The Invisible Hand of Short Selling:
Does Short Selling Discipline Earnings Management?
Massimo Massa, Bohui Zhang, and Hong Zhang☆
Current Version: October 2014
Review of Financial Studies, Forthcoming
____________________________
☆
Massimo Massa ([email protected]) is from INSEAD, Boulevard de Constance, Fontainebleau
Cedex 77305, France. Bohui Zhang ([email protected]) is from the School of Banking and Finance,
Australian School of Business, University of New South Wales, Sydney, NSW 2052, Australia. Hong Zhang
([email protected]) is from PBC School of Finance, Tsinghua University and INSEAD, 43
Chengfu Road, Haidian District, Beijing, PR China 100083. We thank two anonymous referees, Andrew Karolyi
(the Editor), Reena Aggarwal, George Aragon, Douglas Breeden, Michael Brennan, Murillo Campello, Henry
Cao, Gu Chao, Hui Chen, Bernard Dumas, Philip Dybvig, Alex Edmans, Vivian Fang, Nickolay Gantchev,
Mariassunta Giannetti, Zhiguo He, Pierre Hillion, Soren Hivkiar, Albert (Pete) Kyle, Ting Li, Bryan Lim, Jun
Liu, Mark Maffett, Ronald Masulis, David Ng, Lilian Ng, Marco Pagano, Stavros Panageas, Neil Pearson, Lasse
Pedersen, Joel Peress, Yaxuan Qi, David Reeb, Amit Seru, Philip Strahan, Kumar Venkataraman, Jiang Wang,
Yajun Wang, Wei Xiong, Feifei Zhu, and participants of numerous seminars for valuable comments. We also
thank the 2013 China International Conference in Finance’s program committee for awarding us the TCW Best
Paper Award, the 2013 Asian Finance Association’s program committee for awarding us the JUFE Best Paper
Award, the 2013 Northern Finance Association Conference’s program committee for awarding us the CFA
Society Toronto Award. We are grateful to Russell Investments for generously providing data on the Russell
index components.
The Invisible Hand of Short Selling:
Does Short Selling Discipline Earnings Management?
Abstract
We hypothesize that short selling has a disciplining role vis-à-vis firm managers that forces them to
reduce earnings management. Using firm-level short-selling data for 33 countries collected over a
sample period from 2002 to 2009, we document a significantly negative relationship between the
threat of short selling and earnings management. Tests based on instrumental variable and exogenous
regulatory experiments offer evidence of a causal link between short selling and earnings management.
Our findings suggest that short selling functions as an external governance mechanism to discipline
managers.
Keywords: Short selling, earnings management, international finance, governance.
JEL Codes: G30, M41
Short selling has traditionally been identified as a factor that contributes to market informational
efficiency.1 However, short selling has also been regarded as “dangerous” to the stability of financial
markets and has even been banned in many countries during financial crises. 2 Notably, these two
seemingly conflicting views are based on the same traditional wisdom that short selling affects only
the way in which information is incorporated into market prices by making the market reaction either
more effective or overly sensitive to existing information but does not affect the behavior of firm
managers, who may shape, if not generate, information in the first place.
However, short selling may also directly influence the behavior of firm managers. To understand
the intuition, consider a manager who can manipulate a firm’s earnings to reap some private benefits
but who faces reputational or pecuniary losses if the public uncovers this manipulation. The manager
will be confronted with a trade-off between the potential benefits and losses. The presence of short
sellers affects this trade-off. As short sellers increase price informativeness and attack the misconduct
of firms (e.g., Hirshleifer, Teoh, and Yu, 2011, Karpoff and Lou, 2010), their presence, by increasing
the probability and speed with which the market uncovers earnings management, reduces managers’
incentives to manipulate earnings. We call this view the disciplining hypothesis.
On the other hand, the downward price pressure of short selling may increase the negative impact
of failing to meet market expectations. Therefore, any additional downward price pressure arising
from short selling may incentivize firms to manipulate earnings. In other words, the threat of potential
bear raids may drive managers to manipulate earnings to avoid the attention of short sellers and thus
the confounding impact associated with the downward price pressure of their trades. We call this view
the price pressure hypothesis. These considerations, together with the aforementioned traditional
wisdom implying that managers may simply ignore the existence of short sellers (which can thus be
labeled the ignorance hypothesis), suggest that short selling may have conflicting effects in the real
1
Please see Miller (1977), Diamond and Verrecchia (1987), Duffie, Garleanu, and Pedersen (2002), Bris, Goetzmann, and
Zhu (2007), Boehmer, Jones, and Zhang (2008), Boehmer and Wu (2013), Saffi and Sigurdsson (2011), and Akbas et al.
(2013).
2
The general public concern is the potential that short selling is inherently speculative and exerts downward price pressure
that may destabilize the market. The SEC, for instance, believes that the adoption of a short sale-related circuit breaker is
beneficial as it avoids the price impact of manipulative or abusive short selling (http://www.sec.gov/rules/final/2010/3461595.pdf.)
1 economy. Distinguishing among these competing hypotheses is critical to elucidate the real impact of
short selling, which is the aim of this paper.
To detect the potential impact of short selling, we focus on the ex ante “short-selling potential”
(SSP)—i.e., the maximum potential impact that short sellers may have on firm behavior or stock
prices 3 —as opposed to the ex post actions taken by short sellers in response to observed firm
manipulation. The main proxy for SSP is the total supply of shares that are available to be lent for
short sales (hereafter, Lendable). This variable is directly related to the theory on the ex ante impact of
short selling. Diamond and Verrecchia (1987), for instance, demonstrate that short-sale constraints
reduce informative trades and the speed of adjustment to private information. A limited supply of
lendable shares imposes precisely this type of constraint (Saffi and Sigurdsson, 2011). Thus, a high
fraction of shares lendable to short sellers implies a high degree of SSP that may either discipline
managers or exert price pressure. Moreover, more active shareholders are also less likely to lend
shares to short sellers on a large scale (e.g., Prado, Saffi, and Sturgess, 2013).4 This unique property
will also help us to identify the passive supplies of lendable shares as an instrument to control for the
spurious impact of internal monitoring.
We focus on earnings management because it represents one of the “most tangible signs” of
distorted information in global markets (e.g., Leuz, Nanda, and Wysocki, 2003). Moreover, earnings
management has important normative and policy implications in numerous countries that have fallen
under regulatory scrutiny, following Regulation Fair Disclosure and the Sarbanes-Oxley Act in the US
(Dechow, Ge, and Schrand, 2010). In line with the literature (e.g., Jones, 1991, Dechow, Sloan and
Sweeney, 1995, Dechow, Ge, and Schrand, 2010, Hirshleifer, Teoh, and Yu, 2011), we use
discretionary accruals as the main proxy for earnings management. In this context, the disciplining
3
Even with limits to arbitrage, small short sellers can also affect stock prices of hard-to-short companies by using media
campaigns (Ljungqvist and Qian, 2014).
4
Activist investors have less incentive to lend out their shares because the ownership and voting rights of lendable shares will
be transferred because of the short sale–and the lack of voting rights is known to discourage the participation of active
institutional investors (e.g., Li, Ortiz-Molina, and Zhao, 2008). Indeed, lending may occur precisely to transfer voting rights
rather than exercising voting rights (e.g., Christoffersen et al. 2007), and majority lenders do not seem to actively exercise the
voting power of their lendable shares, evident by the fact that only less than 2% of shares on loan are called back on the
proxy voting record date (Aggarwal, Saffi, and Sturgess 2013).
2 hypothesis posits that SSP reduces discretionary accruals, while the price pressure hypothesis posits
the opposite. No effect is expected under the ignorance hypothesis.
We test these hypotheses by using a worldwide sample of short selling covering 17,555 firms from
33 countries over the 2002-2009 period. We begin by documenting a strong negative correlation
between the SSP of a stock and the extent of the firm's earnings management. This effect is both
statistically significant and economically relevant. A one-standard-deviation increase in SSP is
associated with 5.12% standard deviation less earnings management. This relationship is robust to the
use of fixed effects and the adoption of a dynamic-panel generalized method of moments (GMM)
estimator (Arellano and Bond, 1991). These findings offer the first evidence supporting the
disciplining hypothesis.
To address issues of potential endogeneity and spurious correlation, we adopt a twofold approach.
First, we use an instrumental variable approach based on the ownership of exchange-traded funds
(ETFs) that fully replicate benchmarks. On the one hand, fully replicating ETFs are passive investors.
These funds typically do not monitor firms or blow the whistle on corporate fraud (Dyck, Morse, and
Zingales, 2012), as they thrive on a low-fee strategy, which makes active monitoring unlikely, if not
impossible. On the other hand, the same low-fee strategy also induces ETFs to supply lendable shares
to the short-selling market, which enables them to further reduce fees. In this regard, the astonishing
40% annual growth rate of the ETF industry over the last decade, driven by investor demand for index
investment, provides large exogenous variation in the amount of shares that are available for short
selling. In line with our expectations, ETF ownership significantly explains the SSP variations in our
sample. All of these features—the passive nature of ownership, the supply of lendable shares
motivated by fees, and the time series variations in ETF ownership attributable to investor flows
focusing on benchmarks—make ETF ownership an ideal instrument for the share of SSP unrelated to
earnings management. To further control for unobservable firm characteristics, we use both firm-level
and industry-wide ETF ownership in our tests.
3 We find that instrumented SSP also significantly reduces earnings management. Moreover, when
we directly link ETF ownership to earnings management, we find that ETF ownership does not reduce
earnings management when SSP is included in the full-sample regressions or when SSP is low or
prohibited in subsample regressions. These results suggest that ETF ownership affects earnings
management through its effect on short selling.
Second, we consider an event-based approach that explores two regulatory experiments: the SEC
Regulation SHO in the US and the gradual introduction of (regulated) short selling on the Hong Kong
Stock Exchange. The US experiment began in 2005 and lasted until 2007. The SEC established a pilot
program that exempted one-third of the stocks on the Russell 3000 Index from price restrictions that
were related to short selling. The choice of the stock was purely random across average daily trading
volume levels within the NYSE, NASDAQ, and AMEX stock exchanges (e.g., Diether, Lee and
Werner, 2009, Grullon, Michenaud, and Weston, 2012). We find compelling evidence that lifting
short-selling restrictions—i.e., Regulation SHO—reduced earnings management by between 6.57%
and 7.88%, on average, depending on the specifications.
In Hong Kong, short selling was prohibited until 1994, when the Hong Kong Stock Exchange
introduced a pilot scheme allowing short selling for a list of 17 stocks. Since then, the list of firms that
are eligible for short selling has changed, creating both time-series and cross-sectional variations with
respect to short-selling restrictions for firms listed in Hong Kong. Similarly to the case of Regulation
SHO in the US, we find that stocks for which short selling has been allowed experience dramatic
reductions in earnings management.
For both experiments, we design “placebo” tests to further confirm that changes in earnings
management are only related to the regulatory changes in short-selling restrictions. Overall, these tests
support a causal interpretation of the relationship between SSP and reduced earnings management,
which is consistent with the disciplining hypothesis as opposed to the alternative hypotheses.
In addition to the above main tests, we also implement a series of additional tests to further enrich
our economic intuition. First, we show that, worldwide, regulations that restrict short selling (such as
4 country-wide short-selling bans) are typically associated with greater earnings management. These
results suggest that the importance of short-selling regulations in affecting earnings management
incentives is not limited to a few selected markets. Second, in line with the observation that shortselling activity grew tremendously in our sample period with the emergence of hedge funds (e.g., Saffi
and Sigurdsson, 2011), we document that the disciplining impact of short selling on earnings
management also increases over time. Third, we show that the disciplining effect is robust to the use
of alternative SSP proxies and that it applies to a wide spectrum of earnings management measures,
including not only additional discretionary accruals but also a list of target-beating, earnings
persistence, and earnings misstatement measures. Finally, based on the framework of Morck, Yeung,
and Yu (2000) and Jin and Myers (2006) in general and Bris, Goetzmann, and Zhu (2007) in particular,
we not only confirm a positive relationship between short selling and stock price informativeness
(Saffi and Sigurdsson, 2011), but also find that this positive relationship is more pronounced when the
potential impact of short selling on earnings management is high, suggesting that short selling may
increase price efficiency by reducing the incentives for firms to manage their earnings.
Overall, these results offer evidence of a beneficial, rather than detrimental effect of the shortselling market on the corporate market. They are closely related to Hirshleifer, Teoh, and Yu (2011)
and Karpoff and Lou (2010). These authors show that short sellers attack firms that manipulate
earnings or exhibit misconduct; we show that the very possibility of such potential attacks reduces
earnings management. This finding has important normative implications because it shows that short
selling—which is generally considered a source of the problem of deceptive market information—
does in fact contribute to solve such a problem. In a contemporaneous paper, Fang, Huang, and
Karpoff (2014) confirm our conclusions focusing on the SHO experiment. Our paper differs by
providing extensive evidence related to lendable shares and their passive suppliers. This approach not
only allows us to measure the disciplining impact of short selling using a concrete proxy, but also
enables us to identify a pivotal economic channel—i.e., ETFs—that affects the prosperity and
efficiency of the short selling market. The link to ETF investment enriches our knowledge on how
5 passive and active investors intertwine in affecting the real economy. Moreover, our focus is broader,
considering the impact of short selling on the global market, not merely the US market. Jointly,
therefore, our results provide both concrete channels and unique international experience that policy
makers, especially those in emerging markets, can rely on to improve firm efficiency through both the
adoption of better short selling-related regulations and the development of passive (ETF-alike) and
long-term investors.
We contribute to different strands of the literature. First, we provide the first analysis—to the best
of our knowledge—of the real impact of the short-selling market on corporate behavior in general and
earnings management in particular. While the standard short-selling literature links short sellers’
activities to stock returns (Senchack and Starks, 1993, Asquith and Meulbroek, 1995, Aitken et al.,
1998, Cohen, Diether, and Malloy, 2007, Boehmer, Jones, and Zhang, 2008, Boehmer and Wu, 2013,
Saffi and Sigurdsson, 2011), we contribute by directly linking short sellers’ activity—or more
specifically, the threat of their activity—to managerial behavior.
Second, we contribute to the corporate governance literature, which has studied the trade-off
between “voice and exit” (Maug, 1998, Kahn and Winton, 1998, Faure-Grimaud and Gromb, 2004).
This stream of literature has focused on “voice” as the primary disciplining device, though recent
studies also show that “exit” is a governance mechanism in itself (e.g., Admati and Pfleiderer, 2009,
Edmans, 2009, Edmans and Manso 2011, and Edmans, Fang, and Zur 2013). Unlike the previously
discussed governance mechanisms, the disciplining force of the short-selling channel identified in our
paper arises from the outside (i.e., from the external market) as opposed to the inside (i.e., from
existing shareholders). Thus, the “invisible hand” of the market affects and disciplines managers.
Third, our results contribute to the literature on the determinants of earnings management, which
has focused on firm operating and financial characteristics (see DeFond and Park, 1997, Watts and
Zimmerman, 1986, Nissim and Penman, 2001), auditing quality and financial reporting practices
(DeAngelo, 1981, Barth, Landsman, and Lang, 2008), market pressure (Das and Zhang, 2003,
Morsfield and Tan, 2006), as well as investor protection and regulations (Leuz, Nanda, and Wysocki,
6 2003, Dechow, Ge, and Schrand, 2010). Our evidence on the role of SSP provides another external
channel to mitigate managers’ incentives to manage accounting earnings.
1. Data, Variable Construction, and Preliminary Evidence
1.1 Data Sample and Sources
The sample of short selling covers the period between 2002 and 2009. We begin with all publicly
listed companies worldwide for which we have accounting and stock market information from
Datastream/WorldScope. This sample is then matched with short-selling information data from
DataExplorers and data on institutional investors’ stock holdings from FactSet/LionShares.
We obtain equity-lending data from DataExplorers, a research company that collects equity- and
bond-lending data directly from the securities lending desks at the world’s leading financial
institutions. Information detailed at the stock level is available from May 2002 to December 2009. In
particular, the dataset provides unique information on the value of shares that are on loan to short
sellers and on the value of shares that are available to be lent to short sellers; both sets of information
are important for the analysis in this paper. More detailed descriptions of the data can be found in Saffi
and Sigurdsson (2011) and Jain et al. (2013). DataExplorers provides monthly information in 2002 and
2003 (weekly information from July 2004 on and daily information after 2006). Because of
DataExplorers’ low coverage during the first two years, we also show the robustness of our findings
by focusing on a shorter period from 2004 to 2009 or from 2006 to 2009 in Section 4 to address
concerns regarding data quality in the early years of the period considered.
The data on institutional investor ownership are from the FactSet/LionShares database, which
provides information on portfolio holdings for institutional investors worldwide. Ferreira and Matos
(2008) and Aggarwal et al. (2011) provide a more detailed description of this database. Because
institutional ownership represented over 40% of total global stock market capitalization during our
sample period, we control for institutional ownership in all our regressions. FactSet/LionShares also
provides us with data on ETF ownership of stocks. The identity and replicating methods of ETFs (i.e.,
whether an ETF physically replicates its index), however, are provided by Morningstar. We match
7 data from Morningstar with data from the FactSet/LionShares database and identify ETFs that fully
replicate the indices they track using Morningstar and then use the latter database to aggregate ETF
stock ownership.
We combine Datastream data with the short-selling and institutional holdings data by using
SEDOL and ISIN codes for non-US firms. We use CUSIP to merge short-selling data with US
security data from Datastream. The initial sample from the matched datasets of Datastream and
DataExplorers covers 22,562 unique firms. After the match with Factset/Lionshare, the sample was
reduced to 20,128 firms over the period considered. Countries like China, India, Malaysia, and
Thailand, for instance, have been excluded due to the lack of short selling information. We further
require stocks to have nonmissing financial information on firm size, book-to-market ratio, financial
leverage, annual stock return, and stock return volatility. These requirements reduce the number of
stocks to 17,555 in 33 countries. Appendix B tabulates the number of stocks covered by each of these
33 countries over the sample period, from 3,637 non-US firms and 1,193 US firms in the year 2002 to
7,878 for non-US firms and 4,031 for US firms in December 2009. In the year 2008, for instance, we
cover 13,082 stocks, a number comparable to the sample of 12,621 stocks in the same year in 26
countries in Saffi and Sigurdsson (2011). Regarding the coverage of market capitalization, the sample
includes more than 90% of global stocks.
1.2 Main Variables
Following the literature we use accruals as the main proxy for earnings management (e.g., Dechow,
Ge, and Schrand, 2010). Total accruals (Accruals) are calculated from balance sheet and income
statement information. In particular,
∆
∆
∆
∆
∆
,
where ∆CA is the change in current assets, ∆Cash is the change in cash and equivalents, ∆CL is the
change in current liabilities, ∆SD is the change in short-term debt included in the current liabilities,
∆TP is the change in income tax payable, and DP denotes depreciation and amortization expenses. All
of the numbers are scaled by lagged total assets. Total accruals include discretionary and
nondiscretionary components. Because nondiscretionary components depend on the economic
8 performance of a firm—such as changes in revenues and the depreciation of fixed assets—the
discretionary component can measure managerial discretion in reported earnings more precisely.
Therefore, to measure the discretionary component of accruals, we rely on Dechow, Sloan, and
Sweeney’s (1995) modification of Jones's (1991) residual accruals (AccrualMJones) as the main measure.
AccrualMJones denotes the residuals obtained by regressing total accruals on fixed assets and revenue
growth, excluding growth in credit sales, for each country and year.5
We proxy for our main measure of SSP by using Lendable—i.e., the annual average fraction of
shares of a firm that are available to be lent to short sellers. We rely on Saffi and Sigurdsson (2011)
and compute the ratio between the values of shares that are supplied to the short-selling market (as
reported by DataExplorers) and the market capitalization of the stock (as reported by Datastream), and
we then define the time-series average of the monthly (weekly or daily) ratio as the annual Lendable
ratio. We primarily consider the annual frequency because earnings management variables are defined
annually. In addition to our main dependent and independent variables, we have also constructed
alternative measures of earnings management and SSP. These variables will be detailed in subsequent
sections when we conduct robustness checks.
Our control variables are the logarithm of firm size (Size), the logarithm of the book-to-market
ratio (BM), financial leverage (Leverage), the logarithm of annual stock return (Return), stock return
volatility (STD), American Depositary Receipts (ADR), MSCI country index membership (MSCI), the
number of analysts following the firm (Analyst), closely held ownership (CH), institutional ownership
(IO), and Amihud's (2002) illiquidity (Illiquidity). 6 Institutional ownership denotes the aggregate
equity holdings by domestic and foreign institutional investors as a percentage of the total number of
outstanding shares. Similarly, we also construct ETF ownership (ETF), which is defined as the
percentage of the total number of outstanding shares that are invested by ETFs. Industry-level ETF
5
Our results are also robust to regressions by industry and year, by country, or by year only. We do not run regressions by
country, year, and industry, as many countries do not have sufficient observations to support regressions at the industry level.
6
Our results are also robust to alternative illiquidity measures, such as the proportion of zero daily returns in a year, the
turnover ratio, proportional effective spread, and proportional relative spread. Here, we primarily rely on the Amihud
measure because of its importance in the global market (e.g., Karolyi, Lee, and van Dijk, 2012). We tabulate the results for
the alternative liquidity measures in Table IA11 in the Internet Appendix.
9 ownership (ETFIndustry) is computed as the equally weighted average of ETF ownership in any industry
excluding the corresponding firm. A detailed definition of these variables is provided in Appendix A.
Table 1 presents the summary statistics for the main dependent, independent, and control variables.
The summary statistics for all the other variables used in later sections are provided in Table IA2 in
the Internet Appendix. Since now on, we will define the tables contained in the Internet Appendix with
the prefix “IA”. Panel A reports the number of observations and the mean, median, standard deviation,
and decile (90% and 10%) and quartile (75% and 25%) distributions of the variables. Panel B reports
the correlation coefficients among the main variables. We calculate both the Spearman correlation
coefficients and the Pearson correlation coefficients. The former are reported in the upper right part of
the table, whereas the latter are reported in the bottom left part of the table.
We can see that our dependent and independent variables have reasonable variation. For example,
the mean (6.5%) of Lendable is close to the mean (8.0%) of the lending supply variable in Saffi and
Sigurdsson (2011) for firms with reasonable financial information. The remaining difference arises
from two sources. First, we require firms to have valid annual earnings management variables to be
included in our sample, while Saffi and Sigurdsson (2011) require weekly stock return information.
Second, our final sample focuses on the testing period from 2002 to 2009, while their sample is from
2004 to 2008. Broadly speaking, the two sources explain approximately two thirds and one third of the
residual difference, respectively. Table IA3 tabulates the average value of Lendable for each year and
provides a more detailed comparison. Later sections will show that our results are robust to the
inclusion/exclusion of the early years. Our results are also robust whether we include or exclude the
firms for which no shares are available to be sold short (i.e., zero Lendable).
Panel B illustrates that a negative correlation exists between discretionary accruals and SSP,
suggesting a disciplining effect of short selling on earnings management. For example, the Pearson
(Spearman) correlation coefficient between AccrualMJones and Lendable is -0.030 (-0.050), with a tstatistic of 8.23 (9.56), and its absolute magnitude is the fourth (second) largest among the Pearson
(Spearman) correlation coefficients between other control variables and accruals. Although this result
10 provides preliminary evidence of such a correlation, this correlation may be spurious. Thus, the next
step of the analysis is to examine the relationship in a multivariate framework.
2. Short-selling Potential and Earnings Management: Initial Evidence
We rely on the following regression as a baseline for our multivariate analyses:
,
where
,
,
,
,
,
, 1
refers to our main earnings management proxy for firm in year
is the fraction of lendable shares for the same firm in the previous year ; and
,
1;
refers
to a list of lagged control variables, including firm size, book-to-market ratio, financial leverage,
annual stock return, stock return volatility, American Depository Receipts, MSCI country index
membership, number of analysts following the firm, closely held ownership, institutional ownership,
and Amihud's (2002) illiquidity. All the control variables are as of the previous year.
Table 2 reports the results of the regression with various econometric specifications. Model (1)
presents our baseline specification, in which we include industry, country, and year fixed effects (ICY)
and cluster standard errors at the firm level. This regression specification is the standard one in the
literature when accruals are used as the dependent variable (e.g., Yu, 2008, Francis and Wang, 2008,
Francis, Michas, and Seavey, 2013).7 The results show a strong negative correlation between SSP and
earnings management. Specifically, a one-standard-deviation increase in SSP is associated with 5.12%
standard deviation less earnings management. 8 This impact is both statistically significant and
economically relevant. Models (2), (3), and (4) remove the year fixed effect, control for firm and year
fixed effects, and control for firm fixed effects, respectively. Our main conclusions are robust across
all the different specifications.
Next, Models (5) and (6) apply the dynamic-panel GMM estimator of Arellano and Bond (1991).
This method exploits the lagged explanatory variables as instruments and is especially suitable for
7
Compared to Saffi and Sigurdsson (2011), we further control for industry-level fixed effects to remove industry-specific
factors that affect earnings management (e.g., Yu, 2008, Dechow, Ge, and Schrand, 2010). However, Saffi and Sigurdsson
(2011) adopt firm-year double clustering following Petersen (2009) and Thompson (2010). Tables IA4 and IA13 will show
that our results are robust to the use of double clustering.
