitored companies exhibit signi?cant positive announcement e?ects (upwards of six percent for write-o? companies with small boards, strong shareholder protection, and large percentage of outside directors. I conclude from these results that ?rms with e?ective monitoring mechanisms take value-enhancing write-o?s.

The paper is constructed as follows. Section II reviews the relevant literature. I describe my sample in Section III. Section IV examines the link between CEO turnover, pay-performance sensitivity, corporate governance, and the write-o? decision. Section V looks at how the market reacts to write-o? announcements. Section VII concludes. Appendix A discusses the tax issues related to write-o?s.

1.2 Literature Review
Other studies look at write-o?s, but from di?erent perspectives and with results that are not directly comparable to mine. The write-o? literature focuses on three main areas, the e?ects of write-o?s on returns, the relationship of earnings and write-o?s, and the impact of SFAS 121 on write-o? announcements. However, my primary focus is to explore the relations between the governance of a ?rm and the motivation to take write-o?s as a business related decision.

2

The ?rst branch of literature, which looks at e?ects from write-o? announcements, has mixed results. Some papers ?nd that write-o?s generate no abnormal returns, while others ?nd that write-o? announcements generate both positive and negative abnormal returns depending on the segmentation of the sample. Francis, Hanna, and Vincent (1997) collect and analyze write-o?s from 1989 to 1992. Their analysis shows that on average the market views write-o?s as negative news, although it is possible to explain some of the dispersion in market reactions by identifying di?erent types of write-o?s, such as inventory or restructuring. Their study provides evidence that both earnings management and asset impairment drive a write-o? decision. Although Francis et al. use the same collection technique as I use, their sample spans fewer years, and contains fewer announcements. Their results motivate me to examine what role earnings management and asset impairment have in write-o? decisions of good governance companies.

Meyer and Strong (1987) identify a sample of 78 write-o? ?rms from the Wall Street Journal Index during 1981 - 1985. They construct a picture of a typical write-o? ?rm; it has weak prior performance, changes in top management, and is highly leveraged. They also analyze announcement e?ects and report negative and insigni?cant abnormal returns, although the returns are widely dispersed. This paper relates to my study in that we both consider the impact of CEO turnover on a write-o? decision, and it leads me to the yet untested hypothesis that the type of governance structure and CEO turnover a?ect the write-o? decision.

Bartov, Lindahl, and Ricks (1998) use a key word search from Dow Jones News to compose a sample of write-o? ?rms. They attempt to explain why the stock price changes around write-o? announcements are so small relative to the average write-o? amount. They suggest that the market under-reacts to the write-o? announcement and ?nd that abnormal returns are negative by as much as 21 percent after the announcement. Brickley and Van Drunen (1990) ?nd a positive and signi?cant average abnormal return around the announcement of restructuring charges. Kross, Park, and Ro (1996) also ?nd a positive market reaction to the announcement of an initial re-

3

structuring charge, as well as increases in trading volume and market return variability. Alciatore, Easton, and Spear (2000) examine the timeliness of write-o?s for oil and gas ?rms under the SEC’s full-cost ceiling test. These authors ?nd that write-o?s have a signi?cant negative association with contemporaneous quarterly returns and an even more negative association with prior quarter returns. They conclude that such impairments are not timely insofar as they are re?ected in returns before the announcement of a write-o?. Zucca and Campbell (1992) ?nd no signi?cant di?erence in stock performance from 60 days prior to 60 days after a write-o?. He?in and War?eld (1995) ?nd that the returns for write-o? ?rms during the write-o? year are negatively correlated to the amount of the charge. These papers all focus on the abnormal returns associated with write-o?s. They do not consider what motivates ?rms to take write-o?s, and the relation between governance and write-o?s, a main purpose of this paper.

Another branch of the write-o? literature looks at the relation between write-o?s, earnings, and performance. Kinney and Trezevant (1997) examine a large sample of Compustat data spanning the ten-year period 1981 through 1991 and ?nd that write-o?s are consistent with earnings management. They report that ?rms with large changes in reported earnings recognize signi?cantly negative income from special items. This ?nding is consistent with dampening large increases to produce a smooth, upward trend in earnings. Elliot and Hanna (1996) study the information content of earnings conditional on the presence of write-o?s. They also look at the incremental information content of these write-o?s. Their main ?nding is a signi?cant decline in the weight attached to unexpected earnings in quarters following write-o?s. They conclude that this shows evidence that write-o?s create noise in the information environment. These papers concentrate on how write-o?s a?ect the information environment. They do not consider how endogenous factors such as corporate governance might a?ect the value of the information contained in the write-o? announcements. My study adds to this branch of literature by examining how cross sectional characteristics (such as CEO turnover, governance provisions, and pay performance sensitivity) in?uence the information environment surrounding a write-o?.

4

The last branch of literature is tied to the impact of FASB’s 1995 issuance of SFAS 121, Accounting for the Impairment of Long Lived Assets. SFAS 121 was intended to reduce managerial ?exibility and enhance the reporting of long-lived asset write-downs.1 Kim and Kwon (2001) examine the di?erence in market reaction for early versus late adapters of the new FASB standard. They ?nd that early adapters have a positive market reaction, but late adapters have a negative market reaction to a write-down announcement. Riedl (2002) compares the types of write-downs taken before and after SFAS 121. He ?nds that write-downs reported prior to the standard have a greater association to economic factors than do write-downs reported after the standard. Lindbeck, Rezaee, and Smith (1996) ?nd that write-downs increase in magnitude following the adoption of SFAS 121. These papers all focus on one type of write-o? and one main event, whereas my research covers a broader period, as well as a more extensive array of write-o? types.2

1.3 Data
To generate my sample, I collect write-o? information, focusing on announcement behavior between 1980 and 2000 made by companies in the 2000-2999 SIC code, which are primarily consumer manufacturing companies. I focus on this particular industry because it is a mature industry that is asset intensive, and therefore might have greater incentives to take write-o?s compared to other industries. Using a CRSP generated perm and SIC code list, I search Lexis-Nexis and Dow Jones Retrieval services for speci?c key words. For each company, I search for articles that match key words. The key words I use are write down, write-o?, restructure, charge against earnings, layo?s, and severance. When the query results in a match, I take the ?rst article in the series of articles that refers to a current write-o? that the company is announcing. I use the date of the article as
1 This standard addresses (1) the criteria for when to test for the existence of an impairment, (2) the level at which to group assets in testing for impairment, (3) the measurement basis for determining the existence of an impairment, (4) the measurement of the impairment, and (5) the presentation of the recognized amount. The standard only applies to write-downs. 2 I test the robustness of my results for the impact of SFAS 121 in two ways. First, I segment my sample across time and do not ?nd any signi?cant changes in my analysis. Second, I remove the subset of write-downs from my sample, and ?nd no signi?cant changes to my analysis.

5

the announcement date of the write-o?. I obtain the following information from the article: the amount of the write-o?; whether the write-o? was generated by an asset write-down, employee layo?s, or both; the purpose of the write-o? (restructure, write-down, plant closing, etc.); the justi?cation cited by the company; and whether the write-o? amount is stated on a before-tax or after-tax basis. The sample contains asset-based and layo?-based write-o?s. I ?nd 2,429 companies within the consumer manufacturing industry. From this sector, 803 companies (33 percent) had a write-o?, giving a combined 3,738 write-o? announcements.

Write-o?s represent either a write-down of assets, charge due to corporate restructuring, or charge due to lay-o? events. SFAS 5 requires a ?rm to write down or expense asset values that will not be recoverable from future operations. SFAS 121 clari?es these circumstances for write-downs. SFAS 5 and APB Opinion 30 require ?rms to report restructuring charges, including charges from the sale or acquisition of a business in the year incurred. The disposition of a complete business segment must be reported as a separate line item called discontinued operations. Other write-o?s can appear in the footnotes of ?nancial statements. Appendix A provides a more complete explanation of the way write-o?s are handled in ?nancial reporting. In this study, I focus on write-o?s that include the partial disposition of a segment, discontinued operations, restructuring charges, plant closings, costs of employee terminations, and other special charges that are either unusual or infrequent, but not considered an extraordinary item.

I do not include write-o?s due to litigation costs, bankruptcy, goodwill, or capital structure re?nancing in this data set. By including only write-o?s that are related to operational decisions, I can examine the impact on future performance and avoid the legal and accounting peculiarities that are associated with other types of write-o?s. Only write-o?s that are announced singularly are included in the dataset, so that I can attempt to isolate both the reasons companies take write-o?s and the market’s reaction to the announcement.

6

Although COMPUSTAT has data on write-o?s, I opt to use the hand-collected data set for the following reasons. Information on write-o?s can be found in Compustat data item #17. Compustat does not report charges that it deems inconsequential, but these can be important to establish a history of write-o?s. Compustat also understates most write-o?s. I compare the charge amounts listed in the write-o? announcements to the charges recorded in Compustat. Overall, the public announcement of the write-o? charge averages $3.41 million more than the COMPUSTAT write-o? charges. All of the write-o?s identi?ed in COMPUSTAT are also listed in my sample, but there are 352 write-o?s from my sample that are not listed in the COMPUSTAT sample. The di?erences in my sample versus COMPUSTAT are similar to the di?erences reported in an earlier study by Fried, Schi?, and Sondhi (1989).

To ensure that write-o?s in my sample are not extensions of earlier events, I set an arbitrary standard under which I assume that any write-o? announcements occurring within six months of earlier write-o? announcements are related. This exercise is also performed for break o? points of one month, three months, four months, eight months, and twelve months. Although doing so a?ects the sample size, it does not a?ect the analysis or ?ndings. Therefore, I only describe results using the 6-month break point.

It is important to determine which write-o? is a ?rst-time event or a subsequent event. To de?ne multiple write-o?s, I need to establish an arbitrary time interval. The standard most researchers use de?nes multiple write-o?s as any write-o? event that occurs within 16 quarters of a prior write-o? event.3 To identify a company’s ?rst write-o?, I look at all write-o?s that occur during the ?rst ?ve years of the sample: 1980-1985. I require an initial period of 16 ?scal quarters with no write-o?s before I add a ?rm to the sample. I denote the write-o? following this break as a ?rst time write-o?. Because the original sample begins in 1980, the ?rst reported write-o? in the sample occurs in the ?rst quarter of 1985. To test the sensitivity of this break point, I also
3 See

Elliot and Hanna (1996).

7

use ?ve other quarter break points to de?ne ?rst time write-o?s, (8, 12, 18, and 20 quarters) to separate consecutive write-o?s. My conclusions become more robust with the longer measures and weaken slightly with the short-term de?nitions. Since the inference changes only marginally, I use 16 quarters. This procedure leaves me with 767 ?rms and 1,798 write-o? events to evaluate. After I identify the ?rst time write-o? for a company, write-o?s that follow are labeled as second, third, fourth write-o?s, etc. These subsequent write-o?s must occur within 16 quarters after the prior write-o?. If the write-o? occurs after 16 quarters, I label it as another ?rst time write-o?.

To compare write-o? company characteristics, I construct a sample of non-write-o? ?rms. Out of the 2,429 ?rms in the 2000-2999 SIC codes for 1980-2000, there are 1,626 ?rms that do not have a write-o?. I sort these non-write-o? ?rms into their primary 4-digit SIC code. Each write-o? ?rm in the sample is matched to at least two ?rms in the non-write-o? group. This match is based on 4-digit SIC codes and similar total assets. If there are no ?rms that match the 4-digit SIC code of the identi?ed write-o?, I use the 3-digit or 2-digit SIC code. To match by size, I also pair write-o? ?rms to non-write-o? ?rms of similar total assets.4 The matching results in a sample size of 2,037 write-o? events composed of 767 write-o? ?rms and 995 non-write-o? ?rms.

Using the PERM numbers for my write-o? sample, I merge COMPUSTAT and CRSP data into my sample. I measure the abnormal return measure as the market-adjusted returns. The reported results use the di?erence between daily CRSP returns and daily returns on the CRSP equally weighted market portfolio. I also use the value-weighted, beta-weighted, and marketcapitalized CRSP market portfolios, and the results remain similar. I use the quarter prior to the write-o? announcement to match COMPUSTAT data, such as book value, earnings per share, sales, shares outstanding, and total assets, for the write-o? sample.

To calculate abnormal returns, I de?ne the event window as the day of the write-o? annote that the non write-down ?rms have a total asset value that is at most 10 percent greater than the write-o? companies are, or at most 10 percent less than the write-o? companies.
4I

8

nouncement in the ?nancial press, which acts as the date in which the information concerning the write-o? becomes public. I use the three-day horizon surrounding the announcement date (t = ?1 to t = 1) to calculate the announcement e?ects. The abnormal return is the actual ex post return of the security over the event window minus the normal return of the ?rm over the event window (Brown and Warner, 1985). For any company i in month t,

ARit = Rit ? E (Rit),

(1.0)

where Rit is the realized return on day t, and E is the expectations operator. I estimate the expected return E (Rit) for each ?rm as the return on equal-weighted size portfolio model. I estimate ¯it ) for each day in the sample as follows: the average abnormal return (A

N

¯it = 1/N A
i=?1

ARit,

(1.1)

¯it is a cross sectional average.5 where N is the number of securities. A

Table 1, Panel A shows the distribution of write-o?s over the 16-year sample period. The sample contains 604 pure asset related transactions, 1,175 write-o?s that combine both assets and layo?s, and 258 write-o?s related to lay-o?s. The number of write-o?s per year triples from the beginning of the sample (49 in 1985 to 212 in 2000). Panel B of Table 1 shows the frequency of write-o?s by ?rm from 1985-2000. Out of the 767 ?rms that take one write-o?, 61 percent take an additional write-o?, and 42 percent have at least two additional write-o?s.

Table 2 describes the write-o? sample. Restructures are by far the most common type of event, occurring in more than 56 percent of write-o? incidents. Discontinued operations are the second most common write-o? event, occurring 14.28 percent of the time. The table also displays the average charge for each type of write-o?. Partial asset write-downs, and restructuring charges have
5 I also calculate returns using the market model, with an estimation period of -90 to -61 days before the write-o? announcement. I choose the estimation period to minimize the problems associated with estimating parameters with data in?uenced by the write-o? event. The results are comparable to the reported results.

9

the greatest magnitudes, averaging $78 million and $72 million, respectively. Write-o? amounts range from $56 thousand to $2.1 billion, with a mean of $45 million. The distribution is positively skewed; the median write-o? is $22 million. This skewness is also observed for each write-o? category. The table also shows the average write-o? charge total assets (TA). When adjusted by BV and TA, discontinued operations-based write-o?s are the largest, followed by restructure-based write-o?s. On average, discontinued operations charges are over 13 percent of a write-o? ?rm’s total assets. Restructuring charges were second with charges over 6 percent of a write-o? ?rm’s assets. The total amount of ?rm value written o? over the sample amounts to over $98 billion dollars. Due to the potential shareholder welfare implications from a loss this large, it is important to understand the cross sectional characteristics of write-o? companies, and the resultant impact on shareholder value.

Table 3 describes the mean and median of ?rm speci?c variables used in subsequent probit models. I also report univariate signi?cance tests to determine whether there is any di?erence between the write-o? values and the non write-o? values. The variables shown in the table include: • MV = the size of the ?rm, measured as the log of market value one quarter before the write-o? was announced. • SHROWNPC = the percent of the new (if replacement) or old (if not replaced) stock holdings in the company for the CEO. • SALARY = the dollar salary of the CEO in the year of the write-o?, in 100k. • BONUS = the dollar bonus for the CEO in the year of the write-o?, in 100k. • OPTIONS = the aggregate dollar value of all options granted to the executive during the year as valued by the company, in 100k. • SIZEBD = the number of directors, both inside and external. • PER OUT = the percentage of directors who have no relationship with the company.

10

• DIROPT = the number of options, which each non-employee director received during the year in thousands. • DIRSTK = the number of shares, which each non-employee director received during the year in thousands. • GOV INDEX = Governance index, the lower the number the better the shareholder protection, the higher the number, the worse the managerial entrenchment. This index is based on charter provisions listed by IRRC publications • DOLLAR SENSITIVITY = the dollar sensitivity of compensation to performance. • RETURN SENSITIVITY = the return sensitivity of compensation to performance. • ROA = the return of assets for the current quarter. Firm size is related to the likelihood of write-o?s occurring, as discussed in Meyer and Strong (1989). The larger the ?rm, the more assets it can divest. Table 3, Panel A, shows that non-write-o? ?rms have the lowest market value, with an average MV of $191 million. One-time write-o? ?rms are on average $15,367 million, and are signi?cantly larger than the benchmark (signi?cant at 5 percent), while multiple write-o? ?rms are the largest with a market value of $ 265,667 million (signi?cant at 5 percent).6 One-time write-o? executives own four percent of their company’s stock, followed by multiple write-o? ?rms at three percent, and then non-write-o? ?rms at two percent. Only one time write-o? company CEOS have signi?cantly di?erent stock ownership as compared to the benchmark (signi?cant at 10 percent). I ?nd that CEOS of the one-time write-o? ?rms and the multiple write-o? ?rms are paid $666 thousand, and $629,000 respectively, which is less than non-write-o? CEOs pay, $657,000. Likewise, non-write-o? ?rms receive larger dollar bonuses than write-o? ?rms ($848,000, $534,000, and $548,000 on average for non write-o?s, one-time write-o?s, and multiple write-o?s, respectively). First-time write-o? ?rms have option grants of $2,385,000, which are the highest value of option grants, followed multiple at
6 All of the t-values presented test for whether there is a signi?cant di?erence between the write-o? ?rms and the benchmarks.

