Is sin always a sin? The interaction effect of social norms and financial incentives on ma

Description
Using alcohol, tobacco, and gaming consumption data and people’s attitudes toward these
sin products to proxy for social norm acceptance levels, we show a strong interaction effect
between social norms and financial incentives, which significantly influence the behavior
of market participants. Specifically, institutional investors’ shareholdings and analyst coverage
of sin companies increase with the degree of social norm acceptance. The association
between shareholdings/coverage and social norm acceptance is less pronounced for firms
with higher future expected performance. Our results show that social norms and financial
incentives have a powerful interaction effect in determining the behavior of market participants,
suggesting that social norms can be crossed when motive and opportunity exist.

Is sin always a sin? The interaction effect of social norms and
?nancial incentives on market participants’ behavior
Yanju Liu
a
, Hai Lu
b,?
, Kevin Veenstra
c
a
Singapore Management University, Singapore
b
University of Toronto, Canada
c
McMaster University, Canada
a b s t r a c t
Using alcohol, tobacco, and gaming consumption data and people’s attitudes toward these
sin products to proxy for social norm acceptance levels, we show a strong interaction effect
between social norms and ?nancial incentives, which signi?cantly in?uence the behavior
of market participants. Speci?cally, institutional investors’ shareholdings and analyst cov-
erage of sin companies increase with the degree of social norm acceptance. The association
between shareholdings/coverage and social norm acceptance is less pronounced for ?rms
with higher future expected performance. Our results show that social norms and ?nancial
incentives have a powerful interaction effect in determining the behavior of market partic-
ipants, suggesting that social norms can be crossed when motive and opportunity exist.
Ó 2014 Elsevier Ltd. All rights reserved.
Introduction
Debate over the trade-off between private and social
value of economic activities has continued for at least the
past century. This debate reached a new climax in the past
decade when a number of corporate scandals broke out
due to the unethical behavior of economic agents. As a
result, people began to re-think the ultimate objectives of
agents and regulations (e.g., Bayou, Reinstein, & Williams,
2011; Cassidy, 2009; Pigou, 2005). Extant literature shows
that economic agents maximize short-term pro?ts, but
also suggests that socially responsible investors care about
non-?nancial incentives. Since the ?nancial crisis of 2007/
2008, the public has increasingly voiced their strong
interest in knowing whether the social value of economic
activities are sacri?ced due to the incentives of market
participants chasing ?nancial rewards.
Experimental studies in the accounting literature have
explored the interaction effect of ?nancial incentives with
professional ethics (Babcock, Loewenstein, Issacharoff, &
Camerer, 1995; Blanthorne & Kaplan, 2008; Hunton &
Rose, 2008; Thompson & Loewenstein, 1992). However,
to date few empirical studies directly address how ?nan-
cial incentives interact with social norms (de?ned as a col-
lection of ethical standards) in determining the behavior of
other economic agents, including investment decisions of
institutional investors and coverage decisions of ?nancial
analysts. While the accounting profession has introduced
signi?cant guidance (or professional norms) on ethical
behavior, we still see accountants/auditors/executives
with economic incentives violating these norms and
making self-serving decisions. Such evidence is not rare.
For example, Hechter (2008), through an analysis of the
rise and fall of the Arthur Andersen accounting ?rm, shows
that what appeared to have been an unshakable commit-
ment to professional norms was highly contingent uponhttp://dx.doi.org/10.1016/j.aos.2014.04.001
0361-3682/Ó 2014 Elsevier Ltd. All rights reserved.
?
Corresponding author. Address: Rotman School of Management,
University of Toronto, 105 St. George St, Toronto, ON M5S 3E6, Canada.
Tel.: +1 416 946 0677; fax: +1 416 971 3048.
E-mail address: [email protected] (H. Lu).
Accounting, Organizations and Society 39 (2014) 289–307
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
competing norms, macroeconomic conditions, and internal
con?icts. Blanthorne and Kaplan (2008) ?nd that taxpay-
ers’ ethical beliefs do not arise independently from eco-
nomic considerations, but instead are strongly in?uenced
by opportunities to evade income taxes, i.e., the opportu-
nity to evade in?uences the formation of one’s ethical
beliefs, which, in turn, affects one’s intentions and deci-
sions to evade. Other papers, such as Thompson and
Loewenstein (1992) and Babcock et al. (1995), have also
found an in?uence of self-interest on one’s ethical beliefs.
In a recent experiment, Hunton and Rose (2008) ?nd that
directors with multiple directorships are less willing to
support required ?nancial statements restatements due
to the potential adverse effects of restatements on their
reputation capital (namely, future income generating
ability).
Motivated by the interests of both academics and regu-
lators, in this paper we examine whether social norms and
?nancial incentives have a substitution effect in in?uenc-
ing the behavior of market participants regarding those
?rms in the alcohol, tobacco and gaming industries. These
?rms are called sin ?rms because there are social norms
attached to their products, i.e., excessive consumption of
alcohol, tobacco, and gambling generates negative exter-
nalities to society. Speci?cally, we examine the following
market participants and their behavior: (1) institutions
and their stock ownership decisions, and (2) ?nancial ana-
lysts and their coverage decisions. Hong and Kacperczyk
(2009) suggest social norms are priced and show that sin
stocks in the alcohol, tobacco and gaming industries, on
average, have less institutional ownership and analyst cov-
erage compared to other non-sin stocks. We assert that
social norms may not be adhered to when ?nancial
rewards are too enticing for some market participants. In
other words, we explore whether ‘‘money talks’’ when
market participants have the dilemma of choosing
between social norms and ?nancial rewards.
To answer our research questions, we ?rst investigate
how social norms evolve over time. We use the consump-
tion of alcohol, tobacco, and gaming as the proxies for the
acceptance levels of the sins related to these products.
Using Gallup survey data, we ?nd these proxies are highly
correlated with people’s attitude toward these sin prod-
ucts. We also ?nd that the consumption of the sin products
changes substantially over time and follow different evolu-
tion processes. Social norm acceptance levels for alcohol
and tobacco have become lower while acceptance levels
for gaming have increased. The dynamic nature of these
proxies enables us to study the main effect of social norms
and the interaction effect of social norms with the
expected ?nancial performance of sin ?rms on the invest-
ment decisions of institutional investors and the coverage
decisions of ?nancial analysts.
We then conduct both univariate and multivariate
regression analyses. We show that institutional ownership
and analyst coverage of sin stocks are positively associated
with the strength of social norm acceptance, i.e., sin stocks
are shunned more when social norm acceptance is low
(consumption of sin products is low). Such calibration
extends the ?nding in Hong and Kacperczyk (2009) that
sin stocks, compared to the wide universe of stocks, are
held less by institutions and followed less by ?nancial ana-
lysts. Furthermore, our results suggest a strong interaction
effect between social norms and ?nancial incentives of
market participants. When a stock is expected to perform
poorly, the price of obeying social norms is relatively
cheap, leading to additional shunning by institutions and
analysts. In contrast, when the stock is expected to perform
well and the relative price of obeying social norms
becomes high, institutional investors and analysts are
shown to choose ?nancial rewards over social norms by
shunning fewer sin ?rms. Our results survive a number
of additional robustness checks such as using time-varying
sin exposure, alternative measures of expected ?nancial
performance, alternative proxies for social norms, and
change analysis.
Our study contributes to both the economics and
accounting literature. First, we provide strong empirical
support for a substitution effect between ?nancial and
non-?nancial incentives among economic agents. We show
that when social norms interact with ?nancial consider-
ations, market participants will sacri?ce their adherence
to social norms for ?nancial rewards. While such a ?nding
sounds intuitive, large sample empirical evidence is
scarce in the literature outside of numerous experimental
studies dealing mostly with accounting professionals,
audit committee members, auditors, and taxpayers. This
?nding adds to the current debate on why there could be
a gap between the investment practices of Wall Street
and the ethical standards of Main Street. Some ethical
standards consistent with corporate social responsibility
considerations are not necessarily pro?table to follow. As
such, a gap between what is right to do (following ethical
standards) versus what is pro?table to do (investment
practices) can form. Our ?nding is thus of particular
interest to academics, investors, regulators, and various
other stakeholders.
Second, we add a level of richness to the previous liter-
ature by investigating how social norms evolve over time
through the adoption of social norm proxies. Our research
design and results highlight the importance of having
direct empirical measures of social norms. Our use of
changes in consumption of sin products as a proxy for
the evolution of social norms towards sin stocks over-
comes the drawback of assuming a constant social norms
level over time; and thus extends previous studies by
showing how social norms are priced in a dynamic setting.
Therefore, our results potentially shed light on why social
norm effects are found to be weak, in aggregate, for certain
periods in the previous literature (Hong & Kacperczyk,
2009).
Finally, our ?ndings have signi?cant implications and
importance in practice, especially in view of the aftermath
of the recent ?nancial crisis in which numerous individuals
and companies have been accused of sacri?cing social
standards for ?nancial gains. Extending our hypotheses
to a more general context, one may predict that when
existing social norms are interacted with a strong counter-
acting force, i.e. ?nancial considerations, a real risk exists
that these ‘‘compromises’’ become part of acceptable
future social norms (Prentice & Miller, 1996). A ‘‘compro-
mised’’ set of social norms may result in regulations and
290 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
laws needing to be enforced to maintain social order. Such
was the case with the failure of Arthur Andersen, and the
subsequent passing of the Sarbanes–Oxley Act which was
extremely expensive to both implement and enforce
(Hechter, 2008).
The remainder of the paper is organized as follows. Sec-
tion ‘Background and hypotheses’ discusses the back-
ground and develops the hypotheses. Section ‘Sample
selection and variables’ describes the sample selection
and methodology. Section ‘Results’ discusses the empirical
results. Section ‘Conclusion’ concludes.
Background and hypotheses
Social norms are rules and standards understood by
members of a group that guide and constrain social behav-
ior. More precisely, there are two types of social norms:
descriptive norms and injunctive norms. Descriptive
norms refer to the prevalence of a given behavior (i.e. the
number of people in a given population that smoke ciga-
rettes) whereas injunctive norms refer to the degree of
actual or perceived approval of a given behavior
(Neighbors et al., 2007). Descriptive and injunctive norms
should be highly correlated because a popular behavior
implies that many people approve of the behavior. These
social norms develop as a result of interaction with others.
Sanctions for deviating from them come from social net-
works as opposed to the legal system (Cialdini & Trost,
1998). In the economics literature, the impact of social
norms on economic behavior and market outcomes was
?rst studied in the context of the labor market. In the dis-
crimination model of Becker (1957), agents pay for the dis-
cretionary tastes arising from community norms. They
bear ?nancial costs from their decisions to not enter into
contracts with particular types of people. In an unemploy-
ment setting, Akerlof (1980) examined social norms and
claimed that although social norms can be costly, they con-
tinue to exist because of the perceived loss of reputation to
followers for diverting from these norms.
Applied to socially responsible investing (SRI), it is gen-
erally understood that SRI encourages investors to avoid
sin companies such as those companies involved in the
production of alcohol, tobacco, and gaming. It is believed
that investors with a socially responsible investing philos-
ophy can somehow affect the practices of the ?rms in
which they invest and thus improve ‘‘the ef?ciency of the
economic system (in the broad sense of satisfaction of indi-
vidual values)’’ (Elster, 1989). As a result, shares of sin
stocks should be held in smaller proportions by institu-
tions subject to social norm pressures or with socially
responsible investing objectives (Geczy, Stambaugh, &
Levin, 2006). Moreover, since sell-side analysts who pro-
duce ?nancial reports and analyses tend to cater to institu-
tional investors, sin stocks should also be followed less by
these analysts (Hong & Kacperczyk, 2009).
Previous literature, through its use of a dummy variable
to proxy for ‘‘sin’’, implicitly assumes that strong social
norms against sin stocks are constant over time and across
different sin types. However, as documented by the
sociology and economics literature, through a process of
adaptation, social norms are not constant. Rather, they
evolve over time in several different dimensions, including
age groups, social classes, and social groups (Azar, 2004;
Kolstad, 2007; Ostrom, 2000). To overcome this drawback,
our study uses consumption data for sin products or ser-
vices to proxy for the level of social norms in both cross-
section and across time. The use of consumption data to
proxy for social norms is consistent with the de?nition of
descriptive norms and the proxies are highly correlated
with injunctive norms, as we show later. In addition to
raw consumption being the de?nition of a descriptive
norm, previous literature has shown, for example, that
injunctive social norms are strong predictors of gambling
behavior (i.e. descriptive norms) and gambling-related
negative consequences (Larimer & Neighbors, 2003;
Moore & Ohtsuka, 1997; Moore & Ohtsuka, 1999; Takushi
et al., 2004). Our choice of consumption data relies on
the assertion that the undesirable social consequences of
alcohol, tobacco, and gaming, when consumed excessively,
re?ect the consensus social norms against consuming
these products. Indeed, we ?nd that whereas social norms
against gaming have moderated signi?cantly over the per-
iod from 1980 to 2007, the opposite can be said about alco-
hol and tobacco usage. Such new direct proxies allow us to
conduct our study within each sin industry, avoiding the
dif?culty in choosing the appropriate control group as in
the previous literature.