8
The economic impact is computed as the regression coefficient multiplied by the one standard deviation change in Lendable,
which is scaled by the standard deviation of discretionary accruals in the sample.
11 small time-series and large cross-sectional panels, providing unbiased and consistent estimates. The
results show that the dynamic GMM estimator supports a negative correlation between SSP and
earnings management.9
Overall, our results provide consistent multivariate evidence that a higher level of SSP is
associated with less earnings management in the future. Before we move on to develop a causal
interpretation, it is also worthwhile to note that the parameters of the other variables are consistent
with the existing literature on earnings management. For example, large-sized firms have aggressive
accruals because of income-increasing accounting method choices (Watts and Zimmerman, 1986).
Being listed in the US market (i.e., ADR) is negatively and significantly associated with a firm’s
accruals. This observation provides the first supporting evidence—in terms of earnings management–
of the bonding hypothesis that cross-listings on US stock exchanges strengthen corporate governance
and protect outside investors (Doidge, Karolyi, and Stulz, 2004).
3. Endogeneity Issues
The previous results, although favorable to the disciplining hypothesis, may be subject to the issue of
endogeneity. We have already addressed the issue of spurious correlation through the omission of
relevant firm-specific information with alternative fixed effect and dynamic panel data specifications.
In this section, we address this issue using a twofold approach: we employ an instrumental variable
specification and provide two regulatory experiments in which short selling is exogenously determined.
3.1 An Instrumental Variable Approach
We begin with an instrumental variable specification. Relying on the findings of Hirshleifer, Teoh,
and Yu (2011), we argue that the ownership of ETFs that fully replicate benchmarks can be used as an
ideal instrument.10 On the one hand, ETFs are among the main contributors to the short-selling market.
Because the ETF industry thrives on its low-fee reputation, ETFs often lend out shares to the short 9
Based on the Arellano-Bond test for both AR(1) and AR (2) specifications and the Hansen’s test of overidentifying
restrictions, our estimation satisfies the zero-autocorrelation request for the residuals.
10
Hirshleifer, Teoh, and Yu (2011) use institutional ownership as an instrument for the amount of lendable shares to proxy
for the ease of short arbitrage. Our approach is in the same spirit, except we further require the instrument to be uncorrelated
with internal governance. This consideration motivates us to use ETFs to identify the passive component of lendable supply
as a general control to alleviate concerns related to internal governance.
12 sellers to generate additional income that allows them to reduce fees. For instance, iShares Russell
2000 Index Fund (IWM), a $15 billion ETF with an expense ratio of 23 bps, generated 21 bps from
security lending in a one-year period. Overall, iShares made $397 million in securities lending fees in
2011. On the other hand, ETFs are not typically concerned with enjoying active control over the
managers of the firm because their goal is simply to replicate benchmarks. Moreover, precisely
because ETFs replicate benchmarks rather than paying attention to the performance of individual
stocks, the time-series variations in ETF ownership can only be attributed to investor flows related to
benchmark characteristics, as opposed to stock-specific information.
These characteristics make the fraction of stock ownership held by ETFs a suitable instrument
because it reasonably meets both the exclusion restriction (i.e., it is unrelated to earnings management
except through the short-selling market) and the inclusion restriction (i.e., ETFs make shares available
to short sellers). Moreover, the exogenous and high growth rate of the ETF industry over the past
decade suggests that the instrument is likely to have power. In addition to ETF ownership at the stock
level (ETF), we also utilize industry-level (excluding the specific firm) ETF ownership (ETFIndustry).
The latter instrument has lower cross-sectional variation, but is less related to firm-specific
characteristics.
Based on these two instruments, we perform a two-stage IV regression as follows. We regress SSP
on either instrument in the first stage and then regress our earnings management measure
(AccrualMJones) on ETF ownership (ETF)-instrumented SSP in the second stage, together with firmlevel control variables (X) and industry, country, and year fixed effects:
1 2
:
,
:
,
,
,
,
,
,
,
,
,
. 2
The results are tabulated in Table 3. In Panel A, Models (1) and (3) regress lendable shares on
ETF ownership and industry-level ETF ownership, respectively. Models (2) and (4) regress earnings
management on predicted lendable shares. If we focus on the first-stage regressions, we observe that
SSP is strongly positively related to the fraction of ETF ownership at both the firm and the industry
13 level. A one-standard-deviation increase in firm-specific (industry-specific) ETF ownership is
associated with 19.89% (38.53%) higher lendable shares. The t-statistics are always above 5,
translating into an F-test of 29.78 in Model (1) and 28.13 in Model (3), both of which are well above
the threshold of weak exogeneity provided by Staiger and Stock (1997).
The second-stage regressions show a strong negative correlation between instrumented SSP and
earnings management. A one-standard-deviation increase in ETF-instrumented lendable shares is
correlated with 9.61% lower discretionary accruals. Note that in all the regressions, we control for the
level of a firm’s institutional ownership, effectively controlling for any monitoring role played by
institutional investors. Additional tests, reported in Table IA4, confirm that our results are robust to
firm-year double clustering following Petersen (2009) and Thompson (2010), to later periods (20052009), to stocks that are members of the MSCI country index, and to the presence of country-level
characteristics as additional control. In unreported tests, we further orthogonalize ETF ownership with
respect to a list of attention and liquidity variables, such as analyst following, news coverage, and
Amihud's (2002) illiquidity measure, and the results remain identical.
As we have argued, the features of the ETF industry—e.g., the passive nature of ownership, the
supply of lendable shares motivated by fees, the time-series variations in ownership attributable to
benchmark-related investor flows—imply that ETFs do not directly affect managerial behavior.
Although this implication is widely supported (e.g., Dyck, Morse, and Zingales, 2012, show that ETFs
do not blow the whistle on corporate fraud, although short sellers do), we provide additional evidence
that the ETF ownership seems to affect earnings management only through its impact on short selling.
Models (1) and (6) in Panel B report the regression of accruals on the two ETF instruments without
SSP. If ETF ownership indirectly affects earnings management through SSP, we would expect ETF
ownership to be significantly (and negatively) related to earnings management in the absence of SSP.
However, if the inclusion of SSP removes the significance of the effect of ETF ownership on earnings
management, it would provide further evidence supporting the exclusion restriction of the
14 instruments.11 Models (2) and (7) show that the inclusion of SSP does indeed remove the significance
of the effect. While the two instruments are correlated with the dependent variable in general, they
lose their explanatory power when the specific channel, SSP, is presented.12
To further illustrate this point, we reestimate our specifications within the subsamples of firms for
which short selling is more constrained and report the results for Models (3) to (5) and (8) to (10). In
Models (3) to (5), short selling is either prohibited owing to regulation (Legality=0 or SSban=1) or
difficult to implement owing to the low supply of lendable shares (0<Lendable<0.5%). We see again
that firm-level ETF ownership loses its power in affecting earnings management, suggesting that short
selling is the necessary channel for ETF ownership to affect earnings management. Similar subsample
regressions for industry-level ETF ownership are estimated in Models (8) to (10), from which we
obtain the same conclusion.13 The key message is therefore that ETFs provide short sellers with the
“ammunition” that they use to “discipline” managers. However, this provision of shares to short sellers
alone is unlikely to directly affect earnings management.
In Table IA6, we also consider a further instrument: the degree of concentration of institutional
ownership. Prado, Saffi, and Sturgess (2013) demonstrate that because high ownership concentration
is related to investor activism, it typically introduces additional constraints into the lending market. In
this regard, a lower degree of ownership concentration provides an instrument for the supply of
lendable shares that is less related to investor activism. When this instrument is used alone or jointly
with ETF ownership, our main findings remain unchanged. These instrumental variable regressions,
therefore, provide the first evidence supporting the causal disciplining impact of SSP on earnings
management.
11
Following Conley, Hansen, and Rossi (2012), when the dependent variable , the independent variable , and the
instrument are included in the same regression as
, the exclusion restriction is equivalent to the
condition of
0. We also conduct over-identification tests based on the Hansen J statistic in Table IA6.
12
We also apply the plausible exogeneity test of Conley, Hansen, and Rossi (2012) to compute the confidence interval of
when the coefficient may take nonzero values ( and are specified in footnote 9). For instance, when firm-level ETF
ownership is used as the instrument, the 95% confidence interval for the impact of SSP can be estimated as
.169, .035 under the assumption that may take values from the support of 2 , 2 ], where is the estimated
standard deviation of . The results further confirm that the negative impact of SSP is statistically robust.
13
ETF ownership, however, is not a necessary condition for SSP to affect earnings management. We show Table IA5 that the
disciplining effect of short selling is not attenuated when the level of ETF ownership is low. This result is logical because
other (passive) institutional investors, such as pension funds and insurance companies, may also be willing to lend shares to
short sellers. 15 3.2 An Event-based Approach: the US SHO Experiment
We now consider two regulatory “experiments” that have exogenously affected the ability to short sell.
The advantage of this approach is that these policy events can create shocks and variations in shortselling costs that are orthogonal to firm-specific spurious correlation and endogeneity. We begin with
the changes in short-sale price restrictions under Regulation SHO.
In this US experiment, the SEC established a pilot program to exempt one-third of the stocks in
the Russell 3000 Index from uptick rules and other price restrictions (e.g., Diether, Lee and Werner,
2009, Grullon, Michenaud, and Weston, 2012).14 The stocks were selected at random. As described in
SEC Release No. 50104, the regulator “sorted the securities into three groups – Amex, NASDAQ, and
NYSE – and ranked the securities in each group by average daily dollar volume over the year prior to
the issuance of the order from highest to lowest for the period. In each group, we then selected every
third stock from the remaining stocks.” 15 Thus, the SEC essentially generated a randomized
experiment that we can exploit to assess whether a relaxation in short-selling restrictions, which
exogenously enhances SSP, translates into more effective disciplining. We therefore relate earnings
management to an indicator of whether the restrictions have been lifted for the specific stock.
We begin by directly relating earnings management to the removal of the uptick restrictions.
Specifically, we estimate the following annual panel regressions with firm and year fixed effects (FY):
,
2005
2009
where
,
,
2007
,
, (3)
is modified Jones' (1991) residual accruals,
variable, which equals one if the stock is selected as a SHO pilot firm, and
control variables. The variable “
2005
2008
refers to the dummy
,
refers to a list of
2007 ” takes a value of one for the period between
2005 and 2007, when pilot firms face fewer short-selling restrictions and thus higher SSP than the
14
The regulation was announced in 2004 and implemented in 2005. Because firms may have begun reducing earnings
management immediately after the announcement of the policy, the important change in management for our purposes is
from 2003 to 2005, not from 2004 to 2005.
15
The details are available at http://www.sec.gov/rules/other/34-50104.htm. The experiment was performed by the Office of
Economic Analysis.
16 2008
control group. The variable “
2009 ” takes a value of one for the period from 2008
to 2009, in which the regulatory difference between pilot firms and the control group vanishes. The
latter period is selected to provide placebo tests of our analyses: if the difference in earnings
management between pilot and control firms is indeed driven by the difference in short-selling
restrictions between the two groups of firms in the first period of 2005-2007, the difference in earnings
management should vanish when the regulatory difference evaporates in the latter period.
We report the results in Table 4, Models (1)-(3). The testing period in Model (1) is from 2001 to
2007, in which the announcement year (2004) of Regulation SHO is removed from the sample. In
Model (2), the sample period is expanded to 2009 to allow for the placebo test. In Model (3), the
sample period is from 2001 to 2009, excluding the event period of 2004-2007. Models (1) and (2)
clearly show that lifting the short-selling restrictions is associated with a lower level of earnings
management by the group of pilot firms relative to the control group. Exemption from the restrictions
is associated with a 7.88% lower level of earnings management. It is interesting to note that in Models
(2) and (3), the effect disappears after 2007, confirming that the aforementioned earnings management
difference between pilot firms and control firms is specifically related to the relatively lower level of
short-selling restrictions faced by the pilot firms.
In Models (4) and (5), we focus on a specification based on changes and estimate the following
cross-sectional regression:
Δ
where Δ
,
,
Δ
,
,
, (4)
refers to the difference between the three-year average value of
AccrualMJones after year 2004 (from 2005 to 2007 in Model (4) and from 2008 to 2009 in Model (5))
and that before 2004 (from 2001 to 2003) and
,
refers to changes in the average value of the
control variables over the same periods.
We find a significant impact of the relative difference in short-selling restrictions on earnings
management for the 2005 to 2007 period, but not afterward. The cross-sectional regression indicates
that pilot firms subject to less restrictive regulation reduce their earnings management (relative to the
17 control firms) by as much as 6.57%. When the regulations on the pilot and control firms converge, the
difference between the two groups dissipates. This test and the previous tests (although different in
nature) generate the same conclusion—i.e., lifting short-selling restrictions reduces earnings
management. The random nature of the experiment makes it impossible for any spurious crosssectional correlation to dominate the negative correlation.
3.3 An Event-based Approach: the Hong Kong Experiment
We now consider the introduction of regulated short selling into the Hong Kong Stock Exchange. The
Hong Kong Stock Exchange provides a different experiment in which short selling was gradually
introduced in the market (see Chang, Cheng, and Yu, 2007). The most interesting feature of this
experiment is that the list of firms eligible for short-selling changes over time, which creates both
time-series and cross-sectional variation in terms of short-selling restrictions for firms listed in Hong
Kong. Stocks were added at the discretion of the regulator as a function of “changing market
conditions,” initially at irregular frequency and subsequently, after February 12, 2001, on a quarterly
basis according to a set of criteria primarily based on market capitalization, turnover, index
membership, and derivative contracts written on shares. Although these selection conditions make the
experiment less clean than the SHO experiment, the selection remains unlikely to create spurious
correlation because we explicitly control for the relevant variables. Moreover, the use of firm fixed
effects helps us reduce the effects of any other omitted firm-specific characteristics that may have led
to the introduction of short selling.
We therefore regress
,
on a dummy variable,
,
, which equals one if
stock is eligible for short selling in year t in a panel regression with firm and year fixed effects and
standard errors clustered at the firm level. We report the results in the first three models of Table 5.
The results indicate that short-selling eligibility reduces earnings management by 15.53%, which is
even stronger in terms of economic magnitude, compared to the SHO experiment. In a placebo test,
we further define
,
as a dummy variable that equals one if a stock is eligible for short
selling in year t-1 but becomes ineligible for short selling beginning in year t. We find that, once short
18 selling becomes unavailable again—even when it was feasible one year beforehand—the firm no
longer exhibits lower earnings management. This placebo test further confirms that short selling—not
persistent firm characteristics—reduces earnings management.
Next, in Models (4) and (5), we consider a difference-in-differences specification in which we
regress accrual changes on two variables, namely,
,
, which refers to net inclusion and equals
one (negative one) if a firm in included (excluded) in the eligible list, and Δ
,
, which
is a dummy variable for exclusion. We find that the inclusion (exclusion) of a firm in (from) the
eligible list reduces (increases) management by 16.23% (25.41%).16 These results are consistent with
the results from the US regulation.
We further verify that our results are robust when we exclude the observations of eligible firms for
the year prior to their inclusion to the short selling list and when we apply the tests to the propensity
score-matched sample (Table IA7). The first test aims to reduce the potential contamination when
firms can, to some extent, anticipate the inclusion of the firm in the eligible list (this contamination
works only against us in our main tests), and the second test confirms that disciplining effect can be
observed even among firms with close characteristics. Thus, although different in nature, the Hong
Kong experiment leads to a similar conclusion to that of the US experiment: short selling is important
in curbing the incentives for earnings management. By contrast, we do not find evidence supporting
alternative hypotheses that short selling does not affect or may further distort firm incentives.
4. Extensions and Additional Tests
4.1 The Disciplining Implications of Global Short-selling Regulations
The previous section shows that regulatory restrictions on short selling in the US and Hong Kong
affect earnings management. Given the importance of regulation in the global financial market, we
first explore whether the impact of short-selling regulation on earnings management can also be
observed in other markets.
16
We divide the coefficient of HK SS (
the magnitude.
) by the standard deviation of discretionary accruals in Hong Kong to obtain
19 We therefore construct a list of dummy variables that describe the ease of short selling at the
country level. We assign a value of one if short selling is legal (Legality), if short selling is feasible
(Feasibility), if put option trading is allowed (Put Option), and if short selling is feasible or if put
option trading is allowed (F or P). The difference between legality and feasibility is that the latter
requires not only short selling to be legal but also the existence of institutional infrastructures that
support short selling, including a low cost of short selling and the availability of market makers who
are willing to trade on short positions. Whenever we use these variables, we also further control for a
set of commonly used, country-specific variables in addition to firm characteristics. Additional details
on the country regulation variables and the additional control variables are provided in Table IA1.
To save space, we also tabulate the regressions in Table IA8 and only report the main results here.
Our tests reveal a strong negative correlation between market-level SSP and earnings management.
For example, in countries in which short selling is banned (unfeasible), earnings management appears
to be 11% (14%) higher than in countries in which short selling is legal. Thus, the impact of shortselling regulations on earnings management applies to the global market.
Next, we further conduct the market-wide SSP tests on the sample of ADR firms. This sample is
particularly interesting because of the nature of the ADR market. All ADR firms are exposed to the
US regulatory environment, which is known to promote firm value and corporate governance (e.g.,
Doidge, Karolyi, and Stulz, 2004, 2007; Karolyi, 2004, 2006). If US regulations are perfectly enforced,
the link between a firm’s earnings management incentives and its home-country short-selling
regulation (as we have just observed) should be completely suppressed. With the growing importance
of ADRs in the global market, it is important to determine whether concurrent US regulations suffice
to achieve this goal: different answers to this question may lead to drastically different policy
implications.
We therefore reestimate the panel regressions within the sample of ADR stocks. The results, also
tabulated in Table IA8, show that the negative correlation between market-level SSP and earnings
management is not reduced by the ability of the firms to bond to the US regulatory environment
20 through ADR listing. In other words, the disciplining impact of short selling in the global market is a
strong and distinct effect, enough to survive the additional regulatory requirements that the US market
may impose or any additional governance improvements that ADR firms might experience. This
surprising result adds to our understanding of the complexity of financial regulation in the global
market.
Apart from country-level bans, other forms of short-selling restrictions may appear even in
markets allowing for short selling. According to Jain et al. (2013), short selling restrictions in general
include specific trading mechanisms, pre-borrowing requirements (e.g., naked short selling), and bans
on shorting selected stocks. Due to the lack of data on naked short selling as well as the specialty of
bans on selected stocks (e.g., financial stocks), we follow Cumming, Johan, and Li (2011) and
consider one specific and one general form of trading rules. Also, we extend our cross-country
analysis to broader regulations that aim to improve the quality of information supplied to the market—
i.e., disclosure requirements and investor protections. We use the disclosure requirement index of La
Porta, Lopez-de-Silanes, and Shleifer (2006) to proxy for accounting regulations, and the anti-director
index of Pagano and Volpin (2005) to proxy for investor protections.
These results are reported in Table IA9. More specifically, we divide the whole sample into two
subsamples according to these disclosure and trading rules, and conduct our main tests in these
subsamples. Our main result is that markets with weaker accounting regulation and investor
protections are typically associated with higher impact of SSP, suggesting that more regulated
information disclosure/investor protection and SSP substitute each other in disciplining firms.
Furthermore, low trading restrictions may enhance both the disciplining impact and the price pressure
of SSP. Our empirical results document that the disciplining impact dominates. These cross-country
tests, therefore, complement the stock-level tests and reinforce our major conclusions.
4.2 Time-series and Subsample Analyses
In this section, we examine the main results that we presented in Table 2 by considering various
subsamples for robustness. We begin with the important observation made by Saffi and Sigurdsson
21 (2011) that short-selling activity grew tremendously from 2004 to 2008 (their testing period) because
of the boom of hedge funds who short sold stocks at a large scale. We would therefore expect that the
disciplining impact of short selling on earnings management also increases over time. Saffi and
Sigurdsson (2011) also illustrate that the short-selling market experienced drastic changes during the
global financial crisis. Finally, DataExplorers only provides monthly information for 2002 and 2003
(weekly information from July 2004 on and daily information after 2006) and within this period the
coverage is low, which may give rise to data quality concerns in the early years of the period
considered. These considerations lead us to explore the impact of SSP over time.
The results are presented in Panel A of Table 6. In Model (1), we consider the sample period
beginning in 2005 (inclusive) (“≥2005”). The results indicate that the exclusion of earlier years does
not change the disciplining impact of SSP. This result is expected, as the inclusion of earlier years in
our main test would only work against our hypothesis if the sample were less reliable. In addition, we
also find that, as reported in Table IA10, the disciplining role in general applies to both the crisis and
the non-crisis periods.
In Model (2), we include the interaction between lendable shares and a time-sequence variable T,
which equals the year minus 2001, in the baseline regression. We find that this interaction negatively
and significantly correlates with earnings management, suggesting that the impact of SSP on earnings
management increases over time. A robustness check is also provided in Table IA 10, in which we
decompose the impact of lendable shares in various subperiods, and find that the impact of Lendable is
significant in all subperiods and that its magnitude increases over time.
Next, we consider different regions/subsamples and report the results in Models (3) and (4). The
samples “US” and “NUS” refer to firms from the US and non-US countries, respectively. We again
find that the link between SSP and earnings management holds across different subsamples. It is
interesting to note that the disciplining impact is relatively stronger in the US, suggesting that the real
impact of the short-selling market in the US is more prominent. This result can likely be explained by
22 its well-developed institutional infrastructures to support short selling in the US. Nonetheless, the
disciplining impact of short selling applies to the global market and is not limited to the US alone.17
Finally, we investigate whether the impact of short selling concentrates in firms with aggressive
(i.e., inflated) earnings, as opposed to conservative (i.e., deflated) earnings, and whether short selling
disciplines small stocks for which investors have less information to a greater extent. Therefore, in
Panel B, we divide the full sample into the accrual and firm size subsamples. “AccrualMJones≥0” refers
to firms with positive AccrualMJones as a proxy for aggressive earnings, whereas “AccrualMJones<0”
refers to firms with negative AccrualMJones as a proxy for conservative earnings. “Small firms” and
“large firms” are firms with market capitalization below and above the median value of market
capitalization for each country and year, respectively. We find that short selling affects only firms with
aggressive earnings, not those with conservative earnings; this result is exactly what should be
expected from the disciplining effect. Also, the disciplining effect applies to both large and small firms,
but its impact on the latter is greater in magnitude. This observation is consistent with the notion that
the disciplining impact of short selling, which arises from the additional informativeness introduced by
the short-selling market, should be more pronounced for the sector of the market that has less public
information.
4.3 Robustness Checks
Next, we consider a series of robustness checks to determine whether the disciplining effect is robust
to the use of alternative SSP proxies and different earnings management measures, including not only
various types of discretionary accruals but also a list of target-beating, earnings persistence, and
earnings misstatement measures. We further examine whether the disciplining effect is robust to
alternative discipline channels and clustering specifications.
Panel A of Table 7 considers alternative SSP measures. These variables are more endogenous than
Lendable in describing the ex ante impact of short selling; hence, we only consider them as a
robustness check. The first variable, On Loan, is the annual average fraction of the shares of a firm
17
We also examine the impact of short selling on earnings management in subsamples of countries sorted by the feasibility of
short selling and by controlling for firm-level investments. The results are tabulated in Table IA10.
23 that are lent out (or short interest). A high level of realized historical short interest confirms a high
level of short seller attention, which should discipline the earnings management incentives of firms in
the future.
Next, a high short-selling fee implies lower SSP, which should be associated with a less effective
disciplining impact. We can assess this potential association by using three variables: Fee is the annual
value-weighted average loan fee expressed as a percentage; STDFee is the standard deviation of
monthly value-weighted average loan fee expressed as a percentage; and Specialness is a dummy
variable that proxies for very high short-selling costs and equals one if the average loan fee is above
one percent.
The problem with relying on the short-selling fee is that it can be substantially affected by the
demand side of the short-selling market (Cohen, Diether, and Malloy 2007; Kolasinski, Reed, and
Ringgenberg, 2013)—and hence is less exogenous. To overcome this issue, we also define a new
variable: Constraint, which is a dummy variable that equals one if the stock's average number of
shares on loan is in the bottom quartile and its average loan fee is in the top quartile (quartiles are
sorted by country and year). That is, when Constraint takes a value of one, the stock will be very
difficult to short sell because of its limited supply and high cost—and we expect earnings management
to increase as a consequence. Constraint is related to the variable for supply shocks in Cohen, Diether,
and Malloy (2007). The difference between these variables is that supply shocks are defined by using
changes in short interest and fees, while Constraint is related to the levels of short interest and fees.
We focus on Constraint because the first-order impact of disciplining should be related to the level of
short selling, though unreported tests using supply shocks lead to a similar conclusion.
For all these variables, we find a strong negative correlation between SSP and earnings
management. A one-standard-deviation increase in SSP is associated with 2.32% (2.11%, 2.64%,
2.42%, and 1.19%) less earnings management in the case of On Loan (Fee, STDFee, Specialness, and
Constraint, respectively). Thus, fewer restrictions in the short-selling market are associated with lower
earnings management incentives.