11

$2,039,000, and non write-o? ?rms,$1,346,000. First-time write-o? ?rms have the smallest boards, with 10.17 members, followed by multiple, and non write-o? ?rms (11 members, and 12 members respectively). Indeed, not only are the boards smaller for write-o? ?rms, but they are also dominated by outsiders (76 percent, for multiple write-o? ?rms, 73 percent for ?rst-time write-o? ?rms, and 69 percent for non-write-o? ?rms. These results suggest that the boards of write-o? ?rms are better monitors than are the boards of non-write-o? ?rms. GOV INDEX, a measure of the level of shareholder protection measures from the IRRC database, show moderately stronger protection measures for write-o?s as compared to non write-o? ?rms. The ?rst time and multiple write-o? companies show an average ROA of 4.25 percent and 4.7 percent, respectively and are both signi?cantly less than the benchmark ?rm’s ROA of 7.3 percent.

I measure pay-performance sensitivity in two ways. The ?rst measure is the dollar sensitivity of CEO compensation, de?ned as the change in the dollar value of the CEO’s stock and option holdings for a dollar change in ?rm equity value. The second is the change in the dollar value of the CEO’s stock and option holdings for a one percent change in a company’s stock price (the return sensitivity). Baker and Hall (1998) argue that the return sensitivity measure is the appropriate one to use when CEO actions a?ect ?rm percentage returns through their control of ?rm strategy. I calculate these measures using the Core and Guay (1999a, 1999b) method, which allows me to compute incentives using the one-year data on a CEO’s stock option portfolio contained in the annual proxy statements. I estimate the regressions using both measures of incentives and ?nd similar results. Pay-performance sensitivity is de?nes as follows, where W denotes CEO wealth in options and stocks held, and V denotes ?rm value,

ReturnSensitivity = r = 0.01 ? dW/(dV /V )

and DollarSensitivity = (dW/dV ) = 100 ? r/V. (1.2)

12

These ratios act as proxies for the degree of pay-performance sensitivity of a CEO. This sensitivity shows the percentage by which pay increases (decreases) when company performance increases (decreases) by one percent. Table 3 shows the summary of these two sensitivity measures. First-time write-o?s show the greatest sensitivity, followed by multiple and then non-write-o? ?rms.

In addition, I calculate the TLCF, the tax-loss-carry-forwards of the company, in the writeo? quarter.7 Two possible relationships between TLCFs and write-o?s are if a company has had prior poor performance, it is possible that they have had a tax-loss-carry-forward, and the TLCF would act as a proxy for poor performance, which would lead to a positive relationship with writeo?s. Second, if the company already has a TLCF, there are fewer tax incentives to take a write-o?, and so one would expect a negative relationship between write-o?s and the TLCF. In this paper, TLCF is a dummy variable that is one if the company has an unused tax-loss-carry-forward, and zero otherwise. I also calculate the debt ratio, DEBT RATIO, de?ned as total debt over total assets. I expect that due to tax incentives, there will be a negative relationship between write-o?s and debt. If a company has more debt, it has the tax shelter from the interest expense, which would o?set a tax advantage from taking write-o?s.

Table 3, Panel B, shows the governance characteristics broken up by year. This table includes all ?rms in the sample over all years of the sample, regardless of whether the ?rm took a write-o? in that particular year. There is very little change in the governance variables from year to year. Since the governance variables are sticky, they act like ?rm speci?c e?ects in the following regressions. Since there is very little change in board size, percent of outside directors, and the IRRC index over time, these variables can be considered given; that is, they are determined outside the write-o? decision and are not impacted by short term changes in performance. However, CEO turnover does present a self-selection issue in relation to the write-o? decision and performance changes.
7 See

Plesko (1999) for a description of the calculation of the TLCF.

13

1.4 Characteristics of Write-o? Companies
Companies with e?ective monitoring mechanisms are not immune to problems. Certain circumstances, such as negative economic shocks, or increased product market competition can negatively a?ect the company’s performance. However, these well monitored companies quickly recognize the problems and take actions to ?x the problem areas.8 This argument suggests that good governance ?rms have smaller multiple write-o?s, while poorly governed companies have fewer, but much larger write-o?s. One would then expect to see a positive correlation between governance quality and the probability of taking write-o?s, and an improvement in future earnings for the write-o? company.

In this paper, I look at how four factors might a?ect the write-o? decision, and in?uence short run consequences or long run bene?ts. These factors are CEO turnover, pay-performance sensitivity, board composition, and managerial entrenchment. I ?rst test each hypothesis individually to see how the ?rm characteristic is related to the write-o? decision and then examine the hypotheses jointly to see how the characteristics interact in relation to the write-o? decision.

1.4.1 Corporate Cleanup Hypothesis Borokovich, Parrino, and Trapani (1996) show that a turnover announcement is normally succeeded by a positive market reaction for outside replacement, and a negative reaction for insider replacement, especially when the replacement is not voluntary. Prior corporate governance research also emphasizes that a critical element of corporate governance mechanisms is an ability to identify and terminate poorly performing executives (Kaplan (1994); Co?ee (1999); Murphy (1999); Volpin (2002); Berger, Ofek, and Yermack (1997)). For example, Macey (1997) observes that a necessary
8 See, e.g., Paul (2003), which describes how all managers make mistakes, but that companies with good governance are the ?rst to correct these mistakes.

14

condition for competent corporate governance systems is the removal of poorly performing managers. Gibson (1999) asserts that a primary purpose of corporate governance mechanisms is to ensure that poorly performing managers are removed. The importance of replacing un?t CEOs is also consistent with Shleifer and Vishny (1989, 1997), who speculate that the most important form of managers expropriating shareholder wealth are unquali?ed managers who remain with the company. Jensen and Ruback (1983) also support this position by arguing that poorly performing managers who resist removal might be the costliest manifestation of the agency problem.

Whether CEO replacement is external or internal could also a?ect the write-o? decision. An external replacement would be more likely to result in a write-o? than an internal replacement. An external replacement typically indicates di?erent agenda for the company, which can result in the new CEO cleaning up the internal problems prior to pursuing new goals. An internal replacement is a part of the earlier CEO’s agenda and has less incentive to take a write-o?. In addition, if the new CEO expects to remain in place for a longer horizon, then she has an incentive to see the company’s pro?tability to improve in the long term. Formally stated:

H1Corporate Cleanup Hypothesis) The probability of a write-o? is greater for ?rms that have recently had CEO turnover, especially if the replacement is external .

Table 4 Panel A shows the characteristics of executive turnover associated with a write-o? announcement. I obtain my turnover data from Execucomp. After matching the two datasets, my combined sample comprises of 886 write-o? events for the 1992 to 2000 period. I label CEO replacement that occurs within a year prior to the write-o? announcement as a related event. Execucomp lists the reason the CEO leaves the company. If applicable, I mark one of the following options: resigned, retired, deceased, or unknown. The table shows two types of turnovers, those with external replacements, and those with internal replacements. I de?ne a replacement as external or internal by comparing the date the CEO entered o?ce and the date the CEO joined the

15

company. If these coincide, the CEO is external, otherwise internal.

Table 4 Panel A describes the turnover statistics for non write-o?, ?rst-time write-o?, and multiple write-o? ?rms. Non-write-o? ?rms have less CEO turnover than do one-time write-o? ?rms, or multiple write-o? ?rms. There are 130 CEO turnovers for multiple write-o? companies, 59 for one-time write-o? companies, and only 35 for the benchmark companies over the eight years of the sample. The majority of replacements came from outside the company (65 percent, and 81 percent, and 77 percent for multiple, ?rst, and non write-o? ?rms respectively). These results suggest that there is a link between write-o?s and turnover.

As discussed above, the ?rm that decides to terminate its CEO may do so because of unobserved information that is potentially concealed with the information that leads to a writeo?. This leads to self-selection in the CEO turnover decision. In order to control for this source of self-selection bias, I run a two-stage sample selection model. In the ?rst stage, I run the following probit estimate:

pr(EXT U RN OV ER) = ?1 + ?2 LOGM Vi + ?3 ROAi + ?4 Ri + i,

(1.3)

where EXTURNOVER is a dummy variable that is equal to one if there was CEO turnover with an external replacement, and zero otherwise. R , the unadjusted cumulative monthly stock return for the ?rm over the past 12 months, is the independent variable, which measures CEO performance . The speci?cation follows Parrino 1997. Table 4 Panel B gives the results of the estimate. Consistent with prior work, I ?nd that the probability of forced CEO turnover is estimated to be negatively and signi?cantly related to the prior stock return. ROA shows the accounting profitability of the ?rm one year prior. ROA, and the log of market value are also negatively related to EXTURNOVER. The results are similar to those represented by Parrino (1997).

In the second stage, I model the decision to take a write-o? using a probit model. A ?rm 16

takes a write-o? if latent variable WO? 0 and no write-o? if WO ? 0. WOi is empirically speci?ed as:

pr(W O) = ?1 + ?2 SIZEi + ?3 SHROW N P Ci + ?4 CEO OU T + ?5 ROA+

?6 T LCF + ?6 DEBT RAT IO + ?i ,

(1.4)

where ?i is standard normal. CEO OUT is the inverse mills ratio from the CEO turnover probit equation and it corrects for the self-selection in the decision to replace the CEO. ROA is return on assets, TLCF is a dummy variable that is one if a company has tax-loss-carry-forwards, and DEBT RATIO is the debt ratio. and ? is a standard normal error term.

Table 5, Model 1, shows the probit results for Equation (1.4). I ?nd that the probability of a write-o? occurring increases if a CEO turnover with an outside replacement occurs within the one-year period prior to the write-o? (signi?cant at 5 percent). The control variables in Equation (1.4) have the right signs. The results con?rm a signi?cant positive relation between the probability of a write-o? and the size of a ?rm. There is a negative correlation between performance and the write-o? decision. ROA is negatively related to write-o?s, while TLCF is positively related to write-o?s. I also ?nd that CEO shareholdings are related to the write-o? decision, despite the fact that the CEO turnover and the percent ownership are strongly negatively related. In addition, debt is negatively related to the write-o? decision. Table 6, Model 1, shows the marginal e?ects for the probit model. The share ownership, the size, and CEO replacement from outside the company have the largest impact on the write-o? decision, respectively. In another words, the larger the ?rm, the more likely it is that it will take a write-o?. The impact of ?rm size on the write-o? decision is similar to the results in Meyer and Strong (1989). In addition, the results are consistent with companies undergoing a period of poor performance. These results are robust to time period speci?c random e?ects.

17

1.4.2 Executive Compensation and the Write-o? Decision It has become common practice for executive compensation to be tied to the company’s performance. Coughlan and Schmidt (1985), Murphy (1985, 1986), Abowd (1990), Jensen and Murphy (1990) and Leonard (1990) study the relation between executive compensation contracts, incentives and ?rm performance. These papers show that ?rm performance is largely positively related to pay-performance sensitivity, after controlling for the risk, i.e., the variance of performance (Aggarwal and Samwick, 1999). Audt, Cready, and Lopez (2003) ?nd that after controlling for the growth in annual in?ation adjusted CEO cash compensation, CEOs are not protected from the adverse e?ects of charges on earnings on their own utility.

If a CEO’s actions are closely tied to ?rm performance, then the CEO will hesitate to take unnecessary actions that a?ect his compensation. Therefore, it is plausible the CEOs with high pay performance sensitivity will not take a write-o? unless it is necessary to improve future performance. There is a trade-o? between short-term and long-term utility for the CEO. In the short term, write-o?s can reduce stock price, which can reduce compensation. In the long term, write-o?s can improve future performance, which can increase compensation. The future bene?ts would dominate if there were a longer horizon for the CEO (e.g. for a new CEO), or compensation is more dependent on future performance (e.g. stock options).

Formally stated: H2Executive Compensation Hypothesis)The probability of taking a write-o? is positively related to the pay-performance sensitivity of a CEO. The probability of taking a write-o? is also positively related to the actual compensation package.

To test this possibility, I use two di?erent measures of compensation: actual compensation, and pay-performance sensitivity. I use the following probit model to test hypothesis H2. I observe a write-o? is one if latent variable WO?0 and no write-o? if WO ? 0. WOi is empirically speci?ed

18

as:

pr(W O) = ?1 + ?2 SIZEi + ?3 SALARYi + ?4 BON U Si + ?5 SHROW N P Ci + ?6 OP T ION Si

+?7 RET Y RSi + ?8 ROAi + ?9 T LCFi + ?10 DEBT RAT IOi + i,

(1.5)

where

i

is standard normal. I control for the size of the company and the tenure of the CEO

(RETYRS), as well as performance. I expect that higher compensation will be positively related to write-o?s. I also expect that the tenure of the CEO will be negatively related to write-o?s. This could be either because the CEO is entrenched, or because the CEO has not made any mistakes and has no need for write-o?s.

Table 5, Model 2 (A), shows the estimates of Equation (1.5). CEOs with lower salaries are more likely to take write-o?s than are CEOs with higher salaries (coe?cient = -0.001, signi?cant at 5 percent), and CEOs with a greater percentage of shares are more likely to take write-o?s (coe?cient = 4.23, signi?cant at 5 percent). These results suggest that compensation packages, which tie CEO incentives to performance, are related to write-o?s. The control variables in Equation (1.5) have the right signs. The market value of a ?rm is positively related to the write-o? decision, while the tenure of a CEO is negatively related to the write-o? decision. ROA is negatively related to write-o?s and TLCF is positively related to write-o?s. In addition, the debt ratio is negatively related to the write-o? decision. These results suggest that CEOs who are less entrenched are more likely to take write-o?s.

As discussed above, the pay-performance sensitivity might have implications in a write-o? decision. By using the following probit model, I test whether the pay-performance sensitivity of managers and the level of entrenchment a?ect a company’s write-o? decision. I observe a write-o? is one if latent variable WO?0 and no write-o? if WO ? 0. WOi is empirically speci?ed as:

19

pr(W O) = ?1 + ?2 SIZEi + ?3 IN T ERLOCKi + ?4 RET Y RSi + ?5 P P Si + ?6 ROAi

+?7 T LCFi + ?8 DEBT RAT IO + i,

(1.6)

where

i

is standard normal and PPS is the degree of pay-performance sensitivity. I run

the regression using both dollar and return sensitivity as a measure of PPS, as discussed in Core and Guay (1999a, 1999b). INTERLOCK is a dummy variable equal to one if the management is entrenched and zero if it is not, as de?ned by Execucomp. Entrenchment generally involves one of the following situations: the o?cer serves on the board committee that makes his compensation decisions, or serves on the board (and possibly compensation committee) of another company that has an executive o?cer serving on the compensation committee of the indicated o?cer’s company. Hallock (1997) describes the use of this variable as a proxy for managerial entrenchment. RETYRS is the number of years the CEO has been in o?ce. I expect that INTERLOCK and RETYRS will be negatively related to write-o?s.

Table 5, Model 2 (B) shows the estimate of Equation (1.6) for the period 1990-2000. The results indicate a positive relation between the probability of a write-o? and the pay-performance sensitivity of the CEO(t-value = 1.95, signi?cant at 10 percent).9 As before, the control variables in Equation (1.6) have the right signs. Market value, and TLCF are positively and signi?cantly related to write-o?s, while the entrenchment variable, debt ratio, and ROA are negatively related to write-o?s. Table 6, Model 2 (B), shows the results of the marginal e?ect of the probit estimation.

1.4.3 Monitoring Mechanism Hypothesis The board of directors decides on both CEO compensation packages, and CEO turnover replacements. In addition, the board of directors acts as a monitoring mechanism for CEOs. If the board
5 shows the estimation for the return sensitivity measure. Results for the dollar sensitivity measure were comparable.
9 Table

20

is a pro?cient monitor, then there are fewer agency issues with management and the CEO has better incentives to take actions that bene?t the company and shareholders. When non-performing assets a?ect a company, then a write-o? is a tool that management can use to alleviate these operational problems. If a link exists between performance and monitoring mechanism quality, then it is plausible that companies with boards that are good monitors will have a higher probability of taking a write-o?.

The governance literature ?nds strong evidence that board composition (size of board and percentage of insiders on the board) is related to the degree of agency problems (Byrd and Hickman (1992); Wasatch (1988); Borokovich, Parrino, and Trapani (1996); Bhagat and Black (1999); Core, Larcker, and Holthausen (1999); Hermalin and Weisbach (1991); and Yermack (1996). Larger boards with more inside directors tend to have more agency problems. Conversely, ?rms with small boards and a high percentage of outsiders will be more concerned about shareholder welfare and ?rm performance.

Formally stated: H3: The probability of a write-o? increases when there are quality governance mechanisms in place (Monitoring Mechanism Hypothesis).

The size and percentage of outsiders on the board act as proxies for monitoring quality. I also include the number of options and the percentage of equity that the directors own. If the board’s compensation is attached to the performance of the company, the incentive to be quality monitors increases, as discussed in Mayers, Shivdasani, and Smith (1994). I also include the governance index, GOV INDEX. Gompers, Ishii, and Metrick (2003) create a governance database drawn from Investor Responsibility Research Center (IRRC) publications, an organization that has tracked the provisions for about 1,500 ?rms per year since 1990. I merge the write-o? sample to the governance database using ticker symbols and year. G A higher GOV INDEX indicates a

21

?rm with less shareholder rights. GOV INDEX was available for 756 write-o? events.