Consistent with our support for the evolution of social
norms, we argue that institutions and analysts have a
lower demand for sin companies when social norms
against the ‘‘sin’’ are stronger. Hong and Kacperczyk
(2009) provide evidence that, on average, institutional
ownership and analyst coverage are lower for sin compa-
nies than for non-sin companies in their full sample period.
However, in several sub-periods, such a difference is not
signi?cant.
1
Our argument potentially sheds light on the
reason: at the time when social norms are weak in aggre-
gate, it is possible that sin is not priced. We thus have the
following hypotheses:
Hypothesis 1A. Institutional ownership of sin stocks is
associated positively with the strength of social norm
acceptance.
Hypothesis 1B. Analyst coverage of sin stocks is associ-
ated positively with the strength of social norm
acceptance.
Previous studies on sin stocks implicitly assume that
the effect of social norms is unconditional, i.e., it is inde-
pendent of the ?rm’s ?nancial performance. Extant studies
document that institutional ownership and analyst cover-
age have a strong positive correlation with ?nancial perfor-
mance (McNichols & O’Brien, 1997; O’Brien and Bhushan,
1990; Sias, Starks, & Titman, 2006). Speci?cally, institu-
tional investors invest when expected stock returns are
high (e.g., Cai & Zheng, 2004; Grif?n, Harris, & Topaloglu,
2003) and analysts are more likely to provide forecasts
and recommendations for stocks about which their true
1
For example, see Panel D of Table 3 in Hong and Kacperczyk (2009).
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 291
expectations are favorable (McNichols & O’Brien, 1997).
Akerlof (1980) develops a model expressing utility as a
function of consumption, reputation, obedience/disobedi-
ence of a community’s code of behavior, and belief/disbe-
lief in a community’s code of behavior. He ?nds that a
custom that is too costly to follow, in terms of lost utility,
will not be followed; while a custom that is fairly costless
to follow will, once established, continue to be followed
because persons lose utility directly by disobeying the
underlying social code and also because disobedience of
social custom results in loss of reputation. Applied to sin
stocks, social norms against investing in these stocks are
relatively easy to follow when the ?nancial performance
of alternative investments is strong. However, when the
?nancial performance of all other stocks is falling relative
to stocks that promote vice, as was the case in the early
2000s, this custom of not holding sin stocks becomes
expensive to obey and will not always be followed.
As Bayou et al. (2011) discuss at length, one of the crit-
ical roles of accounting is to provide corporations with the
ability to assume a moral identity. To the extent that stud-
ies (already described) show that accounting professionals
can succumb to ?nancial incentives, it is not surprising
that the bene?ts of complying with social norms are some-
times insuf?cient to overcome the costs of foregone high
expected future returns in sin stocks. The assertion that
?nancial incentives can have a strong interaction effect
with social norms for other ?nancial market participants
such as institutional investors and ?nancial analysts,
therefore, seems to be a reasonable conjecture.
These arguments lead to the following hypotheses,
where we include ?nancial incentives and hypothesize
that the impact of sin varies with the relative expected
?nancial performance of sin stocks.
Hypothesis 2A. When expected ?nancial performance of
sin companies is strong, institutional ownership of sin
stocks is less likely to be associated with the level of social
norms.
Hypothesis 2B. When expected ?nancial performance of
sin companies is strong, analyst coverage of sin stocks is
less likely to be associated with the level of social norms.
Sample selection and variables
Sample selection
Following Hong and Kacperczyk (2009), we identify a
list of sin ?rms that are publicly traded and are involved
in the alcohol, tobacco, and gaming industries.
2
We start
with the Fama and French (1997) classi?cation of stocks
based upon their SIC codes into 48 industries. Stocks in the
Fama–French industry group 4 (beer and alcohol – SIC codes
2100–2199) and industry group 5 (smoke or tobacco – SIC
codes 2080–2085) are classi?ed as sin stocks. Firms with
NAICS codes: 7132, 71312, 713210, 71329, 713290, 72112
and 721120 are identi?ed as gaming stocks. In addition,
we use Compustat Segments data to augment our sample
by including ?rms that have segments operating in any of
these SIC or NAICS groups. A company is identi?ed as a sin
stock if any of its segments has an SIC code in either the beer
or the smoke group or an NAICS code in the gaming group.
Since Compustat segments data is only available after
1985, our augmented search is limited to stocks still in exis-
tence as of 1985. For these stocks, we back-?ll our sample,
i.e., those ?rms with any segments operating in sin indus-
tries are characterized as sinful for the years before 1985
as well.
We obtain daily closing stock prices, daily shares out-
standing and all other return-related data from CRSP.
Annual information on a variety of accounting variables
is obtained from Compustat. We restrict our study to com-
panies with CRSP share codes of 10 and 11. Institutional
ownership data is from Thomson Reuters’ database of
13-F ?lings, including institutions that manage at least
$100 million in assets. Analyst coverage data is obtained
from IBES. Since institutional ownership and analyst cover-
age data only became available after 1980 and 1975
respectively, we restrict our analysis to the period com-
mencing from 1980. GDP and unemployment data are
obtained from the websites of the World Bank and the
US Bureau of Labor Statistics respectively.
Unique to our paper is the construction of a social norm
proxy for each of alcohol, tobacco, and gaming industries
over time. Each proxy is based on per capita consumption
data obtained from several different sources. Alcohol sta-
tistics are obtained from the National Institute on Alcohol
Abuse and Alcoholism. Tobacco data is obtained from the
United States Department of Agriculture. Gaming data,
based on visitor volume to Las Vegas as % of total US
population, is obtained from the Las Vegas Convention
and Visitors Authority.
3
Variables
Measures of social norms
Berkowitz (2004), in a review paper of the social norm
approach to changing undesirable social norms, argues
that actual drinking/gambling/smoking behavior (i.e.
consumption levels) are best predicted by perceptions
of drinking/gambling/smoking attitudes (i.e. perceived
injunctive norms). For example, Larimer and Neighbors
(2003), in a college gambling setting, show that social
norms are strong predictors of gambling behavior and
gambling-related negative consequences. In this study,
we assert that changes in consumption levels of alcohol,
2
Hong and Kacperczyk (2009) dataset is available up to 2003. We thank
them for the use of their data set to verify our classi?cation.
3
See Table 1 for the details of the sources. One might argue that the use
of Las Vegas visitor volume is not an optimal measure as it does not
consider % of visitors who are foreigners and it implicitly assumes that
casinos are the only available gaming option as it excludes other gaming
options such as on-line gambling. However, this is the only measure
available over such an extensive period of time. Data for the average
number of annual trips to casinos per US gambler, available for the sub
period 1993–2007 from www.americangaming.org, reveals that our gam-
bling data trends are highly correlated with the alternative measures in the
shorter sub-periods.
292 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
gaming, and tobacco over time serve as useful proxies for
social norm acceptance of these activities. We also verify
the close link between the consumption of these sin prod-
ucts and the attitude of survey participants toward these
sins, based on survey data from Gallup Corporation, the
Inter-University Consortium for Political and Social
Research (ICPSR), and the American Gaming Association.
4
Variables of Interest
Institutional ownership and analyst coverage – institu-
tional ownership is determined as of the end of the year
and is calculated as the fraction of a ?rm’s shares held by
institutions. Analyst coverage is de?ned as the natural log-
arithm of one plus the number of analysts covering ?rmi at
the end of year t. If a stock is missing from IBES or does not
have analyst forecast data as of the end of the year, the ?rm
is recorded as having no analyst following.
Expected ?nancial performance – We measure ?rm
expected ?nancial performance as the market-adjusted
return over a one-year period multiplied by negative one,
measured from the end of year t to the end of year t+ 1.
The market-adjusted return, which is calculated by sub-
tracting market return from a ?rm’s buy-and-hold return,
is multiplied by negative one to construct a performance
weakness measure. Higher values correspond to weaker
?nancial performance. In robustness tests, we use analyst
earnings forecasts and expected costs of capital as alterna-
tive proxies for expected ?nancial performance.
Control variables
Since our empirical tests are designed to capture the
relationship between social norm strength and market par-
ticipants’ behavior, we control for ?rm characteristics that
are known to be correlated with institutional ownership
and analyst coverage in our multivariate tests. Consistent
with Hong and Kacperczyk (2009) and based on evidence
related to predictors of institutional ownership compiled
by Del Guercio (1996) and Gompers and Metrick (2001),
our control variables include size, market to book value,
beta, stock price inverse, standard deviation of stock
return, exchange dummy, and S&P 500 dummy. Moreover,
following other previous literature (e.g. Bhushan, 1989;
Dechow & Dichev, 2002; Kasznik, 1999; Yu, 2008), we also
include lagged return on assets, growth rate of assets,
external ?nancing activities, a dividends dummy, interest
rate and market volatility. To control for factors affecting
the macro economy and thus the consumption of sin prod-
ucts, we include variables for the unemployment rate and
GDP growth. Finally, we also control for ?rm ?xed effects.
De?nitions of all variables are provided in Appendix A.
Results
Social norms and consumption of alcohol, tobacco, and
gaming
We ?rst examine the change in consumption of alcohol,
tobacco, and gaming over time and then verify the correla-
tion between the consumption of these products (descrip-
tive norms) and the attitude of survey participants toward
these products (injunctive norms). A good proxy for a
social norm should indicate a high consistency between
these two social norm types.
Fig. 1 shows the differences in consumption/social
norms for each of these three products for the period
1980–2007. Panels A, B, and C represent the results from
alcohol, tobacco, and gaming, respectively. While the con-
sumption measures for tobacco and alcohol have been
decreasing for most of the period, there has been an
uptrend for alcohol starting from the middle of the
1990s. With respect to gaming, it is clear that the social
norm acceptance level has been increasing steadily over
the years. Increasingly, the government has been advertis-
ing the bene?ts from gaming revenues on social programs
and decreased unemployment (www.safeandsecureig.org).
These different evolutionary processes highlight the
dynamic nature of social norms and the importance of per-
forming separate analyses on alcohol, tobacco, and gaming
stocks. For comparison purposes, we also examine the time
series of ?nancial performance of sin stocks relative to
market returns. Fig. 1 shows the difference between the
sin stock portfolio value weighted returns and market
portfolio returns for the years 1980–2007. This ?gure sug-
gests that the returns of all three sin stock subgroups are
volatile. However, these returns in general move closely
together and perform better than the broader market.
The only exception is the late 1990s when technology
stocks were booming.
We next verify whether our measures of sin product
consumption are reliable measures of social norms in
two different ways. Using data from Gallup Corporation,
the Inter-University Consortium for Political and Social
Research (ICPSR), and the American Gaming Association,
we obtain alternative measures for both descriptive and
injunctive norms. As Table 1 illustrates, we collect survey
results related to descriptive norms for the three sins as
follows: (1) number of drinks in the last 7 days, (2) have
had a drink within the past 30 days, (3) over the last two
weeks, how many times have you had at least ?ve drinks
in a row, (4) % of people who have smoked cigarettes in
the past week, (5) % of people who have smoked at least
one pack per day, and (6) casino visitation (average num-
ber of annual trips to casinos per US gambler). Table 2,
Panel A presents the correlation between the macro-con-
sumption data (our social norm proxies) used in our main
tests with these survey results of people’s consumption of
sin products. The high correlation between our macro-level
consumption and the alternative sources of consumption
4
Gallup Corporation has studied human nature and behavior for more
than 75 years and employs many of the world’s leading scientists in
management, economics, psychology, and sociology in identifying and
monitoring behavioral economic indicators. The ICPSR is funded by the
United States Department of Health and Human Services. There are two
major ICPSR studies of interest in this paper: (1) the National Survey on
Drug Use and Health (NSDUH) – this provides quarterly and annual
estimates on the use of alcohol and tobacco among members of United
States households aged 12 and older; and (2) Monitoring the Future (MTF)
– this explores changes in important values, behaviors, and lifestyle
orientations of contemporary American youth i.e. 12th-grade students.