24 Next, we consider a set of additional types of earnings management that are widely used in the
literature to proxy for managerial distortion. The first type concerns “target-beating measures” (e.g.,
Burgstahler and Dichev, 1997, Degeorge, Patel, and Zeckhauser, 1999), which capture the incentives
for managers to avoid reporting small losses relative to their heuristic target of zero. Such incentives
lead to a well-known “kink” in the distribution of reported earnings near zero—i.e., a statistically
small number of firms with small losses and a statistically large number of firms with small profits
(e.g., Burgstahler and Dichev, 1997). This type of earnings management is especially powerful to
complement our existing tests of the price pressure hypothesis, as the concerns that the downward
price pressure of short selling may amplify the negative impact of not meeting analyst or market
expectations could incentivize firms to engage in more target-beating actions. Therefore, we directly
examine whether short selling can still discipline this type of earnings management incentive.
We use three proxies to capture such distortion. The first proxy is target beating on small positive
forecasting profits (SPAF), based on Degeorge, Patel, and Zeckhauser (1999). This variable is a
dummy that equals one if the difference between reported earnings per share and forecasted earnings
per share scaled by stock price is between 0 and 1%. The variable captures managers’ incentives to
meet or beat analyst forecasts by a small margin. The second proxy is target beating on small positive
past-earnings profits (SPDE) based on Burgstahler and Dichev (1997). This variable is a dummy that
equals one if the change in net income scaled by lagged total assets is between 0 and 1%. The third
proxy is target beating on small positive profits (SPE). Based on Burgstahler and Dichev (1997), this
variable is a dummy that equals one if net income scaled by lagged total assets is between 0 and 1%.
The last two variables proxy for managers’ incentives to meet or beat market expectations by a small
margin, where market expectations are measured by the previous year’s earnings or a general request
for firms to not report losses.
Models (1) to (3) in Panel B test the impact of our main proxy for SSP on these alternative
earnings management measures. We find that the presence of SSP reduces the incentives to beat
analyst or market expectations across all three measures of target-beating behavior. Thus, SSP not
25 only disciplines the incentives to inflate earnings but also exerts a similar impact on the incentives to
meet or beat market expectations. In other words, the disciplining effect dominates the potential
concerns of downward price pressure in the real corporate world.
The second alternative earnings management proxy is earnings persistence. As Dechow, Ge, and
Schrand (2010) have shown, pretending to be capable of generating “sustainable” earnings is another
motivation for a firm to engage in earnings management (in addition to the desire to inflate earnings
captured by our accruals variable) because superior business fundamentals may lead to sustainable
earnings. By contrast, in the absence of earnings management, earnings will be less stable for all firms
except for perhaps the very best group of firms in the economy. Although short selling should not
affect firms with truly superior fundamentals, it reduces the incentives for bad firms to mimic good
firms by manipulating earnings sustainability. We therefore expect that SSP will reduce earnings
persistence.
Models (4) and (5) in Panel B test this effect by regressing earnings (operating income scaled by
lagged total assets) or modified Jones’ (1991) residual accruals on the interaction between SSP and the
lagged dependent variable. The interaction terms are significantly negative for both variables.
Therefore, lendable shares reduce both earnings and accrual persistence.
Panel C examines three alternative accrual measures, namely, Jones's (1991) residual accruals,
Kothari, Leone, and Wasley's (2005) residual accruals, and Dechow and Dichev's (2002) residual
accruals, as well as the probability that firms are involved in earnings misstatements or corporate
scandals. Kothari, Leone, and Wasley's (2005) model further controls for firm fundamentals by
matching a firm with another from the same country, industry, and year with the closest ROA; and
Dechow and Dichev's (2002) further controls for operating performance by regressing results on past,
current, and future cash flows. SSP disciplines all these alternative accrual and misbehavior
variables.18 These results, together with the test on market-beating expectations, demonstrate that short
18
Our results are also robust to other accrual variables, such as Francis et al.'s (2005) and Allent, Larson, and Sloan's (2013)
residual accruals. To save space, these results are tabulated in Table IA12.
26 selling disciplines managerial incentives to manipulate accruals and apply other forms of earnings
management.
We also consider the effect of alternative discipline channels based on the quality of a firm’s
corporate governance and accounting standards, including the quality of the firm’s auditors, the
quality of the firm’s accounting standards, the quality of the firm’s corporate governance (as defined
by the ISS index), the transparency of the firm (dispersion of analysts), and press coverage by news
agencies. We use the following variables: the ISS corporate governance index (ISS), big N auditor
(BigN), international accounting standard (IAS), news coverage (NewsCoverage), and analyst
dispersion (Disp). A higher value for any of these variables typically indicates better governance,
except for Disp, for which a lower value helps mitigate bad managerial incentives.
All of these variables provide alternative means of disciplining managers or improving the ability
of the market to obtain information about them. For example, the quality of governance has been used
by Doidge, Karolyi, and Stulz (2007) and represents the standard governance metric based on the bylaws and statutes of the firm. Additionally, transparency—through improved accounting standards,
better auditors, or a lower dispersion of their forecasts—improves the awareness of uninformed
shareholders. In Panel D, we separately control for these variables because the addition of these
alternative controls drastically reduces the size of the sample. The results are qualitatively and
quantitatively similar to the main results.
An alternative interpretation of Panel D is that these alternative disciplining variables could be
spuriously related to short selling. For instance, large firms may have both more lendable shares and
news coverage. Controlling for these alternative variables helps to reduce the impact of spurious
correlation.
4.4 Earnings Management and Stock Price Synchronicity
Finally, we link to Saffi and Sigurdsson (2011) and explore the extent to which short selling can
increase the informativeness of the stock price specifically through its impact on earnings management.
Specifically, we follow Morck, Yeung, and Yu (2000) and Jin and Myers (2006) in spirit, and Bris,
27 Goetzmann, and Zhu (2007) in particular, to construct a proxy for firm-specific information based on
the idiosyncratic risk of the stock. The measure, downside-minus-upside R2 (R2DMU) , is constructed as
the difference between a firm’s downside R2 and its upside R2, where downside (upside) R2 is
estimated by regressing weekly individual stock on weekly positive (negative) local and US market
returns.
As indicated by Bris, Goetzmann, and Zhu (2007), short selling restrictions will reduce firmspecific information and thus price efficiency, especially during the downside of the market (compared
to the upside of the market), leading to a higher value of R2DMU. As a robustness check, we also
construct a second proxy, downside-minus-upside nonsynchronicity (NonsynDMU), where downside
(upside) nonsynchronicity is the logarithm of (1-downside (upside) R2) divided by downside (upside)
R2. Since nonsynchronicity is high when firm-specific information is abundant, a higher value of
NonsynDMU implies a higher degree of stock price informativeness.
In Models (1) and (3) of Table 8, proxies for stock price informativeness are regressed on
Lendable, firm-level control variables, and the unreported industry, country, and year fixed effects.
The results confirm the finding of Saffi and Sigurdsson (2011) that short selling is in general
associated with more stock price informativeness.
In Models (2) and (4), we further interact Lendable with its potential degree of disciplining impact
based on the finding (reported in Panel C of Table 6) that this impact is the highest among small-cap
stocks that have positive AccrualMJones. More specifically, we define a dummy variable, SSPImpact,
which takes a value of one for small-cap stocks with positive AccrualMJones. We find that the
interaction between Lendable and SSPImpact greatly enhances the positive relationship between SSP and
stock price informativeness, confirming that greater stock price efficiency can be achieved when the
disciplining effect of SSP on earnings management is more pronounced. This analysis completes our
analyses regarding the disciplining role of short selling in reducing managers’ incentives to engage in
earnings management.
28 These results are important. Thus far, we have shown that SSP reduces earnings management.
Table 8 further suggests that short selling increases the informativeness of the stock price by reducing
earnings management. This finding is consistent with existing evidence (e.g., Saffi and Sigurdsson,
2011) indicating that short selling improves price efficiency. However, the channel is different: price
efficiency is enhanced not by improved market conditions but by lower earnings management by firms.
Conclusion
In this paper, we study whether the potential for short selling has a disciplining impact on earnings
management incentives. We argue that short selling affects the behavior and incentives of managers
because its presence can accelerate the pace at which information is incorporated into the market and
thus allows the market to uncover potential earnings management with a higher probability and at a
higher speed. Thus, we expect SSP—the maximum potential impact that short selling may have on
firm behavior—to significantly reduce firms’ incentives to engage in earnings management.
Alternatively, firms may simply ignore the short-selling market or manipulate earnings to a greater
extent when they are concerned about the downward price pressure that may be associated with
potential short selling.
We test these hypotheses by using data on worldwide short selling detailed at the stock level for
the period from 2002 to 2009. Our results show a strong negative correlation between SSP and
earnings management that is statistically significant and economically relevant. Endogeneity tests
based on instrumental variables (ETF ownership) and two experiments (the SHO experiment in the US
and the introduction of short selling into the Hong Kong stock market) inform a causal interpretation
of this negative relationship that SSP reduces earnings management. We show that the disciplining
effect of short selling applies to various types of earnings management, and our results are robust to
the use of alternative proxies for SSP. Moreover, alternative disciplining channels do not absorb the
power of short selling.
These results confirm the disciplining hypothesis and offer evidence of the beneficial effects of the
short-selling market on the corporate market. In this regard, short selling not only contributes to the
29 efficiency of the information environment of the stock market but also may improve the contracting
institutions of the real economy.
30 References
Admati, A.R., P. Pfleiderer, 2009, The Wall Street Walk and Shareholder Activism: Exit as a Form of Voice,
Review of Financial Studies 22, 2645–2685.
Aggarwal, R., I. Erel, M. Ferreira, and P. Matos, 2011, Does Governance Travel around the World? Evidence
from Institutional Investors, Journal of Financial Economics 100, 154-181.
Aggarwal, R., P. A. C. Saffi, and J. Sturgess, 2013, The Role of Institutional Investors in Voting: Evidence from
Changes in Lendable, Working Paper.
Akbas, F., E. Boehmer, B. Erturk, and S. Sorescu, 2013, Short Interest, Returns, and Fundamentals, Working
Paper.
Allent, E., C. Larson, R. Sloan, 2013, Accrual Reversals, Earnings and Stock Returns, Journal of Accounting
Economics 56, 113-129.
Amihud, Y., 2002, Illiquidity and stock returns: cross-section and time-series effects, Journal of Financial
Markets 5, 31–56.
Aitken, M., A. Frino, M. McCorry, and P. Swan, 1998, Short sales are almost instantaneously bad news:
Evidence from the Australian Stock Exchange, Journal of Finance 53, 2205-2223.
Arellano, M. and S. Bond, 1991, Some tests of specification for panel data: Monte Carlo evidence and an
application to employment equations, Review of Economic Studies 58, 277-297.
Asquith, P., and L. Meulbroek, 1995, An empirical investigation of short interest, Working Paper, M.I.T.
Barth, M., Landsman, W., and Lang, M., 2008, International Accounting Standards and Accounting Quality,
Journal of Accounting Research 46, 467-498.
Boehmer, E., C. Jones and X. Zhang, 2008, Which Shorts are Informed? Journal of Finance 63, 491-527.
Boehmer, E. and J. Wu, 2013, Short selling and the price discovery process, Review of Financial Studies 26,
2013, 287-322.
Bris, A., W. Goetzmann, and N. Zhu, 2007, Efficiency and the Bear: Short Sales and Markets Around the World,
Journal of Finance 62, 1029-1079.
Burgstahler, D., and I. Dichev, 1997, Earnings management to avoid earnings decreases and losses, Journal of
Accounting and Economics 24, 99–126.
Chang, E. C., J.W. Cheng, and Y. Yu, 2007, Short-Sales Constraints and Price Discovery: Evidence from the
Hong Kong market, Journal of Finance 62, 2097-2121.
Christoffersen, S.K., C. Geczy, D. K. Musto, and A. V. Reed, 2007, Vote Trading and Information
Aggregation, Journal of Finance 62, 2897–2929.
Cohen, L., K. Diether, and C. Malloy, 2007, Supply and Demand Shifts in the Shorting Market, Journal of
Finance 62, 2061-2096.
Conley, T. G., C. B. Hansen, and P. E. Rossi, 2012, Plausibly Exogenous, Review of Economics and Statistics
94, 260-272.
Cumming, D.J., S.A. Johan, and D. Li, 2011, Exchange Trading Rules and Stock Market Liquidity, Journal of
Financial Economics 99, 651-671.
Das, S., and H., Zhang, 2003, Rounding-up in reported EPS, behavioral thresholds, and earnings management,
Journal of Accounting and Economics 35, 31-50.
DeAngelo, L., 1981, Auditor independence, ‘low balling’, and disclosure regulation, Journal of Accounting and
Economics 3, 113-127.
Dechow, P., and I. Dichev, 2002. The quality of accruals and earnings: The role of accrual estimation errors.
The Accounting Review 77, 35-59.
Dechow, P., W. Ge, and C. Schrand, 2010, Understanding Earnings Quality: A Review of the Proxies, their
Determinants and their Consequences, Journal of Accounting and Economics 50, 344-401.
31 Dechow, P., R. Sloan, and A. Sweeney, 1995, Detecting earnings management. The Accounting Review 70,
193-225.
DeFond, M., Park, C., 1997, Smoothing income in anticipation of future earnings, Journal of Accounting and
Economics 23, 115-139.
Degeorge, F., J., Patel, and R. Zeckhauser, 1999, Earnings management to exceed thresholds, Journal of
Business 72, 1–33.
Diamond, D. W., and R. E. Verrecchia, 1987, Constraints on short-selling and asset price adjustment to private
information, Journal of Financial Economics 18, 277–311.
Diether, K. L, K. H. Lee, and I.M. Werner, 2009, It is SHO time! Short-sale price tests and market quality. The
Journal of Finance, 64, 37-63.
Doidge, C., G. A. Karolyi, and R. M. Stulz, 2004, Why are foreign firms listed in the U.S. worth more? Journal
of Financial Economics 71, 205-238.
Doidge, C., G. A. Karolyi, and R. M. Stulz, 2007, Why Do Countries Matter So Much for Corporate
Governance?, Journal of Financial Economics 86, 1-39.
Duffie, D., N. Garleanu, and L. H. Pedersen, 2002, Securities lending, shorting, and pricing, Journal of
Financial Economics 66, 307–339.
Dyck, A., A. Morse, and L. Zingales, 2012, Who Blows the Whistle on Corporate Fraud? Journal of Finance,
forthcoming.
Edmans, A., 2009, Blockholder Trading, Market Efficiency, and Managerial Myopia, Journal of Finance, 64,
2481–2513.
Edmans, A., Fang, V. W., Zur, E., 2013. The effect of liquidity on governance. Review of Financial Studies 26,
1443–1482.
Edmans, A., and G. Manso, 2011, Governance Through Trading and Intervention: A Theory of Multiple
Blockholders, Review of Financial Studies 24, 2395–2428.
Fang, V., A. Huang, and J. Karpoff, 2014. Short selling and earnings management: A controlled experiment.
Working Paper.
Faure-Grimaud, A., and D. Gromb, 2004, Public Trading and Private Incentives, Review of Financial Studies 17,
985–1014.
Ferreira, M., and P. Matos, 2008, The Colors of Investors’ Money: The Role of Institutional Investors Around
the World? Journal of Financial Economics 88, 499-533.
Francis, J., LaFond, R., Olsson, P., Schipper, K., 2005, The market pricing of accruals quality, Journal of
Accounting and Economics 39, 295-327.
Francis, J., Wang, D., 2008, The joint effect of investor protection and Big 4 audits on earnings quality around
the world. Contemporary Accounting Research 25, 157–191.
Francis, J., P. Michas, and S. Seavey, 2013, Does Market Concentration Harm the Quality of Audited Earnings:
Evidence from Audit Markets in 42 Countries, Contemporary Accounting Research 30, 325-355.
Grullon, G., S. Michenaud, and J. P. Weston, 2012, The Real Effects of Short-Selling Constraints, Working
Paper.
Hirshleifer, D., S.H. Teoh, and J. Yu, 2011, Short Arbitrage, Return Asymmetry and the Accrual Anomaly,
Review of Financial Studies 24, 2429-2461.
Jain, A., P. Jain, T. McInish, and M. McKenzie, 2013, Worldwide Reach of Short Selling Regulations, Journal
of Financial Economics 109, 177-197.
Jin, L. and S. Myers, 2006, R-squared Around the World: New Theory and New Tests, Journal of Financial
Economics 79, 257-92.
Jones, J., 1991, Earnings management during import relief investigations, Journal of Accounting Research 29,
193-228.
32 Kahn, C., and A. Winton, 1998, Ownership Structure, Speculation, and Shareholder Intervention, Journal of
Finance 53, 99–129.
Karolyi, G. A., 2004, The Role of American Depositary Receipts in the Development of Emerging Equity
Markets, Review of Economics and Statistics 86, 670–690.
Karolyi, G. A., 2006, The World of Cross-listings and Cross-listings of the World: Challenging Conventional
Wisdom, Review of Finance 10, 99–152.
Karolyi, G. A., Lee, K., and Van Dijk, M. A., 2012, Understanding Commonality In Liquidity Around The World,
Journal of Financial Economics 105, 82-112.
Karpoff, J.M., and X. Lou, 2010, Short sellers and Financial Misconduct, Journal of Finance 65, 1879-1913.
Kothari, S., Leone, A., and Wasley, C., 2005, Performance matched discretionary accrual measures, Journal of
Accounting and Economics 39, 163-197.
Kolasinski, A.C., A. V. Reed, and M. C. Ringgenberg, 2013, A Multiple Lender Approach to Understanding
Supply and Search in The Equity Lending Market, Journal of Finance 68, 559-595.
La Porta, R., F. Lopez-de-Silanes, and A.Shleifer, 2006. What Works in Securities Laws? Journal of Finance 61,
1-32.
Leuz, C., D. Nanda, and P. D. Wysocki, 2003, Earnings management and investor protection: an international
comparison, Journal of Financial Economics 69, 505–527.
Li, K., Ortiz-Molina, H., Zhao, X., 2008. Do voting rights affect institutional investment decisions? Evidence
from dual-class firms. Financial Management 37, 713-745.
Ljungqvist, A., Qian, W., 2014. How constraining are limits to arbitrage? Working Paper.
Maug, E., 1998, Large Shareholders as Monitors: Is There a Tradeoff Between Liquidity and Control?, Journal
of Finance 53, 65–98.
Miller, E. M., 1977, Risk, Uncertainty, and Divergence of Opinion, Journal of Finance 32, 1151–68.
Morck, R., B. Yeung, and W. Yu, 2000, The Information Content of Stock Markets: Why Do Emerging Markets
Have Synchronous Stock Price Movements? Journal of Financial Economics 59, 215–260.
Morsfield, S., Tan, C., 2006, Do venture capitalists influence the decision to manage earnings in initial public
offerings? The Accounting Review 81, 1119-1150.
Nissim, D., Penman, S., 2001, Ratio analysis and equity valuation: From research to practice, Review of
Accounting Studies 6, 109-154.
Pagano, M., P. F. Volpin, 2005, The Political Economy of Corporate Governance, American Economic Review
95, 1005–1030.
Petersen, M. A., 2009, Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches, Review
of Financial Studies 22, 435–480.
Prado, M.P., P. A. C. Saffi, and J. Sturgess, 2013, Ownership Structure, Limits to Arbitrage and Stock Returns:
Evidence from Equity Lending Markets, Working Paper.
Saffi, P., Sigurdsson, K., 2011. Price Efficiency and Short Selling, Review of Financial Studies, 24, 821-852.
Senchack, A.J., Starks, L., 1993, Short-sale restrictions and market reaction to short-interest
announcements, Journal of Financial and Quantitative Analysis 28, 177-194.
Staiger, D., and J. H., Stock, 1997, Instrumental Variables Regression with Weak Instruments, Econometrica 65,
557-586.
Thompson, S., 2010, Simple Formulas for Standard Errors That Cluster by Both Firm and Time, Journal of
Financial Economics 99, 1–10.
Watts, R., Zimmerman, J., 1986, Positive accounting theory, Prentice-Hall Inc.
Yu, F., 2008, Analyst coverage and earnings management, Journal of Financial Economics 88, 245–271.
33 Appendix A: Variable Definitions
Variable
Acronym
Definition
Data Source
A. Short selling variables
Lendable shares
Shares on loan
Loan fee
Loan fee volatility
Specialness
Short-selling constraint
Lendable
On Loan
Fee
STDFee
Specialness
Constraint
Dataexplorers
Dataexplorers
Dataexplorers
Dataexplorers
Dataexplorers
Dataexplorers
US SHO pilot stock
US SHO
HK short-selling list
HK SS
Legality of short selling
Short-selling ban
Legality
SSBan
ETF ownership
Industry-level ETF ownership
ETF
ETFIndustry
Annual average fraction of shares of the firm available to lend
Annual average fraction of shares of the firm lent out
Annual value-weighted average loan fee as a percentage
Standard deviation of monthly value-weighted average loan fee as a percentage
A dummy variable that equals one if average loan fee is above one percent
A dummy variable that equals one if the stock's average shares on loan is in the bottom
quartile and its average loan fee is in the top quartile; quartiles are sorted by country and year
A dummy variable that equals one if the US stock is included in the pilot program under
US Regulation SHO
A dummy variable that equals one if the Hong Kong stock is eligible for short selling
on the Hong Kong Stock Exchange
A dummy variable that equals one if short selling is legally allowed in a country
A dummy variable that equals one if a short selling ban is imposed on the stock during the
global financial crisis
Annual average holdings by ETF as a percentage of the total number of outstanding shares
Annual average holdings by ETF as a percentage of the total number of outstanding shares
in an industry
Based on Dechow, Sloan, and Sweeney's (1995) modification of Jones's (1991) model,
residual accruals are obtained by regressing accruals on revenue growth excluding growth in
credit sales and fixed assets for each country and year; all numbers are scaled by lagged
total assets
Based on Jones's (1991) model, residual accruals are obtained by regressing accruals on
revenue growth and fixed assets for each country and year; all values are scaled by
lagged total assets
Matches a firm with another from the same country, industry, and year with the closest ROA;
calculate the difference between the firm's modified Jones's (1991) residual accruals and
that of the matched firm
Worldscope
B. Earnings management variables
Modified Jones's (1991) residual
AccrualMJones
accruals
Jones's (1991) residual accruals
AccrualJones
KLW's (2005) residual accruals
AccrualKLW
34
US S.E.C.
H.K.S.E.
Charoenrook and Daouk (2005)
Beber and Pagano (2011)
FactSet
FactSet
Worldscope
Worldscope
Appendix A: Variable Definitions - Continued
Variable
Acronym
Definition
Data Source
DD's (2002) residual accruals
AccrualDD
Worldscope
Small positive forecasting profits
SPAF
Small positive past-earnings profits
SPDE
Small positive profits
SPE
Based on Dechow and Dichev's (2002) model, residual accruals are obtained by regressing
changes in working capital on past, current, and future cash flows for each country and year;
all values are scaled by lagged total assets
A dummy variable that equals one if (reported earnings per share-forecasted earnings
per share)/price is between zero and 1%
A dummy variable that equals one if the change in net income scaled by lagged total assets is
between zero and 1%
A dummy variable that equals one if net income scaled by lagged total assets is between
zero and 1%
C. Control variables
Firm size
Book-to-market ratio
Financial leverage
Annual stock return
Stock return volatility
American Depository Receipts
MSCI country index membership
Size
BM
Leverage
Return
STD
ADR
MSCI
Log of market capitalization denominated in US $.
Log of book-to-market equity ratio
Ratio of total debt to total assets
Log of annual stock return
Annualized standard deviation of monthly stock returns
An ADR dummy equals one if the firm was cross-listed on a US stock exchange
An MSCI index member dummy that equals one if the firm is included in an MSCI
country index and zero otherwise
Datastream
Datastream
Worldscope
Datastream
Datastream
Multiple sources**
Datastream
Number of analysts following the
firm
Closely held ownership
Institutional ownership
Analyst
CH
IO
IBES
Worldscope
Worldscope
Number of financial analysts following a firm
IBES
Fraction of shares closely held by insiders and controlling shareholders
Worldscope
Aggregate equity holdings by institutional investors as a percentage of the total number
FactSet
of outstanding shares
Amihud's (2002) illiquidity
Illiquidity
Log of the average of daily Amihud's (2002) measure calculated as the absolute value
Datastream
of stock return divided by dollar trading volume on a given day
** The information on US cross-listings is gathered from three data sources: depository banks (such as the Bank of New York), US stock exchanges, and Datastream.
35
Appendix A: Variable Definitions - Continued
Variable
Acronym
Definition
Data Source
D. Other variables
ISS corporate governance index
Big N auditor
International accounting standard
News coverage
Analyst dispersion
Earnings misstatement
ISS
BigN
IAS
NewsCoverage
Disp
Misstatement
ISS
Compustat & Worldscope
Compustat & Worldscope
RavenPack
IBES
RavenPack
Fraud and Scandal
Scandal
Earnings
Downside-minus-upside R2
Earnings
R2DMU
Downside-minus-upside
nonsynchronicity
NonsynDMU
Firm-level corporate governance index
A dummy variable that equals one if the firm is audited by any of the Big 4 or Big 5 auditors
A dummy variable that equals one if the firm adopts international accounting standards
Log of one plus the number of news releases recorded in Dow Jones Newswire
Standard deviation of analyst forecasts scaled by stock price
A dummy variable that equals one if the firm has been reported to have an earnings
misstatement
A dummy variable that equals one if the firm has been reported to have executive scandal,
fraud, corruption, or embezzlement events
Operating income scaled by lagged total assets
Difference between a firm’s downside R2 and its upside R2, where downside (upside) R2 is
estimated by regressing weekly individual stock on weekly positive (negative) local and US
market returns
Difference between a firm’s downside nonsynchronicity and upside nonsynchronicity, where
downside (upside) nonsynchronicity = Log of (1-downside (upside) R2) divided by
downside (upside) R2
36
RavenPack
Worldscope
Datastream
Datastream
Appendix B: Number of Stocks by Country and Year
This table summarizes the number of stocks in our sample for each country over the 2002- 2009 sample period. The first column reports the name of the country.