Following Hypothesis 3A, I estimate the following probit speci?cation:

pr(W O) = ?1 + ?2 SIZEi + ?3 SIZEBDi + ?4 DIROT Pi + ?5 GOV IN DEXi + ?6 P ERC OU Ti

+?7 DIROSKi + beta8 ROAi + ?9 T LCFi + ?9 RET Y RSi + ?10 DEBT RAT IO + i,

(1.7)

where SIZEBD is the size of the board of directors, PERC OUT is the percent of outside directors, GOV INDEX is the degree of shareholder protection, DIROPT is the value of options owned by the directors, and DIROSK is the percent stock ownership of the directors. Since

Table 5, Model 3, shows the results of the test for whether better monitoring boards are more likely to take write-o?s. The results indicate that boards that are more independent have a greater likelihood of taking a write-o?. SIZEBD is negatively related to the probability of a write-o? (signi?cant at 5%), and the percentage of outsiders is positively related to the probability of a (write-o? signi?cant at 5%). GOV INDEX is negatively related to the probability of taking a write-o? (signi?cant at 5%). The percent of directors’ option ownership is positively related to the likelihood of a write-o?, while the number of shares is not signi?cantly linked to the tendency to take write-o?s. These results suggest that ?rms with smaller boards, more outside directors, and shareholder protection are more likely to take a write-o?. The coe?cients of the control variables are consistent with the expected signs. The market value of the company (SIZE), and TLCF are positively related to the write-o? decision, while ROA, debt ratio, and RETYRS are negatively related to the write-o? decision. The signs and signi?cance of ROA and TLCF are consistent with companies having poor performance both in the write-o? quarter, and in recent past quarters. The options owned are positively related to the tendency to take write-o?s, while the percent stock ownership is negatively related to the tendency to take write-o?s. Table 6, Model 3, looks at the marginal e?ects of the independent variables on the write-o? decision. Overall, these results con-

22

?rm that companies with desirable board composition and strong shareholder protection measures have a tendency to take write-o?s.

1.4.4 Multivariate Analysis In the previous sections, I test the one-on-one relations between CEO turnover and write-o?s, pay-performance sensitivity and write-o?s, board composition, shareholder protection, and writeo?s. In this section, to examine the interaction of the independent variables, and to test which characteristics are of the most importance, I combine these separate speci?cations.

In addition to ?rm characteristics described above, I include an industry shock variable. Industry shocks and recessions are two possible factors that can a?ect a write-o? decision. By including these variables in the probit estimation, I can test whether these factors are related to the write-o? decision.

Using the Bartelsman and Gray (2002) dataset on the NBER website, I create a measure of industry demand. I use the industry shipments at the 4-digit SIC code level de?ated by a 1987 industry price de?ator, and then aggregate this data at the 3-digit SIC code level. I detrend the data by regressing the actual value of industry shipments on a yearly time trend variable. I then calculate the industry shock is then calculated as the di?erence between the predicted and the actual value of shipments. I use the detrended real industry shipments for the same reasons cited in Maksimovic and Phillips (2002). The reasons include the growth of an industry, which a?ects the value of the capital in the industry, and ?rms’ cash constraints can depend on industry conditions.

To test whether the market’s reaction to the write-o? is in?uenced by the relationship between productivity and segment growth, I create dummy variables for recessionary and expansionary periods. I also classify years as recession or expansion years for an industry, and determine

23

the recession and expansion years by looking at the relationship between the aggregate, and the aggregate detrended production in each industry. If these two variables are negative, then the year is a recession year. If they are both positive, it is an expansion year. Detrended production is the actual production minus the predicted production, where the predicted production is calculated as the production regressed on a time variable.

Table 5, Model 4, looks at the results of the following two regressions, which examine the combined impact of the ?rm quality characteristics while controlling for industry e?ects.

P (W O) = ?1 + ?2 SIZEi + ?3 SHOCKi + ?4?8GOV V ARSi + ?9?12CON T ROLV ARSi + i, (1.8)

P (W O) = ?1 + ?2 SIZEi + ?3 RECESSIONi + ?4 EXP AN SIONi + ?5?9 GOV V ARSi

+?10?13CON T ROLV ARSi + i ,

(1.9)

where GOVVARS are CEO turnover with external replacement(using IMR to control for endogeneity), pay-performance sensitivity, board size, percent of outsiders on the board, and shareholder protection. I also include several control variables, such as percent of shares owned, RETYRS, CEO tenure, ROA, TLCF, and an interlocking relationship dummy.

The results in Table 5, Model 4 show that if a negative shock a?ects the ?rm, the probability of a write-o? increases, especially when I control for governance quality (signi?cant at 5 percent). Likewise, a recession year for the company increases the probability of a write-o? occurring (signi?cant at 5 percent), while an expansion year is negatively related to the write-o? decision. Equations (10) and (1.9) permit me to simultaneously examine the impact of the ?rm characteristics on the write-o? decision. Shareholder protection, board size, and percent of outside directors continue to remain signi?cant. CEO turnover, and pay-performance sensitivity do not have as signi?cant a role when combined with the other governance factors. One reason pay-performance sensitivity 24

may not be signi?cant is because GOV INDEX encompasses compensation plans.

These results suggest that well governed companies that operate in industries impacted by negative shocks or recessions have a tendency to take write-o?s. Table 6 shows the marginal effects for the combined probit estimation of Model 4 (A) and (B). I ?nd that for the independent variables conditional on the write-o? decision, the number of outsiders on the board and the size of the company have the greatest impact on the write-o? decision. The size of the board and the level of shareholder protection also have a highly signi?cant e?ect on the probability of a write-o?.10

1.4.5 Weak Shareholder Protection and Write-o?s So far, I have found evidence that companies with strong monitoring mechanisms have a tendency to take write-o?s. However, these results do not answer any questions about weakly monitored companies. It seems plausible that there is also a link between weak monitoring mechanisms and write-o?s. Companies with less e?ective governance structures continually collect problems and only take write-o?s when there is no other alternative. An example of a poor governance company and write-o?s is Tyco Corporation. In an e?ort to hide slowing growth in its core divisions, Tyco kept on diversifying into new areas. These diversi?cation strategies were not successful, and so it would diversify into yet another area. Eventually the problem became so huge that Tyco was left with little alternative other than to take a write-o? (Symonds, 2002).

I ?rst determine ?rms with weak shareholder protection measures that take write-o?s. I break the sample into three segments: weakly monitored governance companies, neutral governance companies and strong governance companies. I break the sample into three categories based on their GOV INDEX. I sort the sample by GOV INDEX and label the lowest 10 percent of the sample well monitored, and the top 10 percent weakly monitored companies. Table 7, Panel A
also include the industry e?ects in this estimate, but do not ?nd any signi?cant results. This is because I only focus on one main industry - the consumer-manufacturing sector.
10 I

25

shows the results of this segmentation. The t-values test for whether there is a di?erence between the average sizes of good versus bad governance characteristics. It becomes evident that there is wide dispersion between the write-o? ?rms in the sample, based on governance. The weakly monitored companies have an average board of 18 people versus the good governance companies with an average board size of six people (signi?cant at ?ve %). In addition, the poorly monitored companies have signi?cantly fewer outsiders on the board, and have signi?cantly worse shareholder protection measures (t-value=2.98 and 2.31, respectively.)

To formally test whether there is a relationship between poor governance and write-o?s, I isolate worst 50 percent governance ?rms in my sample, both write-o? ?rms and benchmarks, based on GOV INDEX. To do so, I sort the sample based on the GOV INDEX, and then drop the top 50 percent governance ?rms in the sample. I then re-estimate Equation (10) with only the bad governance ?rms. Table 7, Panel B, shows the results. The most important factors in determining whether bad governance companies take write-o?s are shareholder protection measures, and board size. Both GOV INDEX and BDSIZE are positively and signi?cantly related to the write-o? decisions. The other governance variables show the predicted signs but are not signi?cant. The control variables show the predicted signs discussed in earlier sections.

1.4.6 Governance and Size of Write-o?s I have found evidence that suggests both well and poorly monitored companies are subject to write-o?s. Even the best corporations are not immune from mistakes. However, these good governance companies quickly recognize the mistake, and take actions to repair the problem. If this is true, then it is expected that write-o?s would be relatively small for these quality companies. Poorly monitored companies, on the other hand are slow to admit to mistakes, and even slower to take actions to improve the situation. They collect mistakes until it becomes inevitable that a write-o? should occur. Following this argument, it is plausible that poorly monitored companies

26

will have relatively large write-o?s.

In this section, I test whether the size of a write-o? is in?uenced by the quality of the governance. If this is the case, then it supports the story that well governed companies are ?rst to repair problems, whereas poorly monitored companies are reluctant to repair problem areas. Table 8, Panel A, shows the univariate results of write-o? size, segmented by write-o? quality. I segment the sample into three di?erent groups: companies with weak monitoring structures, companies with average governance, and companies with strong governance. I sort the companies on the following items: GOV INDEX, board size, and percent of outsiders. I denote weak as the bottom 10 percent of write-o?s, and strong as the top 10 percent of write-o?s. I adjust write-o?s by the total assets of a company. The average size of all write-o?s is 0.03. The average adjusted size of well-monitored companies’ write-o?s is 0.02, versus 0.06 for poorly monitored companies. I regress the size of the write-o? on the following Equation to test whether governance a?ects write-o? size:

W O/T A = ?1 + ?2?5CON T ROLV ARS i + ?5?10GOV V ARi + i ,

(1.10)

where GOVVARS are CEO turnover with external replacement, pay-performance sensitivity, board size, percent of outsiders on the board, and GOV INDEX. CONTROLVARS are MV, ROA, and TLCF.

Table 8, Panel B, shows the results of this regression, which controls for heteroscedascity in standard errors. The results suggest that companies with larger boards take larger write-o?s (t-value = 2.76). Likewise, companies with more outside directors also take larger charges (t-value = 1.98). Although not signi?cant, the results also suggest that companies with worse shareholder protection measures and lower PPS also take larger write-o?s. These results show that companies with worse governance take larger write-o?s, while good governance companies take small write-o?s. The control variables show the expected relationship to write-o?s. These results are consistent with the story that good governance companies are ?rst to act when problems arise, 27

hence the size of the write-o? is smaller. Bad governance companies wait to take write-o?s and collect problems over an extended period, hence the size of the write-o? is comparably larger.

1.5 Market Reaction and Write-o? Announcements
Having shown that corporate governance impacts in what manner write-o?s are used, I now examine how investors react to write-o?s, taking into account the quality of governance of the announcing company.

Table 9 looks at the abnormal returns surrounding the write-o? announcement for the onetime and multiple events, and for the combined sample. For the full sample, the average market reaction to write-o? announcements is -1.10 percent, and is not signi?cant. These ?ndings are comparable with those in earlier studies, e.g. Meyer and Strong (1987), Elliot and Shaw (1988), Bartov, Lindahl, and Ricks (1998). When looking at the pooled average of 15 years, only one-time write-o?s show any signi?cant abnormal returns. The one-time write-o?s display a signi?cant -1.82 percent return around the announcement day (signi?cant at 5%). I interpret this result as meaning that the market only considers ?rst-time write-o?s to be signi?cant to the company’s performance.

I look at the combined impact of CEO turnover, pay-performance sensitivity, and board composition on the announcement e?ects of write-o? ?rms:

ARi = ?1 + ?2 SIZEi + ?3 SIZEBDi + ?4 P ERC OU Ti + ?5 CEO OU Ti

+?6 RECESSIONi + ?7 P P Si + ?8 ROAi + ?9 W O T Ai

+?10 W O #i + ?11 T Y P Ei + ?12DEBTi + ?13 GOV IN DEXi + i.

(1.11)

In addition to the governance variables I have used for the probit estimates, I include the

28

size of the write-o? (W O T A), the type of write-o? and the number of write-o?s a ?rm has taken (W O #), which includes the current write-o? (a ?rst time write-o? would be equal to one, etc.). I would expect that larger write-o?s would have a more negative impact on returns. In addition, I expect that companies with less write-o? history will see a greater market reaction and stronger monitoring mechanisms will lead to higher returns.

Table 10 shows the results of these estimates. I ?nd that when controlling for negative industry shocks such as recession, companies with strong governance measures actually experience more positive abnormal return. Larger ?rms have a 0.2 percent increase in abnormal returns. Strong shareholder protection leads to a 2.4 percent increase in abnormal returns. Small boards lead to a 1.5 percent increase in abnormal returns. A larger percentage of outside directors leads to a two percent increase in abnormal returns, while high PPS leads to a 2.6 increase in abnormal returns. CEO turnover leads to a 0.2 percent increase in abnormal returns. However, larger writeo?s lead to a one percent drop in returns. The debt ratio, the type of write-o?, and the number of write-o?s. In aggregate, companies with strong monitoring mechanisms have over 6 percent abnormal returns following a write-o? announcement.

Next, I consider companies with a write-o? following CEO turnover. As Borokovich, Parrino, and Trapani (1996) show, the turnover announcement is normally succeeded by a positive market reaction for outside replacement, and a negative reaction for insider replacement when the replacement is not voluntary. These results give evidence that the market di?erentiates between inside and outside replacement, and that outside replacement are good for the future of the ?rm. Hence, I test whether write-o?s generate similar reactions. Are write-o?s from CEO turnover where the replacement is external associated with positive announcement day e?ects, or is the write-o? anticipated following the turnover? Table 11 segments the write-o? announcement e?ects by CEO turnover, and GOV INDEX. I ?nd that there is more than a 6 percent positive write-o? announcement return for good governance companies that have had a recent CEO turnover using

29

an external replacement. These results suggest that an investor who owns a write-o? company, which has good governance and recently had a CEO turnover could make a six percent return over the market.

1.6 Conclusion
This paper examines the relation between write-o?s and corporate governance measures. Companies that have smaller boards with a higher percent of outside directors, stronger shareholder protection measures, high pay-performance sensitivity, or CEO turnover are positively related to the write-o? decision. The write-o?s these good governance companies take are linked to industry speci?c factors, such as industry shocks, or recessions. By segmenting out the good governance companies from the sample, I am able to test whether there is a link between poor governance and the write-o? decision. I ?nd that companies with poor shareholder protection measures, and large boards are also positively related to the tendency to take write-o?s. In addition, lower quality governance leads to larger write-o?s. Well-monitored companies are the ?rst to act when they realize that a problem has arisen and write-o?s are one tool that management can use to clean up the problem area. Conversely, poorly monitored companies wait to amend the companies’ problems, until the magnitude of the problem cannot be ignored. This explains why the size of the write-o?s from poorly monitored companies is signi?cantly larger than the size of write-o?s from well-monitored companies.

I also look at the impact of write-o?s on investors. By segmenting the write-o?s based on the governance quality, I determine whether investors di?erentiate between the di?erent companies taking write-o?s and the types of write-o?s. It becomes evident that companies with quality monitoring mechanisms take write-o?s that result in a positive stock market reaction, while companies with poor monitoring mechanisms take write-o?s that result in a negative stock market reaction. The ?ndings suggest that investors may understand the information content in write-

30

o?s, and are able to di?erentiate between write-o?s that will improve performance and write-o?s that will not. The stock market recognizes the quality of management and the board of directors and an investor’s reaction is based on this knowledge. By looking at the cross-sectional dispersion of governance measures and based on the quality of the monitoring mechanisms, an announcement day e?ect could have more than a six percent positive return in the short term. The evidence in this paper indicates that companies with good governance use write-o?s in a way that is consistent with enhancing shareholder value. In addition, I ?nd evidence that is consistent with the idea that governance matters.

31

1.7 Tables
Table I Sample Information

Panel A shows, by year, the number of write-o? announcements for layo? based, asset based, and combined write-o?s. Panel B shows the number of ?rms in the sample that take a ?rst time write-o? and then breaks into the percent of these ?rms that go on to take another write-o?. For instance, 61 percent of ?rst time write-o? ?rms take a second write o?, and 42 percent of ?rst time write-o? ?rms take two more write-o?s, etc. Panel B also shows the breakdown of the types of write-o?s, whether they are layo? based, asset based, or a combination of the two. Panel A: Number of Write-o?s by Year and Type Year Asset Layo? Both All 1985 20 4 25 49 1986 34 7 45 86 1987 48 6 48 102 1988 34 18 40 92 1989 34 10 45 89 1990 27 7 58 92 1991 40 18 51 109 1992 28 8 63 99 1993 43 20 72 135 1994 35 14 99 148 1995 36 18 91 145 1996 39 17 100 156 1997 42 27 99 168 1998 41 31 98 170 1999 42 22 121 185 2000 61 31 120 212 All 604 258 1175 2037 Panel B: Frequency of Write-o?s Write-o?s # Firms Percent Layo?s Assets Both 1 767 100 17% 37% 46% 2 468 61 14% 39% 47% 3 319 42 13% 34% 53% 4 216 28 9% 41% 50% 5 150 20 14% 33% 53% 6 116 15 11% 27% 62% 7 79 10 17% 15% 68% 8 61 8 10% 18% 72% 9 45 6 8% 28% 64% 10 31 4 7% 35% 58% 11 20 3 10% 35% 55% 12 16 2 23% 40% 38% 13 9 1 13% 20% 67% 14 7 1 14% 43% 43% 15 3 0.5 0% 67% 33% 16 3 0.4 0% 33% 67% 17 2 0.26 0% 0% 100%

32

Table II

Write-o? Characteristics of Sample

This table summarizes the di?erent types of write-o?s that ?rms report. There are several di?erent ways that a ?rm can write employees and assets o? the books. I discuss these methods in Appendix A. The average charge is the average write-o? the ?rm reported for each of the di?erent types of write-downs. Write-o?/Book Value is the total charge divided by book value of shareholders equity one month prior to the announcement. Write-o?/Total Assets is the total charge divided by total assets one quarter prior to the write-o? announcement. Write-o?/Book Value Mean Median 0.02 0.01 0.13 0.01 0.03 0.01 0.06 0.01 0.02 0.01 0.05 0.01 0.04 0.01 0.04 0.01 Write-o?/Total Assets Mean Median 0.09 0.02 0.41 0.02 0.08 0.01 0.23 0.03 0.04 0.02 0.17 0.03 0.09 0.03 0.13 0.02

Type Asset impairment charge Discontinued operations Layo? charge Restructure(asset and layo? based) Severance Partial write down Write-o? of assets Total

Percent 7.61 14.28 8.86 56.35 4.09 3.6 5.22 2472

Average Charge $67,100,000 $18,900,000 $69,700,000 $72,100,000 $38,200,000 $78,800,000 $23,400,000 $59,400,000