Finally, the American Gaming Association is the industry’s ?rst national
information clearinghouse, providing the public with timely, accurate
gaming industry data.
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 293
related data (ranging from 0.73 to 0.96) suggests that our
variables effectively capture the actual consumption of
alcohol, tobacco, and gaming.
As Table 1shows, we also collect survey results relatedto
injunctive norms for alcohol, tobacco, and gaming compa-
nies as follows (all in%of respondents): (1) has drinkingever
been a cause of problems in your family, (2) disapproval of
people over age of 18 who try one or more drinks of an alco-
holic beverage, (3) disapproval of people over age 18 who
have ?ve or more drinks once or twice each weekend, (4)
harmfulness to try one or two drinks of an alcoholic bever-
age (% of respondents who say great risk), (5) harmfulness
to have ?ve or more drinks once or twice each weekend (%
of respondents who say great risk), (6) perceived danger of
second hand smoke, (7) smoking should be made illegal,
(8) risk if one smokes one or more packs of cigarettes per
day, (9) disapproval of people over age 18 smoking one or
more packs of cigarettes per day, and (10) how much do
you think people risk harming themselves by smoking one
or more packs of cigarettes per day (% of respondents who
- 40%
- 30%
- 20%
- 10%
0%
10%
20%
30%
40%
50%
60%
2.0
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3.0
1980 1985 1990 1995 2000 2005
F
i
n
a
n
c
i
a
l

P
e
r
f
o
r
m
a
n
c
e
C
o
n
s
u
m
p
t
i
o
n
Year
Alcohol
Consumption
Financial Performance
- 60%
- 40%
- 20%
0%
20%
40%
60%
80%
1000
1200
1400
1600
1800
2000
2200
2400
2600
2800
3000
1980 1985 1990 1995 2000 2005
F
i
n
a
n
c
i
a
l

P
e
r
f
o
r
m
a
n
c
e
C
o
n
s
u
m
p
t
i
o
n
Year
Tobacco
Consumption
Financial Performance
-
60%
- 40%
- 20%
0%
20%
40%
60%
80%
4%
5%
6%
7%
8%
9%
10%
11%
12%
13%
14%
1980 1985 1990 1995 2000 2005
F
i
n
a
n
c
i
a
l

P
e
r
f
o
r
m
a
n
c
e
C
o
n
s
u
m
p
t
i
o
n
Year
Gaming
Consumption
Financial Performance
(a)
(b)
(c)
Panel A: Alcohol
Panel B: Tobacco
Panel C: Gaming
Fig. 1. Plots of social norm measures and time series of ?nancial performance of sin stocks. Alcohol consumption is the recorded adult (15+) per capita
consumption of alcohol in gallons. Tobacco consumption is the domestic per capita consumption of tobacco in pieces. Gaming consumption is the number
of gaming visitors to Las Vegas as a % of the total population. Financial performance is de?ned as the differences between the ‘‘sin’’ stock portfolio return and
the market portfolio return. These plots cover the years 1980–2007.
294 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
say great risk). Table 2, Panel B presents the correlation
between the main consumption measures used and the
survey results of people’s attitude towards sin. Although
the surveys do not cover our entire sample period, the
correlation between our main macro-consumption data
and the survey results of people’s attitude towards sin is
consistently negative and high, ranging from À0.42 to
À0.93. This ?nding suggests that our consumption data
re?ects social norms toward sin products.
Descriptive statistics
Table 3 presents the descriptive statistics for our vari-
ables of interest and control variables. The mean institu-
tional ownership for alcohol, tobacco, and gaming stocks
is 18%, 34% and 27% respectively. These results are compa-
rable to the 28% documented previously by Hong and
Kacperczyk (2009) for sin stocks as a group. The mean
analyst following for these same stocks is 1.1, 2.0 and 1.5
respectively. Again, this ?nding is comparable to the 1.7
previously documented by Hong and Kacperczyk (2009).
Also of note is the positive mean and abnormal positive
returns realized by holding sin stocks (0.6%, 0.9%, and
0.1% for alcohol, tobacco, and gaming stocks respectively)
as well as the relatively low market risk beta for alcohol
and tobacco stocks (means of 0.657 and 0.741 respec-
tively). Finally, in unreported analysis, the correlation
between institutional ownership (analyst coverage) and
size is high at 0.462 (0.750) as is the correlation between
institutional ownership and analyst coverage (0.536).
These observations are consistent with previous literature
which shows that institutional investors tend to hold large
?rms and that analysts tend to cater to institutional inves-
tors by providing ?nancial reports and analyses on compa-
nies that have high institutional ownership. In each
regression model, we winsorize variables at the 1st and
99th percentiles.
Table 1
Descriptive statistics of social norm measures.
Mean Median Standard
deviation
Source and period
Alcohol
Alcohol consumption (US per capita in gallons) 2.37 2.29 0.22 www.niaaa.nih.gov 1980–2007 Note:
this is the primary measure used in
regressions
Number of drinks in last 7 days (around date of survey) 4.03 4.00 0.21 www.gallup.com 1996–2007
Have had a drink within past 30 days (%) 59.50 58.00 3.63 Michigan,
a
NSDUH database 1982–2007
Over the last two weeks, how many times have you had at least ?ve
drinks in a row (% of respondents)
32.31 30.50 4.94 Michigan, MTF database 1980–2007
Has drinking ever been a cause of problem in your family (% of
respondents)
26.14 23.50 5.96 www.gallup.com 1980–2007
Disapproval of people over age of 18 who try one or two drinks of an
alcoholic beverage (% of respondents)
24.90 26.20 4.59 Michigan, MTF database 1980–2007
Disapproval of people over age 18 who have ?ve or more drinks once or
twice each weekend (% of respondents)
64.08 64.85 4.11 Michigan, MTF database 1980–2007
Harmfulness of trying one or two drinks of an alcoholic beverage (% of
respondents who say great risk)
6.95 7.45 1.89 Michigan; MTF database 1980–2007
Harmfulness of having ?ve or more drinks once or twice each weekend
(% of respondents who say great risk)
43.44 43.30 3.77 Michigan, MTF database 1980–2007
Tobacco
Tobacco consumption (US per capita in pieces) 2087.08 2033.22 444.35 www.usda.gov 1980–2007 Note: this is
the primary measure used in regressions
% of people who smoked cigarettes in the past week 28.71 29.00 5.05 www.gallup.com 1980–2007
Smoke at least one pack per day (as % of those individuals who have
smoked at least once in the past 30 days)
48.37 49.75 9.42 Michigan, NSDUH database 1982–2007
Perceived danger of second hand smoke (% of respondents) 49.36 51.50 6.36 www.gallup.com 1994–2007
Smoking should be made illegal (% of respondents) 13.22 13.00 1.70 www.gallup.com 1990–2007
Risk if smoke one or more pack of cigarettes per day (% of respondents
who say little or no risk)
b
8.42 8.55 1.85 Michigan, NSDUH database 1985–2007
Disapproval of people over age 18 smoking one or more packs of
cigarettes per day (% of respondents)
72.45 71.95 3.72 Michigan, MTF database 1980–2007
How much do you think people risk harming themselves by smoking
one or more packs of cigarettes per day (% of respondents who say
great risk)
69.10 68.65 4.55 Michigan, MTF database 1980–2007
Gaming
Gaming consumption (visitor volume to Las Vegas as % of total US
population)
9.46 10.71 2.95 www.lvcva.com 1980–2007 Note : this is
primary measure used in regressions
Casino Visitation (average number of annual trips to casinos per US
gambler)
5.45 5.70 0.97 www.americangaming.org 1993–2007
a
Source Michigan stands for www.icpsr.umich.edu/icpsrweb/SAMHDA/sda.
b
To be consistent with other questions, this measure is converted to % of respondents who consider smoking risky by subtracting the percentage from 1
in correlation Table 2, Panel B.
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 295
The effect of social norms and ?nancial incentives on
institutional ownership and analyst coverage
The primary effect of social norms
To test hypotheses H1A and H1B, we estimate the fol-
lowing regressions within each of the three sin stock
subgroups:
Dependent Variable
it
¼ a
0
þa
1
SocialNorm
t
þa
2
Controls
it
þe
it
ð1Þ
where the dependent variables consists of institutional
ownership and analyst coverage.
5
Social norm is de?ned
as the raw consumption of alcohol, tobacco, or gaming.
Unlike Hong and Kacperczyk (2009) and Kim and
Venkatachalam (2011), who use non-sin stock ?rms as a
control group, we run regressions within each sin stock sub-
group. The bene?ts of our approach are twofold: ?rst, we
avoid the controversy of choosing the appropriate control
group; and second, this approach facilitates the investiga-
tion of our research question – how social norms and ?nan-
cial incentives interplay to determine the behavior of market
participants. Such an innovation is necessary as Fig. 1 shows
the evolution process differs across the three sin stock sub-
groups and ?nancial performance of sin stocks varies widely
during our sample period. The effects of social norms on
institutional ownership/analyst coverage may be cancelled
out in a pooling regression.
We predict that institutional ownership and analyst
coverage are higher when social norm acceptance levels
are higher. Thus, the coef?cient a
1
in the regression model
(1) is expected to be signi?cantly positive. Table 4, Model
1A presents the regression results for institutional owner-
ship. It shows that, across all three sin subgroups, strong
social norm acceptance levels have a signi?cantly positive
effect on institutional ownership, consistent with our pre-
dictions. The t-statistics for institutional ownership are
1.74, 3.05, and 5.32 for the social norms related to alcohol,
tobacco, and gaming products, respectively. The coef?cient
on social norms implies that a one standard deviation
increase in social norm acceptance results in a 7%, 17%,
and 14% increase in institutional ownership for alcohol,
tobacco, and gaming industries, respectively.
Table 4, Model 1B presents the regressions for analyst
coverage. The z-statistics for analyst coverage are 2.30,
3.03, and 2.19 for social norms related to alcohol, tobacco,
and gaming companies. In addition, size is signi?cantly
positive across all three subgroups, suggesting that institu-
tional investors/analysts are more likely to invest/follow
larger ?rms. Finally, stocks with higher standard deviation
of return and stocks in the S&P 500 index have less institu-
tional ownership, consistent with Falkenstein (1996) and
Hong and Kacperczyk (2009).
The interaction effect of social norms and ?nancial incentives
To test the interaction effect, we ?rst conduct a univar-
iate analysis followed by a multivariate analysis. Fig. 2,
Panels A and B show the results for institutional ownership
and analyst coverage for alcohol, tobacco and gaming sub-
groups separately. Expected ?nancial performance, rather
Table 2
Correlations between consumption of sin products and Gallup/SAMHSA/American Gaming Association survey results. This table presents the correlation
between the consumption of sin products (used as primary social norm measure) and the survey responses collected by Gallup Corporation, the Of?ce of
Applied Studies, Substance Abuse and Mental Health Services Administration (SAMHSA), of the United States Department of Health and Human Services, based
at the Inter-University Consortium for Political and Social Research (ICPSR), and the American Gaming Association in different sample periods.
Survey questions Sin products
Alcohol Tobacco Gaming
Panel A. Correlations between macro-consumption data and the survey results of people’s consumption of sin products
Number of drinks in last 7 days (around date of survey) 0.74
Have had a drink within past 30 days (%) 0.73
Over the last two weeks, how many times have you had at least ?ve drinks in a row (% of respondents) 0.89
% of people who smoked cigarettes in the past week 0.96
Smoke at least one pack per day (as % of those individuals who have smoked at least once in the past 30 days) 0.93
Average number of annual trips to casinos per US gambler 0.80
Panel B. Correlations between macro-consumption data and the survey results of people’s attitude towards sin
Has drinking ever been a cause of problem in your family (% of respondents) À0.66
Disapproval of people over age of 18 who try one or two drinks of an alcoholic beverage (% of respondents) À0.78
Disapproval of people over age of 18 who have ?ve or more drinks once or twice each weekend (% of respondents) À0.71
Harmfulness of trying one or two drinks of an alcoholic beverage (% of respondents who say great risk) À0.76
Harmfulness of having ?ve or more drinks once or twice each weekend (% of respondents who say great risk) À0.70
Perceived danger of second hand smoke (% of respondents) À0.77
Smoking should be made illegal (% of respondents) À0.57
Risk if smoke one or more pack of cigarettes per day (% of respondents) À0.93
Disapproval of people over age 18 smoking one or more packs of cigarettes per day (% of respondents) À0.42
How much do you think people risk harming themselves by smoking one or more packs of cigarettes per day (% of
respondents who say great risk)
À0.92
5
For analyst coverage, we adopt a negative binomial regression model
using maximum likelihood estimation (rather than an OLS regression)
because analyst coverage consists of count data. In a negative binomial
regression model, the variance is not constrained to equal the mean, which
is consistent with our data where the variance is larger. An alternative is
using a Poisson regression model, where the mean and variance are equal.