The second column indicates whether a country is a developed country (DEV) or an emerging market (EMG). The column “N” reports the total number of stocks
across all sample periods for each country. The remaining columns report the number of stocks in each year.
Country
Australia
Austria
Belgium
Brazil
Canada
Denmark
Finland
France
Germany
Greece
Hong Kong
Indonesia
Ireland
Israel
Italy
Japan
Mexico
Netherlands
New Zealand
Norway
Philippines
Poland
Portugal
Singapore
South Africa
South Korea
Spain
Sweden
Switzerland
Taiwan
Turkey
United Kingdom
United States
All
DEV/EMG
DEV
DEV
DEV
EMG
DEV
DEV
DEV
DEV
DEV
EMG
DEV
EMG
DEV
EMG
DEV
DEV
EMG
DEV
DEV
DEV
EMG
EMG
EMG
DEV
EMG
EMG
DEV
DEV
DEV
EMG
EMG
DEV
DEV
Total
N
1,148
66
110
109
1,158
127
109
583
606
63
544
38
54
57
314
2,776
71
134
62
186
24
31
39
303
199
509
146
290
259
234
97
1,536
5,573
17,555
2002
170
19
27
2003
268
27
40
2004
334
31
53
179
21
34
190
137
2
86
8
22
1
101
1,489
19
59
12
28
4
238
31
47
236
169
22
119
7
23
10
131
1,600
32
73
19
44
6
351
45
64
251
240
3
166
12
28
15
161
1,793
33
79
25
59
8
12
51
48
30
60
64
84
17
6
657
1,193
4,830
14
63
64
67
69
105
127
25
6
690
3,552
7,924
16
90
70
105
86
128
159
52
11
680
3,774
8,922
37
2005
389
39
66
2
585
67
67
304
361
4
195
18
28
19
199
2,003
38
93
29
83
8
7
24
105
89
144
91
148
180
58
23
815
4,039
10,320
2006
557
45
79
11
722
94
85
387
385
33
260
24
32
18
220
2,195
43
101
29
99
9
11
29
142
128
332
105
198
192
51
39
911
4,073
11,639
2007
856
50
93
53
836
108
95
455
459
35
400
20
44
36
240
2,333
52
107
43
121
17
2
30
219
143
420
111
232
207
76
69
949
4,101
13,012
2008
819
54
94
91
826
102
94
437
429
44
430
23
40
44
256
2,340
58
97
40
129
15
17
30
240
139
422
118
224
211
145
81
875
4,118
13,082
2009
475
51
85
72
720
69
80
335
357
43
388
11
36
47
235
2,152
59
85
45
107
10
28
33
176
136
410
114
207
208
215
83
706
4,031
11,809
Table 1: Summary Statistics
This table presents the summary statistics and Spearman (Pearson) correlation coefficients of the main variables that are
used in this study. The variables are modified Jones's (1991) residual accruals (AccrualMJones), lendable shares (Lendable),
log of firm size (Size), log of book-to-market ratio (BM), financial leverage (Leverage), log of annual stock return (Return),
stock return volatility (STD), American Depository Receipts (ADR), MSCI country index membership (MSCI), number of
analysts following the firm (Analyst), closely held ownership (CH), institutional ownership (IO), and Amihud's (2002)
illiquidity (Illiquidity). All of the variables are defined in Appendix A. Panel A reports the number of observations (N) and
the mean, median, standard deviation (STD), and decile (10% and 90%) and quartile (25% and 75%) distributions of the
variables. Panel B reports the correlation coefficients among the above variables, where the highlighted upper-right part
(bottom-left part) of the table refers to the Spearman (Pearson) correlation matrix. The sample period is between 2002 and
2009.
Panel A: Summary Statistics
Variable
N
Mean
STD
10%
25%
Median
75%
90%
AccrualMJones
61,624
0.002
0.081
-0.081
-0.034
0.003
0.037
0.080
Lendable
61,624
0.065
0.094
0.000
0.004
0.023
0.086
0.211
Size
61,624
12.939
1.821
10.725
11.673
12.803
14.108
15.369
BM
61,624
-0.622
0.879
-1.676
-1.133
-0.593
-0.067
0.419
Leverage
61,624
0.198
0.180
0.000
0.024
0.170
0.320
0.450
Return
61,624
0.023
0.649
-0.752
-0.241
0.094
0.373
0.678
STD
61,624
0.448
0.316
0.189
0.258
0.372
0.548
0.776
ADR
61,624
0.036
0.185
0.000
0.000
0.000
0.000
0.000
MSCI
61,624
0.663
0.473
0.000
0.000
1.000
1.000
1.000
Analyst
61,624
5.053
5.998
0.000
1.000
2.833
7.500
13.417
CH
61,624
0.311
0.241
0.002
0.109
0.285
0.491
0.653
IO
61,624
0.248
0.299
0.000
0.020
0.110
0.379
0.785
Illiquidity
61,624
-3.430
2.965
-7.353
-5.641
-3.371
-1.284
0.407
Panel B: Correlation Coefficients (Spearman for the upper-right part, highlighted; Pearson for the bottom-left part)
Variable
AccrualMJones
AccrualMJones
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
-
-0.050
0.024
0.002
0.015
-0.001
-0.042
-0.053
-0.011
-0.014
-0.029
0.051
-0.019
Lendable
-0.030
-
0.501
-0.170
0.013
-0.026
-0.135
0.074
0.344
0.577
-0.219
0.552
-0.538
Size
0.026
0.319
-
-0.319
0.157
0.143
-0.341
0.161
0.598
0.732
-0.066
0.383
-0.864
BM
0.015
-0.114
-0.294
-
0.085
-0.204
-0.080
-0.042
-0.103
-0.285
0.097
-0.215
0.258
Leverage
0.026
0.008
0.122
0.016
-
-0.001
-0.099
0.022
0.114
0.095
0.007
-0.017
-0.154
Return
0.023
-0.028
0.146
-0.199
-0.023
-
-0.002
0.005
0.077
-0.007
0.035
0.014
-0.034
STD
-0.040
-0.110
-0.309
-0.097
-0.055
0.068
-
-0.021
-0.168
-0.178
-0.039
-0.072
0.243
ADR
-0.046
0.053
0.193
-0.036
0.017
0.003
-0.019
-
0.062
0.143
-0.063
0.001
-0.116
MSCI
-0.004
0.246
0.559
-0.090
0.099
0.082
-0.166
0.062
-
0.430
-0.027
0.298
-0.625
Analyst
-0.021
0.348
0.728
-0.222
0.064
-0.013
-0.174
0.196
0.344
-
-0.112
0.443
-0.693
CH
-0.016
-0.255
-0.084
0.073
0.007
0.039
-0.038
-0.055
-0.049
-0.134
-
-0.191
0.209
IO
0.037
0.493
0.352
-0.207
-0.005
-0.007
-0.070
-0.048
0.281
0.355
-0.275
-
-0.435
Illiquidity
-0.022
-0.384
-0.868
0.235
-0.132
-0.038
0.249
-0.123
-0.609
-0.675
0.239
-0.469
-
38
Table 2: Short Selling and Earnings Management
This table examines the baseline effect of short selling on earnings management. The main specification is based on a
panel regression of a firm's modified Jones's (1991) residual accruals (AccrualMJones) on lendable shares (Lendable) and
firm-level control variables (X), as well as unreported industry, country, and year fixed effects (ICY). The regression
model is
,
,
,
, , where
, includes firm size (Size), book-to-market
ratio (BM), financial leverage (Leverage), annual stock return (Return), stock return volatility (STD), American
Depository Receipts (ADR), MSCI country index membership (MSCI), number of analysts following the firm (Analyst),
closely held ownership (CH), institutional ownership (IO), and Amihud's (2002) illiquidity (Illiquidity). The construction
of these variables is detailed in Appendix A. Models (1) and (2) control for industry, country, and year fixed effects (ICY)
and industry and country fixed effects (IC), respectively. The t-statistics reported in parentheses are based on standard
errors adjusted for heteroskedasticity and firm-level clustering. Models (3) and (4) control for firm and year (FY) and firm
(F) fixed effects, with standard errors clustered at the year level. Models (5) and (6) apply the Arellano-Bond dynamic
panel GMM estimation to the same relationship while controlling for industry, country, and year fixed effects (ICY) and
industry and country fixed effects (IC), respectively. Obs denotes the number of firm-year observations, and AdjRsq is
adjusted R2. The sample period is from 2002 to 2009.
Lendable
Baseline Model
Model
Model
(1)
(2)
Firm Fixed Effects
Model
Model
(3)
(4)
-0.044
(-7.88)
-0.032
(-7.10)
-0.017
(-2.33)
-0.052
(-8.39)
0.004
(7.10)
0.004
(6.96)
0.010
(4.51)
0.003
(3.63)
-0.006
(-3.85)
-0.013
(-5.82)
-0.006
(-5.57)
-0.001
(-10.29)
-0.004
(-2.19)
0.003
(1.26)
-0.001
(-2.48)
0.005
(8.09)
0.005
(7.60)
0.010
(4.52)
0.001
(1.99)
-0.004
(-2.81)
-0.014
(-6.08)
-0.006
(-5.85)
-0.001
(-10.51)
-0.004
(-2.26)
0.003
(1.35)
-0.000
(-1.35)
0.022
(12.44)
0.013
(8.15)
0.070
(10.04)
-0.000
(-0.26)
-0.009
(-3.54)
0.002
(0.32)
0.021
(12.14)
0.013
(8.63)
0.071
(10.22)
-0.000
(-0.49)
-0.007
(-2.90)
0.005
(0.69)
-0.001
(-3.63)
0.002
(0.42)
-0.005
(-0.83)
-0.003
(-4.16)
-0.000
(-1.52)
0.000
(0.09)
-0.006
(-0.87)
-0.001
(-2.12)
ICY
61,624
2.9%
IC
61,624
2.9%
FY
61,624
19.4%
F
61,624
19.1%
Lagged AccrualMJones
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
39
Arellano-Bond
Model
Model
(5)
(6)
-0.041
(-7.38)
-0.010
(-1.02)
0.003
(5.98)
0.004
(7.02)
0.011
(4.65)
0.003
(3.90)
-0.008
(-4.55)
-0.013
(-5.96)
-0.006
(-5.62)
-0.001
(-10.33)
-0.004
(-2.10)
0.003
(1.54)
-0.001
(-3.76)
-0.032
(-7.01)
-0.009
(-1.00)
0.004
(6.99)
0.005
(7.74)
0.011
(4.71)
0.002
(2.42)
-0.006
(-3.59)
-0.014
(-6.19)
-0.006
(-5.70)
-0.001
(-10.35)
-0.004
(-2.21)
0.004
(1.67)
-0.001
(-2.61)
ICY
59,446
IC
59,446
Table 3: Instrumental Variable Approach
Panel A addresses the endogeneity problem by using ETF ownership (ETF) or industry-level ETF ownership (ETFIndustry)
as an instrumental variable and presents a panel regression of a firm's earnings management measure (AccrualMJones) on
predicted lendable shares (
) and firm-level control variables (X) as well as unreported industry, country, and
year fixed effects (ICY) on the variation of the following models:
:
,
,
,
,
,
;
:
,
,
,
, ,
where
, refers to lendable shares and
, includes the list of standard control variables. Models (1) and (3)
regress lendable shares on ETF ownership and industry-level ETF ownership, respectively. Models (2) and (4) regress
modified Jones's (1991) residual accruals on predicted lendable shares. Panel B provides the diagnostic analyses on the
impact of ETF and ETFIndustry on AccrualMJones. Models (1) and (6) directly regress AccrualMJones on the two instruments.
Models (2) and (7) also include lendable shares in the same regression. The remaining models regress AccrualMJones on the
two instruments for subsamples of the stocks for which short selling is either prohibited owing to regulation (Legality=0
or SSban=1) or low—when very few shares could be lent out (0<Lendable<0.5%). The t-statistics reported in parentheses
are based on standard errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year
observations, and AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
A. ETF and Industry-level ETF as Instrumental Variables
Instrument=ETF
Dep. Variable=
Instrument
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Lendable
(1st Stage)
Model
(1)
AccrualMJones
(2nd Stage)
Model
(2)
0.847
(5.46)
AccrualMJones
(2nd Stage)
Model
(4)
3.039
(5.30)
-0.004
(-9.08)
0.009
(20.86)
0.007
(3.90)
0.004
(6.89)
-0.006
(-5.88)
0.004
(2.48)
0.015
(15.71)
0.001
(11.55)
-0.008
(-5.80)
0.115
(20.03)
-0.007
(-20.37)
-0.102
(-4.69)
0.004
(6.51)
0.005
(7.43)
0.011
(4.66)
0.003
(3.86)
-0.006
(-4.02)
-0.013
(-5.67)
-0.005
(-4.48)
-0.001
(-9.56)
-0.004
(-2.44)
0.011
(2.93)
-0.001
(-3.36)
-0.004
(-6.85)
0.008
(19.33)
0.004
(2.10)
0.002
(2.99)
-0.005
(-4.92)
-0.002
(-1.23)
0.017
(18.21)
0.001
(9.70)
-0.013
(-8.16)
0.132
(40.23)
-0.007
(-26.14)
-0.041
(-1.86)
0.004
(6.97)
0.004
(6.61)
0.010
(4.50)
0.003
(3.60)
-0.006
(-3.83)
-0.013
(-5.84)
-0.006
(-5.37)
-0.001
(-10.05)
-0.004
(-2.17)
0.002
(0.61)
-0.001
(-2.17)
ICY
61,624
65.6%
ICY
61,624
2.8%
ICY
61,624
67.4%
ICY
61,624
2.9%
40
Instrument=ETFIndustry
Lendable
(1st Stage)
Model
(3)
Table 3: Instrumental Variable Approach - Continued
B. Tests on Exclusion Restrictions (Regress Accruals on ETFs)
Accruals on ETFs when SSP is Low
ETF
Legality=0
SSban=1
0<Lendable<0.5%
Model
(1)
Model
(2)
Model
(3)
Model
(4)
Model
(5)
-0.086
(-3.57)
-0.016
(-0.75)
-0.371
(-0.50)
-0.300
(-1.73)
0.010
(0.21)
ETFIndustry
Lendable
Firm Controls and Constant
Fixed Effects
Obs
AdjRsq
Yes
ICY
61,624
2.9%
Yes
ICY
1,025
8.7%
Yes
ICY
3,159
5.0%
Yes
ICY
16,578
2.7%
41
SSban=1
0<Lendable<0.5%
Model
(6)
Model
(7)
Model
(8)
Model
(9)
Model
(10)
-0.124
(-2.01)
0.011
(0.14)
-0.044
(-7.03)
-0.857
(-0.23)
0.885
(0.31)
0.266
(1.29)
Yes
ICY
61,624
2.8%
Yes
ICY
61,624
2.9%
Yes
ICY
1,025
8.7%
Yes
ICY
3,159
4.9%
Yes
ICY
16,578
2.8%
-0.040
(-6.99)
Yes
ICY
61,624
2.9%
Accruals on Industry ETF when SSP is Low
Legality=0
Table 4: The US Regulation SHO Experiment and Earnings Management
This table examines Regulation SHO in the US, in which the SEC randomly selected a sample of pilot firms announced in
2004 and formally removed their uptick restrictions from 2005 to 2007. Models (1)-(3) estimate the following annual
panel regressions with firm and year fixed effects (FY):
,
2005
2007
2008
2009
,
,
,
where
is modified Jones's (1991) residual accruals, refers to a dummy variable that equals
,
one if the stock is selected as a SHO pilot firm, and , refers to a list of control variables. The testing period in Model (1)
is from 2001 to 2007, in which the announcement year (2004) of Regulation SHO is removed from the sample. In Model
(2), the sample period is from 2001 to 2009, excluding 2004, and in Model (3), the sample period is from 2001 to 2009,
excluding 2004-2007. Models (4) and (5) estimate the following cross-sectional regression:
Δ ,
Δ
,
, .
refers to the difference between the three-year average value of AccrualMJones after year 2004
where Δ
,
(from 2005 to 2007 in Model (4) and from 2008 to 2009 in Model (5)) and that before 2004 (from 2001 to 2003) and ΔX ,
refers to changes in the average value of the control variables over the same periods. Control variables are detailed in
Appendix A. The t-statistics reported in parentheses are based on standard errors adjusted for heteroskedasticity and firmlevel clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2.
Dep. Variable=
US SHO×Dummy (2005-2007)
AccrualMJones
2001-2009
Ex. 2004
2001-2009
Ex. 2004-2007
Model
(1)
Model
(2)
Model
(3)
-0.006
(-2.20)
0.013
(3.27)
0.008
(2.54)
0.055
(4.84)
0.006
(2.12)
-0.018
(-3.52)
-0.002
(-5.16)
-0.001
(-0.08)
-0.017
(-1.86)
-0.006
(-3.63)
-0.006
(-2.16)
-0.004
(-0.99)
0.011
(3.72)
0.006
(2.26)
0.043
(4.58)
0.004
(1.70)
-0.018
(-4.57)
-0.002
(-5.87)
-0.003
(-0.45)
-0.017
(-2.27)
-0.006
(-4.16)
-0.003
(-0.79)
0.006
(1.69)
0.001
(0.44)
0.027
(2.29)
0.001
(0.40)
-0.020
(-4.18)
-0.002
(-4.71)
-0.007
(-0.86)
-0.010
(-0.89)
-0.005
(-2.92)
FY
12,597
15.4%
FY
16,347
13.3%
FY
10,020
13.8%
US SHO×Dummy (2008-2009)
Size
BM
Leverage
Return
STD
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
∆AccrualMJones (After-Before)
2001-2007
Ex. 2004
42
After=
Before=
2005-2007
2001-2003
2008-2009
2001-2003
Model
(4)
Model
(5)
US SHO
-0.005
(-2.05)
-0.003
(-0.88)
∆Size
0.000
(0.06)
0.001
(0.32)
0.008
(0.67)
0.018
(3.79)
-0.026
(-3.67)
-0.001
(-1.81)
0.009
(0.86)
-0.013
(-1.27)
-0.007
(-3.68)
0.001
(0.20)
-0.004
(-1.31)
0.001
(0.10)
0.003
(0.56)
-0.026
(-3.65)
-0.001
(-2.17)
0.000
(0.03)
-0.013
(-1.13)
-0.005
(-2.21)
I
2,218
8.1%
I
2,105
6.6%
∆BM
∆Leverage
∆Return
∆STD
∆Analyst
∆CH
∆IO
∆Illiquidity
Fixed Effects
Obs
AdjRsq
Table 5: Hong Kong Short-selling List and Earnings Management
This table explores the unique regulatory setting in the Hong Kong market in which regulators changed the list of stocks
eligible for short selling on a quarterly frequency from 1994 to 2005. Models (1)-(3) estimate the following panel
regression with firm and year fixed effects (FY) and clustered standard errors at the firm and industry levels:
,
,
,
,
, .
where
is modified Jones's (1991) residual accruals,
, is a dummy variable that equals one if a
,
stock is eligible to short selling in year t, and
is
a
dummy
variable
that equals one if a stock is eligible for
,
short selling in year t-1 but becomes ineligible for short selling beginning in year t. Models (4) and (5) estimate the
following panel regressions:
Δ , Δ Δ ,
Δ
,
,
, .
where
, refers to net inclusion and equals one (negative one) if a firm is included in (excluded from) the eligible
list and Δ , is a dummy variable for exclusion. The control variables are detailed in Appendix A. The tstatistics reported in parentheses are based on standard errors adjusted for heteroskedasticity and firm-level clustering. Obs
denotes the number of firm-year observations, and AdjRsq is adjusted R2. Dep. Variable=
Model
(1)
HK SS
BM
Leverage
Return
STD
ADR
Analyst
CH
Illiquidity
Fixed Effects
Obs
AdjRsq
Model
(4)
∆HK SS
-0.024
(-2.25)
0.002
(0.12)
0.049
(5.48)
0.015
(1.84)
0.146
(3.18)
-0.003
(-0.41)
0.007
(0.66)
-0.151
(-3.23)
-0.001
(-1.03)
0.033
(1.14)
0.001
(0.42)
∆HK SSExclusion
0.045
(6.05)
0.013
(2.00)
0.136
(3.65)
-0.001
(-0.18)
0.003
(0.37)
-0.148
(-5.10)
-0.000
(-0.06)
0.018
(0.78)
0.002
(0.92)
-0.009
(-0.77)
0.045
(6.05)
0.014
(2.01)
0.136
(3.66)
-0.001
(-0.20)
0.003
(0.37)
-0.148
(-5.14)
-0.000
(-0.05)
0.018
(0.78)
0.002
(0.91)
FY
4,411
10.8%
FY
4,411
10.7%
FY
3,483
10.3%
Fixed Effects
Obs
AdjRsq
43
∆AccrualMJones
-0.022
(-2.18)
HK SSPost
Size
AccrualMJones
HK SS=0
Model
Model
(2)
(3)
∆Size
∆BM
∆Leverage
∆Return
∆STD
∆ADR
∆Analyst
∆CH
∆Illiquidity
Model
(5)
-0.023
(-1.98)
0.086
(8.41)
0.048
(4.17)
0.329
(5.62)
-0.001
(-0.10)
0.011
(0.92)
-0.127
(-1.28)
-0.001
(-0.89)
0.020
(0.65)
0.001
(0.37)
0.036
(2.25)
0.086
(8.43)
0.047
(4.16)
0.328
(5.62)
-0.000
(-0.04)
0.011
(0.93)
-0.125
(-1.25)
-0.001
(-1.07)
0.020
(0.65)
0.001
(0.34)
CY
3,528
6.2%
CY
3,528
6.2%
Table 6: Subsample Analyses
This table examines the impact of short selling on earnings management in several important subsamples. Panel A
explores the impact of short selling in subperiods, when it is interacted with various time dummies, and different regions.
In the first two columns, the variable “≥2005” refers the sample period from 2005 to 2009, and T equals the year minus
2001. In Models (3) and (4), “US” and “NUS” refer to the subsamples of US and none-US firms, respectively. Panel B
explores the source of the effects of short selling on earnings management by dividing the full sample into the accrual and
firm size subsamples. AccrualMJones≥0 refers to firms with positive AccrualMJones, whereas AccrualMJones<0 refers to firms
with negative AccrualMJones. Small firms are firms with market capitalization below the median value of market
capitalization for each country and year, whereas large firms are firms with market capitalization greater than the median
value of market capitalization for each country and year. The t-statistics reported in parentheses are based on standard
errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year observations, and
AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
A. Subsample Analyses on Different Time Periods and Regions
>=2005
Full Sample
US
Model
Model
Model
(1)
(2)
(3)
Lendable
-0.045
(-6.37)
-0.003
(-0.14)
-0.006
(-2.17)
Lendable×T
-0.064
(-7.39)
NUS
Model
(4)
-0.044
(-3.52)
Firm Controls and Constant
Yes
Yes
Yes
Yes
Fixed Effects
ICY
ICY
ICY
ICY
Obs
44,171
61,624
21,825
39,799
AdjRsq
2.7%
2.9%
3.3%
2.8%
B. Subsample Analyses on Different Degrees of Earnings Management
AccrualMJones≥0
AccrualMJones<0
Small Firms
Large Firms
Small Firms
Large Firms
Model
Model
Model
Model
(1)
(2)
(3)
(4)
Lendable
Firm Controls and Constant
-0.097
(-4.27)
Yes
-0.025
(-3.66)
Yes
-0.038
(-1.55)
Yes
0.004
(0.55)
Yes
Fixed Effects
Obs
AdjRsq
ICY
6,022
6.9%
ICY
26,446
8.7%
ICY
6,224
13.3%
ICY
22,932
10.0%
44
Table 7: Robustness Checks
This table presents a series of robustness checks based on alternative short selling and earnings management measures,
alternative controls for corporate governance, and alternative clustering specifications of the main regression model. Panel
A presents the results for regressions using alternative short-selling potential (SSP) measures that include shares on loan
(On Loan), loan fee (Fee), loan fee volatility (STDFee), specialness (Specialness), and short-selling constraint (Constraint).
Panel B presents the results for regressions using alternative earnings management measures, including three targetbeating measures, namely, small positive forecasting profits (SPAF), small positive past-earnings profits (SPDE), and
small positive profits (SPE), in Models (1) to (3) and two earnings persistence specifications in Models (4) and (5). In
particular, Models (4) and (5) estimate the following regression model:
Accrual
α β Lendable ,
β Lendable , Earnings , Accrual
Earnings ,
,
,
β
X
β
X
Earnings
Accrual
ε
,
β Earnings , Accrual
,
,
,
,
,
,
is operating income scaled by lagged total assets and
is modified Jones's (1991)
where
,
,
residual accruals. Panel C provides alternative discretionary accrual measures in Models (1) to (3), including Jones's (1991)
residual accruals (AccrualJones), KLW's (2005) residual accruals (AccrualKLW), and DD's (2002) residual accruals
(AccrualDD). Models (4) and (5) proxy for manipulation practices by the likelihood that earnings misstatements or scandals
occur. Panel D adds alternative discipline channels as controls. These alternative discipline channels include the ISS
corporate governance index (ISS), big N auditor (BigN), international accounting standard (IAS), news coverage
(NewsCoverage), and analyst dispersion (Disp). All these tests control for industry, country, and year fixed effects (ICY).