33

Table III

Summary of Estimation Variables

This table shows the summary mean and medians of the independent variables used in the probit estimations. Panel A is across the whole sample. MV is the size of the ?rm, de?ned as the log of the market value. SHROWNPC is the CEO’s percentage ownership in the company, shown as a percent. SALARY is the dollar annual CEO salary in 100ks, BONUS is the dollar annual CEO bonus in 100ks, and OPTIONS are the aggregate dollar value of all options granted to the executive during the year as valued by the company in 100ks. DIRSTK is the number of shares, which each non-employee director received during the year in thousands. SIZEBD is the number of directors, both inside and external. PER OUT is the percentage of directors who have no relationship with the company. . GOV INDEX is the Governance index, the lower the number the better the shareholder protection, and the higher the number, the higher the level of managerial entrenchment. GOV INDEX is an index number formed from charter provisions listed in the IRRC publications. It does not replace board composition. DOLLAR SENSITIVITY = is the dollar sensitivity of compensation to performance. RETURN SENSITIVITY = is the return sensitivity of compensation to performance. The signi?cance test uses a two-sided test to determine whether there is a statistical di?erence between the non write-o? benchmark ?rms, and the write-o? ?rms. Panel B shows the summary of governance variables on a per year basis for all ?rms in the sample. It also shows the percent of ?rms that decreased, remained unchanged , and increased their governance variables. A * denotes signi?cance at the 5 percent level, and ** denotes signi?cance at the 10 percent level. Panel A Non-Write-o? One-Time Write-o? Mean Median Mean Median t-value 5.25 5.64 9.64 10.53 13.92* 2.00 0.00 4.00 0.00 1.94** 666.21 650.00 629.55 599.07 1.130 848.11 413.27 534.5 351.00 2.24* 1,346.08 353.44 2,385.19 451.26 1.99** 0.06 0.00 0.05 0.00 0.350 12.05 12.00 10.17 10.00 -11.97* 68.00 70.00 73.00 75.00 -2.87* 9.06 10.00 8.10 9.00 1.99** 3,286.00 61 20,948.00 57.00 -2.98* 231.00 43.00 1,972.00 34.00 -2.33* 7.3 7.33 4.25 4.38 3.85* Panel B Board Size Mean -1 11 9% 11.86 11% 12.51 7% 11.52 10% 11.76 8% 12.24 10% 11.58 12% 11.83 14% 11.83 18% 10.61 13% Percent Outsiders 0 1 Mean 68% 23% 0.73 74% 15% 0.76 86% 7% 0.74 81% 9% 0.76 75% 16% 0.76 75% 14% 0.77 71% 17% 0.75 64% 22% 0.75 68% 14% 0.75 74% 12% 0.76 Multiple Write-o? Mean Median t-value 12.490 13.45 5.62* 3.00 0.00 0.22 657.95 667.51 0.32 548.31 439.46 3.67* 2,039.56 663.81 0.89 0.13 0.00 1.07 11.05 11.00 -3.67* 76.00 78.00 -7.11* 9.06 10.00 2.62* 8,842.00 68.00 -1.91** 864.00 55.00 -1.98** 4.7 5.83 3.71*

MV SHROWNPC SALARY BONUS OPTIONS DIRSTK SIZEBD PERC OUT GOV INDEX RETURN SENSITIVITY DOLLAR SENSITIVITY ROA GOV Mean 9.66 9.6 9.62 9.52 9.6 9.46 9.24 9.63 9.13 8.95

Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

INDEX -1 0 2% 93% 3% 94% 2% 94% 7% 78% 2% 97% 7% 76% 2% 96% 2% 95% 5% 83% 2% 96%

1 5% 3% 4% 15% 1% 18% 2% 2% 12% 2%

-1 7% 10% 5% 7% 6% 9% 10% 16% 11% 10%

0 66% 65% 64% 64% 64% 57% 57% 62% 64% 69%

1 26% 26% 30% 29% 30% 34% 34% 22% 26% 21%

34

Table IV

CEO turnover and the Write-o? Decision

Panel A shows the characteristics of executive turnover associated with a write-o? announcement. The turnover data is from EXECUCOMP. If the write-o? occurs up to one year after the CEO turnover, I include the turnover in the sample. There are two types of turnovers shown; those that occur with the replacement coming from inside the ?rm; and those with replacements coming from outside the ?rm. The table shows the percentage of non write-o?, one-time write-o?, and multiple write-o? ?rms that have experienced CEO turnover. The t-value tests the hypothesis that write-o?s have statistically signi?cant di?erent number of CEO turnovers to the industry benchmark. Panel B shows the probit estimation of CEO turnover with external replacement controlling for ROA, market value, and ROE. The results of this estimate are used to calculate the Inverse Mill’s Ratio. Panel A Non-Write-o? One-Time Write-o? Inside Outside Total Inside Outside Total 0% 0% 0% 2% 7% 9% 0% 0% 0% 0% 2% 2% 0% 0% 0% 0% 2% 2% 23% 77% 100% 17% 70% 87% 8 27 35 11 48 59 -1.99** Panel B Constant MV R ROA Coe?. -1.07 -.15 -0.03 -0.02 T-value -5.65 * -2.01* -3.66 * -2.96* Multiple Write-o? Inside Outside Total 2% 8% 10% 2% 0% 2% 0% 0% 0% 31% 57% 88% 46 85 130 -2.07*

Reasons Retires Resigns Dies Not Listed Total Number of Firms t -value

35

Table V

Probit Model Results

Panel A shows the results of the probit estimations for various models. Model 1 tests for a relation between CEO turnover and the write-o? decision as shown in Equation (5). Model 2 (A) tests the probit estimate shown in Equation (6). Model 2 (B) tests the probit estimation discussed in Equation (7). Model 3 shows the results for the estimation of Equation (8), which tests for the relationship between the board’s independence and the probability of a write-o? occurring. Model 4 (A) tests the probit estimate of the impact of industry shocks on the write-o? decision, while controlling for the quality of the company, as described in Equations (9) and (10). MV is the size of the ?rm; SHROWNPC is the percentage ownership in the company. CEO OUT is a dummy variable, zero for ?rms without turnover, and one for ?rms with turnover and replacement from outside the ?rm. ROA is the return on assets, and TLCF is a dummy variable that is one if a ?rm has tax loss carry forwards. SALARY is the annual CEO salary, BONUS is the annual CEO bonus, and OPTIONS are the aggregate value of all options granted to the executive during the year as valued by the company. INTERLOCK is one if the management is entrenched and zero if it is not, as de?ned by Execucomp. Entrenchment generally involves one of the following situations: the o?cer serves on the board committee that makes his compensation decisions, or serves on the board (and possibly compensation committee) of another company that has an executive o?cer serving on the compensation committee of the indicated o?cer’s company. Hallock (1997) describes the use of this variable as a proxy for managerial entrenchment. RETYRS is the number of years the CEO has been in o?ce. PPS is either the return sensitivity or the dollar sensitivity. The results shown are for the dollar sensitivity, although I perform the analysis for both measures. SIZEBD is the number of directors, both inside and external. PER OUT is the percentage of directors who have no relationship with the company. DIROPT is the number of options, which each non-employee director received during the year in thousands. DIRSTK is the number of shares, which each non-employee director received during the year in thousands. GOV INDEX = Governance index, the lower the number the better the shareholder protection. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. Independent Constant MV CEO OUT DEBT RAT SHROWNPC ROA TLCF SALARY BONUS OPTIONS PPS RETYRS INTRLOCK SIZEBD PERC OUT DIROTP DIRSTK GOV INDEX SHOCK RECESSION EXPANSION ?2 (d.f.) Model 1 -2.54 0.41 * 0.04 * -1.77 * -0.63 ** -0.14 * 0.99 * Model 2(A) -3.49 * 7.37 * -2.71 * 4.23* -0.07* 0.08* -0.01 * 0.00 -0.02 Model 2 (B) -3.46 * 0.59 * -2.36 * -0.08 * 1.38 * Model 3 -5.65 * 0.58 * -0.44 * -0.06 * 1.07 * Model 4 (A) -7.12 * 0.81 * 1.13 * -5.98 * 3.21 -0.96 * 3.57 * Model 4 (B) -7.20 * 0.78 * -1.36 * -5.01 * 1.62 -0.09 * 3.49 *

0.01 * -0.02 * -1.53 * -0.08 * 1.84 * 0.10 * -0.08 -0.03 *

0.91 -0.29 -4.83 * -0.27 * 4.42 * 0.33 * -0.01 -0.21 * 0.17 *

0.01 -0.03 -4.74 * -0.24 * 4.49 * -0.02 * -0.01 -0.16 * 0.58 * -0.12 385.00

451.90

417.00

370.00

629.00

438.00

36

Table VI

Marginal E?ects of Probit Estimates

This table shows the marginal e?ects of the probit estimations for various models. Model 1 tests for a relation between CEO turnover and the write-o? decision as shown in Equation (5). Model 2 (A) tests the probit estimate shown in Equation (6). Model 2 (B) tests the probit estimation discussed in Equation (7). Model 3 shows the results for the estimation of Equation (8), which tests for the relationship between the board’s independence and the probability of a write-o? occurring. Model 4 (A) tests the probit estimate of the impact of industry shocks on the write-o? decision, while controlling for the quality of the company, as described in Equations (9) and (10). MV is the size of the ?rm; SHROWNPC is the percentage ownership in the company. CEO OUT is a dummy variable, zero for ?rms without turnover, and one for ?rms with turnover and replacement from outside the ?rm. ROA is the return on assets, and TLCF is a dummy variable that is one if a ?rm has tax loss carry forwards. SALARY is the annual CEO salary, BONUS is the annual CEO bonus, and OPTIONS are the aggregate value of all options granted to the executive during the year as valued by the company. INTERLOCK is one if the management is entrenched and zero if it is not, as de?ned by Execucomp. Entrenchment generally involves one of the following situations: the o?cer serves on the board committee that makes his compensation decisions, or serves on the board (and possibly compensation committee) of another company that has an executive o?cer serving on the compensation committee of the indicated o?cer’s company. Hallock (1997) describes the use of this variable as a proxy for managerial entrenchment. RETYRS is the number of years the CEO has been in o?ce. PPS is either the return sensitivity or the dollar sensitivity. The results shown are for the dollar sensitivity, although I perform the analysis for both measures. SIZEBD is the number of directors, both inside and external. PER OUT is the percentage of directors who have no relationship with the company. DIROPT is the number of options, which each non-employee director received during the year in thousands. DIRSTK is the number of shares, which each non-employee director received during the year in thousands. GOV INDEX = Governance index, the lower the number the better the shareholder protection. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. Independent MV DEBT RAT CEO OUT SHROWNPC ROA TLCF SALARY BONUS OPTIONS PPS RETYRS INTERLOCK SIZEBD PERC OUT DIROTP DIRSTK GOV INDEX SHOCK RECESSION EXPANSION Model 0.05 -0.20 4.49 0.07 -0.02 0.11 1 * * * * * * Model 2(A) 0.06 * -0.03 * 0.57 * -0.07 * 0.08 * 0.01 * -0.03 -0.02 Model 2 (B) 0.01 * -0.04 * Model 3 0.05 * -0.01 * Model 4 (A) 0.07 * -0.02 * 0.02 * 0.47 -0.08 * 1.15 * Model 4 (B) 0.07 * -0.03 * 0.03 ** 0.35 -0.09 * 1.52 *

-0.08 * 0.89 *

-0.06 * 1.07 *

0.09 * 0.00 * -0.13 -0.01 * 0.22 * 0.01 ** -0.01 -0.01 *

0.01 0.00 -0.19 -0.01 * 0.29 * 0.00 ** 0.00 -0.01 * -0.01 *

0.00 0.00 0.61 * -0.01 * 0.30 * 0.00 ** 0.00 -0.01 * 0.01 ** -0.02

37

Table VII

Corporate Governance Measures and Write-o?s

Panel A shows the univariate results for the governance quality of the weakest and strongest governance write-o? ?rms. The signi?cance tests whether good governance variables are di?erent from bad governance variables. Panel B shows the estimation of Equation (9) for the 50 percent worst governance ?rms. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. Panel A: Strong Monitors vs. Weak Monitors Weak Monitors Strong Monitors t-value Mean Median Mean Median t-value BDSIZE 18.00 18.00 6.05 6.00 36.01* PERC OUT 0.46 0.48 0.80 0.75 2.98* GOV INDEX 10.12 11.00 5.94 6.00 2.31* Panel B: Multivariate Analysis of Governance Independent Model 4 (A) t-values MV 1.35 2.88* 0.02 0.06 CEO OUT SHROWNPC 0.47 0.35 ROA -0.05 -0.65 TLCF 2.63 1.91** PPS 0.00 -0.37 RETYRS -0.04 -0.98 INTERLOCK -0.19 0.61 SIZEBD 2.35 2.01* -0.03 -0.99 PERC OUT DIRSTK 0.38 1.38 0.47 2.69* GOV INDEX SHOCK 0.00 0.09

38

Table VIII

Corporate Governance Measures and Write-o?s

Panel A shows the univariate results for the size of the write-o? charges based on governance quality. The signi?cance tests whether good governance charges are di?erent from bad governance charges. Writeo? charges are adjusted by the total assets. Panel B shows the estimation of Equation (12), W O/T A = ?1 + ?2?7 GOV V ARSi + i. (12)

This robust regression tests whether governance a?ects the size of the write-o?. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. Panel A: Univariate Estimate Mean Median t-value Median test Well-monitored 0.02 0.01 Poorly-monitored 0.06 0.03 1.53 0.09** Panel B: Robust Regression Coe?cient t-value MV 0.01 2.21* ROA -0.03 -0.9 TLCF 0.04 1.92** SIZEBD 0.02 2.76** -0.18 -1.98* PER OUT -0.03 -0.40 CEO OUT GOV INDEX 0.02 0.25 PPS 0.01 0.18 CONSTANT 0.12 0.91 0.09 R2

39

Table IX

Abnormal Return Breakdown, by Year and Firm Type

This table reports the breakup of the type of ?rm year by year, based on write-o?s in the period +1 of 1985-2000. I compute abnormal returns as ARi = T T ?1 Ri,t ? Rs,i,t , where Ri,t is the return on date t for ?rm i, and Rs,i,t is the return on date t, of the equally weighted index of the size portfolio s to which ?rm i belongs. AR is reported in percentage format. t is the announcement date. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. One-time Write-o? 0.00 1.00 Sum -0.65 -0.36 -1.30 -.62 1.20 -0.81 0.30 -1.34 0.09 -0.70 -0.73 -1.95** -0.98 2.20 1.03 -1.69 0.69 0.72 1.16 -0.51 -0.70 -1.93 -3.67 -1.69 -0.10 0.11 0.75 -0.79 -1.65 1.04 0.85 -1.29 -0.25 -0.29 -0.63 -1.06 -0.63 -0.55 -1.23 0.00 -2.16* -2.55 0.66 -1.98 -1.29 1.77 0.84 1.11 -.69 -1.04 0.14 -0.70 -1.27 0.45 0.43 1.36 -1.95** -1.74 -1.09 -1.31 -1.82 0.58 -2.38 -1.65 -1.95** -1.06 -0.40 -1.82 -2.02* Multiple Write-o? 0 1 Sum -2.26 -1.74 -4.95 -2.28* 0.05 1.07 0.64 -0.56 -0.03 0.82 0.32 -0.89 -1.10 0.85 -0.34 -.65 -0.28 -0.54 -0.61 -1.47 0.71 0.47 1.43 1.95** -0.23 0.07 -0.57 -1.10 -0.26 0.52 0.97 -2.03 0.51 0.16 1.37 -0.08 0.37 0.16 0.73 -1.05 0.04 -0.30 -0.12 -.23 0.43 0.24 0.19 -.41 -0.41 -0.44 -1.04 -1.11 0.35 0.45 0.58 -0.99 -0.78 -0.02 -0.93 -.84 -0.03 0.38 0.50 -0.49 -0.19 -0.17 -0.37 -0.49 All Firms 0 1 -1.46 -1.05 0.29 0.15 -0.72 0.21 -0.55 -0.07 -0.50 0.24 -0.13 -0.14 0.45 -0.42 -0.20 0.02 0.54 -0.63 0.05 0.28 0.57 0.09 0.08 -0.24 0.45 0.11 -0.18 -0.29 0.33 -0.15 -0.19 -0.50 -0.50 -0.29

Year 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 ALL

-1.00 -0.29 -0.09 -0.12 -0.19 -0.25 -1.04 0.74 1.46 -0.09 -0.05 -0.09 0.18 0.20 0.48 1.52 0.15 -0.36

-1 -0.95 -0.48 -0.47 -0.09 0.21 0.25 -0.41 0.71 0.70 0.20 0.14 -0.19 -0.22 -0.13 0.15 -0.01

-1 -0.62 -0.26 -0.59 -0.33 -0.02 0.19 -0.11 -0.74 -0.59 0.07 0.25 0.10 -0.13 -0.43 0.40 0.13 -0.19

Sum -3.13 -1.81 0.08 -1.12 -0.16 -0.56 -0.48 -1.49 0.28 -0.41 -0.28 -0.14 -0.42 -1.53 -0.79 -0.48 -0.24 -1.53 -0.24 -0.54 -0.18 -0.79 0.88 -1.72 -0.70 -0.56 -0.82 -0.79 -0.08 -0.73 0.17 -0.34 -1.10 -1.08

40

Table X

Market Reaction to Write-o?s

This table shows the combined impact of CEO turnover, pay-performance sensitivity, and board composition on the announcement e?ects for Equation (13), which uses only the write-o? ?rms: ARi = ?1 + ?2 SIZEi + ?3 SIZEBDi + ?4 P ERC OU Ti + ?5 CEO OU Ti +?6RECESSIONi + ?7 P P Si + ?8 ROAi + ?9 W O T Ai +?10 W O #i + ?11 T Y P Ei + ?12 DEBTi + ?13 GOV IN DEXi + i .
T +1 T ?1

(13)