These results, which yield the same inferences, are discussed in the
robustness checks section.
296 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
than actual ?nancial performance, is used in this analysis
because institutions and analysts presumably base their
investment/coverage decisions primarily on future expec-
tations. We use actual stock returns in the subsequent year
to proxy for the expected ?nancial performance.
6
The ?rst cluster of results in Fig. 2 (Panels Aand B) repre-
sent the sample ?rm-year observations that are subject to
low social norm acceptance levels for sin stocks and where
the ?rm’s expected market-adjusted ?nancial return is
below the sample median (i.e. these ?rms are expected to
underperform the benchmark in the future). In contrast,
the last cluster of results contains ?rms that are subject to
high social normacceptance levels for sin stocks and where
the ?rm’s expected market-adjusted return is above the
sample median. Of all four clusters, institutional ownership
and analyst coverage are predicted to be the smallest (larg-
est) in the ?rst (last) cluster. This expectation is consistent
with the results. The difference between institutional own-
ership (analyst coverage) between the ?rst and last clusters
is consistently negative across all three sin stock subgroups.
With respect to institutional ownership, the decrease in
ownership going from the last cluster to the ?rst cluster is
8.5%, 19.0%, and 24.4% for alcohol, tobacco, and gaming
stocks respectively. With respect to analyst coverage, going
fromcluster four to cluster one, the decrease in the number
of analysts following a given ?rm is 1.05 for alcohol ?rms,
3.01 for tobacco ?rms, and 2.42 for gaming ?rms. The differ-
ences in institutional ownership and analyst coverage
between the two subgroups are signi?cantly different from
zero at the 5% signi?cance level for all three sin products
except for analyst coverage for tobacco which is signi?cant
at the 10% level (t = 1.94).
For our multivariate analysis, we add the interaction
termof Social Norms * Performance Weakness into regression
Eq. (1). Performance weakness is de?ned as the market-
adjusted expected return multiplied by negative one, and
social norms is as de?ned previously. Social Norms *
Performance Weakness is the main variable of interest,
capturing the impact of social norms on institutional
ownership and analyst coverage conditional on expected
?nancial performance.
We predict that the positive impact of high social norm
acceptance levels on institutional ownership (analyst
coverage) is strengthened when expected ?nancial perfor-
mance is weak. Analogously, we predict that high social
norm acceptance levels moderate the negative association
between expected ?nancial performance weakness and
institutional ownership (analyst coverage). Thus, the coef-
?cient of the interaction term Social Norms * Performance
Weakness is expected to be signi?cantly positive. The
coef?cient on Social Norms is expected to be signi?cantly
positive, consistent with hypotheses 1A and 1B; the coef?-
cient on Performance Weakness, representing the impact of
expected future performance weakness on institutional
ownership (analyst coverage), is expected to be negative.
A negative sign would indicate that institutions (analysts)
avoid ?rms with weaker expected future performance.
Table 5, Models 2A and 2B report the results from the
regression for institutional holdings and analyst coverage,
respectively. Consistent with our expectations, the coef?-
cients on Social Norms and the Social Norms * Performance
Weakness interaction term are all signi?cantly positive
while the coef?cients on Performance Weakness are all sig-
ni?cantly negative. Since performance weakness is de?ned
as market-adjusted return over a one-year period multi-
plied by minus one (measured from the end of year t to
the end of year t + 1), higher expected future performance
translates into a more negative performance weakness
measure. Thus, the results suggest that when a ?rm’s
future performance is expected to be good, institutional
investors (analysts) will be less concerned about social
norms. Analogously, institutional investors (analysts) will
Table 3
Summary statistics. Institutional ownership is the fraction of shares of a ?rm held by institutions. Analyst coverage is the log of one plus the number of analyst
estimates issued on a company at the end of the year. Alcohol consumption is the recorded adult (15+) per capita consumption of alcohol in gallons. Tobacco
consumption is the domestic per capita consumption of tobacco in pieces. Gaming consumption is the number of gaming visitors to Las Vegas as a % of the total
population. Financial performance is the market-adjusted return over one-year calculated by subtracting market return from a ?rm’s buy-and-hold return. Size
is the logarithm of the market capitalization of the company. Beta is the ?rm’s industry market beta. Inverse of stock price is one over the stock price at the end
of the year. Std of return is the daily stock return standard deviation during the year. Return on assets is calculated as earnings before extraordinary items,
divided by lagged total assets. Growth rate of assets is calculated as the change in assets scaled by lagged assets. External ?nancing activities are measured by
the sum of net cash received from equity and debt issuance scaled by total assets.
Variable Alcohol Tobacco Gaming
Mean Median StdDev Mean Median StdDev Mean Median StdDev
Institutional Ownership 0.181 0.175 0.237 0.341 0.412 0.298 0.274 0.289 0.298
Analyst Coverage 1.139 1.014 1.105 1.966 1.914 1.160 1.527 1.479 1.034
Consumption (Social norm proxy) 2.366 2.254 0.215 2087 2012 444 9.457 10.212 2.947
Financial performance 0.006 0.005 0.037 0.009 0.008 0.033 0.001 0.001 0.052
Size 12.619 11.241 2.065 14.391 13.699 2.261 12.409 12.763 1.717
Beta 0.657 0.621 0.280 0.741 0.845 0.315 1.339 1.219 0.497
Inverse of stock price 0.118 0.108 0.220 0.051 0.049 0.080 0.172 0.196 0.267
Std of return 0.026 0.033 0.015 0.021 0.023 0.011 0.036 0.038 0.016
Return on assets 0.053 0.047 0.074 0.103 0.124 0.113 0.033 0.029 0.151
Growth rate of assets 0.137 0.124 0.439 0.125 0.117 0.404 0.373 0.457 0.905
External ?nancing activities 0.005 0.004 0.141 À0.045 À0.022 0.150 0.096 0.101 0.242
6
In robustness tests, we use a future return indicator variable, analyst
earnings forecasts, and expected cost of capital as alternative proxies for
expected ?nancial performance.
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 297
Table 4
Regression analyses of institutional ownership and analyst coverage on social norms. This table presents the OLS (negative binomial) regressions of institutional ownership (analyst coverage) on social norms and
control variables for the full sample. The dependent variables in Model 1A and 1B are institutional ownership and analyst coverage, respectively. All variables are de?ned in Appendix A. Hypothesis 1A is tested in the
?rst six columns. Hypothesis 1B is tested in the next six columns. The t-statistics (z-statistics) in all regression models are adjusted for ?rm-level clustering.
Variable 1A 1B
Dependent Variable: Institutional ownership Dependent Variable: Analyst coverage
Alcohol Tobacco Gaming Alcohol Tobacco Gaming
Estimate t-value Estimate t-value Estimate t-value Estimate z-value Estimate z-value Estimate z-value
Social norms 0.322
*
1.74 0.388
***
3.05 0.049
***
5.32 1.016
**
2.30 0.742
***
3.03 0.082
**
2.19
Size 0.059
***
2.77 0.051
***
2.91 0.059
**
2.53 0.487
***
3.87 0.468
***
3.92 0.461
***
3.89
Beta 0.068 1.35 0.062 1.59 0.007 1.24 0.147 1.38 0.383 1.15 0.098
*
1.76
Inverse of stock price À0.040 À0.38 À0.262 À0.73 À0.086 À1.16 0.078 0.22 0.089 0.13 À0.228 À0.78
Std of return À0.978
***
À3.10 À6.838
*
À1.89 À2.065
**
À2.18 9.365 1.25 1.108 0.68 2.654 1.32
NASD 0.028 0.51 À0.189 À1.54 À0.070 À0.20 À0.575
**
À2.39 À0.529 À1.41 À0.187 À1.53
SP500 À0.089
***
À3.18 À0.030
***
À3.30 À0.100
*
À1.74 0.184
***
2.85 0.490 1.54 0.129
**
2.19
Lagged return on assets 0.049 0.81 0.120 0.77 0.104 1.21 0.084 0.99 0.765 0.97 0.378
*
1.92
Growth rate of assets À0.087 À1.32 À0.123
*
À1.90 À0.030 À1.49 À0.210 À1.43 À0.281 À0.21 À0.119
*
À1.90
External ?nancing activities 0.041 0.29 0.332
**
2.23 À0.021 À0.34 0.427 1.00 À0.777
*
À1.86 À0.004 À0.02
Dividends 0.002
***
4.27 0.017
**
2.02 0.031
***
3.70 0.109 0.18 0.365 0.71 0.176 0.89
Interest rate 0.001 0.28 0.002 0.33 0.001 0.31 0.018 1.22 0.276 1.45 0.162 0.76
Market volatility 0.029 1.12 0.035 0.99 0.022 1.19 À0.031 À1.36 À0.407 À1.05 À0.398 À0.96
GDP 0.000 1.40 0.000 1.01 0.000 0.97 0.000 1.02 0.000
*
1.70 0.000 0.69
Unemployment À0.057 À1.20 À0.041 À0.64 À0.029 À1.28 À0.283
**
À2.36 À0.368
*
À1.90 À0.582
**
À2.22
Intercept À1.938
*
À1.76 À1.574
***
À4.87 À1.193 À0.28 À0.365
***
À4.81 À1.746
***
À3.97 À2.490
***
À3.10
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 452 222 585 452 222 585
Adj. R-square/pseudo R-square 0.260 0.583 0.372 0.441 0.780 0.717
***
Statistical signi?cance at 1% levels (two-tailed).
**
Statistical signi?cance at 5% levels (two-tailed).
*
Statistical signi?cance at 10% levels (two-tailed).
2
9
8
Y
.
L
i
u
e
t
a
l
.
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
9
(
2
0
1
4
)
2
8
9
–
3
0
7
be more concerned about social norms when a ?rm’s
future performance is expected to be weak.
To provide some clarity to the discussion above, we cal-
culate that the marginal effect of social norms on institu-
tional ownership is 32.9% for alcohol ?rms with poor
?nancial performance (lower quartile) versus 27.5% for
?rms with good ?nancial performance (upper quartile).
7
Analogously, we calculate that the marginal effect of perfor-
mance weakness on institutional ownership is À3.5% for
alcohol ?rms with low social norm acceptance levels (lower
quartile) versus À1.8% for ?rms with high social norm
acceptance levels (upper quartile). Similar directional results
apply for tobacco and gaming ?rms as well as for analyst
coverage across all sin stock types.
8
In addition to the main
variables of interest noted above, Table 5, Models 2A and 2B
show that institutional ownership (analyst coverage) is
increasing in ?rm size. These results are consistent with
those found in Table 4, Models 1A and 1B. Finally, the
inclusion of the performance weakness and social norms/
performance weakness interaction terms does not materi-
ally change the sign and signi?cance of the other control/
explanatory variables.
The effect of social norms and ?nancial incentives on
institutional ownership: subgroup analysis
We strengthen our analyses using ?ve different classes
of institutions de?ned in the Thomson Financial 13F
database. We expect type 3 (mutual funds) and type 4
(independent investment advisors) institutions, who are
natural arbitrageurs in the market, to be less constrained
by social norms than other types of institutions whose
positions in stocks are public information, institutions with
diverse constituents, and institutions that can be readily
exposed to public scrutiny, i.e. type 1 (banks), type 2
(insurance companies) and type 5 (all other institutions,
including universities, employee stock ownership plans,
etc.). Therefore, we divide the institutions in our data set
into two subgroups: we group types 1, 2 and 5 in one
group; and types 3 and 4 in a second group. Panel A of
Table 6 shows the regression results of institutional
ownership on social norms and other control variables.
Institutional Ownership
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
poor financial
performance/low
social norm
poor financial
performance/high
social norm
good financial
performance/low
social norm
good financial
performance/high
social norm
alcohol
tobacco
gaming
Analyst Coverage
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
poor financial
performance/low
social norm
poor financial
performance/high
social norm
good financial
performance/low
social norm
good financial
performance/high
social norm
alcohol
tobacco
gaming
(a)
(b)
Fig. 2. Panel A. Univariate analysis – institutional ownership partitioned by social norm measure and ?nancial performance. Institutional ownership is the
fraction of shares of a ?rm held by institutions. Panel B. Univariate analysis – analyst coverage partitioned by social norm measure and ?nancial
performance. Analyst coverage is the number of analyst estimates issued on a company as of year-end.