The t-statistics reported in parentheses are based on standard errors adjusted for heteroskedasticity and firm-level
clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is from 2002
to 2009.
SSP=
SSP
Firm Controls and Constant
Fixed Effects
Obs
AdjRsq
A. Alternative SSP Measures
On Loan
Fee
Model
Model
(1)
(2)
-0.052
(-4.12)
Yes
0.001
(3.74)
Yes
STDFee
Model
(3)
Specialness
Model
(4)
Constraint
Model
(5)
0.003
(4.93)
Yes
0.004
(4.05)
Yes
0.004
(2.35)
Yes
ICY
ICY
ICY
ICY
ICY
61,623
61,575
59,011
61,575
61,574
2.9%
2.9%
2.9%
2.9%
2.9%
B. Alternative Earnings Management Measures on Target Beating and Earnings Persistence
Target Beating
Earnings Persistence
Dep. Variable=
SPAF
SPDE
SPE
Earnings
AccrualMJones
Model
Model
Model
Model
Model
(1)
(2)
(3)
(4)
(5)
Lendable
-0.767
(-3.63)
-1.111
(-3.99)
-0.914
(-2.06)
Earnings
Lendable×Earnings
0.033
(4.50)
0.805
(12.68)
-0.301
(-4.66)
AccrualMJones
0.109
(0.96)
-0.453
(-4.15)
Lendable×AccrualMJones
Firm Controls and Constant
Earnings (AccrualMJones)×Control
Fixed Effects
Obs
AdjRsq (PseRsq)
Yes
No
Yes
No
Yes
No
Yes
Yes
Yes
Yes
ICY
46,381
3.4%
ICY
35,986
4.6%
ICY
19,091
10.2%
ICY
58,302
68.1%
ICY
55,816
5.5%
45
-0.019
(-3.73)
Table 7: Robustness Checks - Continued
C. Alternative Earnings Management Measures on Other Accruals and Earnings Misstatements
Other Accrual Measures
Misstatement and Scandals
AccrualJones
AccrualKLW
AccrualDD
Dep. Variable=
Prob(Misstatement)
Prob(Scandals)
Model
Model
Model
Model
Model
(1)
(2)
(3)
(4)
(5)
Lendable
Firm Controls and Constant
Fixed Effects
Obs
AdjRsq (PseRsq)
-0.038
(-6.95)
-0.035
(-4.15)
-0.019
(-3.83)
-0.818
(-2.23)
-1.209
(-2.58)
Yes
Yes
Yes
Yes
Yes
ICY
30,047
5.8%
ICY
30,047
17.6%
Model
(4)
Model
(5)
ICY
61,562
3.1%
Model
(1)
Lendable
ISS
ICY
ICY
61,015
57,603
0.5%
1.2%
D. Alternative Discipline Channels
Model
Model
(2)
(3)
-0.027
(-3.06)
-0.025
(-2.92)
-0.028
(-3.04)
-0.025
(-2.95)
-0.000
(-0.00)
-0.032
(-3.43)
-0.025
(-2.97)
-0.001
(-0.15)
-0.008
(-3.97)
-0.033
(-3.62)
-0.025
(-2.94)
-0.001
(-0.23)
-0.008
(-4.13)
-0.002
(-3.16)
Yes
Yes
Yes
Yes
-0.022
(-2.32)
-0.023
(-2.52)
0.001
(0.29)
-0.005
(-2.22)
-0.003
(-3.75)
-0.308
(-3.31)
Yes
ICY
16,184
4.7%
ICY
16,149
4.7%
ICY
16,065
4.8%
ICY
16,065
4.8%
ICY
13,396
6.1%
IAS
BigN
NewsCoverage
Disp
Firm Controls and Constant
Fixed Effects
Obs
AdjRsq
46
Table 8: Stock Price Informativeness and Earnings Management
This table presents the results of a panel regression of a firm's stock price informativeness on lendable shares, its potential
impact on earnings management, the interaction between lendable shares and its potential impact on earnings management,
and firm-level control variables (X), as well as unreported industry, country, and year fixed effects (ICY) for the full
sample and different subsamples. The regression model is
,
,
,
,
,
,
, ,
2
2
where
, refers to two proxies of stock price informativeness, downside-minus-upside R (R DMU)
in Models (1) and (2), and downside-minus-upside nonsynchronicity (NonsynDMU) in Models (3) and (4). A higher degree
of price informativeness is associated with lower values of R2DMU and higher values of NonsynDMU.
, is a dummy
variable that takes a value of one when the stock is a small firm with
greater
than
zero,
and zero
,
otherwise. , includes the same list of firm control variables as before. The construction of these variables is detailed in
Appendix A. The t-statistics reported in parentheses are based on standard errors adjusted for heteroskedasticity and firmlevel clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is from
2002 to 2009.
R2DMU
Price Informativeness=
Lendable
Model
(2)
-0.055
(-13.67)
-0.002
(-2.93)
-0.001
(-1.49)
0.003
(1.62)
0.005
(7.58)
-0.005
(-4.97)
-0.003
(-1.55)
-0.001
(-0.91)
0.000
(0.05)
-0.007
(-4.28)
0.013
(11.36)
0.001
(3.08)
-0.053
(-13.08)
-0.078
(-5.23)
0.002
(1.56)
-0.001
(-2.87)
-0.001
(-1.33)
0.003
(1.59)
0.005
(7.41)
-0.005
(-4.95)
-0.003
(-1.52)
-0.001
(-0.84)
0.000
(0.00)
-0.006
(-4.16)
0.013
(11.23)
0.001
(3.13)
0.014
(2.79)
0.006
(1.45)
-0.024
(-1.37)
-0.055
(-8.02)
-0.009
(-0.73)
0.029
(1.77)
0.036
(3.76)
-0.001
(-1.24)
0.074
(4.81)
-0.126
(-11.09)
-0.012
(-3.73)
0.371
(9.42)
0.455
(1.76)
-0.007
(-0.47)
0.014
(2.83)
0.006
(1.34)
-0.024
(-1.36)
-0.054
(-7.91)
-0.009
(-0.73)
0.028
(1.74)
0.036
(3.72)
-0.001
(-1.24)
0.074
(4.75)
-0.125
(-11.04)
-0.013
(-3.76)
ICY
59,953
7.6%
ICY
59,953
7.6%
ICY
59,952
5.4%
ICY
59,952
5.4%
Lendable×SSPImpact
SSPImpact
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
47
NonsynDMU
Model
Model
(3)
(4)
Model
(1)
0.383
(9.37)
Internet Appendix
“The Invisible Hand of Short Selling:
Does Short Selling Discipline Earnings Management?”
This appendix has two purposes. First, it provides additional tests (Part 1). Second, when a table
reported in the main text does not present the regression coefficients of the control variables in the
interest of brevity, this appendix tabulates the full specifications (Part 2). When there is no confusion,
additional tests are labelled adding the extension “IA” for “Internet Appendix” (e.g., Table IA), while
the full specifications of the tables reported in the main text are labelled with the original table name.
Below, we primarily discuss the results of the additional tests.
1. Additional Summary Statistics
Table IA1 tabulates the definition of the additional variables that we use in the Internet Appendix,
including a new instrument, alternative illiquidity and information control variables, country-level
regulations that directly affect short-selling constraints, and country characteristics. Specifically, the
new instrument is the degree of concentration of institutional investors (hereafter, HHI), defined
following Prado, Saffi, and Sturgess (2013). Additional illiquidity and information control variables
include the proportion of zero stock returns (ZRP), effective spread (ESprd), relative quoted spread
(RSprd), stock trading turnover (TV), and the probability of informed trading (PIN). Country-level
regulations are proxied by a list of dummies that take a value of one if short selling is legal (Legality),
if short selling is feasible (Feasibility), if put option trading is allowed (Put Option), and if short
selling is feasible or put option trading is allowed (F or P). When these variables are used, we further
control for a set of commonly used country-specific variables in addition to firm characteristics,
including Bekaert et al.’s (2011) degree of market segmentation of the country (SEG), the antidirector index (ADRI), the market capitalization-to-GDP ratio (MVGDP), the standard deviation of
GDP growth (STDGDPG), the future stock market return (MKTReturn), and the future S&P sovereign
credit rating (MKTCreditRating); all the variables have very wide distributions.
Internet Appendix, Page 1 The country regulation variables are constructed following Charoenrook and Daouk (2005); we
also refer to Bris, Goetzmann, and Zhu (2007) and Beber and Pagano (2011) for more recent periods.
For each country, we rebalance the variables annually for the period from 1990 to 2009, although the
majority of the regulatory changes in short-selling restrictions occur over the 1990-2000 period. The
definitions of these variables, as well as additional country-specific control variables, are reported in
Table IA1. The summary statistics for these variables are reported in Table IA2. We see that, for
instance, short selling is legal in a majority of countries in our sample, whereas there are greater crosscountry variations in terms of Feasibility. This difference implies that some countries may allow short
selling but that their institutional infrastructure may not be sufficiently sophisticated to support largescale short selling.
Table IA2 provides summary statistics for other variables that are not reported in Table 1 in the
main text, including instrumental variables (ETF ownership and HHI), alternative proxies for shortselling potential (SSP) and earnings management, alternative governance variables, sentiment and
scandal variables that are used in the SHO test, and additional liquidity proxies. We find that these
variables have reasonable distributions and are comparable to those in the literature. For instance,
HHI has a mean value of 13.2% in our sample, which is similar to the value of 12.4% reported in
Prado, Saffi, and Sturgess (2013).
Table IA3 provides the mean of lendable shares by year and market for our final sample.
Consistent with Saffi and Sigurdsson (2011), we find that the short-selling market, especially in the
US, exhibits a substantial upward trend over the 2004-2008 period. By contrast, the evolution of the
global short-selling market outside the US exhibits a much smoother pattern. The statistics are
generally consistent with both Aggarwal, Saffi, and Sturgess (2013) for US stocks and Saffi and
Sigurdsson (2011) for international stocks. For instance, the average Lendable value for Russell 3000
firms during the 2007-2009 period in Aggarwal, Saffi, and Sturgess (2013) is approximately 22%
(quoted from their Table 1). This number is comparable to the average Lendable value for all US
firms that are used in our main tests (approximately 20% for our sample of US firms). For
international stocks, the average Lendable value for firms that have reasonable weekly financial data
Internet Appendix, Page 2 over the 2004-2008 period is reported as 8% in Table 3 of Saffi and Sigurdsson (2011). The same
statistic for our full sample period (2002-2009) is 6.5%. Moreover, if we compute the corresponding
statistics for our final sample of international stocks, the mean value is approximately 7% over the
same period (2004-2008). The remaining difference arises because our data needs (annual accounting
and financial information) differ from theirs. Broadly speaking, time period and annual accounting
and financial information account for one-third and two-thirds of the difference, respectively. We note
that the difference is within a very reasonable range.
2. Robustness Checks and Diagnostics for the Instrumental Variable Tests
Table IA4 provides robustness checks and additional diagnostics for the instrumental variable
tests reported in Table 3. Panels A extends the instrumental variable test to double cluster standard
errors at the firm and year levels following Petersen (2009) and Thompson (2009). Panel B employs a
different robustness check and verifies that the instrumental variable tests are robust in the subperiod
from 2005 to 2009, where the early years of the full period are excluded. In both robustness checks,
we find that the results are quantitatively identical, suggesting that the instrumental variable test is
highly robust.
Another potential concern is that, since ETF ownership could be higher for stocks included in
major stock indices, our results may be spuriously correlated with the membership of such indices.
That is, cross-stock variations in ETF ownership may simply reflect the difference between stocks that
are included in major indices and stocks that are not. To address this concern, Panel C applies the
main instrumental variable only to the subsample of stocks that are included in the MSCI country
index—one of the leading indices in global market. We see that our results hold even among the
stocks that are already members of the most popular global index, suggesting that index membership
is not the major driver for our results. Panel D extends the main test by further controlling for countrylevel characteristics, including the control-of-corruption index (Corruption) and the market
capitalization-to-GDP ratio (MVGDP), and the results remain the same. In general, results of our
instrumental variable test are robust to the control of country characteristics.
Internet Appendix, Page 3 Table IA5 further complements the diagnostic analyses provided in Panel B of Table 3. We recall
that, in Panel B of Table 3, the ETF instruments lose their explanatory power when SSP is included in
the regression or the short-selling channel is highly constrained. Table IA5 extends the latter result by
exploring the reverse constraint by regressing AccrualMJones on Lendable on the subsample of stocks
with low ETF ownership. The regression specifications and control variables remain the same as those
in Panel B of Table 3. We find that the disciplining effect of short selling is not attenuated when ETF
ownership is low. This result is in line with the expectation that other (passive) institutional investors,
such as pension funds and insurance companies, may also be willing to lend shares to short sellers.
Combining this result with the results reported in Table 3, we therefore find that short selling appears
to be the necessary condition for ETFs to affect managerial behavior but that ETFs are not a necessary
condition for SSP to affect earnings management. In this regard, ETFs are a good instrument to
highlight the causal effect of short selling on earnings management initiated by lendable shares
insensitive to earnings management in the first place.
Table IA6 further incorporates Prado, Saffi, and Sturgess’ (2013) finding that because high
concentration (HHI) implies investor activism, it typically introduces additional constraints into the
lending market. In this regard, a lower degree of concentration provides an instrument on the supply
of lendable shares less related to investor activism. Thus, we extend our existing instrument variable
test based on HHI alone (Models 1 and 2), jointly with ETF ownership (Models 3-4) or jointly with
industry-level ETF ownership (Models 5-6); the main results remain unchanged. In particular, these
models indicate that HHI reduces lendable shares in the first-stage regression and that instrumented
lendable shares, in all specifications, significantly reduce earnings management. The use of two
instruments has the additional benefit that we can rely on the Hansen test to assuage concerns of
overidentification. The Hansen J statistic for over-identification tests is 0.822 for Models (3) and (4),
for instance, with an insignificant p-value of 0.365 to reject the null that the joint model is properly
identified. These instrumental variable regressions, therefore, provide the first evidence supporting a
causal disciplining impact of SSP on firm manipulations.
Internet Appendix, Page 4 3. Robustness Checks on Hong Kong Experiment
We now move on to provide two robustness checks for the Hong Kong short selling experiment.
One potential concern is that, since the criteria for firms to be selected into the short selling list were
known, firm managers could anticipate the eligibility of short selling and take actions even before the
real inclusion date.1 Note that this concern works only against us in finding any significant reduction
in earnings management after the firms are included in the short selling list—compared to the years
that they are not. Nonetheless, we can conduct a robustness check to eliminate the potential
anticipation effect by excluding the observations of eligible firms for the year prior to their inclusion
to the short selling list.
The results are reported in Panel A of Table IA7. We see that our results are robust to this
additional adjustment. Moreover, in Models (1), (2), and (4), both the magnitude of the regression
parameter of HK SS and the level of statistical significance get enhanced slightly, consistent with the
notion that managers could start to reduce earnings management even before short selling really starts.
This early action also illustrates the disciplining impact of short selling – of course, the major
disciplining impact still seems to occur when short selling becomes eligible.
In our second robustness check, we create a control group for eligible stocks based on propensity
score matching. More specifically, we match each eligible stock with a non-eligible stock of the same
period based on Size and TV—two of the most important elements to affect the short selling list
(including other characteristics such as book-to-market does not affect the results)—and then apply
the tests reported in Table 5 to this propensity score-matched sample. The goal of this test is twofold.
First, it creates a more balanced distribution between eligible and non-eligible firms. Second, it also
utilizes the subgroup of firms that are the most similar to eligible firms in the economy, thereby
eliminating the potential concern that the results of Table 5 could be driven by a subset of very small
1
The following types of stocks were selected: all constituent stocks of indices that are the underlying indices of equity index
products traded on the Exchange, all constituent stocks of indices that are the underlying indices of equity index products
traded on the Hong Kong Futures Exchange, all underlying stocks of stock options traded on the Exchange, all underlying
stocks of Stock Futures Contracts traded on the Hong Kong Futures Exchange, stocks that maintain a public float
capitalization of not less than HK$1 billion for either (i) a period of 60 consecutive trading days during which dealings in
such stocks have not been suspended or (ii) a period of no more than 70 consecutive trading days comprising 60 trading days
during which dealings in such stocks have not been suspended, stocks with market capitalizations of not less than HK$1
billion and an aggregate turnover during the preceding 12 months to a market capitalization ratio of not less than 40%, and
the tracker Fund of Hong Kong and other ETFs approved by the Board in consultation with the Commission.
Internet Appendix, Page 5 and illiquid stocks. The results are tabulated in Panel B of Table IA7. We see that the disciplining
impact remain significant even among firms with similar characteristics.
In addition, an unreported univariate analysis also confirms our results. In this test, we compute
the diff-in-diff statistics as the changes in earnings management of eligible firms minus those of the
control group. The change in earnings management for a specific eligible firm is computed as the
difference between earnings management during the eligible period and that during the ineligible
period—and the change in earnings management for its control firm is computed over the same
periods. The mean value of the diff-in-diff earnings management is -0.024 with a t-statistics of 4.49.
Overall, the Hong Kong experiment also provides supporting evidence to the disciplining impact of
short selling.
4. Countrywide Tests
Table IA8 reports the general impact of countrywide short-selling constraints on earnings
management. To do so, Models (1) to (4) regress earnings management on a list of dummy variables,
including Legality, Feasibility, Put Option, and F or P, which are defined above. To alleviate the
concerns related to country-specific characteristics, in all the tests, we control for industry, country,
and year fixed effects (ICY) and a set of country-level variables, including Bekaert et al.’s (2011)
degree of market segmentation of the country (SEG), the anti-director index (ADRI), the market
capitalization-to-GDP ratio (MVGDP), the standard deviation of GDP growth (STDGDPG), the future
stock market return (MKTReturn), and the future S&P sovereign credit rating (MKTCreditRating). As
discussed in the main text, the results demonstrate a strongly significant impact of country-level shortselling regulation on earnings management, which is in line with the disciplining hypothesis.
Specifically, in Models (1) and (2), Legality and Feasibility yield similar impacts to earnings
management. Put Option in Model (3) appears to only have marginal explanatory power on crosscountry manipulation. In Model (4), when Feasibility and Put Option are jointly considered, the
explanatory power seems to be primarily attributable to Feasibility.
Models (5) to (8) further refine the tests to the sample of firms that issue ADRs in the US. The
interesting, if not surprising, observation is that the impact of home-country SSP seems to survive the
Internet Appendix, Page 6 bonding effect imposed by US regulation (e.g., Doidge, Karolyi, and Stulz, 2004, 2007). Earnings
management, for instance, is 20% (17%) higher for ADR firms from countries in which short selling
is banned (unfeasible). This evidence can be compared to the findings of Jain et al. (2013) that homecountry short-selling restrictions can affect short selling on ADRs, a phenomenon referred to as
regulatory reach in their paper. Using their terminology, we can also conclude that the home-market
disciplining effect of short-selling regulation “reaches” the ADR market. Note that the country
distribution of ADR firms is quite different from that of the full sample. For ADR firms, both
Feasibility and Put Option are negatively associated with earnings management. These results suggest
that the effectiveness of financial regulation in the global market could be quite complex.
Next, we extend our analyses from regulations that directly affect the legality of short selling to
more subtle regulations that may affect the information environment and the flexibility of short selling.
We first ask to what extend market-wide regulations aiming to supply more information to the
market—i.e., accounting rules—affect the disciplining impact of short selling. To do so, Models (1)
and (2) of Table IA9 implement the previous model specification that regresses a firm's earnings
management measure (Accruals) on lendable shares (Lendable), conditioning on the strength of a
country’s disclosure regulation index (La Porta, Lopez-de-Silanes, and Shleifer 2006). Here, we
classify the countries with below (above) median values of the disclosure index as “low” (high). The
row “Difference” further reports the difference between the two sensitivity parameters of Accruals
with respect to SSP, i.e., Model (1) sensitivity minus Model (2) sensitivity, followed by the p-value of
the difference.
We document two important facts in this table. First, the disciplining impact of short selling is
robust regardless of accounting rules. This result is reasonable, as manipulation incentives can still
occur even in countries with the most strict disclosure regulations. In both cases, short selling
represents the invisible hand of the market to potentially uncover information above and beyond what
publically available accounting numbers can tell, and thus still disciplines managers ex ante. Second,
the markets with weaker regulations in terms of accounting disclosure are associated with a stronger
impact of short selling. This positive correlation is in line with the expectation that the disciplining
Internet Appendix, Page 7 role of short selling is more effective in the markets with poor public information. Of course, we need
to interpret this result with caution, because accounting regulations may create more earnings
management in the first place, in which case the stronger sensitivity of earnings management with
respect to short selling does not necessarily mean that short sellers do more in these markets. But
(better) accounting regulations and (higher) SSP do seem to substitute each other in reducing the
incentives for firms to be involved in earnings management. This substituting effect seems to be
reasonable, as earnings management arises due to insufficient information.
Models (3) and (4) of Table IA9 further apply the previous regression analysis to stocks in
markets with weak and strong investor protections, as proxied by the anti-director index of Pagano
and Volpin (2005). We documents two findings. First, the disciplining impact of short selling is
robust across different degrees of investor protection. Second, we would expect that manipulation
incentives should be higher in countries with weak investor protection and therefore high SSP should
be associated with a stronger disciplining effect. And indeed, the empirical results confirm the sign
of this prediction. However, the significance level is between 10% and 20%. This level of
significance suggests that the impact of investor protection regulations is less direct than that of
accounting regulations in affecting manipulation incentives.
Next, we perform the previous analysis conditioning on the strength of trading regulations of the
market in which the stocks are listed. More flexible trading rules could reduce the cost of short
selling, which enhances the ex ante impact of SSP on manipulation incentives according to the
disciplining hypothesis. However, more flexibility in trading could also encourage more speculations
and enhance price pressure, which encourages firms to manipulate even more according to the price
pressure hypothesis. Hence, trading rules provide an interesting market-wide testing ground to
differentiate the two competing hypotheses.
To conduct these tests, we follow Cumming, Johan, and Li (2011) and use as trading regulation
the price manipulation index and the market manipulation index. The latter measure sums up a list of
more specific trading rules, including the price manipulation index, the volume manipulation index,
the spoofing index, and the false disclosure index. The price manipulation index and the market
Internet Appendix, Page 8 manipulation index allow us to examine the impact of SSP in more explicit and more general setups.
For both trading rules, we use a dummy taking the value of one if the index value is below (above)
the median value of the index to proxy for low (high) trading flexibility or low (high) price pressure.
Models (5) and (6) and Models (7) and (8) report the impact of SSP on discretionary accruals in
these different subsamples of markets. As in the previous case, we find two results. First, the
disciplining impact of short selling is robust across different sets of trading rules. Second,
interestingly, more flexible trading rules, even when they may encourage more price manipulation,
are associated with a stronger impact of SSP on earnings management. Hence, even at the market
level, the evidence supports the disciplining hypothesis as opposed to the price pressure hypothesis.
This test complements our main test conducted at the stock level.
5. Additional Robustness Checks
Next, we extend the main analyses (Table 2) and subsample analyses (Table 6) by examining the
impact of short selling on earnings management in subsamples of countries sorted by the feasibility of
short selling and by controlling for firm-level investments. Even though our main tests include all the
countries in which we can find short selling data from Data Explorer, short selling is not actively
practiced in some countries, such as Finland, Indonesia, Philippines, South Korea, and Taiwan,
according to Maffett, Owens, and Srinivasan (2014). These countries provide an ideal Placebo test to
examine whether our main analyses have the proper power to detect the real disciplining impact of
short selling. Note that our goal here differs from that of the previous legality and feasibility tests,
because we are not exploring the impact of cross-country regulation variations in the current test.
In Panel A of Table IA10, “GFC” in Model (1) refers to the global financial crisis period from
2007 to 2008, whereas “Ex.GFC” in Model (2), excludes the global financial crisis period. We
conduct our baseline regression in these subperiods and find that our main results essentially hold in
both of them. Thus, the disciplining role in general applies to both the crisis and the noncrisis periods.
In Model (3), the sample “Ex.Zeros” includes only firms with nonzero short-selling values. Our main
results remain the same.
Internet Appendix, Page 9 Model (4) of Panel A provides a robustness check to Model (4) of Table 6, which suggests that
the impact of SSP on earnings management increases over time. Instead of interacting Lendable with
a time-sequence variable T, we decompose the impact of lendable shares by interacting Lendable with
a list of subperiod dummies, including Dummy (2004-2006), Dummy (2007-2008), and Dummy
(2009). These dummy variables are equal to one in the years specified in the parentheses and zero
otherwise. We find that the impact of Lendable is significant in all subperiods and that its magnitude
increases over time.