I compute abnormal returns as ARi = Ri,t ? Rs,i,t , where Ri,t is the return on date t for ?rm i, and Rs,i,t is the return on date t, of the equally weighted index of the size portfolio s to which ?rm i belongs. A * denotes signi?cance at the 10 percent level, and ** denotes signi?cance at the 10 percent level. Independent Variable MV ROA DEBT RATIO GOV INDEX SIZEBD PERC OUT TYPE WO NUM RECESSION WO TA PPS CEO OUT CONSTANT Coe?cient 0.002 0.001 -0.030 -0.024 -0.016 0.020 -0.001 0.001 0.010 -0.011 0.026 0.002 -0.026 t-Value 2.51* -0.16 -1.65 -2.63* -1.97** 2.56 * -0.16 1.39 1.99** -3.41 * 11.49* 2.42* -2.15*

41

Table XI

Announcement Day Returns Sorted by Governance Measures

This table breaks up the announcement day e?ects into portfolios based on the governance quality, and CEO turnover. There are ten portfolios of governance qualities, with 1 housing the best governance companies, and 10 housing the worst. The governance data is from IRRC, and the turnover data from Execucomp. In parentheses is the t-values that test whether the returns are signi?cantly di?erent from +1 0. I compute abnormal returns as ARi = T T ?1 Ri,t ? Rs,i,t , where Ri,t is the return on date t for ?rm I, and Rs,i,t is the return on date t, of the equally weighted index of the size portfolio s to which ?rm I belongs. A * denotes signi?cance at the 5% level, and ** denotes signi?cance at the 10 percent level. One-Time Write-o? Multiple Write-o? No CEO Turnover CEO Turnover No CEO Turnover AR t-value AR t-value AR t-value 1.98 2.65* 6.10 2.70* 1.40 2.41* 1.90 4.90* 1.50 2.91* 1.10 2.11* 0.20 0.02 1.60 1.03 0.70 0.46 1.50 -0.02 1.90 1.80 -0.80 -1.11 -1.70 -0.83 -2.10 -1.79 1.10 2.66* 1.70 1.18 2.00 0.61 0.00 0.01 -3.10 -2.21* 7.00 0.94 -0.30 -0.41 -0.50 -0.43 -0.60 1.03 0.30 0.46 0.30 0.26 1.50 0.01 0.80 0.95 0.00 0.09 -1.40 -1.22 -1.60 -1.95**

GOV Port 1 2 3 4 5 6 7 8 9 10

CEO Turnover AR t-value 1.70 1.55 2.20 1.99** 1.00 1.30 0.40 0.30 -0.80 -1.18 0.70 0.78 0.90 0.72 0.70 0.62 0.40 0.33 0.40 0.15

42

Chapter 2 Write-o?s and Liquidity

43

2.1 Introduction
Write-o?s are fast becoming a prominent event in U.S. ?nancial markets. The number of write-o?s for consumer manufacturing ?rms increased from 1980 to 2000 by 140 percent. With this increased usage of write-o?s, it has become increasingly important to understand what, if any, impact write-o?s have. In this paper, I analyze the e?ect of write-o? announcements on stock market liquidity. If there is a high level of asymmetric information, spreads will increase to re?ect this knowledge gap. Likewise, if there is a decrease in asymmetric information, spreads will decrease to re?ect this improved information environment. Information asymmetry between investors and management can hurt a ?rm’s market value (Myers and Majluf, 1984). Informed traders thrive in a less transparent environment and pro?t more from their private information, which creates an adverse selection problem for investors. O’Hara (2003) argues that reducing the amount of hidden private information can favorably a?ect asset prices due to improved price-discovery process and liquidity. Prior research provides a framework for this study of investigating the relationship between write-o? announcements and secondary market liquidity. Write-o?s may convey speci?c information about operating performance and strategies. When ?rms announce write-o?s, two types of information might be disclosed to the public. The write-o? announcement could uncover a problem not known to exist before the announcement. The announcement can also show how the ?rm is taking actions to repair the problem area. This voluntary disclosure of information has the potential to reduce information asymmetry by making private information acquisition more readily available to potential traders. The spread could therefore be a?ected by a decrease in perceived adverse selection risk that is not re?ected by an observable decrease in volumes. I use several tests to determine what impact write-o?s have on secondary market liquidity. Using univariate analysis, I compare the absolute and relative spreads before a write-o? announcement days to the write-o? window and ?nd a signi?cant improvement in liquidity following a write-o? announcement. I run the same analysis for trading volume and total number of transactions and ?nd that both increase following the write-o? announcement. In addition, I test whether the liquidity impact of write-o?s is di?erent from any other announcement. I ?nd that write-o? announcements show a greater liquidity improvement than earnings announcements. I also use multivariate analysis to test whether the liquidity e?ect (as shown by absolute spread, relative spread, and total number of transactions) is robust to the inclusion of price, volume, and volatility control variables. Both absolute and relative spreads decrease following a write-o? announcement. The number of transactions improves following a write-o? announcement. Taken as a whole, the ?ndings demonstrate that write-o? announcements generate a bene?t to investors in the form of improved liquidity. Next, looking only at write-o? ?rms, I test whether the liquidity bene?t of write-o?s is greater for companies with good corporate governance versus companies with bad corporate governance. Minnick (2004) show that the market reacts di?erently to write-o? announcements, based on the quality of the company’s governance. If a company has e?ective monitoring mechanisms, then traders may trust the quality of the information to a greater extent than the information from a poor governance ?rm, leading to a greater reduction in the asymmetric information. I ?nd that governance does a?ect the liquidity e?ects of write-o?s. I ?nd the number of transactions increases and spreads decrease more for high governance ?rms versus poor governance ?rms consistent with a larger reduction in asymmetric information for high governance ?rms. Lastly, I decompose the bid-ask spread in order to measure changes in the adverse selection component resulting from write-o?s. A reduction in information asymmetry, resulting from the write-o? announcement should generate a decrease in the adverse selection component of the spread. In fact, it is the decrease in this component that is expected to produce both narrower spread and greater transactions volume. The positive relation between adverse selection and bid-ask spreads is well documented in the literature. See Brockman and Chung (1999) and He?in and Shaw (2000) for evidence of the inverse relation between adverse selection and secondary market liquidity. I ?nd that adverse selection costs decrease following a write-o?, and that this decrease is greater for companies with stronger monitoring mechanisms. These ?ndings paint an economically intuitive picture of managerial and investor behavior in the secondary market. Write-o?s convey private information that managers possess, but that outside market

44

participants do not observe. Market participants understand that write-o?s convey some information. When the write-o?s occur, they enter the market, thereby decreasing the bid-ask spreads and increasing number of transactions. This is especially important for companies with good monitoring mechanisms. This liquidity provision dynamic is important because higher liquidity can lead to lower costs of capital and higher ?rm values (see Amihud and Mendelson (1986), Barclay and Smith (1988), and Jacoby, Fowler, and Gottesman (2000)). Secondary-market investors adjust spreads, adverse selection costs, and number of transactions in a manner consistent with a reduction in information asymmetry. The remainder of this paper is structured as follows. In section II, the methodology and data selection methods are discussed. In Section III, the liquidity e?ects of write-o?s are tested by comparing average bid-ask spreads, trading volumes, and non-trading days before and after write-o?s while controlling for the behavior of non -write-o? ?rms. Section IV looks at corporate governance and write-o?s, while Section V looks at the adverse selection costs, and Section VI concludes the paper.

2.2 Sample and Methodology
To generate my sample, I collect write-o? information, focusing on NYSE listed companies from 1980 to 2000 in the 2000-2999 SIC code. Using the original company list, I search Lexis Nexis and Dow Jones News Retrieval services for speci?c key words. For each company, I search for articles that match key words for my sample. The key words I used are write down, write-o?, restructure, charge against earnings, layo?s, and severance. When the query results in a match, I take the ?rst article in the series of articles that refers to a current write-o? that the company is announcing. I use the date of the article as the announcement date of the write-o?. I obtain the following information from the article: the amount of the write-o?; whether the write-o? was generated by assets, layo?s, or both; the purpose of the write-o? (restructure, write-down, plant closing, etc.); the justi?cation cited by the company; and whether the write-o? amount is stated on a before-tax or after-tax basis. The sample contains asset-based and layo?-based write-o?s (see Minnick (2004) for a more in depth description of the data collection process). To ensure that write-o?s in my sample are not extensions of earlier events, I set an arbitrary standard under which I assume that any write-o? announcements occurring within six months of earlier write-o? announcements are related. This exercise is also performed for break o? points of one month, three months, four months, eight months, and twelve months. Although doing so a?ects the sample size, it does not a?ect the analysis or ?ndings. Therefore, I only describe results using the 6-month break point. I determine which write-o? is a ?rst-time event or a subsequent event. To de?ne multiple write-o?s, I need to establish an arbitrary time interval. The standard most researchers use de?nes multiple write-o?s as any write-o? event that occurs within 16 quarters of a prior write-o? event. To identify a company’s ?rst write-o?, I look at all write-o?s that occur during the ?rst ?ve years of the sample: 1980-1985. I require an initial period of 16 ?scal quarters with no write-o?s before I add a ?rm to the sample. I denote the write-o? following this break as a ?rst time write-o?. Because the original sample begins in 1980, the ?rst reported write-o? in the sample occurs in the ?rst quarter of 1985. To test the sensitivity of this break point, I also use ?ve other quarter break points to de?ne ?rst time write-o?s, (8, 12, 18, and 20 quarters) to separate consecutive write-o?s. My conclusions become more robust with the longer measures and weaken slightly with the short-term de?nitions, and since the inference changes only marginally, I use 16 quarters. After I identify the ?rst time write-o? for a company, write-o?s that follow are labeled as second, third, fourth write-o?s, etc. These subsequent write-o?s must occur within 16 quarters after the prior write-o?. If the write-o? occurs after 16 quarters, I label it as another ?rst time write-o?.1 The data collection and cleansing process leaves me with 230 companies that announced 1,075 write-o?s from 1985-2000. To examine write-o? company characteristics, it is important to have a benchmark to compare the write-o? ?rms. Out of the 390 NYSE listed ?rms in the 2000-2999 SIC codes for 1985-2000, there are 160 ?rms that have never had a write-o?. I match the announcement date of each write-o? to the 160 non write-o? ?rms. This results in 172,000 non write-o? matched ?rms. I then average across these ?rms to create a benchmark measure for every write-o? event.
1 See

Minnick (2004) for more details on the data collection.

45

The distribution of write-o?s over the 15-year sample period appears in Table 1. The number of write-o?s more than doubles from 1985 to 2000 (25 versus 90 write-o?s). Write-o?s that combine both assets and lay-o?s, such as restructuring, have the largest charges ($100.2 million on average), followed by layo?s ($68.7 million on average), and assets ($63.5 million on average). There is no clear trend in the charge amounts over time. When adjusting write-o?s by total assets, no ratio is greater than ?ve percent. Table 2 provides summary statistics for all of the NYSE listed consumer manufacturing companies from 1993 - 2000 in the consumer manufacturing sector a month prior to the write-o?. Approximately 25 percent of all the NYSE consumer-manufacturing ?rms took a write-o? over my sample. Although writeo?s are representative of the population, they tend to have lower price levels, trading volumes, and daily returns than the non-write-o? ?rms. Market capitalization is larger for write-o? ?rms, as compared to non-write-o? ?rms. Twenty-one percent of the write-o? companies in my sample have taken two write-o?s, and 14 percent have taken three or more write-o?s.

2.2.1 Liquidity Data
Raw trading data is collected from the New York Stock Exchange Trade and Quote database (TAQ). This database reports every round lot trade and every quote from 1993 onwards on the New York Stock exchange.2 I match the TAQ data to my write-o? dataset using perm numbers. This leaves 594 remaining write-o?s. The loss in data comes from excluding NASDAQ and AMEX ?rms from the sample and limiting the write-o?s to 1993 to 2000. I begin my analysis by looking at the liquidity trends surrounding write-o? announcements, both in the short and long terms. Using two separate measures for liquidity, I look at the trend from 500 trading days before the write-o? to 500 trading days after the write-o?. The variables for liquidity include relative bid-ask spread, absolute bid-ask spread, turnover, and number of transactions. Relative spread is de?ned as follows, RSPi,t = APi,t ? BPi,t , 0.5 ? (APi,t + BPi,t ) (2.1)

where APi,t is the closing ask price on day t for ?rm i, BPi,t is the closing bid price on day t for ?rm i, and RSPi,t is the relative spread on day t for ?rm i. Absolute spread is de?ned as follows, ASPi,t = APi,t ? BPi,t , (2.2)

where ASPi,t is the absolute spread on day t ?rm i. I ?lter out quotations for which the ask is smaller than or equal to the bid price (crossed markets), as well as all spreads greater than $5.00 and spreads that represent more than 20% of the quote midpoint (outliers). These ?lters a?ect less than one percent of the observations. Turnover is de?ned as the total monthly volume divided by number of shares outstanding. The daily average trading volume is from TAQ, and the shares outstanding are from Compustat. The number of transactions is de?ned as the number of round lots(100 shares)available to trade at the bid price plus the number of shares available to trade at the ask price as follows: T ransaction = BidT ransactions + AskT ransactions (2.3)

Market makers reduce the number of transactions when they are wary of the informational environment. Lower transaction numbers give market makers an opportunity to adjust prices quickly. Bacidore, Battalio, and Jennings (2002) suggest that each measure of liquidity is de?cient in properly assessing the level of liquidity. Having a composite measure is especially helpful in empirical analysis, especially if spreads and transactions point in di?erent directions. To alleviate this issue, I calculate a composite measure of liquidity, called transaction/spread ratio as described below. This measure is similar to both
2 See Hvidkjaer (2004) for an explanation of the database, and the technique used to aggregate the data to daily data.

46

Bacidore et. al. (2002) and Jain, Kim and Rezaee (2004). It depicts market liquidity as a function of both higher number of transactions and lower quoted spreads: T /S = T ransactionN umber/AbsoluteSpread (2.4)

I analyze changes in liquidity using two tests. The ?rst is a t-test comparing the cross sectional mean from the pre-announcement period to the cross sectional mean after the write-o? announcement. The second, more powerful test calculates for each stock, the di?erence between the mean before the announcement date and the mean after the announcement date. I then compare the frequencies of the increases and decreases between the write-o? and non-write-o? ?rms using a chi-square test under the null hypothesis that the relative frequencies are the same (Gibbons (1976)).

2.3 Liquidity E?ects
Table 3 shows the summary statistics for the di?erent liquidity measures for the write-o? ?rms six months before and after the write-o? announcement, compared to 25 days following the write-o? announcement. I de?ne the write-o? period as 25 days following the write-o? announcement, so the write-o? window is t=0 to t=25. The non write-o? window is de?ned as any time 120 days before the write-o? announcement, or 120 days after the write-o? announcement. The period after one write-o? is not mutually exclusive with respect to the period before another write-o?. Because there is no clear interpretation of before and after periods, I rely only on the surrounding non write-o? period as my benchmark. I calculate the means and medians for various measures across write-o? periods and surrounding non-write-o? periods for each sample ?rm. Table 3 provides summary statistics for the write-o? and surrounding non-write-o? periods, along with paired t-test and sign test results. Volume is the total trading volume per day. Price is the average daily price transaction, and Returns is the average daily return. Volatility measures the variance of returns. Absolute Spread, Relative Spread, Total number of transactions, Ask number of transactions and Bid number of transactions are daily averages for the absolute dollar spread, relative spread, total number of transactions, ask-side number of transactions, and bid-side number of transactions, respectively. The univariate test results show that average daily trading volumes are 9 percent higher during the write-o? period, as compared to the surrounding non write-o? period. The average stock price is 12 percent lower during the write-o? period as compared to the surrounding non-write-o? period. The average absolute (relative) spread of the write-o? ?rms is $0.28 (1%) while the average absolute (relative) spread of the surrounding non write-o? period is $0.37(2%). The average number of transactions of write-o? ?rms is slightly higher than the surrounding non-write-o? period. The liquidity measures yield strong evidence. Spreads tend to decrease, thereby increasing liquidity, while turnover and number of transactions tend to increase, which also increase liquidity. Overall, the write-o?s appear to have improved liquidity as compared to the surrounding non write-o? periods. I next test whether this liquidity improvement is unique to write-o?s or if it occurs for any quarterly announcement. To test for this unique reaction, I compare the liquidity changes the write-o? announcement to the liquidity changes of earnings announcements in the write-o? quarter. Using a univariate analysis, I compare the mean di?erence in the liquidity changes. The results are shown in Table 4. I ?nd that the liquidity improvement for write-o?s is signi?cantly di?erent than it is for earnings announcements. These ?ndings suggest that write-o?s have a greater impact on liquidity than other types of announcements. I have established that actual write-o? periods are associated with signi?cant changes in price, volume, and volatility. The univariate tests have also shown that liquidity improves in two ways, number of transactions, and spread. I next focus on measuring the impact of write-o?s on liquidity after controlling for changes in price, volume, and volatility. These three independent variables are widely used in the market microstructure literature to control for the trading e?ects on ?rm liquidity. Tinic and West (1974), Benston and Hagerman (1974), and Weston (2000) have shown that spread is positively related to share price. We include the return volatility measure since the risk of the security is a component of dealer risk and dealer inventory carrying costs. Several theoretical studies include risk as a factor that positively a?ects the

47

spread, including Garman(1976), Stoll (1989), and Ho and Stoll (1981). Table 5 presents the results from the following regression model: Liquidityi = ? + ?1 W Oi + ?2 V olumei + ?3 P ricei + ?4 V olatilityi + i , (2.5)

where Liquidityi , the dependent variable, represents three liquidity measures: absolute spread, relative spread, and total number of transactions. V olumei , P ricei , and V olatilityi are the independent control variables. W Oi is a dummy variable that is one if the day falls in the write-o? window and zero otherwise. All variables represent daily averages, and all but the dummy variable are transformed by taking the log. I adjust the t-statistics for heteroscedasticity, serial correlation, and arbitrary cross-correlations by using the Newey and West (1987) procedure. Table 5 provides the results from estimating equation (2.5) for each of the four liquidity measures, absolute spread, relative spread, total number of transactions, and the transaction/spread ratio. The coe?cients for all of the control variables are signi?cant at 5 percent or less. The signs of the coe?cients are consistent with microstructure theory. Higher volumes and prices are associated with higher liquidity, while higher volatility levels are associated with lower liquidity. The estimated volume coe?cients are negatively related to absolute and relative spreads, while positively related to total number of transactions. Increased stock prices are related to wider absolute spreads, narrower relative spreads, and increased number of transactions. Higher volatility is positively related to spreads and negatively related to number of transaction. The most important result of this estimate is the coe?cients for WO, the dummy variable for the write-o? period. The negative and highly signi?cant write-o? period coe?cients for both the absolute and relative spread regression demonstrate that bid-ask spreads decrease following a write-o? announcement, even after controlling for changes in price, volume, and volatility. The write-o? coe?cient is positive and signi?cant for the transaction regression, which shows the write-o? activity increases ?rm transactions.3 I interpret these results as evidence of the asymmetric-information hypothesis. When traders are a?ected by a decrease in the asymmetric information, they increase liquidity by reducing bid-ask spreads and increasing number of transactions. I show that spread decrease signi?cantly in both univariate and multivariate testing.