7
To calculate the marginal effect of social norms on institutional
ownership for alcohol ?rms with poor ?nancial performance, we use the
mean of performance weakness for ?rms belong to that lower quartile
(which equals 0.015) multiplied by the coef?cient for ‘‘Social norms * per-
formance weakness’’ in Table 5 (which equals 1.289), plus the coef?cient
for ‘‘Social norms’’ (i.e. 0.015 * 1.289 + 0.310 = 0.329). Similarly, the mean of
performance weakness for ?rms belong to the upper quartile (which equals
À0.027) is used to calculate the marginal effect of social norms on
institutional ownership for alcohol ?rms with good ?nancial performance
(i.e. À0.027 * 1.289 + 0.310 = 0.275).
8
For analyst coverage, the marginal effect of social norms on analyst
coverage is calculated by holding independent variables at their mean (this
is due to the fact that we adopt a negative binomial regression model using
maximum likelihood estimation). The marginal effect of social norms for
alcohol ?rms with poor (good) ?nancial performance is 0.210 (0.144).
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 299
For group 1 (types 1, 2 and 5), across all sin stock types,
institutional ownership is increasing in social norm accep-
tance. The coef?cients on social norms are statistically sig-
ni?cant at two-tailed 5% (alcohol) and 1% (tobacco and
gaming) levels. The coef?cient on beta is signi?cantly neg-
ative, indicating that more risky stocks are held less by
banks, insurance companies, and other institutions. Insti-
tutional ownership of this group for alcohol, tobacco and
gaming stocks is also associated negatively with inclusion
in the S&P 500 and daily stock return standard deviation
(except for alcohol). For group 2 (types 3 and 4), the impact
of social norms on institutional ownership is insigni?cant
for two out of three types of ‘‘sin’’ (alcohol and gaming).
The t-test results on differences in coef?cients on Social
Norms between the two subgroups are reported at the bot-
tom of Table 6, Panel A. For alcohol companies, the differ-
ence in regression coef?cients on Social Norms between
group 1 and group 2 is 0.262 with a t-statistic of 5.24. This
t-statistic suggests a signi?cantly different impact of social
norms on institutional ownership for these two groups.
The results are similar for tobacco and gaming companies.
These ?ndings are consistent with our expectation that
mutual funds and independent investment advisors are
less concerned about social norms than other types of
institutions and that they are more likely to play the role
of arbitrageurs and buy sin stocks.
Table 6, Panel B presents the regression results of insti-
tutional ownership on social norms and performance
weakness by subgroups of institutional ownership. Consis-
tent with Hypothesis 2A and 2B, Panel B shows that type 1,
2, and 5 institutional investment decisions are more likely
to be driven by a comprehensive consideration of social
norms and ?nancial incentives (compared to type 3 and
4 institutional investors). These results are con?rmed by
the signi?cantly positive coef?cients on Social Norms and
the interaction term Social Norms * Performance Weakness,
and a signi?cantly negative coef?cient on Performance
Weakness for types 1, 2, and 5 but mostly insigni?cant
coef?cients for types 3 and 4. The sign and signi?cance of
all control/explanatory variables are consistent with the
regressions in the previous section; namely, institutional
ownership for types 1, 2, and 5 are increasing in ?rm size
and social norms and decreasing in beta and standard devi-
ation of return. The t-test results on differences in coef?-
cients on Social Norms and on Social Norms * Performance
Weakness between the two subgroups are reported at the
bottom of Table 6, Panel B. Consistent with our predictions,
there is a signi?cant difference in coef?cients between the
two subgroups for all three sin industries.
In summary, our ?ndings in Section ‘The effect of social
norms and ?nancial incentives on institutional ownership
and analyst coverage’ suggest that the behavior of market
Table 5
Regression analyses of institutional ownership and analyst coverage on social norms and performance weakness. This table presents the OLS (negative
binomial) regressions of institutional ownership (analyst coverage) on social norms, performance weakness and control variables for the full sample. The
dependent variables in Model 2A and 2B are institutional ownership and analyst coverage, respectively. All variables are de?ned in Appendix A. Hypothesis 2A
is tested in the ?rst six columns. Hypothesis 2B is tested in the next six columns. The t-statistics (z-statistics) in all regression models are adjusted for ?rm-level
clustering.
Variable 2A 2B
Dependent variable: Institutional ownership Dependent variable: Analyst coverage
Alcohol Tobacco Gaming Alcohol Tobacco Gaming
Estimate t-value Estimate t-value Estimate t-value Estimate z-value Estimate z-value Estimate z-value
Social norms 0.310
**
2.37 0.461
***
3.48 0.067
***
5.11 1.183
***
2.66 0.746
***
3.81 0.127
**
2.25
Performance weakness À5.731
***
À4.29 À5.837
***
À11.21 À3.672
***
À5.89 À3.849
***
À2.78 À1.647
**
À2.02 À0.650
*
À1.64
1.289
***
7.28 0.462
**
2.18 0.056
**
2.35 0.273
***
3.29 0.178
*
1.72 0.038
***
3.67
Social norms
*
performance weakness
Size 0.110
***
4.29 0.031
**
2.18 0.102
***
4.28 0.748
***
7.31 0.628
***
5.90 0.389
***
3.54
Beta 0.070 1.43 0.067 1.07 À0.006 À0.29 0.181 0.84 0.338 1.58 À0.079 À0.68
Inverse of stock price À0.034 À0.34 À0.276 À1.01 À0.014 À0.22 0.222 0.70 À0.516 À0.99 À0.482 À1.41
Std of return À1.159
*
À1.79 À7.283
**
À2.01 À4.283
***
À5.28 12.374
**
2.06 À3.227 À0.49 À2.139 À0.49
NASD À0.052 À0.81 À0.134 À1.41 À0.029 À0.29 À0.628
**
À2.38 À0.709
*
À1.89 À0.190
*
À1.67
SP500 À0.099 À1.17 À0.218
***
À2.60 À0.079 À1.63 0.190 0.61 0.320 1.05 0.040 0.36
Lagged return on assets 0.028 0.53 0.176 1.15 0.048 0.40 À0.098 À0.20 0.803 1.48 0.410
*
1.83
Growth rate of assets À0.070 À1.13 À0.110
*
À1.78 À0.039
*
À1.83 À0.221 À1.18 À0.067 À0.33 À0.109
*
À1.77
External ?nancing activities 0.067 0.51 0.390
**
2.09 0.011 0.14 0.572 1.34 À0.520 À1.29 0.061 0.30
Dividends 0.013
*
1.74 0.016
**
2.00 0.047
***
4.95 0.147 0.37 0.339 0.62 0.158 0.68
Interest rate 0.001 0.28 0.002 0.33 0.001 0.32 0.027 1.41 0.303 1.62 0.179 0.83
Market volatility 0.028 1.10 0.034 0.98 0.024 1.30 À0.056 À1.46 À0.399 À1.00 À0.467 À1.23
GDP 0.000 1.53 0.000 1.14 0.000 1.03 0.000 1.28 0.000
*
1.67 0.000 0.72
Unemployment À0.017 À1.31 À0.011 À0.71 À0.031 À1.55 À0.171
***
À3.17 À0.374
**
À2.37 À0.422
***
À5.01
Intercept À2.37
***
À3.37 3.83
***
8.37 2.98 1.08 À1.56 À1.33 À2.55
***
À5.79 À3.67
***
À6.90
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 447 220 554 447 220 554
Adj. R-square/pseudo R-square 0.259 0.590 0.396 0.447 0.779 0.745
***
Statistical signi?cance at 1% levels (two-tailed).
**
Statistical signi?cance at 5% levels (two-tailed).
*
Statistical signi?cance at 10% levels (two-tailed).
300 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
Table 6
Regression analyses of institutional ownership for subgroups. This table presents the OLS regressions of institutional ownership on social norms, performance weakness, and control variables for two subgroups. For the
?rst six columns, the dependent variable is the fraction of shares held by type 1 (banks), type 2 (insurance companies), and type 5 (others including pension plans, endowments, and employee-ownership plans)
institutions. For the next six columns, the dependent variable is the fraction of shares held by type 3 (mutual funds) and type 4 (independent investment advisors) institutions. All variables are de?ned in Appendix A.
The t-statistics in all regression models are adjusted for ?rm-level clustering.
Variable Dependent variable: Institutional ownership (Type 1 + 2 + 5) Dependent variable: Institutional ownership (Type 3 + 4)
Alcohol Tobacco Gaming Alcohol Tobacco Gaming
Estimate t-value Estimate t-value Estimate t-value Estimate t-value Estimate t-value Estimate t-value
Panel A. OLS regressions of institutional ownership on social norms
Social norms 0.276
**
2.20 0.689
***
3.11 0.034
***
3.74 0.014 0.22 0.147
**
2.41 0.004 1.07
Size 0.059
***
3.64 0.028 1.59 0.049
***
3.20 0.010 1.37 0.080
***
3.30 0.019 0.76
Beta À0.021
**
À2.00 À0.117 À1.63 À0.049
**
À2.28 0.084
**
2.28 0.192
***
3.29 0.058
***
3.67
Inverse of stock price À0.026 À0.45 À0.048 À0.21 À0.019 À0.32 À0.047 À0.48 À0.213 À1.01 À0.054
**
À2.39
Std of return À0.987 À1.28 À5.320
*
À1.78 À1.367
**
À1.97 0.090 0.19 À0.721 À0.59 À0.765 À1.17
NASD À0.004 À0.10 À0.127 À1.28 À0.005 À0.14 À0.029 À0.78 À0.038 À0.87 À0.031 À0.65
SP500 À0.048
**
À2.11 À0.152
**
À2.04 À0.033
*
À1.92 À0.040 À1.15 À0.216
***
À2.75 À0.019 À0.33
Lagged return on assets 0.029 0.37 À0.028 À0.19 0.100 1.42 0.020 1.19 0.154 1.44 0.006 0.15
Growth rate of assets À0.098 À1.39 À0.029 À0.87 À0.009 À0.65 À0.036
*
À1.78 À0.088
**
À2.23 À0.017
*
À1.66
External ?nancing activities 0.049 0.39 0.196
*
1.83 À0.064 À1.61 0.021 0.48 0.219
**
2.30 0.028 1.10
Dividends 0.003
***
5.89 0.021
**
2.44 0.030
***
3.18 0.001 1.58 0.014
**
2.10 0.033
***
3.41
Interest rate 0.001 0.27 0.002 0.34 0.001 0.30 0.001 0.49 0.002 1.10 0.001 0.34
Market volatility 0.034 1.37 0.041 1.10 0.049 1.29 0.026 1.16 0.033 1.17 0.018 0.89
GDP 0.000 1.49 0.000
*
1.66 0.000 0.82 0.000 1.45 0.000 0.64 0.000 0.74
Unemployment À0.095 À1.38 À0.038 À1.10 À0.041 À1.30 À0.026 À0.92 À0.010 À0.36 À0.026 À0.84
Intercept 2.647
**
2.48 2.183
***
3.64 1.173 0.38 2.913 0.29 1.986
***
2.99 2.176 0.28
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 452 222 585 452 222 585
Adj. R-square 0.302 0.574 0.391 0.111 0.415 0.127
T-test of difference in coef?cients (Social norms) 0.262
***
5.24 0.542
***
8.16 0.030
***
7.03
Panel B. OLS regressions of institutional ownership on social norms and performance weakness
Social norms 0.279
**
2.22 0.365
***
3.96 0.035
***
3.29 0.015 0.27 0.042
*
1.87 0.003 0.54
Performance weakness À5.268
**
À2.17 À3.765
***
À3.27 À0.328
***
À2.87 À2.453 À0.79 À0.521 À0.17 À0.165 À0.68
Social norms * performance weakness 1.367
***
2.67 0.276
*
1.83 0.048
**
2.27 0.963 1.36 0.152 1.42 0.004 1.18
Size 0.060
***
4.02 0.024
***
3.11 0.038
***
2.81 0.018 1.28 0.030
***
2.95 0.006 0.52
Beta À0.019
*
À1.90 À0.110
*
À1.91 À0.063
***
À2.83 0.114
***
2.75 0.170
***
3.73 0.068
***
3.92
Inverse of stock price À0.017 0.27 À0.083 À0.48 0.011 0.29 À0.029 À0.54 À0.190 À1.03 À0.034 À1.37
Std of return À1.156 À1.43 À5.295
**
À2.01 À2.291
***
À2.62 0.081 0.21 À0.890 À0.84 À1.349
*
À1.94
NASD À0.012 À0.39 À0.084 À1.03 À0.003 À0.09 À0.030 À0.91 À0.057 À1.28 À0.010 À0.49
SP500 À0.049
***
À3.80 À0.161
**
À2.02 À0.062 À0.44 À0.052 À1.41 À0.121
**
À2.41 À0.020 À0.34
Lagged return on assets 0.015 0.31 À0.010 À0.12 0.079 0.79 0.025 1.08 0.164
*
1.65 À0.011 À0.37
Growth rate of assets À0.054 À1.38 À0.038 À0.85 À0.028 À1.26 À0.020 À0.68 À0.066
**
À2.31 À0.017
*
À1.75
External ?nancing activities 0.061 0.58 0.231
*
1.84 À0.048 À1.15 0.022 0.37 0.218
**
2.19 0.060 1.31
Dividends 0.002
***
2.71 0.016
**
2.18 0.027
***
2.82 0.021
***
3.17 0.017
**
2.21 0.056
***
4.97
Interest rate 0.001 0.29 0.002 0.31 0.001 0.29 0.001 0.27 0.002 1.21 0.001 0.38
Market volatility 0.031 1.29 0.035 1.20 0.038 1.32 0.025 1.07 0.032 1.03 0.020 1.02
GDP 0.000 1.51 0.000
*
1.73 0.000 1.09 0.000 1.60 0.000 0.79 0.000 0.54
Unemployment À0.087 À1.55 À0.029 À1.14 À0.039 À1.48 À0.019 À1.02 À0.003 À0.22 À0.022 À1.01
Intercept 3.649
***
2.90 2.364
***
3.07 1.226 0.27 1.163 1.11 2.191
*
1.90 1.922 0.28
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 447 220 554 447 220 554
Adj. R-square 0.313 0.581 0.433 0.130 0.429 0.142
T-test of difference in coef?cients (Social norms) 0.264
***
5.31 0.323
***
4.20 0.032
***
7.23
T-test of difference in coef?cients (Social norms * performance weakness) 0.404
***
2.78 0.124
**
2.10 0.044
**
2.19
***
Statistical signi?cance at 1% levels (two-tailed).