In Panel B of Table IA10, Model (1) tabulates the results of the baseline test conducted at the
whole sample as a benchmark. Models (2) to (5) apply the same test to the subsets of countries. More
specifically, “Five Countries” in Model (2) refer to the subsample of five countries in which short
selling is not actively practiced according to Maffett, Owens, and Srinivasan (2014). “Ex. Five
Countries” in Model (3) refer to the remaining countries. In Models (4) and (5), “Feasibility= 0”
refers to the subsample of countries in which, according to Charoenrook and Daouk (2005), short
selling is not feasible, whereas “Feasibility= 1” refers to the rest of countries. We find that the
disciplining impact of short selling concentrates in countries in which short selling is important and
feasible. By contrast, in countries that short selling is not well practiced, the disciplining impact
disappears. The sharp difference between the two subsamples demonstrates that our tests have the
proper power to detect the real impact of short selling. Of course, since short selling is well practiced
in majority countries, our main tests reported in the text focus on the whole sample of countries.
Grullon, Michenaud, and Weston (2013) show that short selling potential affects firms’ investment
and growth. This raises the possibility that the relation between lendable shares and discretionary
accruals does not come through a disciplining channel, but rather, through a real impact on investment
and growth. To address this issue, Models (6) and (7) extend the whole-sample baseline test to include
variables describing the growth opportunity and investment behaviour of firms. These variables are
CapEx, the ratio of capital expenditures scaled by total assets, R&D, the ratio of research and
development expenses to total assets, and SalesGrowth, defined as the log of changes in net sales. We
Internet Appendix, Page 10 find that the disciplining impact of short selling is robust once we control for firm-level growth
opportunities and investment policies.
Given the importance of liquidity in the global market (e.g., Karolyi, Lee, and van Dijk, 2012), in
Table IA11, we provide further robustness checks of our main specification by using alternative
illiquidity and information measures as controls. To do so, in Models (1), (2), (3), (4), and (5), we
replace the Amihud illiquidity measure, our main liquidity control, with the proportion of zero stock
returns (ZRP), effective spread (ESprd), relative quoted spread (RSprd), stock trading turnover (TV),
and the probability of informed trading (PIN), respectively. We find that these alternative measures do
not affect the explanatory power of SSP on earnings management. Note that some of these measures,
such as PIN, also control for the potential impact of informed trading. Thus, our main conclusion
regarding the disciplining impact of short selling is robust to various liquidity and informational
environments.
Table IA12 extends Panel C of Table 7 in constructing two additional discretionary accrual
measures. The first is based on Francis et al.'s (2005) residual accruals (AccrualFLOS), where residual
accruals are obtained by regressing changes in working capital on past, current, and future cash flows,
revenue growth, and fixed assets for each country and year. The second measure is based on Allent,
Larson, and Sloan's (2013) residual accruals (AccrualALS), where residual accruals are obtained by
regressing changes in working capital on past, current, and future cash flows, revenue growth, and
employee growth for each country and year. The specifications are the same as Panel C of Table 7.
Finally, Table IA13 considers alternative clustering specifications. Models (1) and (2) cluster
standard errors at the country and year levels, respectively. Models (3) to (5) follow Petersen (2009)
and Thompson (2010), and adopt firm-year, country-year, and industry-year double clustering. These
specifications help us address the concerns of biased standard errors when residuals are also
correlated at the country, industry, and year levels. The results are basically the same, confirming that
the disciplining impact of short selling on earnings management is very robust in the real economy.
Internet Appendix, Page 11 References
Bekaert, G., C. Harvey, C. Lundblad, and S. Siegel, 2011, What Segments Equity Markets? Review
of Financial Studies 24, 3841-3890.
Charoenrook, A. and H. Daouk, 2005, A Study of Market-Wide Short-Selling Restrictions, Working
Paper.
Maffett, M.G., E.L. Owens, and A. Srinivasan, 2014, The Effect of Short-Sale Constraints on Default
Prediction around the World, working paper (available at
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2296992).
Internet Appendix, Page 12 Part 1: Additional Analyses
Table IA1: Definitions of Additional Variables
Variable
Acronym
Definition
Data Source
Concentration of institutional
ownership
Legality of short selling
Feasibility of short selling
Put option trading
Feasibility or Put Option
Market segmentation
Anti-director index
Market capitalization-to-GDP ratio
Standard deviation of GDP growth
Future stock market return
Future S&P sovereign credit rating
Proportion of zero stock returns
HHI
Concentration of institutional ownership using the Hirschman-Herfindahl Index
FactSet
Legality
Feasibility
Put Option
F or P
SEG
Anti-Director
MVGDP
STDGDPG
MKTReturn
MKTCreditRating
ZRP
Charoenrook and Daouk (2005)
Charoenrook and Daouk (2005)
Charoenrook and Daouk (2005)
Charoenrook and Daouk (2005)
Datastream
Pagano and Volpin (2005)
World Development Indicators
World Development Indicators
Datastream
Standard & Poor's
Datastream
Effective spread
Relative quoted spread
Stock trading turnover
Probability of informed trading
ESprd
RSprd
TV
PIN
Capital expenditures
R&D
Sales growth
Disclosure requirement index
Anti-director index
Price manipulation index
Market manipulation index
CapEx
R&D
SalesGrowth
Disclosure
Anti-Director
PMI
MMI
Control of corruption
Corruption
A dummy variable that equals one if short selling is legally allowed in a country
A dummy variable that equals one if short selling is feasible in a country
A dummy variable that equals one if put option trading is feasible in a country
A dummy variable that equals one if either short selling or put option is feasible in a country
Segmentation measure developed by Bekaert, Harvey, Lundblad, and Siegel (2011)
Anti-director index
Ratio of stock market capitalization to GDP
Standard deviation of GDP growth over the last five years.
Log of one-year ahead annual stock market index return
One-year-ahead S&P rating of a country’s government debt scaled by 22
The ratio of the number of days with zero stock returns to the total number of days
with non-missing stock returns in a given year
Average of the daily dollar-volume weighted average of effective spreads in a given year
Average of the daily dollar-volume weighted average of relative quoted spreads in a given year
Total number of shares traded in a given year, scaled by the number of shares outstanding
Probability of informed trading computed following the microstructure model developed by
Easley, Hvidkjaer, and O’Hara (2002)
Ratio of capital expenditures scaled by total assets
Ratio of research and development expenses to total assets
Log of changes in net sales
A country's disclosure requirement index
A country's anti-director index
Sum of dummy variables such as marking the open, marking the close, and so on
Sum of price manipulation rules index, volume manipulation rules index, spoofing rules index,
and false disclosure rule index
An index capturing perceptions of the extent to which public power is exercised for private gain,
including both petty and grand forms of corruption, as well as "capture" of the state by elites and
private interests
Internet Appendix, Page 13 TRTH
TRTH
Datastream
TRTH
Worldscope
Worldscope
Wordscope
La Porta, Lopez-de-Silanes, and Shleifer (2006)
Pagano and Volpin (2005)
Cumming, Johan, and Li (2011)
Cumming, Johan, and Li (2011)
Kaufmann, Kraay, and Mastruzzi (2009)
Table IA2: Summary Statistics of Additional Variables
This table presents the summary statistics for the additional variables used in this study. The variables are
Jones's (1991) residual accruals (AccrualJones), KLW's (2005) residual accruals (AccrualKLW), DD's (2002)
residual accruals (AccrualDD), ETF ownership (ETF), industry-level ETF ownership (ETFIndustry), concentration
of institutional ownership (HHI), shares on loan (On Loan), loan fee (Fee), loan fee volatility (STDFee),
specialness (Specialness), short-selling constraint (Constraint), ISS corporate governance index (ISS), big N
auditor (BigN), international accounting standard (IAS), news coverage (NewsCoverage), analyst dispersion
(Disp), stock price non-synchronicity (Nonsyn), fraud and scandal (Scandal), legality of short selling (Legality),
feasibility of short selling (Feasibility), put option trading (Put), feasibility or put option (F or P), segmentation
(SEG), anti-director index (AntiDirector), market capitalization-to-GDP ratio (MVGDP), standard deviation of
GDP growth (STDGDPG), future stock market return (MKTReturn), future S&P sovereign credit rating
(MKTCreditRating), proportion of zero stock returns (ZRP), effective spread (ESprd), relative quoted spread
(RSprd), stock trading turnover (TV), and probability of informed trading (PIN). All the variables are defined in
Appendix A and Table IA1. The summary statistics include the number of observations (N) and the mean,
median, standard deviation (STD), and decile (10% and 90%) and quartile (25% and 75%) distributions of the
variables.
Internet Appendix, Page 14 Variable
AccrualJones
N
Mean
STD
10%
25%
Median
75%
90%
61,562
0.002
0.080
-0.081
-0.033
0.004
0.037
0.079
AccrualKLW
60,984
-0.002
0.117
-0.130
-0.057
-0.001
0.054
0.125
AccrualDD
ETF
57,603
61,624
0.004
0.010
0.062
0.022
-0.053
0.000
-0.021
0.000
0.001
0.001
0.027
0.011
0.065
0.034
ETFIndustry
HHI
On Loan
Fee
61,624
61,624
61,623
61,575
0.009
0.132
0.018
1.319
0.012
0.202
0.036
1.698
0.000
0.000
0.000
0.119
0.001
0.000
0.001
0.192
0.004
0.052
0.004
0.619
0.013
0.168
0.017
1.914
0.028
0.363
0.049
3.510
STDFee
Specialness
Constraint
ISS
IAS
BigN
NewsCoverage
Disp
Nonsyn
59,011
61,575
61,574
16,184
61,382
60,925
61,624
41,250
59,953
0.477
0.394
0.062
0.561
0.189
0.677
2.497
0.012
1.538
0.710
0.489
0.241
0.133
0.392
0.468
1.727
0.032
1.504
0.020
0.000
0.000
0.366
0.000
0.000
0.000
0.001
0.046
0.062
0.000
0.000
0.439
0.000
0.000
0.693
0.002
0.592
0.215
0.000
0.000
0.561
0.000
1.000
2.708
0.005
1.271
0.614
1.000
0.000
0.659
0.000
1.000
3.850
0.011
2.140
1.256
1.000
0.000
0.732
1.000
1.000
4.585
0.024
3.270
Scandal
16,347
0.005
0.073
0.000
0.000
0.000
0.000
0.000
Legality
166,221
0.948
0.222
1.000
1.000
1.000
1.000
1.000
Feasibility
166,221
0.878
0.327
0.000
1.000
1.000
1.000
1.000
Put Option
F or P
SEG
Anti-Director
MVGDP
STDGDPG
MKTReturn
MKTCreditRating
ZRP
ESprd
RSprd
TV
PIN
166,221
166,221
166,221
166,221
166,221
166,221
166,221
166,221
64,637
44,718
45,761
64,672
36,829
0.925
0.979
0.019
4.309
1.138
0.016
0.114
0.918
0.087
0.011
0.010
1.380
0.211
0.264
0.145
0.014
0.922
0.717
0.015
0.330
0.143
0.115
0.022
0.015
1.741
0.084
1.000
1.000
0.006
3.000
0.466
0.005
-0.300
0.773
0.008
0.001
0.001
0.150
0.114
1.000
1.000
0.010
4.000
0.665
0.008
-0.124
0.864
0.020
0.003
0.002
0.354
0.148
1.000
1.000
0.016
5.000
1.041
0.012
0.115
1.000
0.049
0.006
0.005
0.812
0.200
1.000
1.000
0.024
5.000
1.403
0.017
0.312
1.000
0.107
0.012
0.011
1.733
0.260
1.000
1.000
0.037
5.000
1.744
0.029
0.485
1.000
0.200
0.024
0.024
3.203
0.320
Internet Appendix, Page 15 Table IA3: Mean of Lendable Shares by Year and Market
This table summarizes the mean of lendable shares (Lendable) for all our sample stocks (Total), stocks in the US
(US), and stocks in countries other than the US (NUS) over the 2002-2009 sample period.
Year
2002
2003
2004
2005
2006
2007
2008
2009
Total
0.011
0.013
0.024
0.051
0.077
0.096
0.100
0.097
US
0.004
0.005
0.016
0.065
0.143
0.206
0.207
0.193
Internet Appendix, Page 16 NUS
0.013
0.018
0.028
0.042
0.043
0.046
0.046
0.039
Table IA4: Additional Robustness Tests and Diagnostics of the ETF Instruments
This table presents the results of robustness checks and additional diagnostics for the instrumental variable tests
reported in Table 3. Panels A-D conduct the same instrumental variable test as Panel A of Table 3, using ETF
ownership (ETF) or industry-level ETF ownership (ETFIndustry) as an instrumental variable and presents a panel
regression of a firm's earnings management measure (AccrualMJones) on predicted lendable shares (
) as
follows:
:
,
,
,
,
, ;
:
,
,
,
, ,
where
, refers to lendable shares and
, includes the list of standard control variables. The
difference is that Panel A employs double clustering following Petersen (2009) and Thompson (2009) at both
the firm and year levels, while in Panel B, the tests are conducted in subperiods from 2005 to 2009. Panel C
applies the main instrumental variable to the subsample of stocks that are included in the MSCI country index.
Panel D extends the main test by further controlling for country-level characteristics, including the control-ofcorruption index (Corruption) and the market capitalization-to-GDP ratio (MVGDP).
Internet Appendix, Page 17 Table IA4: Additional Robustness Tests and Diagnostics of the ETF Instruments - Continued
A. ETF and Industry-level ETF as Instrumental Variables (Firm-Year Double Clustering)
Instrument=ETF
Dep. Variable=
Lendable
(1st Stage)
Instrument
AccrualMJones
(2nd Stage)
Instrument=ETFIndustry
Lendable
(1st Stage)
Model
Model
Model
Model
(1)
(2)
(3)
(4)
0.847
3.039
(3.59)
(5.57)
-0.102
-0.041
(-4.83)
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
(-1.80)
-0.004
0.004
-0.004
0.004
(-2.88)
(7.13)
(-4.67)
(7.64)
0.009
0.005
0.008
0.004
(5.55)
(7.98)
(5.43)
(7.02)
0.007
0.011
0.004
0.010
(2.60)
(5.35)
(1.28)
(5.16)
0.004
0.003
0.002
0.003
(1.78)
(3.88)
(0.92)
(3.62)
-0.006
-0.006
-0.005
-0.006
(-1.29)
(-4.12)
(-1.06)
(-3.92)
0.004
-0.013
-0.002
-0.013
(2.37)
(-7.16)
(-0.90)
(-7.37)
0.015
-0.005
0.017
-0.006
(5.22)
(-4.95)
(5.15)
(-5.83)
0.001
-0.001
0.001
-0.001
(4.57)
(-11.20)
(3.50)
(-11.75)
-0.008
-0.004
-0.013
-0.004
(-1.21)
(-2.66)
(-2.22)
(-2.37)
IO
0.115
0.011
0.132
0.002
(3.41)
(3.17)
(3.47)
(0.62)
Illiquidity
-0.007
-0.001
-0.007
-0.001
(-4.29)
(-3.57)
(-6.41)
(-2.29)
ICY
ICY
ICY
ICY
Obs
61,624
61,624
61,624
61,624
AdjRsq
65.6%
2.8%
67.4%
2.9%
Fixed Effects
Internet Appendix, Page 18 AccrualMJones
(2nd Stage)
Table IA4: Additional Robustness Tests and Diagnostics of the ETF Instruments - Continued
B. ETF and Industry-level ETF as Instrumental Variables (2005-2009)
Instrument=ETF
Dep. Variable=
Instrument
Lendable
Instrument=ETFIndustry
(1st Stage)
AccrualMJones
(2nd Stage)
(1st Stage)
Lendable
AccrualMJones
(2nd Stage)
Model
Model
Model
Model
(1)
(2)
(3)
(4)
0.840
2.833
(4.40)
(25.39)
-0.128
-0.060
(-4.92)
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
0.004
-0.004
0.005
(-6.42)
(6.51)
(-7.04)
(6.80)
0.009
0.005
0.009
0.005
(19.27)
(6.90)
(19.28)
(5.93)
0.002
0.009
-0.001
0.009
(0.93)
(3.45)
(-0.26)
(3.45)
-0.001
0.003
-0.002
0.003
(-1.40)
(2.65)
(-3.09)
(2.73)
-0.007
-0.004
-0.007
-0.003
(-6.83)
(-1.96)
(-6.64)
(-1.73)
0.002
-0.013
-0.001
-0.014
(0.84)
(-4.90)
(-0.56)
(-4.99)
0.015
-0.005
0.018
-0.007
(13.23)
(-3.97)
(16.64)
(-4.72)
0.001
-0.001
0.001
-0.001
(8.32)
(-9.38)
(8.68)
(-9.74)
-0.021
-0.007
-0.025
-0.005
(-13.52)
(-2.99)
(-16.26)
(-2.22)
0.144
0.016
0.169
0.004
(19.14)
(3.12)
(51.44)
(0.64)
-0.008
-0.002
-0.009
-0.001
(-20.40)
(-3.67)
(-27.90)
(-2.16)
ICY
ICY
ICY
ICY
Obs
44,171
44,171
44,171
44,171
AdjRsq
74.9%
2.4%
74.9%
2.6%
IO
Illiquidity
Fixed Effects
Internet Appendix, Page 19 (-1.79)
-0.003
Table IA4: Additional Robustness Tests and Diagnostics of the ETF Instruments - Continued
C. ETF and Industry-level ETF as Instrumental Variables for MSCI Indexed Stocks
Instrument=ETF
Instrument=ETFIndustry
Dep. Variable=
Instrument
Size
BM
Leverage
Return
STD
ADR
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Lendable
(1st Stage)
Model
(1)
AccrualMJones
(2nd Stage)
Model
(2)
0.748
(5.17)
AccrualMJones
(2nd Stage)
Model
(4)
3.424
(4.31)
-0.009
(-13.60)
0.011
(17.62)
0.008
(3.46)
0.006
(9.01)
-0.012
(-7.92)
0.007
(3.61)
0.001
(11.79)
-0.015
(-7.57)
0.115
(19.72)
-0.008
(-17.60)
-0.102
(-4.11)
0.002
(3.14)
0.004
(5.37)
0.012
(4.54)
0.001
(0.49)
-0.011
(-5.36)
-0.012
(-5.16)
-0.001
(-6.72)
-0.001
(-0.52)
0.011
(2.52)
-0.001
(-1.66)
-0.008
(-11.51)
0.010
(16.67)
0.006
(2.38)
0.004
(4.61)
-0.012
(-7.77)
0.001
(0.62)
0.001
(11.42)
-0.021
(-9.47)
0.124
(23.67)
-0.009
(-19.85)
-0.058
(-2.82)
0.003
(3.76)
0.004
(4.86)
0.012
(4.43)
0.000
(0.25)
-0.011
(-5.13)
-0.012
(-5.37)
-0.001
(-7.17)
-0.001
(-0.22)
0.005
(1.21)
-0.000
(-0.91)
ICY
40,847
68.7%
ICY
40,847
3.6%
ICY
40,847
71.7%
ICY
40,847
3.9%
Internet Appendix, Page 20 Lendable
(1st Stage)
Model
(3)
Table IA4: Additional Robustness Tests and Diagnostics of the ETF Instruments - Continued
D. ETF and Industry-level ETF as Instrumental Variables with Country-level Controls
Instrument=ETF
Dep. Variable=
Instrument
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Corruption
MVGDP
Fixed Effects
Obs
AdjRsq
Lendable
(1st Stage)
Model
(1)
AccrualMJones
(2nd Stage)
Model
(2)
0.775
(5.34)
AccrualMJones
(2nd Stage)
Model
(4)
2.706
(4.71)
-0.004
(-9.88)
0.008
(20.55)
0.005
(3.07)
0.005
(9.87)
-0.008
(-7.79)
0.005
(2.59)
0.015
(15.46)
0.001
(9.99)
-0.010
(-7.43)
0.112
(21.12)
-0.007
(-22.30)
-0.027
(-27.43)
-0.027
(-28.68)
-0.113
(-4.71)
0.004
(6.39)
0.005
(7.52)
0.010
(4.58)
0.003
(4.02)
-0.007
(-4.18)
-0.013
(-5.61)
-0.005
(-4.32)
-0.001
(-9.68)
-0.005
(-2.62)
0.012
(3.07)
-0.002
(-3.58)
-0.004
(-3.57)
0.001
(0.34)
-0.004
(-7.42)
0.008
(19.39)
0.003
(1.84)
0.003
(4.24)
-0.006
(-5.85)
-0.002
(-0.78)
0.016
(17.91)
0.001
(9.39)
-0.014
(-9.30)
0.131
(44.38)
-0.008
(-25.38)
-0.015
(-4.36)
-0.021
(-11.01)
-0.052
(-1.90)
0.004
(6.83)
0.004
(6.63)
0.010
(4.48)
0.003
(3.62)
-0.006
(-3.88)
-0.013
(-5.79)
-0.006
(-5.11)
-0.001
(-10.04)
-0.004
(-2.27)
0.004
(0.84)
-0.001
(-2.24)
-0.002
(-1.84)
0.002
(1.25)
ICY
61,624
66.9%
ICY
61,624
2.7%
ICY
61,624
67.8%
ICY
61,624
2.9%
Internet Appendix, Page 21 Instrument=ETFIndustry
Lendable
(1st Stage)
Model
(3)
Table IA5: Additional Robustness Tests and Diagnostics of the ETF Instruments
This table provides tests to further complement the diagnostic analyses provided in Panel B of Table 3.
Specifically, Panel B of Table 3 verifies that the ETF instruments lose their explanatory power when the shortselling channel is highly constrained, while this table explores the reverse constraint by regressing AccrualMJones
on Lendable on the sample of stocks with low ETF. The t-statistics reported in parentheses are based on
standard errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year
observations, and AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
Accruals on SSP when ETF is Low
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
0<ETF<0.5%
0<ETFIndustry<0.5%
Model
(1)
Model
(2)
-0.047
(-4.81)
0.006
(8.06)
0.004
(5.64)
0.013
(4.36)
0.004
(4.43)
-0.005
(-2.84)
-0.010
(-3.87)
-0.008
(-5.95)
-0.001
(-7.09)
-0.005
(-2.29)
0.009
(2.89)
-0.000
(-0.38)
-0.046
(-3.54)
0.005
(6.19)
0.005
(5.74)
0.013
(4.13)
0.003
(3.24)
-0.008
(-3.73)
-0.012
(-5.16)
-0.009
(-5.92)
-0.001
(-6.40)
-0.004
(-1.62)
0.019
(3.48)
-0.000
(-0.81)
ICY
39,937
2.7%
ICY
35,664
2.8%
Internet Appendix, Page 22 Table IA6: Additional Instrumental Variables
This table extends the instrumental variable tests reported in Table 3 by using the concentration of institutional
ownership (HHI), both HHI and ETF ownership (ETF), or industry-level ETF ownership (ETFIndustry) as
instrumental variables and presents the results of a panel regression of a firm's earnings management measure
) and firm-level control variables (X), as well as
(AccrualMJones) on predicted lendable shares (
unreported industry, country, and year fixed effects (ICY) on the variation of the following models:
:
,
, , or
,
,
, ;
,
:
,
,
,
,
refers
to
lendable
shares
and
includes
the
list
of
standard
control
variables.
Model (1)
where
,
,
regresses Lendable on HHI. Models (3) and (5) regress Lendable on ETF (ETFIndustry) and HHI, respectively.
Models (2), (4), and (6) regress AccrualMJones on predicted lendable shares. The t-statistics reported in
parentheses are based on standard errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes
the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
Dep. Variable=
HHI
Lendable
(1st Stage)
Model
(1)
AccrualMJones
(2nd Stage)
Model
(2)
-0.017
(-12.90)
Lendable
(1st Stage)
Model
(3)
AccrualMJones
(2nd Stage)
Model
(4)
-0.016
(-12.71)
0.845
(5.45)
ETF
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
3.038
(5.31)
-0.005
(-10.90)
0.009
(20.85)
0.006
(3.43)
0.004
(7.55)
-0.005
(-5.40)
0.005
(3.19)
0.017
(18.63)
0.001
(10.45)
-0.008
(-5.69)
0.146
(60.81)
-0.007
(-25.87)
-0.423
(-3.20)
0.002
(2.60)
0.008
(5.82)
0.012
(5.02)
0.004
(4.54)
-0.008
(-4.59)
-0.011
(-4.45)
0.000
(0.05)
-0.001
(-3.68)
-0.007
(-3.32)
0.058
(3.00)
-0.004
(-3.54)
-0.115
(-5.30)
0.004
(6.41)
0.005
(7.62)
0.011
(4.69)
0.003
(3.92)
-0.006
(-4.06)
-0.013
(-5.63)
-0.005
(-4.29)
-0.001
(-9.43)
-0.005
(-2.50)
0.013
(3.45)
-0.001
(-3.62)
-0.004
(-9.59)
0.009
(21.01)
0.008
(4.22)
0.004
(7.77)
-0.006
(-6.35)
0.004
(2.54)
0.015
(15.73)
0.001
(11.70)
-0.007
(-5.54)
0.116
(20.16)
-0.006
(-19.56)
Hansen J statistics
[P-value]
Fixed Effects
Obs
AdjRsq
ICY
61,624
63.3%
ICY
61,624
2.6%
ICY
61,624
65.7%
-0.050
(-2.33)
0.004
(6.90)
0.004
(6.75)
0.010
(4.53)
0.003
(3.65)
-0.006
(-3.86)
-0.013
(-5.81)
-0.006
(-5.25)
-0.001
(-9.97)
-0.004
(-2.21)
0.004
(0.97)
-0.001
(-2.36)
-0.004
(-7.29)
0.008
(19.48)
0.005
(2.41)
0.002
(3.75)
-0.005
(-5.40)
-0.002
(-1.18)
0.017
(18.24)
0.001
(9.84)
-0.013
(-7.94)
0.133
(40.50)
-0.007
(-25.14)
[0.365]
Internet Appendix, Page 23 AccrualMJones
(2nd Stage)
Model
(6)
-0.017
(-13.46)
ETFIndustry
Size
Lendable
(1st Stage)
Model
(5)
[0.253]
ICY
61,624
2.7%
ICY
61,624
67.5%
ICY
61,624
2.9%
Table IA7: Robustness Checks on Hong Kong Short-selling
This table provides two robustness checks for the tests utilizing the unique regulatory setting in the Hong Kong
market in which regulators changed the list of stocks eligible for short selling on a quarterly frequency from
1994 to 2005. The first robustness check, reported in Panel A, excludes the observations of eligible firms for the
year prior to their inclusion to the short selling list to eliminate potential anticipation effect. In the second test,
reported in Panel B, we create a control group for eligible stocks based on propensity score matching. More
specifically, we match each eligible stock with a non-eligible stock of the same period based on Size and TV.