2.4 Governance and Liquidity
Minnick (2004) shows that the market reacts di?erently to the information content of write-o?s based on the quality of the announcing ?rm’s governance. This relationship between governance and write-o?s can also have implications on the liquidity e?ect of write-o? announcements. If the information ?owing from good governance companies’ write-o?s were more transparent than the information from bad governance write-o?s, then one would expect to see a greater improvement in liquidity for good governance ?rms, as compared to bad governance write-o? ?rms. To test whether this is true, I run the following model, Liquidityi = ? + ?1?5 1GOV V ARSi + ?6V olumei + ?7 P ricei + ?8 V olatilityi +
i,

(2.6)

where Liquidityi , the dependent variable, represents the change in three liquidity measures: absolute spread, relative spread, and total number of transactions. V olumei, P ricei , and V olatilityi are the independent control variables. All variables represent daily averages, and all are transformed by taking the log. GOV V ARS are the various governance variables used in Minnick (2004), including CEO turnover, board size, percent of outsiders on board, and shareholder protection index. NEWCEO is a dummy variable that is one if there was CEO turnover, and zero otherwise. BDSIZE is the number of members on the board,
3 I also run the analysis with three sub samples: companies with strong monitoring mechanisms, companies with mediocre monitoring mechanisms, and companies with weak monitoring mechanisms, based on the IRRC database from Gompers, Ishii, and Metrick (2003). I ?nd that the above results are driven by the strong and mediocre monitored companies. The poor weakly monitored companies do not show any signi?cant liquidity improvement.

48

PERCTOUT is the percent of uniquely independent outside board members, as discussed in Yermack (1999). GOV INDEX is the IRRC metric, where the higher the index, the worse the level of shareholder protection. Conversely, the lower the number, the stronger the shareholder protection. This measure is discussed in detail by Gompers, Ishii, and Metric (2002). I adjust the t-statistics for heteroscedasticity by using a robust regression. The analysis is run only on the write-o? ?rm 25 day window. Table 6 shows the results of the estimate of model (2.6). As before, the control variables in Table 6 are all highly signi?cant and exhibit the expected signs. More important are the results of the governance variables. The results suggest that the better the governance, the higher the improvement in liquidity. The board composition variables, as well as the shareholder protection variable are all signi?cant determinants for both the spread and number of transactions at the ?ve percent signi?cance level. BDSIZE and GOV INDEX are positively related to spreads, and negatively related to number of transactions. PERCTOUT is negatively related to spread and positively related to number of transactions. CEO turnover is negatively related to spreads and positively related to number of transactions. I interpret these results as evidence that companies with good board composition, or strong shareholder protection measures show a greater reduction in asymmetric information from a write-o? announcement than companies with poor governance. The empirical evidence in Tables 5 and 6 supports the following conclusions. When companies divulge private information, such as a write-o?, absolute and relative spreads decrease signi?cantly, and total number of transactions increase substantially. The quality of the information released is better, and less noisy for good governance ?rms, as compared to poor governance ?rms. To test the robustness of my results, I run the above analysis for both ?rst time and multiple writeo?s. Table 7 shows the results of these estimates. Panel A shows the results of model (2.6) for a company’s ?rst time write-o?. Panel B shows the estimate for multiple write-o?s. There is not much di?erence for segmenting the impact of liquidity by ?rst versus multiple write-o?s. I use a Hausman test to see if there is any signi?cant di?erence in liquidity for ?rst and multiple write-o?s. I ?nd that the liquidity bene?t of write-o?s exists regardless of how many write-o?s a company has taken in the past. In addition, I examine whether the size of the write-o? a?ects the changes in liquidity. The results are shown in Table 8. I ?nd that the larger the size of the write-o?, the smaller the impact on liquidity. I ?nd that larger write-o?s lead to smaller changes in spreads, and absolute spreads. The number of transactions is also negatively impacted by write-o? size.

2.5 Adverse Selection
A signi?cant recent advance in the market microstructure literature is the development of models that decompose the bid-ask spread into various components. In these models, the spread generally has three components: order processing, inventory holding, and adverse selection (asymmetric information). Demsetz (1968) and Tinic (1972) identify an order processing cost that is made up of exchange and clearing fees, bookkeeping and back o?ce costs, the market maker’s time and e?ort, and other random business costs. Since a large part of this cost is ?xed, order-processing costs are lower for more heavily traded securities. Inventory holding costs are due to order ?ow imbalances that cause the market maker’s inventory positions to deviate from optimal levels (Stoll (1978) and Ho and Stoll (1983)). Wider bid-ask spreads, and larger inventory holding costs result from increased deviation. Copeland and Galai (1983), Glosten and Milgrom (1985) and Easley and O’Hara (1987) suggest that asymmetric information and its consequent informed trading is a third spread component. Adverse selection costs are included in the spread to cover market participants’ expected losses to informed traders. In prior sections, I show that the write-o? announcement is associated with narrower spreads, and increased number of transactions, particularly after controlling for changes in price, volume, and volatility, and governance measures. In this section, I focus on the adverse selection component of the spread in order to isolate the potential cause of these improvements in liquidity. According to the information-asymmetry hypothesis, increased information in the market will decrease adverse selection costs. The validity of my

49

empirical results, however is dependent on the accuracy of the component estimation technique.4 To control for co-tangential speci?cation errors, I estimate the adverse selection components using several decomposition models, which help to address issues of robustness and accuracy. I intend to measure changes in adverse selection caused by a write-o? announcement. In each of the decomposition models, I introduce the interaction term WO, which takes the value of one for trades associated with write-o? announcements, and zero otherwise. I de?ne WO to include the actual announcement day, and 25 subsequent trading days.5 Alternative de?nitions of the period provide similar results as those reported here within. Positive and signi?cant coe?cients on the WO term would con?rm the hypothesis that write-o?s induce higher adverse selection costs. Measures such as bid-ask midpoints and transaction process are transformed by taking natural logarithms, as in Lin, Sanger, and Booth (1995). Each decomposition model is estimated on a pooled cross sectional basis.6 Lin Sanger and Booth (1995) develop a method of estimating empirical components of the e?ective spread that follows Huang and Stoll (1994), Lin (1993), and Stoll (1989). Lin et al. (1995) de?ne the signed e?ective half spread, zt , as the transaction price at time t, Pt, minus the spread midpoint, Mt . The signed e?ective half spread is negative for sell orders and positive for buy orders. To re?ect possible adverse selection information revealed by the trade at time t, Lin et. al (1995) add ?, which is the adverse selection component of the bid-ask spread. I follow the LSB (1995) decomposition technique. In the model, the parameters are estimated through the following regression equation: ?Mt+1 = ?zt +
t+1 ,

(2.7)

where, Mt = log quote midpoint at time t ? = parameter of regression which estimates the adverse selection component of the spread ? = Change in relative variable from t to t+1 zt = Pt - Mt Pt = log trade price at time t, and t+1 = random error term with zero mean and constant variance. I follow Lin et al. (1995) by using a robust OLS to estimate the following equation: ?Mt+1 = ?zt + ?W O (zt ? W Ot ) +
t+1 ,

(2.8)

where ?W O is the incremental adverse selection component during the write-o? period. Table 9 shows the estimated adverse selection component, ? of 0.15 (t-value = 23.55) is signi?cant at ?ve percent. The result can be interpreted as 15 percent of the bid-ask spread is attributable to information costs. The more important result is the estimated interaction term, ?W O of -0.30 (t-value = -23.44), which is signi?cant at ?ve percent. The interaction term con?rms that write-o? announcements decrease adverse selection costs. During write-o? periods, adverse selection decreases by 15 percent of the bid-ask spread. I also decompose the spread using the empirical model of Huang and Stoll (1997), and implemented by Weston (2000). They derive a simple model that allows a one-step decomposition of the information component as a percentage of the spread. The remaining spread stems from order-processing costs and market maker rents. The model identi?es these components by measuring how the midpoint of the spread, Mt , changes as a function of the direction of trades. They de?ne an indication variable Qt , which takes on the values,
4 See Van Ness, Van Ness, and Warr (2001) for a discussion of the various bene?ts of di?erent adverse selection models. 5 See Brockman and Chung (2001) for a description of this interaction technique. 6 I also run the estimations on a ?rm-by-?rm basis, and ?nd that it does not alter the results.

50

{-1,0,1} based on the direction of trade. If Pt ¡ Mt , then Qt = -1 (sell order), if Pt = Mt , then Qt =0, and if Pt . Mt , then Qt =1 (buy order). The model is speci?ed as ?Mt = ?(St?1 /2)Qt?1 + + t, (2.9)

where ? measures the proportion of the half spread St?1 /2, that stems from information costs. I follow Huang and Stoll (1997) by using a robust OLS to estimate the following equation: ?Mt = ?(St?1 /2)Qt?1 + ?W O (St?1 /2)Qt?1 ? W Ot ) +
t+1 ,

(2.10)

where ?W O is the incremental adverse selection component during the write-o? period. Table 8 shows that the estimated adverse selection component of the model, ? has a coe?cient of 1.43. (t-value = 5.01). This can be interpreted as 143 percent of the bid-ask spread is attributable to information costs. The interaction variable,?W O , is -1.5, (t-value = -4.61). This can be interpreted as evidence that write-o?s decrease adverse selection costs by 7 percent. Overall, the decomposition results are consistent with the liquidity results. Write-o?s reduce the information asymmetry in the market. This results in a situation where market participants react by reducing adverse selection costs, and increasing liquidity.

2.6 Robustness Checks
In addition to the microstructure literature concerning information asymmetry, another strand of literature considers analysts estimates as a good indicator of information asymmetry, where an increase in information about a ?rm tends to lead to a convergence of opinions regarding the ?rms expected future earnings. These papers typically use proxies for asymmetric information derived from consensus analysts forecasts of earnings per share. Krishnaswami and Subramaniam (1998), for example, use the analysts forecast errors to examine the change in the information environment before and after the completion of a spin-o?. In this section, I use analyst forecasts as an additional measure of the level of information asymmetry. I use IBES and First Call data to analyze the relationship between write-o?s and earnings. Table 10, Panel A, shows the summary statistics of the First Call analyst data. The estimates are for the ?rst quarter following the write-o? announcement. Instead of using the average of all of the analyst forecasts for a particular ?rm in a particular quarter, I look at each one individually. Doing so allows me to see if there is less dispersion in forecasts after the write-o?. The results are consistent with our earlier ?ndings. Overall, write-o?s lead to a signi?cant reduction in information asymmetry which is re?ected in the analyst estimates. These results are driven by the better goverened ?rms. The weakly monitored ?rms do not show a signi?cant change in earnings transparency(via a change is surprise). However, both the mediocre and strongly monitored ?rms do show an improvement. In Table 10, Panel B, I formally test the relationship between forecast error, and ?rm speci?c variables. Using the following regression, I estimate what impact write-o?s, governance, and earnings management have on the analyst forecast error: F ORECAST ERROR = ?1 +?2 SIZEi +?3 W Oi +?4?9 GOV V ARSi +?10 ACCRU ALSi +?11 GROW T H t+1 , (2.11) where FORECASTERROR is de?ned as the absolute value of the surprise, GROWTH is de?ned as the change in sales from the same quarter in the previous year. I also control for ACCRUALS, de?ned as GAAP earnings less cash from operations. I include growth because growing ?rms have predictably lower cash ?ows due to higher working capital and long-term capital investments, but less predictable earnings. I include ACCRUAULS as a proxy for earnings management, as a company that utilizes earnings management techniques has more easily predicted future earnings. The results support the above conclusions, where write-o?s lead to a signi?cant decline in forecast error. In addition, ?rms with smaller boards, and a higher percent of outside directors see a signi?cant decline in the forecast error. Overall, the forecast error evidence suggests that write-o?s, especially write-o?s from well monitored companies lead to a reduction

51

in information asymmetry, and an improvement in earnings transparency.

2.7 Conclusion
This study is motivated by the rise in the use of write-o?s over the past two decades. With the increase usage of write-o?s, it has become unclear as to whether write-o?s improve the information environment, or just create more noise. The problem is exacerbated by unclear disclosure policies for the write-o? events. Management has the discretion to decide when to take a write-o? and for how much the write-o? amount should be. This paper attempts to shed some light on the impact of the announcements on the information environment, by analyzing write-o? announcements from 1990 to 2000. The goal of this paper is to measure the impact of write-o?s on liquidity. Using three separate liquidity measures, I ?nd that liquidity increases substantially following a write-o? announcement. In univariate analysis, I ?nd that bid-ask spreads, both relative and absolute, decline, and that the number of transactions increase following write-o?s. Using a multivariate framework, after controlling for changes in price, volume, and volatility, I still ?nd that liquidity improves following a write-o? announcement. These results provide overwhelming evidence in favor of the information-asymmetry hypothesis. Following the test of the information asymmetry hypothesis, it is important to see if there is any di?erence in the liquidity bene?t from good governance write-o? ?rms versus bad governance write-o? ?rms. To test whether governance in?uences write-o?s, I use a multivariate analysis that controls for the impact of price, volume, and volatility on spreads and number of transactions. I ?nd that even when controlling for systematic changes, write-o?s from companies with small boards, larger percent of outside directors, and strong shareholder protection will lead to a greater improvement in liquidity than companies with poor governance. As a robustness check, I also use analyst forecast error as a measure of information asymmetry, and ?nd that write-o?s, especially write-o?s from well-monitored companies are related to a reduction in forecast error. Finally, I decompose bid-ask spreads in order to measure the e?ect of write-o?s on adverse selection costs. The adverse selection results con?rm that write-o?s improve the information asymmetry, and improve liquidity as a response. Adverse selection costs decrease signi?cantly during the write-o? announcement period in all three decomposition models. Overall, spread, number of transactions, and decomposition results suggest a picture where write-o?s, especially those from good governance ?rms, improve the information environment and lead to a liquidity bene?t for investors.