**
Statistical signi?cance at 5% levels (two-tailed).
*
Statistical signi?cance at 10% levels (two-tailed).
Y
.
L
i
u
e
t
a
l
.
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
9
(
2
0
1
4
)
2
8
9
–
3
0
7
3
0
1
participants (institutions and analysts) dynamically
changes with the level of social norms and ?nancial incen-
tives. The evidence extends the insights provided by Hong
and Kacperczyk (2009) that sin stocks have higher
expected returns than otherwise comparable stocks. They
attribute the ?nding to the fact that sin stocks are
neglected by constrained investors and also face greater
litigation risk heightened by social norms. Focusing on
sin stocks themselves, we avoid the measurement errors
in the proxies for litigation risk and investor’s attention
which are different across sin and non-sin stocks. Our ?nd-
ing suggests social norm constrained investors may not
neglect sin stocks; otherwise, they would not adjust their
sin stock holdings and coverage dynamically in response
to changes in social norms.
Additional analyses and robustness checks
We conduct a set of additional analyses to test the
robustness of our conclusions. Results discussed but not
tabulated in this section are available upon request.
Time-varying measure of ?rms’ sin exposure
In the previous section, ?rms’ exposure to sin is assumed
to remainconstant over time. While the social acceptance of
sin is time-varying, ?rms’ exposure to sin could also vary
across time. We extend our analysis and further test
whether the perception of sin companies varies with the
revenue from sin products. We obtain historical segment
data for each ?rm-year and construct a new indicator
variable – Sin Exposure, which equals 1 if the ?rm’s revenue
generated by sin segments as a percentage of the ?rm’s total
revenue is greater than the industry median for the year;
and 0 otherwise. The higher the proportion of a ?rm’s
revenue that is generated by sin segments, the more sin
exposure it faces. We then include this new indicator vari-
able Sin Exposure and the interaction terms of Sin Exposure
with Social Norms and Performance Weakness measures into
our regression models. The empirical results are reported in
Table 7, Panels A and B. Our ex-ante prediction is that for
?rms which have above-median sin exposure (i.e., Sin
Exposure = 1), social acceptance of sin has a stronger effect
on institutional investors’ shareholding/analyst coverage
(Hypothesis 1A/1B) as well as the relation between institu-
tional investors’ shareholding/analyst coverage and future
expected performance (Hypothesis 2A/2B). In other words,
the coef?cient on the interaction term Social Norms * Sin
Exposure (Table 7, Panel A) and Social Norms * Performance
Weakness * Sin Exposure (Table 7, Panel B) is expected to be
signi?cantly positive. Our empirical results are consistent
with our predictions.
Correlation between social norms and ?nancial performance
One potential issue with using consumption data as a
social norm proxy is that consumption data can also be
thought of as a primitive proxy for demand. Unexpected
increases in demand should affect future cash ?ows, which
in turn affects accounting and stock returns. As such, one
may argue that our results will be affected by the following
relationship: increased consumption leads to higher
pro?ts, which in turn leads to increased analyst following
and institutional ownership. In our opinion, this relation
cannot explain why ?nancial performance goes up while
consumption goes down, which has been generally the
case for alcohol and tobacco industries, as Fig. 1 illustrates.
Fig. 1 shows that the social norm and ?nancial perfor-
mance plots are not correlated. The correlations between
ROA and our consumption based social norm proxies are
only 0.126, 0.003, and À0.111 for alcohol, tobacco, and
gaming industries respectively. In addition, to further
address concerns that consumption data is nothing more
than a primitive for demand/net income and that con-
sumption data may be highly correlated with other macro-
economic factors (such as unemployment rate and GDP)
which have nothing to do with social norm acceptance lev-
els, we ?rst regress the consumption data on unemploy-
ment rate, GDP and industry-level ROA as follows:
SocialNorm
t
¼ a
0
þa
1
UnemploymentRate
t
þa
2
GDP
t
þa
3
IndustryROA
t
þe
t
ð2Þ
where industry-level ROA is calculated as total earnings
divided by total assets for alcohol, tobacco and gaming
industries, respectively. We use the residuals from the
above regression, which are orthogonal to economic fac-
tors, to proxy for social norms. We then re-run our regres-
sions presented in the previous sections. The results
remain robust.
Alternative proxies for ?nancial performance
One potential concern about the performance weakness
measure used in our regressions is whether realized return
is a good proxy for ex ante expectations of future returns.
To mitigate this concern, we adopt two alternative proxies
to measure performance. First, we use analyst earnings
forecasts as a proxy for ex ante expectations of future
returns. The performance weakness measure is then calcu-
lated as the difference between the realized earnings in
year t and one-year consensus median analyst forecast
for year t + 1, scaled by the stock price at the end of year t.
Second, we estimate ?rm-year beta using a rolling win-
dow of the past ?ve years of monthly stock returns. Com-
bining this ?rm-year beta estimate with expected market
returns, equal to the past 10 year average equity return,
and the current risk-free rate, we are able to use the CAPM
model to generate expected future returns. Our results are
robust to each of these alternative proxies and are avail-
able from the authors upon request.
Finally, we rerun our regressions using a binary mea-
sure of expected ?nancial performance. Performance
weakness is de?ned as a binary variable equal to 1 if the
?rm’s realized return next year is below the industry med-
ian, and 0 otherwise. This choice not only helps interpret
the interaction effects of social norms and ?nancial perfor-
mance on market participants’ behaviour, it also mitigates
the concern whether institutions can precisely predict a
?rm’s future returns. The untabulated regression results
are similar to the results when expected performance is a
continuous measure.
Alternative social norm proxies
In Section ‘Social norms and consumption of alcohol,
tobacco, and gaming’, we verify that the consumption of
302 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
Table 7
Time-varying measure of ?rms’ sin exposure. This table presents the OLS (negative binomial) regressions of institutional ownership (analyst coverage) on social norms, performance weakness, ?rms’ sin exposure and
control variables for the full sample. Sin Exposure equals 1 if the ?rm’s revenue generated by sin segments, as a percentage of the ?rm’s total revenue, is greater than the industry median for the year; and 0 otherwise.
The dependent variables in Models I and II are institutional ownership and analyst coverage, respectively. All variables are de?ned in Appendix A. The t-statistics (z-statistics) in all regression models are adjusted for
?rm-level clustering.
Variable 1A 1B
Dependent variable: Institutional ownership Dependent variable: Analyst coverage
Alcohol Tobacco Gaming Alcohol Tobacco Gaming
Estimate t-value Estimate t-value Estimate t-value Estimate z-value Estimate z-value Estimate z-value
Panel A. Regression analyses of institutional ownership and analyst coverage on social norms and sin exposure
Social norms 0.271
*
1.65 0.322
***
3.25 0.038
***
3.28 0.812
*
1.65 0.603
**
2.38 0.072
*
1.84
Sin exposure À0.421
***
À3.76 À0.288
***
À4.89 À0.331
**
À2.10 À0.267
*
À1.77 À0.309
***
À3.05 À0.288
***
À2.87
Social norms * Sin exposure 0.389
***
3.11 0.461
***
4.01 0.067
***
5.99 1.287
**
2.18 0.868
***
3.49 0.094
**
2.16
Size 0.059
***
2.77 0.050
***
2.90 0.057
***
2.51 0.494
***
3.90 0.471
***
3.95 0.472
***
3.93
Beta 0.069 1.36 0.063 1.58 0.007 1.22 0.152 1.58 0.391 1.18 0.090
*
1.77
Inverse of stock price À0.042 À0.37 À0.258 À0.75 À0.085 À1.15 0.079 0.22 0.090 0.15 À0.231 À0.83
Std of return À0.980
***
À3.14 À6.841
*
À1.92 À2.067
**
À2.26 9.374 1.26 1.116 0.72 2.543 1.29
NASD À0.029 À0.52 À0.188 À1.54 À0.080 À0.29 À0.577
**
À2.41 À0.531 À1.42 À0.188 À1.54
SP500 À0.088
***
À3.17 À0.030
***
À3.32 À0.105
*
À1.75 0.189
***
2.89 0.490 1.55 0.141
**
2.35
Lagged return on assets 0.049 0.81 0.123 0.78 0.106 1.29 0.085 0.99 0.768 0.99 0.384
**
2.04
Growth rate of assets À0.084 À1.33 À0.124
*
À1.93 À0.030 À1.49 À0.216 À1.47 À0.281 À0.21 À0.129
**
À1.98
External ?nancing activities 0.042 0.30 0.331
**
2.23 À0.020 À0.32 0.428 1.03 À0.782
*
À1.95 À0.004 À0.02
Dividends 0.002
***
4.29 0.018
**
2.15 0.032
***
3.73 0.117 0.26 0.365 0.71 0.176 0.89
Interest rate 0.001 0.31 0.002 0.35 0.001 0.31 0.019 1.23 0.276 1.45 0.162 0.76
Market volatility 0.030 1.14 0.036 0.98 0.026 1.24 À0.031 À1.36 À0.407 À1.05 À0.398 À0.97
GDP 0.000 1.42 0.000 1.01 0.000 0.97 0.000 1.05 0.000
*
1.81 0.000 0.69
Unemployment À0.058 À1.23 À0.045 À0.65 À0.030 À1.29 À0.284
**
À2.39 À0.368
*
À1.92 À0.588
**
À2.24
Intercept À1.919
*
À1.82 1.562
***
4.82 1.184 0.19 0.375
***
4.91 À1.788
***
À3.87 À2.512
***
À3.41
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 452 222 585 452 222 585
Adj. R-square/pseudo R-square 0.262 0.586 0.375 0.442 0.782 0.719
Panel B. Regression analyses of institutional ownership and analyst coverage on social norms, sin exposure and performance weakness
Social norms 0.465
**
2.11 0.401
***
3.02 0.054
***
4.14 1.004
***
2.07 0.615
**
2.06 0.108
**
2.01
Performance weakness À5.062
***
À4.05 À5.736
***
À8.29 À3.619
***
À3.09 À3.593
**
À2.41 À1.322
*
À1.87 À0.549 À1.52
Social norms * performance weakness 1.139
***
5.16 0.449
*
1.78 0.042
*
1.92 0.231
***
2.96 0.138
*
1.71 0.028
**
2.18
Sin exposure À0.399
**
À2.32 À0.251
***
À2.65 À0.219
**
À1.98 À0.201 À1.59 À0.216
**
À2.17 À0.164
*
À1.95
Social norms * Sin exposure 0.501
**
2.11 0.529
***
3.68 0.076
***
4.26 1.292
***
2.89 0.875
***
3.93 0.145
***
2.61
Performance weakness * Sin exposure À6.387
***
À5.95 À5.922
***
À9.16 À3.717
***
À3.59 À4.012
***
À2.87 À1.921
***
À2.98 À0.752
*
À1.89
Social norms * performance weakness * Sin exposure 1.438
***
4.82 0.471
**
2.03 0.069
**
2.41 0.297
***
3.52 0.219
***
2.77 0.049
***
2.72
Size 0.111
***
4.29 0.038
**
2.44 0.138
***
4.68 0.759
***
7.82 0.636
***
5.96 0.399
***
3.61
Beta 0.093
*
1.79 0.068 1.18 À0.006 À0.31 0.193 0.98 0.338 1.58 À0.079 À0.69
Inverse of stock price À0.034 À0.42 À0.282 À1.28 À0.015 À0.23 0.225 0.72 À0.516 À0.99 À0.482 À1.41
Std of return À1.168
**
À1.99 À7.304
**
À2.55 À4.276
***
À5.18 12.362
**
2.18 À3.187 À0.42 À2.287 À0.63
NASD À0.053 À0.82 À0.136 À1.42 À0.026 À0.24 À0.639
**
À2.55 À0.711
*
À1.93 À0.229
**
À2.06
SP500 À0.101 À1.27 À0.219
***
À2.64 À0.085
*
À1.84 0.192 0.62 0.320 1.05 0.044 0.38
Lagged return on assets 0.039 0.89 0.186 1.38 0.048 0.45 À0.098 À0.21 0.803 1.48 0.419
*
1.87
Growth rate of assets À0.070 À1.14 À0.128
*
À1.82 À0.038
*
À1.83 À0.227 À1.25 À0.067 À0.33 À0.109
*
À1.79
External ?nancing activities 0.068 0.52 0.395
**
2.28 0.011 0.14 0.572 1.34 À0.520 À1.29 0.061 0.32
Dividends 0.014
*
1.75 0.019
**
2.18 0.048
***
4.96 0.147 0.37 0.339 0.62 0.158 0.68
Interest rate 0.001 0.28 0.002 0.35 0.001 0.32 0.027 1.41 0.303 1.62 0.179 0.83
Market volatility 0.028 1.15 0.034 0.98 0.025 1.37 À0.056 À1.45 À0.399 À1.03 À0.467 À1.23
GDP 0.000 1.53 0.000 1.14 0.000 1.03 0.000 1.24 0.000
*
1.68 0.000 0.72
Unemployment À0.017 À1.32 À0.017 À0.96 À0.031 À1.56 À0.176
***
À3.29 À0.395
**
À2.46 À0.429
***
À5.16
Intercept À2.310
***
À3.02 3.701
***
8.27 2.510 1.36 1.489 1.28 À2.127
***
À4.29 À3.091
***
À6.37
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 447 220 554 447 220 554
Adj. R-square/pseudo R-square 0.261 0.591 0.397 0.447 0.781 0.746
***
Statistical signi?cance at 1% levels (two-tailed).