Based on the remaining data, Models (1)-(3) estimate the following panel regression with firm and year fixed
effects (FY) and clustered standard errors at the firm and industry levels:
,
,
,
,
, .
where
is modified Jones's (1991) residual accruals,
, is a dummy variable that
,
equals one if a stock is eligible to short selling in year t, and
, is a dummy variable that equals one if
a stock is eligible for short selling in year t-1 but becomes ineligible for short selling beginning in year t.
Models (4) and (5) estimate the following panel regressions:
Δ , Δ Δ ,
Δ
,
,
, .
where
, refers to net inclusion and equals one (negative one) if a firm is included in (excluded from)
the eligible list and Δ , is a dummy variable for exclusion. The control variables are detailed in
Appendix A. The t-statistics reported in parentheses are based on standard errors adjusted for heteroskedasticity
and firm-level clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2.
A. Excluding the Year Before the Inclusion of a Stock to the Short Selling List
Dep. Variable=
AccrualMJones
∆AccrualMJones
HK SS=0
Model
Model
Model
Model
Model
(1)
(2)
(3)
(4)
(5)
HK SS
-0.029
(-2.44)
∆HK SSExclusion
0.047
(6.05)
0.016
(2.34)
0.133
(3.49)
-0.004
(-0.53)
0.006
(0.58)
-0.142
(-5.20)
0.000
(0.23)
0.027
(1.05)
0.002
(0.80)
0.002
(0.12)
0.049
(5.48)
0.015
(1.84)
0.146
(3.18)
-0.003
(-0.41)
0.007
(0.66)
-0.151
(-3.23)
-0.001
(-1.03)
0.033
(1.14)
0.001
(0.42)
FY
4,155
11.3%
FY
4,155
11.3%
FY
3,483
10.3%
Fixed Effects
Obs
AdjRsq
HK SSPost
Size
BM
Leverage
Return
STD
ADR
Analyst
CH
Illiquidity
Fixed Effects
Obs
AdjRsq
∆HK SS
-0.031
(-2.49)
-0.008
(-0.69)
0.047
(6.05)
0.017
(2.35)
0.133
(3.50)
-0.004
(-0.55)
0.005
(0.58)
-0.142
(-5.23)
0.000
(0.24)
0.027
(1.06)
0.002
(0.80)
∆Size
∆BM
∆Leverage
∆Return
∆STD
∆ADR
∆Analyst
∆CH
∆Illiquidity
Internet Appendix, Page 24 -0.034
(-2.19)
0.082
(8.10)
0.046
(4.03)
0.313
(5.26)
-0.003
(-0.40)
0.014
(1.16)
-0.127
(-1.28)
-0.000
(-0.27)
0.023
(0.74)
0.003
(0.75)
0.034
(2.19)
0.082
(8.10)
0.046
(4.03)
0.313
(5.26)
-0.003
(-0.40)
0.014
(1.16)
-0.127
(-1.28)
-0.000
(-0.27)
0.023
(0.74)
0.003
(0.75)
IY
3,345
5.7%
IY
3,345
5.7%
Table IA7: Robustness Checks on Hong Kong Short-selling – Continued
Dep. Variable=
HK SS
B. Robustness Checks based on the Propensity Score-matched Sample
AccrualMJones
∆AccrualMJones
HK SS=0
Model
Model
Model
Model
Model
(1)
(2)
(3)
(4)
(5)
-0.018
(-1.82)
∆HK SSExclusion
0.033
(3.98)
0.002
(0.20)
0.122
(3.35)
-0.001
(-0.17)
0.000
(0.01)
-0.146
(-5.08)
-0.000
(-0.18)
-0.002
(-0.08)
0.001
(0.52)
-0.001
(-0.10)
0.033
(3.16)
0.001
(0.13)
0.122
(2.52)
-0.004
(-0.53)
0.004
(0.32)
-0.143
(-3.08)
-0.001
(-1.04)
0.004
(0.13)
-0.001
(-0.23)
FY
1,870
14.0%
FY
1,870
14.0%
FY
1,870
14.8%
Fixed Effects
Obs
AdjRsq
HK SSPost
Size
BM
Leverage
Return
STD
ADR
Analyst
CH
Illiquidity
Fixed Effects
Obs
AdjRsq
∆HK SS
-0.020
(-1.92)
-0.009
(-0.78)
0.033
(3.97)
0.002
(0.22)
0.122
(3.35)
-0.001
(-0.19)
0.000
(0.01)
-0.145
(-5.12)
-0.000
(-0.17)
-0.002
(-0.08)
0.001
(0.51)
Internet Appendix, Page 25 ∆Size
∆BM
∆Leverage
∆Return
∆STD
∆ADR
∆Analyst
∆CH
∆Illiquidity
-0.021
(-1.80)
0.085
(7.28)
0.044
(4.01)
0.327
(5.68)
-0.000
(-0.05)
0.001
(0.11)
-0.107
(-1.13)
-0.001
(-1.18)
0.019
(0.63)
0.001
(0.31)
0.032
(1.97)
0.085
(7.33)
0.044
(4.02)
0.327
(5.68)
0.000
(0.02)
0.001
(0.12)
-0.105
(-1.10)
-0.001
(-1.35)
0.019
(0.63)
0.001
(0.27)
IY
1,870
6.9%
IY
1,870
6.9%
Table IA8: Market-wide Short Selling and Earnings Management
This table explores how regulations that constrain short selling are, on average, related to earnings management.
Panel A presents the results of a panel regression of a firm's modified Jones's (1991) residual accruals
(AccrualMJones) on market-wide short-selling variables, firm-level control variables (X), and country-level control
variables (C), as well as unreported industry, country, and year fixed effects (ICY) on the variation of the
following model:
,
,
,
,
, .
, includes the legality of short selling (Legality), feasibility of short selling (Feasibility), put option
trading (Put), and feasibility or put option (F or P). , refers to the list of market-level control variables,
including segmentation (SEG), anti-director index (AntiDirector), market capitalization-to-GDP ratio (MVGDP),
standard deviation of GDP growth (STDGDPG), future stock market return (MKTReturn), and future S&P
sovereign credit rating (MKTCreditRating). , includes the same list of firm control variables as before. The
construction of these variables is detailed in Appendix A and Table IA1. Panel B repeats the same regression for
the subsample of firms that have American Depository Receipts (ADRs) traded in the US. The t-statistics
reported in parentheses are based on standard errors adjusted for heteroskedasticity and firm-level clustering.
Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is from 1990
to 2009.
Internet Appendix, Page 26 Model
(1)
Legality
A. Full Sample
Model
Model
(2)
(3)
-0.011
(-2.89)
Feasibility
-0.014
(-3.66)
MVGDP
STDGDPG
MKTReturn
MKTCreditRating
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
Illiquidity
Fixed Effects
Obs
AdjRsq
-0.007
(-1.77)
-0.037
(-3.04)
0.172
(4.45)
0.003
(3.01)
0.002
(1.63)
0.038
(1.19)
-0.010
(-7.07)
0.044
(6.98)
0.004
(11.89)
0.003
(7.40)
0.012
(7.21)
0.007
(12.58)
-0.008
(-8.11)
-0.013
(-7.71)
-0.008
(-10.58)
-0.001
(-16.76)
-0.005
(-4.65)
-0.001
(-4.23)
0.167
(4.32)
0.003
(2.93)
0.002
(1.67)
0.035
(1.08)
-0.010
(-6.81)
0.043
(6.67)
0.004
(11.88)
0.003
(7.31)
0.012
(7.17)
0.007
(12.57)
-0.008
(-8.06)
-0.013
(-7.74)
-0.008
(-10.57)
-0.001
(-16.73)
-0.005
(-4.63)
-0.001
(-4.26)
0.162
(4.19)
0.003
(3.15)
0.002
(1.45)
0.017
(0.54)
-0.010
(-6.98)
0.045
(6.99)
0.004
(11.88)
0.003
(7.36)
0.012
(7.17)
0.007
(12.58)
-0.008
(-8.08)
-0.013
(-7.69)
-0.008
(-10.56)
-0.001
(-16.74)
-0.005
(-4.66)
-0.001
(-4.25)
-0.012
(-3.02)
0.155
(4.01)
0.003
(2.93)
0.002
(1.41)
0.015
(0.46)
-0.010
(-6.89)
0.046
(7.15)
0.004
(11.89)
0.003
(7.40)
0.012
(7.16)
0.007
(12.54)
-0.008
(-8.11)
-0.013
(-7.68)
-0.008
(-10.56)
-0.001
(-16.70)
-0.005
(-4.65)
-0.001
(-4.21)
ICY
166,221
3.5%
ICY
166,221
3.5%
ICY
166,221
3.5%
ICY
166,221
3.5%
Internet Appendix, Page 27 Model
(8)
-0.017
(-2.22)
F or P
Anti-Director
Model
(5)
-0.020
(-1.94)
Put Option
SEG
Model
(4)
B. Firms with ADRs
Model
Model
(6)
(7)
0.255
(1.47)
0.003
(0.67)
0.004
(0.86)
0.252
(1.88)
-0.009
(-1.24)
0.007
(0.24)
0.002
(1.13)
0.001
(0.28)
0.004
(0.42)
0.006
(2.03)
-0.015
(-2.57)
0.255
(1.47)
0.003
(0.66)
0.005
(0.88)
0.215
(1.57)
-0.009
(-1.17)
0.014
(0.45)
0.002
(1.10)
0.000
(0.27)
0.004
(0.39)
0.006
(2.04)
-0.014
(-2.51)
0.265
(1.53)
0.003
(0.89)
0.004
(0.70)
0.197
(1.47)
-0.010
(-1.31)
0.016
(0.54)
0.002
(1.09)
0.001
(0.27)
0.004
(0.44)
0.007
(2.05)
-0.014
(-2.50)
-0.053
(-6.32)
0.254
(1.47)
0.003
(0.76)
0.003
(0.68)
0.194
(1.49)
-0.009
(-1.25)
0.015
(0.51)
0.002
(1.10)
0.001
(0.28)
0.004
(0.45)
0.006
(2.03)
-0.014
(-2.53)
-0.002
(-0.39)
-0.001
(-3.66)
-0.006
(-0.86)
-0.000
(-0.05)
-0.002
(-0.40)
-0.001
(-3.58)
-0.006
(-0.87)
-0.000
(-0.07)
-0.002
(-0.38)
-0.001
(-3.50)
-0.005
(-0.82)
-0.000
(-0.04)
-0.002
(-0.38)
-0.001
(-3.49)
-0.005
(-0.81)
-0.000
(-0.02)
ICY
4,383
11.3%
ICY
4,383
11.3%
ICY
4,383
11.3%
ICY
4,383
11.3%
Table IA9: The Impact of Accounting and Trading Regulations
This table explores the disciplining impact of lendable shares in subsamples of countries with different
accounting and trading regulations. Models (1) and (2) report the regression coefficients when our main tests are
conducted in countries with low and high disclosure requirement index (Disclosure), respectively. The line
“Difference” further reports the difference between the two sensitivity parameters of Accruals with respect to
SSP, i.e., Model (1) sensitivity minus Model (2) sensitivity, followed by the p-value of the difference. Likewise,
Models (3) and (4), Models (5) and (6), and Models (7) and (8) report the results in the subsamples of countries
with low anti-director index (Anti-Director), with weak and strong price manipulation index (PMI), and with
weak and strong market manipulation index (MMI), respectively. The construction of these variables is detailed
in Table IA1. The t-statistics reported in parentheses are based on standard errors adjusted for heteroskedasticity
and firm-level clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The
sample period is from 2002 to 2009.
Lendable
Disclosure
Low
High
Model
Model
(1)
(2)
Anti-Director
Low
High
Model
Model
(3)
(4)
Low
Model
(5)
High
Model
(6)
Low
Model
(7)
High
Model
(8)
-0.083
(-2.72)
-0.061
(-2.52)
-0.075
(-3.42)
-0.050
(-7.70)
-0.073
(-3.39)
-0.049
(-7.60)
Difference
[P-value]
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed
Effects
Obs
AdjRsq
-0.044
(-7.60)
-0.039
[0.051]
-0.043
(-7.24)
-0.018
[0.171]
MMI
-0.025
[0.072]
-0.024
[0.081]
0.001
(0.76)
0.005
(2.37)
0.010
(1.19)
0.012
(3.24)
-0.014
(-1.69)
-0.014
(-2.89)
-0.010
(-2.88)
-0.000
(-1.50)
0.003
(0.50)
0.038
(3.25)
-0.003
(-2.78)
0.005
(7.36)
0.004
(6.57)
0.010
(4.46)
0.002
(2.70)
-0.006
(-3.57)
-0.013
(-5.22)
-0.006
(-4.65)
-0.001
(-11.09)
-0.005
(-2.62)
0.002
(0.67)
-0.001
(-1.94)
0.000
(0.32)
0.005
(2.71)
0.018
(3.00)
0.013
(4.87)
-0.014
(-2.70)
-0.014
(-3.50)
-0.006
(-2.48)
-0.000
(-0.65)
0.002
(0.51)
0.029
(2.99)
-0.001
(-1.69)
0.005
(7.73)
0.004
(6.66)
0.009
(3.58)
0.002
(2.00)
-0.005
(-3.12)
-0.013
(-4.77)
-0.006
(-4.85)
-0.001
(-11.66)
-0.005
(-2.57)
0.002
(0.67)
-0.001
(-1.96)
0.005
(5.38)
0.004
(3.57)
0.007
(1.86)
-0.003
(-2.10)
-0.009
(-2.99)
-0.012
(-3.24)
-0.010
(-5.63)
-0.001
(-6.45)
0.001
(0.30)
0.057
(4.92)
-0.000
(-0.45)
0.003
(4.41)
0.005
(6.74)
0.013
(4.70)
0.006
(5.60)
-0.006
(-3.18)
-0.012
(-4.43)
-0.005
(-3.10)
-0.001
(-7.52)
-0.006
(-2.57)
0.000
(0.03)
-0.001
(-2.60)
0.005
(4.97)
0.003
(3.11)
0.006
(1.62)
-0.003
(-1.97)
-0.009
(-2.97)
-0.015
(-4.09)
-0.009
(-5.39)
-0.001
(-6.24)
0.001
(0.26)
0.054
(4.79)
-0.000
(-0.69)
0.003
(4.66)
0.005
(6.92)
0.013
(4.76)
0.006
(5.59)
-0.006
(-3.14)
-0.011
(-3.94)
-0.005
(-3.16)
-0.001
(-7.56)
-0.006
(-2.53)
0.000
(0.00)
-0.001
(-2.39)
ICY
ICY
ICY
ICY
ICY
ICY
ICY
ICY
6,564
4.8%
55,060
2.8%
11,328
4.1%
50,296
2.9%
19,761
3.2%
41,863
3.3%
20,435
3.3%
41,189
3.2%
Internet Appendix, Page 28 PMI
Table IA10: Additional Subsample Analyses and Controls
This table extends the main analyses (Table 2) and subsample analyses (Table 6) by examining the impact of
short selling on earnings management in subsamples of countries sorted by the feasibility of short selling and by
controlling for firm-level investments. In Panel A, GFC refers to the global financial crisis period from 2007 to
2008, whereas Ex.GFC excludes the global financial crisis period. Ex.Zeros only includes firms with nonzero
short-selling values. The variable “Dummy (2004-2006)” refers to the dummy variable that takes the value of
one during the period from 2004 to 2006, and zero otherwise. Variables Dummy (2007-2008) and Dummy
(2009) are defined similarly. In Panel B, Model (1) tabulates the results of the baseline test conducted at the
whole sample as a benchmark. Models (2) to (5) apply the same test to subsets of countries. More specifically,
“Five Countries” in Model (2) refer to the subsample of five countries in which short selling is not actively
practiced according to Maffett, Owens, and Srinivasan (2014), including Finland, Indonesia, Philippines, South
Korea, and Taiwan. “Ex. Five Countries” in Model (3) refer to the remaining countries. In Models (4) and (5),
“Feasibility= 0” refers to the subsample of countries in which, according to Charoenrook and Daouk (2005),
short selling is not feasible, whereas “Feasibility= 1” refers to the rest of countries. Models (6) and (7) extend
the whole-sample baseline test to include variables describing the growth opportunity and investment behavior
of firms. These variables are CapEx, the ratio of capital expenditures scaled by total assets, R&D, the ratio of
research and development expenses to total assets, and SalesGrowth, the log of changes in net sales. The
construction of these variables is detailed in Table IA1. The t-statistics reported in parentheses are based on
standard errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year
observations, and AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
Internet Appendix, Page 29 Table IA10 – Continued
A. More Subsample Analyses on Different Time Periods
Ex. GFC
GFC
Ex. Zeros
Model
Model
Model
(1)
(2)
(3)
Lendable
-0.049
(-6.73)
-0.029
(-2.69)
-0.042
(-7.40)
0.004
(5.14)
0.004
(4.90)
0.007
(2.81)
0.002
(2.10)
-0.008
(-4.45)
-0.012
(-5.20)
-0.006
(-4.59)
-0.001
(-8.41)
-0.003
(-1.51)
0.005
(2.09)
-0.001
(-2.32)
0.005
(5.46)
0.006
(5.43)
0.018
(4.69)
0.006
(4.14)
-0.000
(-0.08)
-0.017
(-4.17)
-0.007
(-3.22)
-0.001
(-7.05)
-0.006
(-1.82)
-0.006
(-1.25)
-0.000
(-0.69)
0.003
(5.89)
0.004
(6.65)
0.010
(4.45)
0.003
(3.87)
-0.006
(-3.19)
-0.013
(-5.80)
-0.006
(-5.33)
-0.001
(-10.00)
-0.003
(-1.43)
0.003
(1.56)
-0.001
(-3.17)
0.036
(1.08)
-0.070
(-2.18)
-0.081
(-2.42)
-0.092
(-2.72)
0.004
(7.10)
0.004
(6.90)
0.010
(4.51)
0.003
(3.64)
-0.006
(-3.81)
-0.013
(-5.90)
-0.006
(-5.59)
-0.001
(-10.29)
-0.004
(-2.04)
0.003
(1.43)
-0.001
(-2.37)
ICY
42,684
2.8%
ICY
18,940
3.4%
ICY
57,593
3.0%
ICY
61,624
2.9%
Lendable×Dummy (2004-2006)
Lendable×Dummy (2007-2008)
Lendable×Dummy (2009)
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Internet Appendix, Page 30 Full Sample
Model
(4)
Table IA10 – Continued
B. More Subsample Analyses on Short Selling Feasibility and Additional Controls
Full
Five
Ex. Five
Feasibility
Feasibility
Full
Sample
Countries
Countries
=0
=1
Sample
Model
Model
Model
Model
Model
Model
(1)
(2)
(3)
(4)
(5)
(6)
Lendable
-0.044
(-7.88)
0.053
(0.75)
-0.042
(-7.52)
0.055
(0.86)
-0.044
(-7.70)
CapEx
-0.037
(-6.76)
-0.175
(-19.63)
R&D
SalesGrowth
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
0.004
(7.10)
0.004
(6.96)
0.010
(4.51)
0.003
(3.63)
-0.006
(-3.85)
-0.013
(-5.82)
-0.006
(-5.57)
-0.001
(-10.29)
-0.004
(-2.19)
0.003
(1.26)
-0.001
(-2.48)
-0.002
(-0.77)
0.009
(2.85)
0.043
(3.92)
0.013
(2.63)
0.018
(1.72)
-0.026
(-2.75)
0.010
(1.96)
0.000
(0.57)
0.007
(0.85)
0.047
(1.74)
0.001
(0.44)
0.004
(7.57)
0.004
(6.59)
0.009
(3.81)
0.003
(3.20)
-0.007
(-4.10)
-0.011
(-5.11)
-0.007
(-6.12)
-0.001
(-10.78)
-0.004
(-2.29)
0.003
(1.22)
-0.001
(-2.36)
0.000
(0.18)
0.011
(3.88)
0.027
(2.39)
0.008
(1.64)
0.022
(2.41)
-0.013
(-2.18)
0.006
(1.40)
-0.000
(-0.44)
-0.008
(-1.13)
0.017
(0.76)
0.001
(0.35)
0.004
(7.30)
0.004
(6.31)
0.009
(3.91)
0.003
(3.42)
-0.007
(-4.20)
-0.012
(-5.16)
-0.007
(-5.94)
-0.001
(-10.64)
-0.004
(-1.87)
0.003
(1.21)
-0.001
(-2.49)
0.019
(12.63)
0.004
(6.39)
0.005
(7.37)
0.016
(7.34)
0.002
(2.88)
-0.007
(-4.50)
-0.014
(-6.30)
-0.006
(-5.30)
-0.001
(-9.28)
-0.001
(-0.53)
0.002
(1.00)
-0.001
(-3.05)
ICY
61,624
2.9%
ICY
2,776
6.2%
ICY
58,848
2.9%
ICY
3,333
6.0%
ICY
58,291
2.8%
ICY
59,925
5.2%
Internet Appendix, Page 31 Full
Sample
Model
(7)
-0.037
(-6.76)
-0.177
(-19.83)
-0.045
(-6.58)
0.019
(12.49)
0.003
(5.84)
0.004
(5.98)
0.014
(6.32)
0.002
(2.32)
-0.007
(-4.04)
-0.013
(-5.88)
-0.005
(-4.96)
-0.001
(-8.78)
-0.001
(-0.65)
0.002
(0.72)
-0.001
(-3.00)
ICY
59,895
5.3%
Table IA11: Alternative Illiquidity Measures
This table examines the baseline effect of short selling on earnings management with alternative illiquidity
measures as control. The specification is based on a panel regression of a firm's modified Jones's (1991) residual
accruals (AccrualMJones) on lendable shares (Lendable), and firm-level control variables (X) as well as unreported
industry, country, and year fixed effects (ICY). Alternative illiquidity measures include proportion of zero stock
returns (ZRP), effective spread (ESprd), relative quoted spread (RSprd), stock trading turnover (TV), and
probability of informed trading (PIN). The construction of these variables is detailed in Table IA1. The tstatistics reported in parentheses are based on standard errors adjusted for heteroskedasticity and firm-level
clustering. Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is
from 2002 to 2009.
Lendable
ZRP
Model
(1)
-0.042
(-7.71)
-0.017
(-3.41)
ESprd
Model
(2)
-0.063
(-7.96)
Model
(3)
-0.058
(-7.68)
Model
(4)
-0.043
(-7.77)
-0.049
(-2.03)
RSprd
-0.184
(-3.93)
TV
0.001
(3.43)
PIN
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Fixed Effects
Obs
AdjRsq
0.005
(11.88)
0.004
(6.66)
0.010
(4.70)
0.003
(3.46)
-0.007
(-4.34)
-0.013
(-6.21)
-0.006
(-5.89)
-0.001
(-10.37)
-0.005
(-3.17)
0.004
(1.61)
0.005
(10.85)
0.004
(4.85)
0.012
(4.44)
0.001
(0.59)
-0.007
(-3.97)
-0.012
(-4.90)
-0.008
(-6.15)
-0.001
(-9.05)
-0.005
(-2.64)
0.009
(2.79)
0.005
(9.68)
0.004
(5.14)
0.012
(4.61)
0.000
(0.47)
-0.007
(-3.84)
-0.012
(-5.29)
-0.008
(-6.45)
-0.001
(-8.90)
-0.005
(-2.51)
0.008
(2.28)
0.005
(12.82)
0.004
(6.80)
0.010
(4.56)
0.003
(3.48)
-0.008
(-4.96)
-0.012
(-5.90)
-0.006
(-5.65)
-0.001
(-10.65)
-0.005
(-2.81)
0.002
(1.02)
-0.000
(-0.00)
0.005
(9.49)
0.003
(3.85)
0.014
(4.88)
-0.000
(-0.15)
-0.010
(-4.68)
-0.010
(-4.06)
-0.007
(-4.79)
-0.001
(-8.46)
-0.006
(-2.57)
0.008
(2.33)
ICY
64,637
2.8%
ICY
44,718
2.8%
ICY
45,761
2.9%
ICY
64,672
2.8%
ICY
36,829
2.8%
Internet Appendix, Page 32 Model
(5)
-0.057
(-6.90)
Table IA12: Additional Accrual Measures
This table augments Panel C of Table 7 in providing two more alternative measures of accruals, including
Francis et al.'s (2005) residual accruals (AccrualFLOS), where residual accruals are obtained by regressing
changes in working capital on past, current, and future cash flows, revenue growth, and fixed assets for each
country and year; and Allent, Larson, and Sloan's (2013)residual accruals (AccrualALS), where residual accruals
are obtained by regressing changes in working capital on past, current, and future cash flows, revenue growth,
and employee growth for each country and year. The rest of specifications are the same as Panel C of Table 7.