52

2.8 Tables
Table I Sample Information

This table shows, by year, the number of write-o? announcements for layo? based, asset based, and combined write-o?s. It also shows the mean and median write-o? charge by write-o? type. The prior quarter’s total assets to create a ratio adjust the write-o? charge amounts. The mean and median of this ratio, Charge/TA, is shown. The charge amounts are in millions of dollars. Asset Write-o?s Charge Charge/TA Mean Med. Mean Median 10.4 2.7 0.009 0.003 44.8 6.7 0.024 0.012 65.0 12.0 0.041 0.012 11.2 5.7 0.009 0.002 15.0 2.6 0.010 0.003 41.7 13.7 0.010 0.006 79.2 7.0 0.019 0.016 36.5 25.0 0.029 0.006 99.9 21.7 0.029 0.004 134.0 13.6 0.021 0.017 63.3 15.7 0.023 0.021 69.3 42.9 0.058 0.007 191.0 9.0 0.032 0.006 28.6 16.2 0.016 0.005 46.5 15.7 0.007 0.007 79.6 16.5 0.014 0.004 63.5 9.0 0.022 0.007 Layo? Write-o?s Charge Charge/TA Mean Med. Mean Med. 62.5 62.5 0.010 0.013 3.5 3.5 0.034 0.013 61.3 51.0 0.032 0.026 67.4 34.2 0.030 0.009 92.5 51.6 0.021 0.013 147.0 139.0 0.032 0.013 81.2 10.5 0.011 0.002 82.6 48.0 0.021 0.015 88.0 33.0 0.021 0.008 35.1 18.0 0.011 0.005 104.0 21.6 0.029 0.001 40.5 1.9 0.049 0.002 96.7 62.0 0.010 0.005 30.6 15.0 0.049 0.002 5.9 5.8 0.018 0.009 63.2 62.4 0.006 0.007 68.7 20.0 0.012 0.002 Combination Write-o?s Charge Charge/TA Mean Med. Mean Med. 175.0 44.0 0.019 0.007 44.7 13.2 0.013 0.005 234.0 43.5 0.045 0.006 122.0 12.2 0.033 0.007 51.7 16.5 0.012 0.004 88.8 35.0 0.026 0.005 57.8 24.3 0.018 0.007 108.0 49.5 0.022 0.010 132.0 43.4 0.018 0.011 129.0 49.5 0.022 0.012 64.8 16.3 0.015 0.005 109.0 29.8 0.016 0.008 101.0 27.5 0.017 0.008 100.0 30.6 0.019 0.007 111.0 36.2 0.061 0.007 66.2 27.9 0.022 0.011 100.2 29.2 0.023 0.007

Year 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 All

# 25 34 45 44 55 53 75 68 82 82 78 85 86 86 87 90 1075

53

Table II

Summary Statistics

This table shows summary statistics on market and write-o? activity for NYSE listed companies in the consumer manufacturing industry. The sample period spans 1993-2000. Comparative market statistics over the same period are given for the population of all non write-o? ?rms on the NYSE. Average market capitalization Average trading Volume Average daily closing price Average daily returns (with dividends) Average size of write-o?s, adjusted by total assets Percentage of companies with one write-o? Percentage of companies with second write-o?s Percentage of companies with third + write-o?s Write-o? Companies 11,200,000 118,319 43.98 0.02 0.01 25% 21% 14% Non Write-o? Companies 10,600,000 107,331 37.19 0.01

54

Table III

Liquidity ummary Statistics

This table shows summary statistics on liquidity measures for write-o? periods versus surrounding non write-o? periods. The write-o? window is t=0 to t=25. The non write-o? window ends 120 days before an announcement, and begins 120 after a write-o? announcement. Volume is the total trading volume on a trading day. Returns is the returns over each trading day. Volatility measures the variance of returns. Price is the average daily transaction price. Absolute spread, relative spread, total number of transactions, number of ask transactions, and bid transactions are the averages for the absolute dollar bid-ask spread, the relative percent bid-ask spread, total number of transactions, number of ask transactions, and bid transactions. The t statistics are from the paired t-test for the di?erences in means between the write-o? window, and the surrounding non write-o? window. The sign test statistics are from the non-parametric sign test for the di?erences in the median measures between the write-o? window, and the surrounding non write-o? window. All p-values are reported based on two tailed signi?cance. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively. WO Period Mean Median 248.13 160.8 33.74 31.54 0 0 0.4 0.49 0.28 0.19 0.01 0.01 349.26 279.99 164.92 131.96 158.8 146.83 Non WO Period Mean Median 229.27 134.82 35.91 33.09 0 0 0.39 0.46 0.37 0.19 0.02 0.01 280.05 212.68 133.97 100.15 128.86 113 Di?erence Mean Median 18.86 25.98 -2.17 -1.55 0 0 0.02 0.03 -0.09 0 -0.01 0 69.22 67.31 30.95 31.81 29.94 33.83 Signi?cance Tests Paired t-test Sign test -2.8** 0.00** 6.41** 0.00** -0.83 0.21 -7.39** 0.83 12.96 ** 0.56 6.3 ** 0** -15.15** 0** -13.81** 0** -16.11** 0 **

Volume Price Returns Volatility Absolute Spread Relative Spread Total Transactions Ask Transactions Bid Transactions

55

Table IV

Univariate Analysis of Write-o? Liquidity Changes to Earnings Liquidity Changes

This table looks at the univariate statistics for the change in liquidity when a company announces a writeo? and when a company announces its quarterly earnings. Change is calculated as the di?erence between liquidity for a write-o? ?rm, and all other ?rms on the NYSE on the announcement date. he t statistics are from the paired t-test for the di?erences in means between the write-o? window, and the surrounding non write-o? window. I compare the average values of changes in absolute spreads, relative spreads, and total volume for both the earnings and write-o? announcement date. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively. Absolute Spread Relative Spread Turnover WO change -0.08 -0.10 52 Earnings change -0.04 -0.006 30 t-value -8.69 ** -5.97 ** 8.67 **

56

Table V

Multivariate Analysis of Liquidity

This table shows the results of a regression of liquidity measures across write-o? and non write-o? periods, controlling for the e?ects of price, volume, and volatility. Liquidityi = ? + ?W Oi + ?1 V olumei + ?2 P ricei + ?3 V olatilityi + , where Liquidityi is the dependent variable and stands for either the log of absolute spread, relative spread ,total number of transactions, or total number of transactions/absolute spread. Absolute spread is a measure of the average absolute dollar bid-ask spread of a sample ?rm. Similarly, relative spread and total number of transactions are the daily averages for the relative bid-ask spread, and total number of transactions. WO is coded with a one if the day is within 25 days following a write-o? event, otherwise 0. Volume is the total trading volume during the trading day. Price is the average of all transaction prices recoded on the trading day. Volatility is the variance of returns over the trading day. All non-dummy variables are calculated by taking the natural logarithm. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively and all results are presented based on two-tail signi?cance. Absolute Coe?cient -0.098 -0.029 0.004 -0.081 -1.168 165.42 Spread t-value -5.92** -2.83 1.43 ** -23.95 ** -31.63 ** Relative Coe?cient -0.075 0.152 -0.008 -0.227 -4.755 777.96 Spread t-value -2.05 ** 12.43 ** -2.43 ** -55.38 ** -109.35 ** Total Trans. Coe?cient t-value 0.194 11.97 ** -0.014 -12.98 ** 0.005 107.21 ** 0.598 2.73 ** 2.521 55.91 ** 6989.84 Total Trans/Abs. Spread Coe?cient t-value 0.129 3.32 ** -0.203 -7.23 ** 0.621 65.74 ** 0.028 3.89 ** 4.727 48.43 ** 1157.98

WO Price Volume Volatility Constant F(4,7592)

57

Table VI

Multivariate Analysis of Liquidity and Governance

This table shows the results of a regression of liquidity measures across write-o?s and governance measures, controlling for the e?ects of price, volume, and volatility. Liquidityi = ? + ?1?5 1GOV V ARSi + ?6 V olumei + ?7 P ricei + ?8 V olatilityi +
i

where Liquidityi is the dependent variable and stands for either the percent change in absolute spread, relative spread , total transactions, or total transactions/absolute spread. from the non write-o? period to the write-o? period . Absolute spread is a measure of the average absolute dollar bid-ask spread of a sample ?rm. Similarly, relative spread and total transactions are the daily averages for the relative bid-ask spread, and total number of transactions. Volume is the total trading volume during the trading day. Price is the average of all transaction prices recoded on the trading day. Volatility is the variance of returns over the trading day. All liquidity variables are calculated by taking the natural logarithm. BDSIZE is the size of the board of directors. PERCTOUT is the percent of outside directors on the board. GOV INDEX is an index that ranks the level of shareholder protection for the shareholders, where the higher the number, the worse the protection. It is calculated using the Gompers, Ishii, and Metrick (2002) index. NEWCEO is a dummy variable that is one if there was a CEO turnover in the days following. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively and all results are presented based on two-tail signi?cance. Absolute Spread Coe?cient t-value -0.24 -6.23 ** 0.04 4.21 ** -0.04 -2.62 ** 0.03 3.5 ** 0.08 0.51 -0.16 -3.48 ** 0.03 3.41 ** 1.02 5.09 ** 101.29 Relative Coe?cient -2.64 0.12 -0.27 0.11 0.62 -0.43 0.17 11.12 129.41 Spread t-value -18.06 ** 3.39 ** -5.26 ** 3.86 ** 1.03 -2.4 ** 5.22 ** 14.59 ** Total Trans. Coe?cient t-value 0.27 10.97 ** -0.03 -13.41 ** 0.55 9.67 ** -0.12 -5.54 ** 1.66 1.21 0.28 2.99 ** -0.07 -3.15 ** -4.96 -26.66 ** 1218.5 Trans./Abs. Spread Coe?cient t-value 0.08 5.64 ** 0.03 3.41 ** 1.05 104.61 ** -0.02 -8.16 ** 0.049 0.9 0.05 3.02 ** -0.03 -10.9 ** 1.56 22.49 ** 847.67

Price Volume Volatility BDSIZE PERCTOUT NEWCEO GOV INDEX Constant F( 7, 3610)

58

Table VII

Liquidity and Number of Write-o?s

This table shows the results of a regression of liquidity measures across write-o?s and governance measures, controlling for the e?ects of price, volume, and volatility. Liquidityi = ? + ?1?5 1GOV V ARSi + ?6 V olumei + ?7 P ricei + ?8 V olatilityi +
i

where Liquidityi is the dependent variable and stands for either the percent change in absolute spread, relative spread , or total Transactions from the non write-o? period to the write-o? period . Panel A shows the results for ?rst time write-o?s, while Panel B is for multiple write-o?s. Absolute spread is a measure of the average absolute dollar bid-ask spread of a sample ?rm. Similarly, relative spread and total Transactions are the daily averages for the relative bid-ask spread, and total number of transactions. I also run the estimate for transactions divided by absolute spread. Volume is the total trading volume during the trading day. Price is the average of all transaction prices recoded on the trading day. Volatility is the variance of returns over the trading day. All liquidity variables are calculated by taking percent change in liquidity write-o?s compared to the rest of the NYSE on that day. BDSIZE is the size of the board of directors. PERCTOUT is the percent of outside directors on the board. GOV INDEX is an index that ranks the level of shareholder protection for the shareholders, where the higher the number, the worse the protection. It is calculated using the Gompers, Ishii, and Metrick (2002) index. NEWCEO is a dummy variable that is one if there was a CEO turnover in the days following. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively and all results are presented based on two-tail signi?cance. Panel A Absolute Spread Coe?cient t-value -0.24 -6.23 ** 0.04 4.21 ** -0.04 -2.62 ** 0.03 3.50 ** 0.08 0.51 -0.16 -3.48 ** 0.03 3.41 ** 1.02 5.09 ** 101.29 Relative Coe?cient -2.64 0.12 -0.27 0.11 0.62 -0.43 0.17 11.12 129.41 Spread t-value -18.06 ** 3.39 ** -5.26 ** 3.86 ** 1.03 -2.40 ** 5.22 ** 14.59 ** Total Trans. Coe?cient t-value 0.03 1.23 0.17 16.47 ** 0.03 4.62 ** -0.05 -8.78 ** 1.27 10.41 ** 0.22 5.88 ** -0.01 -0.71 -2.50 -17.16 ** 155.36 Trans./Abs. Spread Coe?cient t-value 0.12 4.80 ** 0.03 2.08 ** 0.96 47.87 ** -0.02 -2.37 ** -0.46 -3.82 ** -0.02 -0.56 0.02 2.86 ** 1.62 11.78 ** 819.9

Price Volume Volatility BDSIZE PERCTOUT NEWCEO GOV INDEX Constant F( 7, 3610)

Panel B Absolute Spread Coe?cient t-value -0.03 -2.03 ** -0.05 -7.74 ** 0.01 2.91 ** 0.01 3.09 ** 0.16 2.47 ** -0.05 -2.54 ** 0.01 0.49 -0.19 -2.16 ** 18.94 Relative Coe?cient -0.58 -0.05 0.01 0.01 0.08 -0.09 0.01 1.90 169.58 Spread t-value -27.62 ** -7.19 ** 1.82 1.81 2.47 ** -3.47 ** 1.84 17.45 ** Total Trans. Coe?cient t-value 0.09 3.82 ** 0.42 49.88 ** -0.01 -2.15 ** 0.05 10.55 ** 0.73 7.62 ** 0.02 0.52 -0.04 -7.82 ** -2.80 -22.62 ** 628.53 Trans/Abs. Spread Coe?cient t-value 0.05 3.37 ** 0.02 2.21** 1.06 93.97 ** -0.02 -6.69 ** 0.21 3.49 ** 0.05 2.30 ** -0.05 -14.05 ** 1.59 19.97 ** 4107.66

Price Volume Volatility BDSIZE PERCTOUT NEWCEO GOV INDEX Constant F( 7, 3610)

59

Table VIII

Liquidity and Size of Write-o?s

This table shows the results of a regression of liquidity measures across write-o?s and governance measures, controlling for the e?ects of price, volume, and volatility. Liquidityi = ? + ?1?5 1GOV V ARSi + ?6 V olumei + ?7 P ricei + ?8 V olatilityi + ?9 W O T Ai + ?1 0W O#i
i

where Liquidityi is the dependent variable and stands for either the percent change in absolute spread, relative spread , or total transactions from the non write-o? period to the write-o? period. Absolute spread is a measure of the average absolute dollar bid-ask spread of a sample ?rm. Similarly, relative spread and total transactions are the daily averages for the relative bid-ask spread, and total number of transactions. I also run the estimate for transactions divided by absolute spread. Volume is the total trading volume during the trading day. Price is the average of all transaction prices recoded on the trading day. Volatility is the variance of returns over the trading day. All liquidity variables are calculated by taking percent change in liquidity write-o?s compared to the rest of the NYSE on that day. BDSIZE is the size of the board of directors. PERCTOUT is the percent of outside directors on the board. GOV INDEX is an index that ranks the level of shareholder protection for the shareholders, where the higher the number, the worse the protection. It is calculated using the Gompers, Ishii, and Metrick (2002) index. NEWCEO is a dummy variable that is one if there was a CEO turnover in the days following. WO TA is the size of the write-o? divided by total assets. WO# is the number of write-o?s that the ?rm has taken from 0 to 26. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively and all results are presented based on two-tail signi?cance. Absolute Spread Coe?cient t-value -0.04 -3.04 ** -0.03 -6.37 ** 0.02 4.70 ** 0.01 5.05 ** 0.11 1.82 -0.06 -3.32 ** 0.04 0.98 3.06 5.27 ** -0.01 -4.36 ** -0.11 -1.42 193.28 3839.25 Relative Coe?cient -0.64 -0.05 0.02 0.01 0.04 -0.07 0.03 3.75 -0.02 2.13 26.04 Spread t-value -33.33 ** -6.54 ** 3.71 ** 3.68 ** 0.52 -3.07 ** 0.99 4.99 ** -5.32 ** 21.32 ** Total Trans. Coe?cient t-value 0.01 0.45 0.29 49.71 ** 0.01 1.22 0.04 12.86 ** 0.32 4.70 ** -0.01 -0.28 0.01 -0.84 -0.49 -0.78 0.01 5.12 ** -2.11 -25.44 ** 533.57 Trans./Abs. Spread Coe?cient t-value 0.082 5.93 ** 0.03 4.08 ** 1.02 99.75 ** -0.03 -9.86 ** 0.18 3.10 ** 0.05 3.11 ** -0.04 -11.02 ** -3.56 -6.43 ** 0.00 -0.82 1.614 23.11 **

Price Volume Volatility BDSIZE PERCTOUT NEWCEO GOV INDEX WO TA WO# Constant F( 7, 3610)

60

Table IX

Adverse Selection

This table shows the results of the adverse selection tests to see if write-o?s reduce adverse selection. Lin Sanger and Booth (1995) develop a method of estimating empirical components of the e?ective spread, where the signed e?ective half spread, zt , is de?ned as the transaction price at time t, Pt , minus the spread midpoint, Mt . The signed e?ective half spread is negative for sell orders and positive for buy orders. To re?ect possible adverse selection information revealed by the trade at time t, Lin et. al (1995) add ?, which is the adverse selection component of the bid-ask spread. I follow Lin et al. (1995) by using a robust OLS to estimate the following equation: ?Mt+1 = ?zt + ?W O (zt ? W Ot ) +
t+1 ,

where ?W O is the incremental adverse selection component during the write-o? period. The results are shown in Model 1. I also decompose the spread using the empirical model of Huang and Stoll (1997), and implemented by Weston (2000). The midpoint of the spread is de?ned as, Mt , and changes as a function of the direction of trades. An indication variable Qt takes on the values, {-1,0,1} based on the direction of trade. If Pt¡ Mt , then Qt = -1 (sell order), if Pt = Mt , then Qt =0, and if Pt . Mt , then Qt =1 (buy order). I follow Huang and Stoll (1997) by using a robust OLS to estimate the following equation: ?Mt = ?(St?1 /2)Qt?1 + ?W O (St?1 /2)Qt?1 ? W Ot ) +
t+1 ,

where ? measures the proportion of the half spread St?1 /2, that stems from information costs and ?W O is the incremental adverse selection component during the write-o? period. Signi?cance is indicated at the 0.05 and 0.01 levels by one and two asterisks respectively and all results are presented based on two-tail signi?cance. ?Mt+1 - Model 1 Coe?cient t-value 0.146 23.55 -0.302 -23.34 ?Mt - Model 2 Coe?cient t-value

Z Z*WO (St?1 /2)Qt?1 (St?1 /2)Qt?1 *WO R-squared F( 2, 7623)

0.114 277.3

1.43 -1.45 0.009 5.35

5.01 -4.61

61

Table X

Earnings and Write-o?s

This table examines the earnings in the periods surrounding the write-o? announcement. Panel A shows the earnings surprises, and Panel B shows the regression of the absolute value of forecast error on ?rm speci?c variables, where forecast error is de?ned as the absolute value of the median analyst estimate for the ?rst earnings quarter following the write-o?. Earnings data is from the First Call database. A * denotes signi?cance at the 5 percent level, and ** denotes signi?cance at the 10 percent level. Panel A: Earnings Surprises Earnings Surprise Mean Median t-value Sign-rank test Weak Monitors -0.07 -0.02 -1.03 -1.30 Mediocre Monitors -0.02 0.00 -1.97* -2.26 ** Strong Monitors -0.03 -0.01 -2.54** -2.35 ** All Write-o?s -0.03 -0.01 -2.89** -3.48 ** Panel B - OLSQ Results for Forecast Error Coe?cient t-value WO -0.018 -3.40 ** SIZE 0.033 6.79 ** BDSIZE 0.026 5.16 ** PERCTOUT -0.015 -5.35 ** -0.001 -1.87 GOV INDEX ACCRUALS -0.001 -1.57 ROA 0.001 3.52 ** DEBT RATIO 0.001 0.87 NEWCEO -0.002 -0.51 GROWTH 0.004 0.92 CONSTANT 0.042 4.22 **