**
Statistical signi?cance at 5% levels (two-tailed).
*
Statistical signi?cance at 10% levels (two-tailed).
Y
.
L
i
u
e
t
a
l
.
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
9
(
2
0
1
4
)
2
8
9
–
3
0
7
3
0
3
Table 8
Changes analyses. This table presents the OLS (negative binomial) regressions of change in institutional ownership (change in analyst coverage) on change in social norms, change in performance weakness, and change
in control variables for the full sample. All variables are de?ned in Appendix A. The t-statistics (z-statistics) in all regression models are adjusted for ?rm-level clustering.
Variable 1A 1B
Dependent variable: Change in institutional ownership Dependent variable: Change in analyst coverage
Alcohol Tobacco Gaming Alcohol Tobacco Gaming
Estimate t-value Estimate t-value Estimate t-value Estimate z-value Estimate z-value Estimate z-value
Panel A. Regression analyses of change in institutional ownership and change in analyst coverage on change in social norms
DSocial norms 0.318
*
1.64 0.100 1.33 0.024
**
2.10 0.383
*
1.90 0.273
*
1.78 0.043
*
1.86
DSize 0.078
***
6.89 0.039
**
2.19 0.037
**
2.34 0.292
**
2.19 0.273
*
1.90 0.187
**
2.36
Dbeta 0.021 1.01 0.036 1.41 0.004 1.19 0.072 1.05 0.372 0.98 0.087
*
1.68
DInverse of stock price À0.052 À0.61 À0.219 À0.62 À0.071 À1.26 0.054 0.18 0.067 0.11 À0.171 À0.37
DStd of return À0.473
**
À2.16 À1.038 À1.26 À0.732
*
À1.83 3.092 0.78 0.938 0.37 1.728 1.01
DLagged return on assets 0.021 0.55 0.082 0.43 0.089 1.11 0.064 0.79 0.183 0.25 0.362 1.18
DGrowth rate of assets À0.038 À1.12 À0.042 À1.61 À0.018 À1.20 À0.291 À1.10 À0.239 À0.18 À0.104
*
À1.85
DExternal ?nancing activities 0.020 0.16 0.102
*
1.92 À0.014 À0.26 0.182 0.28 À0.435 À1.42 À0.002 À0.01
DDividends 0.036
*
1.78 0.041
**
2.01 0.010
**
2.20 0.112 0.27 0.173 0.63 0.261
*
1.78
DInterest rate 0.001 0.18 0.001 0.59 0.001 0.66 0.019 1.18 0.028 1.55 0.017 1.28
DMarket volatility 0.016 0.84 0.021 0.90 0.019 0.38 À0.029 À1.46 À0.025
*
À1.66 À0.018 À1.57
DGDP 0.000 1.31 0.000 0.93 0.000 0.67 0.000 1.02 0.000 0.37 0.000 0.68
DUnemployment À0.028 À0.97 À0.031 À0.38 À0.019 À1.17 À0.172
*
À1.80 À0.436 À1.37 À0.253
**
À1.99
Intercept À0.156
**
À2.33 0.173
***
3.76 0.256 0.19 0.036
**
2.30 À1.111
***
À2.62 À0.361 À1.43
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 437 210 553 437 210 553
Adj. R-square/pseudo R-square 0.140 0.212 0.149 0.183 0.132 0.111
Panel B. Regression analyses of change in institutional ownership and change in analyst coverage on change in social norms and change in performance weakness
DSocial norms 0.173
*
1.71 0.110 1.43 0.018
*
1.77 0.291 1.32 0.219
*
1.69 0.040
*
1.70
DPerformance weakness À3.940
**
À2.02 À2.841
***
À3.85 À1.638
***
À4.29 À1.384
**
À2.16 À1.637
**
À1.98 À0.638 À1.39
D(Social norms * Performance weakness) 1.040
*
1.77 0.376
**
2.05 0.060
**
1.96 1.122
**
2.14 0.102
**
2.06 0.018
*
1.90
DSize 0.069
***
5.48 0.022
*
1.72 0.028
*
1.67 0.372
***
3.07 0.326
***
3.15 0.281
***
2.67
Dbeta 0.018 0.94 0.059 1.22 À0.006 À0.29 0.092 1.26 0.317 1.47 À0.065 À1.39
DInverse of stock price À0.027 À0.19 À0.173 À0.89 À0.017 À0.24 0.051 0.16 À0.365 À0.66 À0.110 À0.41
DStd of return À0.283
*
À1.76 À1.273
*
À1.67 À1.029
***
À2.77 3.917 0.62 À1.293 À0.38 À1.837 À1.21
DLagged return on assets 0.043 0.93 0.064 0.38 0.019 0.18 À0.036 À0.21 0.540 1.17 0.191 0.99
DGrowth rate of assets À0.028 À0.89 À0.036 À1.51 À0.028
*
À1.80 À0.286 À1.09 À0.054 À0.31 À0.091
*
À1.79
DExternal ?nancing activities 0.035 0.26 0.100
*
1.88 0.006 0.12 0.101 0.18 À0.573 À1.21 0.053 0.41
DDividends 0.010 1.33 0.035 1.62 0.012
**
2.38 0.103 0.28 0.288 1.24 0.317
*
1.90
DInterest rate 0.001 0.17 0.001 0.53 0.001 0.41 0.029 1.30 0.038
*
1.71 0.022 1.48
DMarket volatility 0.015 0.81 0.023 0.99 0.017 0.35 À0.019 À1.20 À0.020 À1.41 À0.026
*
À1.73
DGDP 0.000 1.38 0.000 0.99 0.000 1.00 0.000 1.28 0.000 1.61 0.000 0.70
DUnemployment À0.014 À0.67 À0.008 À0.27 À0.027 À1.49 À0.114
*
À1.75 À0.547
*
À1.73 À0.263
***
À3.16
Intercept À0.120
***
À5.18 0.181
***
3.83 0.026
**
2.01 0.117
***
2.59 À0.417 À1.56 À0.531
*
À1.94
Firm ?xed effect Yes Yes Yes Yes Yes Yes
N 432 208 522 432 208 522
Adj. R-square/pseudo R-square 0.145 0.182 0.129 0.188 0.127 0.124
***
Statistical signi?cance at 1% levels (two-tailed).
**
Statistical signi?cance at 5% levels (two-tailed).
*
Statistical signi?cance at 10% levels (two-tailed).
3
0
4
Y
.
L
i
u
e
t
a
l
.
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
9
(
2
0
1
4
)
2
8
9
–
3
0
7
sin products is associated with the attitude toward sin
products by using survey results from Gallup Corporation
and the Inter-University Consortium for Political and Social
Research (ICPSR). While the time series availability of Gal-
lup data, ICPSR data, and American Gaming Association
data in some cases is not as extensive as the consumption
data used in our main tables, the sixteen questions cover a
wide variety of both injunctive and descriptive norm prox-
ies. Because the time series correlation between these
measures from Gallup Corporation, ICPSR, and the Ameri-
can Gaming Association are between 42% and 96% (average
correlation of 77%), we include these new measures proxy-
ing social norms in lieu of our main proxies in each of our
multivariate analyses. In most cases, our results are robust
and signi?cant.
Changes analysis
Our main empirical analyses adopt a levels approach. To
mitigate the concern of correlated omitted variables, we
perform changes analyses for our main models. The empir-
ical results of the changes analyses are reported in Table 8.
Table 8, Panel A (changes analysis, Hypothesis 1A/1B) is
analogous to Table 4 (levels analysis, Hypothesis 1A/1B).
With the exception of tobacco for changes in institutional
ownership, the inferences we drawfromthe changes analy-
sis are the same as the levels analysis i.e. the coef?cient on
Social Norms is both positive and statistically signi?cant.
Table 8, Panel B (changes analysis, Hypothesis 2A/2B) is
analogous to Table 5 (levels analysis, Hypothesis 2A/2B).
Again, the inferences we draw from the changes analysis
are the same as the levels analysis i.e. the coef?cient on
the Social Norms * Performance Weakness interaction term
is both positive and statistically signi?cant for all sin types
and for both changes in institutional ownership and analyst
coverage. In general, the signed statistical signi?cance for
most control variables is also consistent with the levels
analyses. The t-values/z-values associated with the coef?-
cients, as well as the adjusted R-square/pseudo R-square
values, are mostly smaller than the levels analysis equiva-
lents. The lower levels of statistical signi?cance associated
with the changes analyses are consistent with other
accounting studies.
Other issues related to model speci?cations
While we highlight the bene?ts of running regressions
within each sin stock subgroup and examining the evolu-
tionary process of social norms in Section ‘The effect of
social norms and ?nancial incentives on institutional own-
ership and analyst coverage’, the approach may suffer from
correlated omitted variables. To mitigate this concern, in
addition to the analyses already performed in Section
‘Changes analysis’ above, we conduct a random sampling
analysis. Speci?cally, we select a random sample of ?rms
fromnon-sin industries where the number of non-sin ?rms
equals the number of sin ?rms in each sin industry for each
year. We then rerun our regressions. We repeat this ran-
dom sampling regression 100 times for each sin industry.
We calculate the mean of the coef?cients and the t-values.
Our analyses show that neither the coef?cients on the
social norm variable nor those on the interaction terms
of social norms and performance weakness are signi?-
cantly different from zero.