Dep. Variable=
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
AccrualFLOS
Model
(1)
AccrualALS
Model
(2)
-0.009
(-1.96)
0.001
(3.35)
0.001
(1.85)
0.006
(3.76)
0.002
(3.07)
-0.006
(-4.61)
-0.002
(-1.16)
-0.003
(-3.45)
-0.000
(-4.03)
0.003
(1.91)
-0.001
(-0.56)
-0.000
(-1.12)
-0.013
(-2.61)
0.000
(0.91)
-0.002
(-3.41)
-0.001
(-0.64)
0.004
(5.81)
-0.007
(-4.72)
-0.003
(-1.74)
-0.003
(-3.28)
-0.000
(-4.31)
0.004
(2.52)
-0.000
(-0.24)
-0.001
(-2.81)
ICY
57,553
0.9%
ICY
52,771
1.1%
Internet Appendix, Page 33 Table IA13: Alternative Clustering Specifications
This table presents a series of robustness checks based on alternative clustering specifications of the main
regression model, allowing for double-clustered standard errors following Pedersen (2009). The t-statistics
reported in parentheses are based on standard errors adjusted for heteroskedasticity and firm-level clustering.
Obs denotes the number of firm-year observations, and AdjRsq is adjusted R2. The sample period is from 2002
to 2009.
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Country
Clustering
Model
(1)
Year
Clustering
Model
(2)
Firm-Year
Clustering
Model
(3)
Country-Year
Clustering
Model
(4)
Industry-Year
Clustering
Model
(5)
-0.044
(-9.00)
0.004
(3.66)
0.004
(4.01)
0.010
(2.85)
0.003
(1.17)
-0.006
(-3.52)
-0.013
(-2.96)
-0.006
(-2.88)
-0.001
(-4.34)
-0.004
(-1.93)
0.003
(0.78)
-0.001
(-1.46)
-0.044
(-4.41)
0.004
(5.14)
0.004
(8.50)
0.010
(1.83)
0.003
(1.52)
-0.006
(-2.50)
-0.013
(-4.84)
-0.006
(-7.06)
-0.001
(-7.12)
-0.004
(-2.08)
0.003
(0.97)
-0.001
(-5.67)
-0.044
(-4.41)
0.004
(4.93)
0.004
(7.77)
0.010
(1.80)
0.003
(1.52)
-0.006
(-2.47)
-0.013
(-4.31)
-0.006
(-6.18)
-0.001
(-6.67)
-0.004
(-1.94)
0.003
(0.90)
-0.001
(-4.54)
-0.044
(-6.02)
0.004
(3.47)
0.004
(4.35)
0.010
(1.79)
0.003
(1.04)
-0.006
(-2.84)
-0.013
(-2.84)
-0.006
(-3.11)
-0.001
(-4.11)
-0.004
(-1.89)
0.003
(0.87)
-0.001
(-1.96)
-0.044
(-3.19)
0.004
(3.09)
0.004
(4.39)
0.010
(1.26)
0.003
(1.24)
-0.006
(-1.58)
-0.013
(-3.14)
-0.006
(-4.92)
-0.001
(-6.95)
-0.004
(-1.27)
0.003
(1.13)
-0.001
(-1.33)
ICY
61,624
2.9%
ICY
61,624
2.9%
ICY
61,624
2.9%
ICY
61,624
2.9%
ICY
61,624
2.9%
Internet Appendix, Page 34 Part 2: Full Specifications of Reported Tables
Table 3 (Panel B): Full Specifications
Panel A addresses the endogeneity problem by using ETF ownership (ETF) or industry-level ETF ownership
(ETFIndustry) as an instrumental variable and presents a panel regression of a firm's earnings management
measure (AccrualMJones) on predicted lendable shares (
) and firm-level control variables (X) as well as
unreported industry, country, and year fixed effects (ICY) on the variation of the following models:
:
,
,
,
,
,
;
:
,
,
,
, ,
refers
to
lendable
shares
and
includes
the
list
of
standard
control
variables.
Models (1)
where
,
,
and (3) regress lendable shares on ETF ownership and industry-level ETF ownership, respectively. Models (2)
and (4) regress modified Jones's (1991) residual accruals on predicted lendable shares. Panel B provides the
diagnostic analyses on the impact of ETF and ETFIndustry on AccrualMJones. Models (1) and (6) directly regress
AccrualMJones on the two instruments. Models (2) and (7) also include lendable shares in the same regression.
The remaining models regress AccrualMJones on the two instruments for subsamples of the stocks for which short
selling is either prohibited owing to regulation (Legality=0 or SSban=1) or low—when very few shares could be
lent out (0<Lendable<0.5%). The t-statistics reported in parentheses are based on standard errors adjusted for
heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year observations, and AdjRsq is
adjusted R2. The sample period is from 2002 to 2009.
ETF
Model
(1)
Model
(2)
-0.086
(-3.57)
-0.016
(-0.75)
B. Tests on Exclusion Restrictions (Accruals regress on ETFs)
Accruals on ETFs when SSP is Low
Accruals on Industry ETF when SSP is Low
Legality=0
SSban=1
0<Lendable<0.5%
Legality=0
SSban=1
0<Lendable<0.5%
Model
Model
Model
Model
Model
Model
Model
Model
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
-0.371
(-0.50)
-0.300
(-1.73)
0.010
(0.21)
ETFIndustry
0.004
(7.34)
0.004
(6.40)
0.010
(4.35)
0.003
(3.45)
-0.006
(-3.67)
-0.013
(-5.91)
-0.007
(-6.06)
-0.001
(-10.85)
-0.004
(-2.02)
-0.001
(-0.25)
-0.001
(-1.85)
-0.040
(-6.99)
0.004
(7.05)
0.004
(6.93)
0.010
(4.47)
0.003
(3.62)
-0.006
(-3.81)
-0.013
(-5.81)
-0.006
(-5.52)
-0.001
(-10.37)
-0.004
(-2.19)
0.004
(1.78)
-0.001
(-2.57)
0.007
(1.52)
0.002
(0.29)
0.081
(3.66)
-0.005
(-0.61)
-0.024
(-1.27)
-0.011
(-0.74)
-0.023
(-2.63)
-0.003
(-2.67)
-0.018
(-1.10)
0.055
(1.37)
-0.003
(-1.10)
0.009
(3.83)
0.014
(5.23)
0.041
(4.39)
-0.000
(-0.15)
-0.003
(-0.47)
-0.030
(-2.93)
-0.013
(-2.91)
-0.002
(-4.20)
0.010
(1.40)
0.091
(3.28)
0.001
(0.42)
0.008
(7.13)
0.006
(4.56)
0.007
(1.84)
0.004
(2.31)
-0.007
(-2.70)
-0.013
(-2.44)
-0.007
(-3.30)
-0.001
(-4.28)
-0.010
(-2.88)
0.001
(0.10)
-0.000
(-0.00)
0.004
(7.37)
0.004
(6.39)
0.010
(4.43)
0.003
(3.52)
-0.006
(-3.71)
-0.013
(-5.80)
-0.007
(-6.24)
-0.001
(-10.73)
-0.003
(-1.88)
-0.003
(-1.52)
-0.001
(-1.60)
0.011
(0.14)
-0.044
(-7.03)
0.004
(7.09)
0.004
(6.96)
0.010
(4.51)
0.003
(3.62)
-0.006
(-3.85)
-0.013
(-5.83)
-0.006
(-5.56)
-0.001
(-10.29)
-0.004
(-2.20)
0.003
(1.27)
-0.001
(-2.49)
ICY
61,624
2.9%
ICY
61,624
2.9%
ICY
1,025
8.7%
ICY
3,159
5.0%
ICY
16,578
2.7%
ICY
61,624
2.8%
ICY
61,624
2.9%
-0.124
(-2.01)
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Internet Appendix, Page 35 -0.857
(-0.23)
0.885
(0.31)
0.266
(1.29)
0.007
(1.50)
0.002
(0.28)
0.081
(3.67)
-0.005
(-0.62)
-0.024
(-1.26)
-0.011
(-0.74)
-0.023
(-2.61)
-0.003
(-2.85)
-0.017
(-1.06)
0.053
(1.31)
-0.003
(-1.05)
0.008
(3.71)
0.014
(5.12)
0.041
(4.30)
-0.000
(-0.10)
-0.003
(-0.47)
-0.031
(-3.07)
-0.012
(-2.80)
-0.002
(-4.55)
0.011
(1.51)
0.087
(3.15)
0.001
(0.56)
0.008
(7.15)
0.006
(4.52)
0.007
(1.78)
0.004
(2.27)
-0.006
(-2.70)
-0.014
(-2.63)
-0.007
(-3.30)
-0.001
(-4.33)
-0.010
(-2.98)
0.002
(0.35)
-0.000
(-0.08)
ICY
1,025
8.7%
ICY
3,159
4.9%
ICY
16,578
2.8%
Table 6 (Full Specifications): Subsample Analyses
This table examines the impact of short selling on earnings management in several important subsamples. Panel
A explores the impact of short selling in subperiods, when it is interacted with various time dummies, and
different regions. In the first two columns, the variable “≥2005” refers the sample period from 2005 to 2009, and
T equals the year minus 2001. In Models (3) and (4), “US” and “NUS” refer to the subsamples of US and noneUS firms, respectively. Panel B explores the source of the effects of short selling on earnings management by
dividing the full sample into the accrual and firm size subsamples. AccrualMJones≥0 refers to firms with positive
AccrualMJones, whereas AccrualMJones<0 refers to firms with negative AccrualMJones. Small firms are firms with
market capitalization below the median value of market capitalization for each country and year, whereas large
firms are firms with market capitalization greater than the median value of market capitalization for each
country and year. The t-statistics reported in parentheses are based on standard errors adjusted for
heteroskedasticity and firm-level clustering. Obs denotes the number of firm-year observations, and AdjRsq is
adjusted R2. The sample period is from 2002 to 2009.
A. Subsample Analyses on Different Time Periods and Regions
>=2005
Full Sample
US
Model
Model
Model
(1)
(2)
(3)
Lendable
-0.045
(-6.37)
-0.003
(-0.14)
-0.006
(-2.17)
-0.064
(-7.39)
-0.044
(-3.52)
0.005
(7.05)
0.004
(6.21)
0.009
(3.45)
0.003
(2.75)
-0.003
(-1.69)
-0.014
(-5.01)
-0.007
(-5.32)
-0.001
(-10.13)
-0.005
(-2.21)
0.002
(0.60)
-0.001
(-2.38)
0.004
(7.08)
0.004
(6.89)
0.010
(4.49)
0.003
(3.64)
-0.006
(-3.78)
-0.013
(-5.87)
-0.006
(-5.62)
-0.001
(-10.37)
-0.004
(-2.07)
0.003
(1.36)
-0.001
(-2.45)
0.002
(1.42)
0.006
(5.37)
0.007
(2.09)
0.007
(4.29)
-0.004
(-1.60)
-0.001
(-0.31)
-0.001
(-7.07)
-0.002
(-0.76)
-0.002
(-0.67)
-0.003
(-3.70)
0.005
(6.45)
0.004
(4.93)
0.012
(4.06)
0.002
(1.77)
-0.007
(-3.56)
-0.013
(-5.44)
-0.009
(-6.51)
-0.001
(-7.27)
-0.003
(-1.18)
0.020
(3.69)
-0.000
(-1.11)
ICY
44,171
2.7%
ICY
61,624
2.9%
ICY
21,825
3.3%
ICY
39,799
2.8%
Lendable×T
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Internet Appendix, Page 36 NUS
Model
(4)
Table 6 (Full Specifications) – Continued
B. Subsample Analyses on Different Degrees of Earnings Management
AccrualMJones≥0
Small Firms
Large Firms
Model
Model
(3)
(4)
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
AccrualMJones<0
Small Firms
Large Firms
Model
Model
(5)
(6)
-0.097
(-4.27)
-0.007
(-3.62)
-0.012
(-7.72)
0.013
(2.21)
-0.002
(-0.94)
0.010
(2.98)
0.010
(0.69)
-0.002
(-0.99)
-0.002
(-2.10)
-0.005
(-1.11)
-0.004
(-0.48)
-0.004
(-4.15)
-0.025
(-3.66)
-0.006
(-9.52)
-0.008
(-11.60)
0.004
(1.65)
-0.001
(-1.35)
0.017
(8.03)
-0.004
(-2.04)
-0.002
(-1.61)
-0.000
(-3.73)
0.005
(2.06)
-0.005
(-2.01)
-0.002
(-3.76)
-0.038
(-1.55)
0.019
(9.75)
0.019
(12.99)
0.018
(3.25)
0.009
(5.40)
-0.014
(-4.35)
-0.007
(-0.38)
-0.006
(-2.24)
0.000
(0.07)
-0.006
(-1.48)
0.010
(1.35)
0.002
(2.00)
0.004
(0.55)
0.007
(10.53)
0.010
(14.28)
0.021
(7.74)
0.002
(2.15)
-0.022
(-10.32)
-0.009
(-3.78)
-0.001
(-0.47)
-0.000
(-0.88)
-0.006
(-2.61)
0.005
(1.69)
0.002
(3.88)
ICY
6,022
6.9%
ICY
26,446
8.7%
ICY
6,224
13.3%
ICY
22,932
10.0%
Internet Appendix, Page 37 Table 7 (Full Specifications): Robustness Checks
This table presents a series of robustness checks based on alternative short selling and earnings management
measures, alternative controls for corporate governance, and alternative clustering specifications of the main
regression model. Panel A presents the results for regressions using alternative short-selling potential (SSP)
measures that include shares on loan (On Loan), loan fee (Fee), loan fee volatility (STDFee), specialness
(Specialness), and short-selling constraint (Constraint). Panel B presents the results for regressions using
alternative earnings management measures, including three target-beating measures, namely, small positive
forecasting profits (SPAF), small positive past-earnings profits (SPDE), and small positive profits (SPE), in
Models (1) to (3) and two earnings persistence specifications in Models (4) and (5). In particular, Models (4)
and (5) estimate the following regression model:
Accrual
α β Lendable ,
β Lendable , Earnings , Accrual
Earnings ,
,
,
β X , β X , Earnings , Accrual
ε, ,
β Earnings , Accrual
,
,
is
operating
income
scaled
by
lagged
total
assets
and
is modified
where
,
,
Jones's (1991) residual accruals. Panel C provides alternative discretionary accrual measures in Models (1) to
(3), including Jones's (1991) residual accruals (AccrualJones), KLW's (2005) residual accruals (AccrualKLW), and
DD's (2002) residual accruals (AccrualDD). Models (4) and (5) proxy for manipulation practices by the
likelihood that earnings misstatements or scandals occur. Panel D adds alternative discipline channels as
controls. These alternative discipline channels include the ISS corporate governance index (ISS), big N auditor
(BigN), international accounting standard (IAS), news coverage (NewsCoverage), and analyst dispersion (Disp).
All these tests control for industry, country, and year fixed effects (ICY). The t-statistics reported in parentheses
are based on standard errors adjusted for heteroskedasticity and firm-level clustering. Obs denotes the number of
firm-year observations, and AdjRsq is adjusted R2. The sample period is from 2002 to 2009.
Internet Appendix, Page 38 Table 7 (Full Specifications) – Continued
A. Alternative SSP Measures
SSP=
SSP
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
On Loan
Model
(1)
Fee
Model
(2)
STDFee
Model
(3)
Specialness
Model
(4)
Constraint
Model
(5)
-0.052
(-4.12)
0.004
(6.67)
0.004
(6.31)
0.011
(4.64)
0.003
(3.38)
-0.006
(-3.48)
-0.013
(-5.75)
-0.007
(-5.84)
-0.001
(-10.53)
-0.003
(-1.68)
-0.002
(-0.74)
-0.001
(-2.33)
0.001
(3.74)
0.005
(8.04)
0.004
(7.01)
0.010
(4.33)
0.003
(3.66)
-0.007
(-4.09)
-0.013
(-5.97)
-0.007
(-6.02)
-0.001
(-10.76)
-0.003
(-1.87)
-0.003
(-1.43)
-0.001
(-1.58)
0.003
(4.93)
0.004
(7.16)
0.004
(6.60)
0.010
(4.48)
0.003
(3.63)
-0.006
(-3.73)
-0.013
(-5.99)
-0.007
(-5.97)
-0.001
(-10.27)
-0.003
(-1.54)
-0.003
(-1.41)
-0.001
(-1.84)
0.004
(4.05)
0.005
(7.98)
0.004
(6.85)
0.010
(4.30)
0.003
(3.56)
-0.006
(-3.99)
-0.013
(-5.93)
-0.007
(-6.23)
-0.001
(-10.62)
-0.004
(-2.02)
-0.003
(-1.50)
-0.001
(-1.58)
0.004
(2.35)
0.004
(7.57)
0.004
(6.55)
0.010
(4.40)
0.003
(3.44)
-0.006
(-3.69)
-0.013
(-5.95)
-0.007
(-6.20)
-0.001
(-10.92)
-0.004
(-1.94)
-0.004
(-1.81)
-0.001
(-1.80)
ICY
61,623
2.9%
ICY
61,575
2.9%
ICY
59,011
2.9%
ICY
61,575
2.9%
ICY
61,574
2.9%
Internet Appendix, Page 39 Table 7 (Full Specifications) – Continued
B. Alternative Earnings Management Measures on Target Beating and Earnings Persistence
Target Beating
Earnings Persistence
Dep. Variable=
Lendable
SPAF
Model
(1)
SPDE
Model
(2)
SPE
Model
(3)
Earnings
Model
(4)
AccrualMJones
Model
(5)
-0.767
(-3.63)
-1.111
(-3.99)
-0.914
(-2.06)
0.033
(4.50)
0.805
(12.68)
-0.301
(-4.66)
-0.019
(-3.73)
Earnings
Lendable×Earnings
AccrualMJones
0.109
(0.96)
Lendable×AccrualMJones
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Earnings (AccrualMJones)×Control
Fixed Effects
Obs
AdjRsq (PseRsq)
0.037
(3.51)
-0.005
(-0.46)
-0.140
(-3.60)
0.170
(10.71)
0.003
(0.09)
-0.086
(-2.33)
0.052
(2.43)
0.006
(3.23)
-0.064
(-1.94)
0.197
(5.83)
-0.011
(-1.59)
0.056
(4.96)
-0.144
(-11.29)
-0.108
(-2.81)
0.335
(16.62)
-0.154
(-4.06)
-0.157
(-3.52)
-0.028
(-1.29)
0.002
(0.98)
-0.102
(-3.03)
0.191
(4.76)
0.013
(1.85)
0.057
(3.02)
-0.046
(-2.28)
-0.291
(-4.49)
0.314
(11.82)
-0.539
(-8.24)
-0.160
(-1.97)
0.007
(0.20)
0.006
(1.64)
-0.149
(-2.62)
-0.338
(-4.80)
-0.004
(-0.38)
0.003
(4.40)
-0.005
(-8.75)
0.019
(7.82)
0.020
(23.55)
-0.016
(-10.05)
-0.004
(-2.17)
0.009
(8.01)
0.001
(4.57)
0.006
(2.84)
0.004
(1.31)
0.003
(8.42)
-0.453
(-4.15)
0.003
(5.08)
-0.001
(-2.36)
-0.024
(-11.78)
0.013
(15.20)
-0.007
(-4.15)
-0.010
(-5.25)
-0.001
(-0.74)
-0.001
(-7.79)
-0.002
(-1.42)
-0.002
(-0.78)
0.000
(0.28)
No
No
No
Yes
Yes
ICY
46,381
3.4%
ICY
35,986
4.6%
ICY
19,091
10.2%
ICY
58,302
68.1%
ICY
55,816
5.5%
Internet Appendix, Page 40 Table 7 (Full Specifications) – Continued
C. Alternative Earnings Management Measures on Other Accruals and Earnings Misstatements
Other Accrual Measures
Misstatement and Scandals
Dep. Variable=
Lendable
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq (PseRsq)
AccrualJones
Model
(1)
AccrualKLW
Model
(2)
AccrualDD
Model
(3)
Prob(Misstatement)
Model
(4)
Prob(Scandals)
Model
(5)
-0.038
(-6.95)
0.004
(7.09)
0.006
(9.93)
0.012
(5.22)
0.001
(1.66)
-0.005
(-3.49)
-0.010
(-4.47)
-0.005
(-4.95)
-0.001
(-9.62)
-0.005
(-2.57)
0.002
(0.95)
-0.001
(-1.68)
-0.035
(-4.15)
0.001
(1.40)
0.004
(4.65)
0.017
(5.58)
-0.003
(-2.39)
0.007
(3.39)
-0.008
(-2.35)
-0.007
(-4.46)
-0.001
(-5.41)
-0.007
(-2.97)
0.002
(0.59)
-0.001
(-1.12)
-0.019
(-3.83)
0.001
(1.30)
-0.002
(-4.08)
0.001
(0.60)
0.004
(6.02)
-0.007
(-4.87)
-0.007
(-4.81)
-0.004
(-4.82)
-0.000
(-5.49)
0.003
(1.89)
0.000
(0.25)
-0.001
(-4.28)
-0.818
(-2.23)
0.028
(0.79)
0.123
(3.58)
0.351
(3.05)
-0.221
(-4.44)
0.254
(3.53)
0.074
(0.73)
-0.056
(-0.70)
0.004
(0.79)
0.002
(0.02)
-0.082
(-0.87)
-0.035
(-1.49)
-1.209
(-2.58)
0.209
(3.93)
0.019
(0.42)
0.060
(0.39)
-0.066
(-0.83)
0.371
(4.42)
0.190
(1.59)
-0.023
(-0.15)
0.007
(1.32)
-0.244
(-1.47)
0.004
(0.03)
-0.049
(-1.20)
ICY
61,562
3.1%
ICY
61,015
0.5%
ICY
57,603
1.2%
ICY
30,047
5.8%
ICY
30,047
17.6%
Internet Appendix, Page 41 Table 7 (Full Specifications) – Continued
D. Alternative Discipline Channels
Model
Model
Model
Model
(1)
(2)
(3)
(4)
Lendable
ISS
-0.027
(-3.06)
-0.025
(-2.92)
-0.028
(-3.04)
-0.025
(-2.95)
-0.000
(-0.00)
-0.032
(-3.43)
-0.025
(-2.97)
-0.001
(-0.15)
-0.008
(-3.97)
-0.033
(-3.62)
-0.025
(-2.94)
-0.001
(-0.23)
-0.008
(-4.13)
-0.002
(-3.16)
0.003
(2.51)
0.005
(4.18)
0.012
(3.02)
0.006
(3.06)
-0.002
(-0.53)
-0.005
(-1.59)
-0.003
(-1.00)
-0.001
(-6.81)
-0.011
(-3.04)
-0.003
(-1.04)
-0.002
(-2.11)
0.003
(2.50)
0.005
(4.18)
0.012
(2.95)
0.006
(3.01)
-0.002
(-0.52)
-0.005
(-1.47)
-0.003
(-1.03)
-0.001
(-6.81)
-0.011
(-3.03)
-0.003
(-1.01)
-0.002
(-2.15)
0.003
(2.72)
0.005
(4.56)
0.012
(2.91)
0.006
(2.90)
-0.002
(-0.47)
-0.006
(-1.65)
-0.001
(-0.49)
-0.001
(-7.05)
-0.011
(-3.05)
-0.002
(-0.66)
-0.002
(-2.47)
0.004
(3.20)
0.005
(4.62)
0.012
(2.95)
0.005
(2.75)
-0.001
(-0.25)
-0.004
(-1.25)
-0.001
(-0.55)
-0.001
(-6.61)
-0.011
(-3.12)
-0.000
(-0.16)
-0.002
(-2.57)
-0.022
(-2.32)
-0.023
(-2.52)
0.001
(0.29)
-0.005
(-2.22)
-0.003
(-3.75)
-0.308
(-3.31)
0.004
(3.56)
0.006
(4.87)
0.012
(2.82)
0.000
(0.07)
0.007
(1.19)
-0.003
(-0.86)
0.003
(1.03)
-0.001
(-6.05)
-0.010
(-2.64)
-0.001
(-0.38)
-0.001
(-1.43)
ICY
16,184
4.7%
ICY
16,149
4.7%
ICY
16,065
4.8%
ICY
16,065
4.8%
ICY
13,396
6.1%
IAS
BigN
NewsCoverage
Disp
Size
BM
Leverage
Return
STD
ADR
MSCI
Analyst
CH
IO
Illiquidity
Fixed Effects
Obs
AdjRsq
Internet Appendix, Page 42 Model
(5)