62

Tax Laws and Write-o?s Restructuring charges have become a popular topic with FASB. These charges are based on the big bath practice where ?rms take one-time charges to clean up their balance sheet. These charges include employee bene?ts, costs associated with discontinued operations, closed plants, product line elimination, and losses incurred from asset disposal and impairment. Many of these charges are not currently addressed by accounting standards, although some such as severance are. From a tax viewpoint, the main issue involved with these write-o?s is the uncertainty of when these costs should be incurred. Employee termination bene?ts or severance, are covered by accounting standards. EITF 94-3 applies to bene?ts to be provided to employees a?ected by layo?s. This liability is recognized in the period management approves the layo?s if the following criteria hold: • Prior to the date of the ?nancial statements, management approves of the termination bene?ts and speci?es the amount to be paid out. • Prior to the date of the ?nancial statements, the details of the layo?s is communicated to the employees in su?cient detail. • The termination plan speci?es the number of layo?s, the job classi?cation of the layo?s, and the speci?c departments. • Changes in the plan are unlikely to occur. Termination bene?ts that fall under the following criteria are not allowable as a write-o?: • Included with a disposal of a segment, which is also charged against earnings. • Paid pursuant to the terms of an ongoing employee bene?t plan. • Paid under the terms of an individual deferred compensation plan. Costs to exit an activity also have stated guidelines. When management commits to exiting an activity, exit costs are incurred. Exit costs include: • Costs that are a direct result of the exit plan and that the ?rm would not incur without the plan. • Costs that existed through a contractual obligation prior to the exit plan, like the penalty to break a lease. The discontinued operations segment of an income statement consists of two parts. The ?rst part is income (loss) from operations, and the second is gain (loss) on disposal of assets. Income (loss) from operations is disclosed for the current year only if the decision to discontinue operations is made after the beginning of the ?scal year. Gain (loss) from sale of assets is a combination of income (loss) from operations during the phase-out period and the gain (loss) from disposal of a segment. The gain (loss) on disposal includes costs arising from the decision, such as severance, additional pension costs, employee relocation expenses, and future rentals or leases. If a loss is expected from the proposed sale or abandonment of a segment, the ?rm should provide for the loss it makes at the time the decision to dispose. If there is a gain, it should be recognized when the gain occurs. The results of discontinued operations appear as an independent line item on the income statement, before extraordinary items. In addition to being disclosed in the ?nancial statements, the footnotes should disclose the following: • The segment of business that has been a?ected. • The expected disposal date. • A description of the remaining assets and liabilities of the segment. • The expected manner of disposal. • The income or loss from operations and any proceeds from the disposal of the segment. Extraordinary items are sometimes used to disclose write-o?s. An item is considered extraordinary if it is both unusual in nature and infrequent in its occurrence. An item is considered unusual in nature if it is unrelated to the line of business. To identify such items one needs to consider:

63

• Type and scope of operations. • Line of business. • Operating Policies. • Industry. • Geographic locale. • Nature and extent of government regulations. Accounting standards speci?cally note some items that can or cannot be included in extraordinary items. The following items are considered extraordinary: • Gains or losses from extinguishments of debt, except for sinking fund requirements. • Pro?ts or losses resulting from the disposal of a signi?cant part of the assets of previously separate companies. • Write-o? of operating rights of motor carriers. • The investors share of an investees extraordinary item when the investor uses the equity method of accounting. • Gains of a debtor due to a troubled debt restructuring. Items that cannot be considered extraordinary are: • Write-down or write-o? of receivables, inventory, equipment leased, or intangible assets. • Foreign currency gains or losses. • Gains or losses from the disposal of a business segment. • Gains or losses from sale or abandonment of property, plant, or equipment. • E?ects of a strike. • Adjustments on accruals on long-term contracts. Extraordinary items are shown independently of ongoing operations and are shown net of taxes in a separate section of the income statement. In 2002, FASB created new guidelines in its Statement 144 for restructuring charges. Unless the disposal activity involves a discontinued operation, costs associated with a disposal activity should be reported in continuing operations before income taxes in the income statement. If the disposal activity does involve a discontinued operation, then those costs should be included in discontinued operations. The total amount of costs incurred and charged to expense should be reported by reportable segment. The accounting model for long-lived assets to be disposed of by sale applies to all long-lived assets, including discontinued operations. Statement 144 requires that those long-lived assets be measured at the lower of either carrying amount or fair value less cost to sell, whether reported in continuing operations or in discontinued operations. Therefore, discontinued operations can no longer be measured at net realizable value or include amounts for operating losses that have not yet occurred. Statement 144 also broadens the reporting of discontinued operations to include all components of an entity with operations that can be distinguished from the rest of the entity and that will be eliminated from the ongoing operations of the entity in a disposal transaction.

64

BIBLIOGRAPHY
[1] Abowd, J., 1990, Does performance-based managerial compensation a?ect corporate performance?, Industrial and Labor Relations Review, 43, 52-73. [2] Aggarwal R. and A. Samwick, 1999, Executive compensation evaluation: Theory and evidence, Journal of Finance, 54, 1999-2043. [3] Alciatore, M., P. Easton, and N. Spear, 2000, Accounting for the impairment of long-lived assets: Evidence from the petroleum industry, Journal of Accounting and Economics, 29, 151-172. [4] Amihud, Y. and H. Mendelson, 1986, Asset Pricing and the bid-ask spread, Journal of Financial Economics 17, 223-49. [5] Audt, D., W. Cready, and T. Lopez, 2003, Restructuring charges and CEO cash compensation: A reexamination, Accounting Review, 78(1), 169-192. [6] Bacidore, J., R. Battalio, and R. Jennings, 2002. Depth improvement and adjusted price improvement on the New York Stock Exchange, The Journal of Financial Mar kets , 5, 169–195. [7] Barclay, M. and C. Smith, 1988, Corporate payout policy: Cash dividends versus open market repurchases, Journal of Financial Economics 22, 61-82. [8] Baker, G. and B. Hall, 1998, CEO incentives and ?rm size, Working paper-6868, NBER. [9] Bartelsman, E. and W. Gray, 2002, The NBER Manufacturing Productivity Database, NBER Working Paper. [10] Bartov, E., F. Lindahl, and W. Ricks, 1998, Stock price behavior around announcements of write-o?s, Review of Accounting Studies, 3/4, 327-346. [11] Benston , G. and R. Hagerman, 1974, Determinations of bid-ask spreads in the over-the-counter market, Journal of Financial Economics 1, 353-64. [12] Berger, P., E. Ofek and D. Yermack, 1997, Managerial entrenchment and capital structure Decisions, Journal of Finance, 52(4), 1411-1446. [13] Bhagat, S. and B. Black, 1999, The uncertain relationship between board composition and ?rm performance, Business Lawyer, 55, 921-963. [14] Borokovich, K., R. Parrino, and T. Trapani, 1996, Outside directors and CEO selection, Journal of Financial and Quantitative Analysis, 31, 337-355. [15] Bradshaw, M and R. Sloan, 2002, GAAP versus the street: An empirical assessment of two alternative de?nitions of earnings, Journal of Accounting Research, 4 0, 4 1-65. [16] Brickley, J. and L. Van Drunen, 1990, Internal corporate restructuring: An empirical analysis, Journal of Accounting and Economics, 12, 251-280. [17] Brockman, P., Chung, D.Y., 1999, An analysis of depth behavior in an electronic, order-driven environment, Journal of Banking and Finance 23, 1861-1886. [18] Brockman, P., Chung, D.Y., 2001, Managerial Timing and Corporate Liquidity: Evidence from actual share repurchases, Journal of Financial Economics, 61, 417-4 [19] Brown, L. and K. Sivakumar, 2001, Comparing the quality of three earnings measures, Working Paper, Georgia State University. [20] Brown, S. and J. Warner, 1985, Using daily stock returns: The case of event studies, Journal of Financial Economics, 14, 3-31. [21] Byrd, J. and K. Hickman, 1992, Do outside directors monitor managers? Evidence from tender o?er bids, Journal of Financial Economics, 32, 95-221. [22] Co?ee, J., 1999, The future as history: The prospects for global convergence in corporate governance and its implications, Northwestern University Law Revie w, 9 3, 641-708. [23] Copeland, T.E., Galai, D., 1983, Information e?ects on the bid-ask spread,Journal of Finance 38, 1457-1470. [24] Core, J. and W. Guay, 1999a, Estimating the value of stock option portfolios and their sensitivities to price and volatility, Working Paper, University of Pen nsyl vania. [25] Core, J. and W. Guay, 1999b, The use of equity grants to manage optimal equity incentive levels, Journal of Accounting and Economics, 28, 151-184.

65

[26] Core, J., R. Holthausen, and D. Larcker, 1999, Corporate governance, chief executive o?cer compensation, and ?rm performance, Journal of Financial Economi cs, 51, 371-406. [27] Coughlan, A. and R. Schmidt, 1985, Executive compensation, management turnover, and ?rm performance: An empirical investigation, Journal of Accounting and E cono mics 7(1-3), 43-66. [28] Demsetz, H., 1968, The cost of transacting,Quarterly Journal of Economics 82, 33-53. [29] Doyle J. and M. Soliman, 2002, Do managers use pro forma earnings to exceed analysts forecasts?, Working Paper, University of Michigan. [30] Doyle, J., R., Lundholm, and M. Soliman, 2003, The predictive value of expense excluded from pro forma earnings, Review of Accounting Studies, 8, 145-74. [31] Easley, D. and M. OHara (1987): Price, Trade Size and Information in Securities Markets, Journal of Financial Economics, 19, pp 69-90. [32] Easley, D., O’Hara, M., 1987, Price, trade size, and information in securities markets, Journal of Financial Economics 19, 69-90. [33] Elliot, J., and J. Hanna, 1996, Repeated accounting write-O?s and the information content of earnings, Journal of Accounting Research, 34, 135-156. [34] Elliot, J., and W. Shaw, 1988, Write-o?s as accounting procedures to manage perceptions, Journal of Accounting Research, 26, 91-119. [35] Francis, J., D. Hanna, and L. Vincent, 1997, Causes and e?ects of discretionary asset write-o?s, Journal of Accounting Research, 34, 117-134. [36] Fried, D., M. Schi?, and A. Sondhi, 1989, Impairments and write-o?s of long-lived assets, Management Accounting, 71, 48-50. [37] Garman, M., 1976, Market Micro-Structure, Journal of Financial Economics, pp. 257-275. [38] Gibbons, J., Nonparametric Methods for Quantitative Analysis. 1976, Columbus, OH: American Sciences Press. [39] Glosten, L., Milgrom, P., 1985, Bid, ask, and transaction prices in a specialist market with heterogeneously informed traders, Journal of Financial Econom ics 14, 71-100. [40] Gibson, M., 1999, Is corporate governance e?ective in emerging markets?, Working Paper, Federal Reserve Board . [41] Gompers, P, J. Ishii and A. Metrick, 2003, Corporate governance and equity prices, Quarterly Journal of Economics, forthcoming. [42] Hallock, K., 1997, Reciprocally interlocking boards of directors and executive compensation, Journal of Financial and Quantitative Analysis, 32, 331-343. [43] He?in, F. and K.W. Shaw, 2001, Adverse selection, inventory holding costs, and depth, Journal of Financial Research 24. [44] He?in, F., and T. War?eld, 1995, Earnings, ?rm value, and unusual charges of impaired assets, Working paper, University of Wisconsin-Madison. [45] Hermalin, B. and M. Weisbach, 1988, The determinants of board composition, The RAND Journal of Economics, 19, 589-606. [46] Hermalin, B. and M. Weisbach, 1991, The e?ects of board composition and direct incentives on ?rm performance, Financial Management, 20, 101-112. [47] Ho, T., Stoll, H.R., 1981, Optimal Dealer Pricing Under Transaction and Return Uncertainty, Journal of Financial Economics, pp. 47-73. [48] Ho, T., Stoll, H.R., 1983, The dynamics of dealer markets under competition, Journal of Finance38, 1053-1074. [49] Huang, R.D., Stoll, H.R., 1997, The components of the bid-ask spread: a general approach, Review of Financial Studies 10, 995-1034. [50] Hvidkjaer, S., 2003, A trade-based analysis of momentum, Working paper, University of Maryland. [51] Jacoby, G., Fowler, D., and A. Gottesman, 2000, The capital asset pricing model and the liquidity e?ect: a theoretical approach, Journal of Financial mar kets 3, 69-81.

66

[52] Jain, P. K., and Z. Rezaee. 2004. The Sarbanes Oxlet Act of 2002 and Market Liquidity. Working Paper, University of Memphis. [53] Jensen, M. and K. Murphy, 1990, Performance pay and top-management incentives, Journal of Political Economy, 98, 225-64. [54] Jensen, M. and R. Ruback, 1983, The market for corporate control, Journal of Financial Economics, 11, 5-50. [55] Johnson, W., and W. Schwartz, 2001, Are investors misled by pro forma earnings, Working Paper, University of Iowa. [56] Kaplan, S., 1994, Top executives, turnover, and ?rm performance in Germany, Journal of Law, Economics & Organization, 10, 142-159. [57] Kim, S. and S. Kwon, 2001, SFAS 121 asset write-down: Early versus late ?rms, Working paper, Rutgers University - Camden. [58] Kinney, M. and R. Trezevant, 1997, The use of special items to manage earnings and perceptions, Journal of Financial Statement Analysis, 3, 45-53. [59] Kross, W., T. Park, and B. Ro, 1996, The impact of operational restructuring announcements on stock price, risk, and trading volume, Working Paper, Purdue Uni vers ity . [60] Leonard, J., 1990, Executive pay and ?rm performance, Industrial and Labor Relations Review 43, 13S-29S. [61] Lin, J., 1993, Order Persistence, Adverse Selection, and Gross Pro?ts Earned by NYSE Specialists, Journal of Finance Proceedings (Abstract)(July), 1108- 9 [62] Lin, J.-C., Sanger, G., Booth, G.G., 1995. Trade size and components of the bid-ask spread, Review of Financial Studies 8, 1153-1183. [63] Lindbeck, R., Z. Rezaee, and J. Smith, 1996, An examination of long lived asset impairments under SFAS No. 121, International Advances in Economic Research, 2, 86-91. [64] Macey, J., 1997, Institutional investors and corporate monitoring: A demand-side perspective, Managerial and Decision Economics, 18, 601-610. [65] May, D., 1995, Do managerial motives in?uence ?rm risk reduction strategies?, The Journal of Finance, 50 , 1291-1308. [66] Mayers, D., A. Shivdasani, and C. Smith, Jr., 1994, Board composition and corporate control: Evidence from the insurance industry, Working Paper, Ohio State Un iver sity . [67] Meyer, J. and J. Strong, 1987, Asset write-downs: Managerial incentives and security returns, The Journal of Finance, 42, 643-661. [68] Minnick, K, 2004, Write-o?s and Corporate Governance, Working paper - University of Maryland. [69] Murphy, K., 1985, Corporate performance and managerial remuneration: An empirical analysis, Journal of Accounting and Economic, 7, 11-42. [70] Murphy, K., 1986, Incentives, learning, and compensation: A theoretical and empirical investigation of managerial labor contracts, Rand Journal of Economics, 17, 59- 76. [71] Murphy, K., 1999, Executive Compensation, in Orley Ashenfelter and David Card, Handbook of Labor Economics, Vol. 3, North Holland, Amsterdam. [72] Myers, S., and N. Majluf, 1984, Corporate ?nancing and investment decisions when ?rms have information that investors do not have, Journal of Financial Econo mics , 13, 187–221. [73] Newey, W. and K. West, 1987, A simple positive semi-de?nite, heteroscedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703-708. [74] OHara, M., 2003. Presidential address: Liquidity and price discovery, Journal of Finance, 59, 1335– 1354. [75] Parrino, Robert, 1997, CEO turnover and outsider succession: A cross-sectional analysis, Journal of Financial Economics 46, 165-197. [76] Paul, D., 2003, Board composition and corrective action: Evidence from corporate responses to bad acquisition bids, Working Paper, Babson College. [77] Phillips, G. and V. Maksimovic, 2002, Do conglomerate ?rms allocate resources ine?ciently across industries? Theory and evidence, Journal of Finance, 57 , 72 1-767.

67

[78] Riedl, E., 2002, An examination of long-lived asset impairments, Working Paper, Pennsylvania State University. [79] Shleifer, A. and R. Vishny, 1989, Management entrenchment: The case of manager-speci?c investment, Journal of Financial Economics, 25, 123-140. [80] Shleifer, A. and R. Vishny, 1997, A survey of corporate governance, Journal of Finance, 52, 737-783. [81] Stoll, H., 1989, Inferring the components of the bid-ask spread: Theory and empirical evidence, Journal of Finance 44, 115-134. [82] Stoll, H.R., 1978, The pricing of security dealer services: an empirical study of NASDAQ stocks, Journal of Finance 33, 1153-1172. [83] Symonds, B., The Long, Hard Road Ahead for Tyco, Business Week, July 2002. [84] Tinic, S., 1972, The economics of liquidity services, Quarterly Journal of Economics 86, 79-93. [85] Tinic, S., West, R., 1974, Marketability of common stocks in Canada and the U.S.A.: a comparison of agent versus dealer dominated markets, Journal of Fina nce 29, 729-746. [86] Van Ness, R. Van Ness, and R. Warr, 2001, ”How well do adverse selection components measure adverse selection?” Financial Management, 30, 3, 77-98. [87] Volpin, P., 2002, Governance with poor investment protection: Evidence from top executive turnover in Italy, Journal of Financial Economics, 64, 61-91. [88] Weisbach, M., 1988, Outside directors and CEO turnover, Journal of Financial Economics, 20, 431460. [89] Weston, J.P., 2000, Competition on the Nasdaq and the impact of recent market reforms, Journal of Finance 55, 2565-2598. [90] Yermack, D., 1996, Higher valuation of companies with a small board of directors, Journal of Financial Economics, 40, 185-212. [91] Yermack, D., 1997, Good timing: CEO stock option awards and company news announcements, Journal of Finance, 52, 449-476. [92] Zucca, L. and D. Campbell, 1992, A closer look at discretionary write-downs of impaired assets, Accounting Horizons, 6, 30-41.

68



doc_304326527.pdf