Following Hong and Kacperczyk (2009), we adopt a
standard OLS regression model for our analysis on institu-
tional ownership. However, as the institutional ownership
variable is bounded at zero and one, there is no guarantee
that the ?tted values will always lie within the feasible
range. As such, we also apply a generalized linear model
with a logit link and the binomial family to our sample.
The new model guarantees that the predicted value of
the dependent variable (i.e., institutional ownership) is
bounded between zero and one. The untabulated results
show that the inferences we draw from the regression
results remain the same.
Following previous literature, we adopt a log transfor-
mation of analyst coverage in an OLS framework for our
analyses on analyst coverage. To check the robustness of
our results, we also applied a Poisson model to our sample.
The inferences remain the same.
Conclusion
Using alcohol, tobacco, and gaming consumption data
and people’s attitudes toward these sin products to proxy
for the social norm acceptance level, we investigate how
the investment decisions of institutional investors and
the coverage decisions of ?nancial analysts are affected
by the ?nancial performance of sin companies and related
social norms. We ?nd that institutional investors’ share-
holdings and analyst coverage of ‘‘sin’’ companies are
increasing in the degree of social norm acceptance and that
the association between shareholdings/coverage and social
norm acceptance is less pronounced for ?rms with higher
future expected performance.
While in the accounting profession, the interaction
between ?nancial incentives and ethics has been debated
for many years, little research has been done in economics
and ?nance due in part to the maintained assumption that
economic agents are modeled as maximizing utilities
based on economic gain. We provide strong empirical sup-
port for a substitution effect between ?nancial and non-
?nancial incentives among economic agents. We show that
when social norms interact with ?nancial considerations,
market participants, namely institutional investors and
?nancial analysts, will forego their adherence to social
norms for ?nancial rewards. This ?nding is important
and should be of interest to academics, investors, standard
setters, and various other stakeholders, given the recent
?nancial crisis. Our research question, research design
and ?ndings highlight the importance of having empirical
measures of social norms. By using changes in consump-
tion of sin products as a proxy for the evolution of social
norms towards sin stocks, we document the distinct evolu-
tionary process of social norms related to different sin
products (alcohol, tobacco, and gaming). We thus over-
come the drawback of assuming a constant social norms
level over time and extend previous studies by showing
how social norms are priced in a dynamic setting.
Future research could examine whether ?nancial and
social norm considerations interact in a similar way in
other countries, and explore the differential impact of eth-
ics and ?nancial incentives on behavior in an organiza-
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 305
tional versus an individual setting. Booth and Schulz
(2004) have started to explore this line of research in an
experimental setting, ?nding that regardless of their level
of individual ethical reasoning, a strong ethical corporate
environment may lead to a general tendency for managers
to act in the interests of their organizations. In the current
version of this paper, both social norms (consumption and
survey based) and institutional ownership and analyst cov-
erage are domestic variables. Due to this data limitation,
we are not able to generalize our results to the global level.
However, we strongly believe that identifying the bound-
ary of social norms and its impact on market stakeholders
is an emerging area of interest; one that will continue to
provide useful insights to our society in the future.
Appendix A
Variable De?nition
Institutional
ownership
The fraction of the shares of
company i held by institutions at
the end of year t. This is
calculated by aggregating the
shares held by all types of
institutions at the end of the
year and then dividing this
amount by shares outstanding at
the end of the year
Analyst coverage Logarithm of one plus the
number of analyst estimates
issued on a company at the end
of the year t. Stocks that do not
appear in IBES are assumed to
have no analyst estimates
Consumption
(social norm
proxy)
Alcohol consumption is the
recorded adult (15+) per capita
consumption of alcohol in
gallons for the year t. Tobacco
consumption is the domestic per
capita consumption of tobacco in
pieces for the year t. Gaming
consumption is the number of
gaming visitors to Las Vegas as a
% of the total population for the
year t
Financial
performance
Market-adjusted return over
one-year calculated by
subtracting market return from a
?rm’s buy-and-hold return for
the year t
Performance
weakness
The expected ?nancial
performance of the ?rm
multiplied by negative one,
where the expected ?nancial
performance is the market-
adjusted return over year t + 1
calculated by subtracting market
return for year t + 1 from a ?rm’s
buy-and-hold return for year
t + 1
Appendix A (continued)
Variable De?nition
Size Logarithm of the market value of
equity of the company in year t
Beta Firm’s systematic risk calculated
using the past ?ve years of
monthly return
Inverse of stock
price
One over the stock price at the
end of year t
Std of return Daily stock return standard
deviation during the past ?scal
year t
NASD One if the stock is listed on
Nasdaq; and zero otherwise
SP500 One if the stock is in the S&P 500
index; and zero otherwise
Lagged return on
assets
Net income for year t À 1divided
by the average of total assets for
year t – 1
Growth rate of
assets
Change in total assets from year
t À 1 to year t scaled by total
assets for year t – 1
External ?nancing
activities
The sum of net proceeds from
equity ?nancing and debt
?nancing in year t scaled by total
assets of the year t. The proceeds
from equity ?nancing are
measured as net cash received
from the sale of common and
preferred stock less cash
dividends paid. The proceeds
from debt ?nancing are
measured as net cash received
from the issuance (and/or
reduction) of short- and long-
term debt
Dividends One if the ?rm issued a dividend
in year t; and zero otherwise
Interest rate Interest rate reported by the
Federal Reserve in year t
Market volatility Monthly value-adjusted stock
market return volatility over
year t
GDP US per capita GDP in thousands
of dollars for the year t; adjusted
for in?ation
Unemployment US unemployment rate for the
year t
Sin exposure One if the ?rm’s revenue
generated by sin segments over
the ?rm’s total revenue is greater
than the industry median for the
year t; and zero otherwise
306 Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307
References
Akerlof, G. (1980). A theory of social custom, of which unemployment
may be one consequence. The Quarterly Journal of Economics (June),
749–775.
American Gaming Association. Casino visitation – Average number of
annual trips per US gambler. .
Azar, O. H. (2004). What sustains social norms and how they evolve?: The
case of tipping. Journal of Economic Behavior & Organization, 54(1),
49–64.
Babcock, L., Loewenstein, G., Issacharoff, S., & Camerer, C. (1995). Biased
judgments of fairness in bargaining. American Economic Review, 85(5),
1337–1343.
Bayou, M., Reinstein, A., & Williams, P. (2011). To tell the truth: A
discussion of issues concerning truth and ethics in accounting.
Accounting, Organizations and Society, 36, 109–124.
Becker, G. (1957, 1971). The economics of discrimination. Chicago, IL:
University of Chicago Press.
Berkowitz, A. (2004). The social norms approach: Theory, research, and
annotated bibliography. .
Bhushan, R. (1989). Firm characteristics and analyst following. Journal of
Accounting and Economics, 11, 255–274.
Blanthorne, C., & Kaplan, S. (2008). An egocentric model of the relations
among the opportunity to underreport, social norms, ethical beliefs,
and underreporting behavior. Accounting, Organizations and Society,
33, 684–703.
Booth, P., & Schulz, A. (2004). The impact of an ethical environment on
managers’ project evaluation judgments under agency problem
conditions. Accounting, Organizations and Society, 29, 473–488.
Cai, F., & Zheng, L. (2004). Institutional trading and stock returns. Finance
Research Letters, 1, 178–189.
Cassidy, J. (2009). An economist’s invisible hand. The Wall Street Journal.
November 28–29, W3.
Cialdini, R., & Trost, M. (1998). Social norms, conformity and compliance.
In D. Gilbert, S. Fiske, & G. Lindzey (Eds.). The handbook of social
psychology (Vol. II, 4th ed., pp. 151–192). Oxford University Press.
Dechow, P., & Dichev, I. (2002). The quality of accruals and earnings: The
role of accrual estimation errors. The Accounting Review, 77(Suppl.),
35–59.
Del Guercio, D. (1996). The distorting effect of the prudent-man laws on
institutional equity investments. Journal of Financial Economics, 40,
31–62.
Elster, J. (1989). Social norms and economic theory. Journal of Economic
Perspectives, 3, 99–117.
Fama, E., & French, K. (1997). Industry costs of equity. Journal of Financial
Economics, 43, 153–193.
Falkenstein, E. (1996). Preferences for stock characteristics as revealed by
mutual fund portfolio holdings. Journal of Finance, 51, 111–135.
Gallup Corporation. Alcohol and drinking poll. .
Gallup Corporation. Tobacco and smoking poll. .
Geczy, C., Stambaugh, R., & Levin, D. (2006). Investing in socially responsible
mutual funds. Working paper. University of Pennsylvania.
Gompers, P., & Metrick, A. (2001). Institutional investors and equity
prices. The Quarterly Journal of Economics, 116, 229–259.
Grif?n, J., Harris, J., & Topaloglu, S. (2003). The dynamics of institutional
and individual trading. Journal of Finance, 58, 2285–2320.
Hechter, M. (2008). The rise and fall of normative control. Accounting,
Organizations and Society, 33, 663–676.
Hong, H., & Kacperczyk, M. (2009). The price of sin: The effects of social
norms on markets. Journal of Financial Economics, 93, 15–36.
Hunton, J., & Rose, J. (2008). Can directors’ self-interests in?uence
accounting choices? Accounting, Organizations and Society, 33,
783–800.
Kasznik, R. (1999). On the association between voluntary disclosure and
earnings management. Journal of Accounting Research, 37, 57–81.
Kim, I., & Venkatachalam, M. (2011). Are sin stocks paying the price for
accounting sins? Journal of Accounting, Auditing, and Finance, 26(2),
415–442.
Kolstad, I. (2007). The evolution of social norms: With managerial
implications. Journal of Socio-Economics, 36, 58–72.
Larimer, M., & Neighbors, C. (2003). Normative misperception and the
impact of descriptive and injunctive norms on college student
gambling. Psychology of Addictive Behaviors, 17, 235–243.
Las Vegas Convention and Visitors Authority (LVCVA) (2009). Historical
Las Vegas visitor statistics (1970–2009). .
McNichols, M., & O’Brien, P. C. (1997). Self-selection and analyst coverage.
Journal of Accounting Research, 35(Suppl.), 167–199.
Moore, S., & Ohtsuka, K. (1997). Gambling activities of young Australians:
Developing a model of behavior. Journal of Gambling Studies, 13,
207–236.
Moore, S., & Ohtsuka, K. (1999). The prediction of gambling behavior and
problem gambling from attitudes and perceived norms. Social
Behavior and Personality, 27, 455–466.
National Institute on Alcohol Abuse and Alcoholism. Alcohol consumption
statistics. .
Neighbors, C., Lostutter, T., Whiteside, U., Fossos, N., Walker, D., & Larimer,
M. (2007). Injunctive norms and problem gambling among college
students. Journal of Gambling Studies, 23, 259–273.
O’Brien, P., & Bhushan, R. (1990). Analyst following and institutional
ownership. Journal of Accounting Research, 28, 55–76.
Ostrom, E. (2000). Collective action and the evolution of social norms. The
Journal of Economic Perspectives, 14(3), 137–158.
Pigou, A. (2005). The economics of welfare (Vol. I). New York: Cosimo
Classics.
Prentice, D., & Miller, D. (1996). Pluralistic ignorance and the perpetuation
of social norms by unwitting actors. Advances in Experimental Social
Psychology, 28, 161–209.
Safe and Secure Internet Gambling Initiative (2010). New government
revenues. .
Sias, R., Starks, L., & Titman, S. (2006). Changes in institutional ownership
and stock returns: Assessment and methodology. Journal of Business,
79, 2869–2910.
Takushi, R., Neighbors, C., Larimer, M., Lostutter, T., Cronce, J., & Marlatt,
G. (2004). Indicated prevention of problem gambling among college
students. Journal of Gambling Studies, 20, 83–93.
Thompson, L., & Loewenstein, G. (1992). Egocentric interpretations of
fairness and interpersonal con?ict. Organizational Behavior and
Human Decision Processes, 51, 176–197.
United States Department of Labor & Bureau of Labor Statistics (2010).
Employment status of the Civilian noninstitutional population, 1940 to
date. .
United States Department of Health and Human Services, Substance
Abuse and Mental Health Services Administration, & Of?ce of Applied
Studies. National survey on drug use and health (NSDUH). Ann Arbor,
MI: Inter-University Consortium for Political and Social Research.
.
World Bank (2010). World Bank national accounts data, and OECD national
accounts data.
Yu, F. (2008). Analyst coverage and earnings management. Journal of
Financial Economics, 88, 245–271.
Y. Liu et al. / Accounting, Organizations and Society 39 (2014) 289–307 307

doc_808810545.pdf