private litigation costs and voluntary disclosure ... · private litigation costs and voluntary...
TRANSCRIPT
Private Litigation Costs and Voluntary Disclosure: Evidence from Foreign Cross-listed Firms
James P. Naughton
Kellogg School of Management, Northwestern University
Tjomme O. Rusticus London Business School
Clare Wang
Kellogg School of Management, Northwestern University
Ira Yeung Kellogg School of Management, Northwestern University
May 2014
Abstract
We use a natural experiment, the Supreme Court Ruling in Morrison v. National Australia Bank and the subsequent Dodd-Frank Act, to examine whether and how expected private litigation costs affect voluntary disclosure behavior. The Morrison decision applied a presumption against extraterritoriality for all securities actions. Congress quickly responded by exempting SEC actions through the Dodd-Frank Act, with the result that Morrison eliminates only private securities actions for shares purchased on non-US exchanges. These events lowered the expected private litigation costs for foreign firms cross-listed on US exchanges. We find a deterioration in our proxies for voluntary disclosure for these firms relative to a matched sample of US firms. The effects we document are stronger for firms with weaker home country institutions and for firms that experienced a larger decline in expected private litigation costs following Morrison. The evidence is consistent with firms responding to a reduction in expected private litigation costs by reducing voluntary disclosure.
Keywords: Voluntary disclosure, litigation risk, cross-listing, Morrison ruling JEL codes: G15, G18, M41
* We appreciate helpful suggestions and comments from Daniel Bens (discussant), Dain Donelson, Luzi Hail, John Jiang (discussant), Ray Ke, Darren Roulstone, George Serafeim (discussant), Aida Wahid, Wendy Wilson (discussant) and workshop participants at the 2013 London Business School Accounting Symposium, the 2013 CAPANA conference, the 2014 International Accounting Section Midyear meeting, and the 2014 PwC Young Scholars Research Symposium at the University of Illinois at Urbana-Champaign. We are grateful for the funding of this research by The Kellogg School of Management and the Lawrence Revsine Research Fellowship.
1. Introduction
We examine the causal effect of expected private litigation costs on voluntary disclosure
using a unique natural experiment, the Supreme Court ruling in Morrison v National Australia
Bank1 (hereafter referred to as Morrison). This unexpected ruling reduced expected private
litigation costs by eliminating the right of shareholders who purchased shares of foreign
companies on a foreign exchange to pursue shareholder lawsuits in US courts under Section
10(b) of the Securities and Exchange Act.2 Morrison was unexpected because it contradicted
more than 40 years of legal precedent that allowed such shareholders to pursue claims in US
courts. In addition, Morrison allows us to isolate the effect of expected private litigation costs on
voluntary disclosure because of provisions in the subsequent Dodd-Frank Act, which ensured
that there was no impact on the public enforcement capabilities of the SEC.
The relation between expected private litigation costs and voluntary disclosure is unclear
in the literature for two reasons. First, prior studies have relied on comparisons between firms
from different legal environments or examined the disclosure choices of firms that are actually
subject to lawsuits. The results from these studies are difficult to interpret since firms in these
settings could endogenously adjust their disclosure to reduce expected litigation costs. Second,
the relation between expected private litigation costs and voluntary disclosure is shaped by two
competing economic forces. On the one hand, a reduction in expected private litigation cost may
1 The full text of the Morrison case is available at http://www.supremecourt.gov/opinions/09pdf/08-1191.pdf 2 Morrison reduced expected litigation costs by both reducing potential recoveries (conditional on a settlement) and reducing the likelihood of a suit being filed. The reduction in recoveries occurs because shareholders who acquire shares outside the US are no longer entitled to compensation. This can lead to a significant reduction in settlement costs. For example, for the 2003 suit against Royal Ahold which was settled for $1 billion in 2006, only 2.4 per cent of trading during the class period was in the US. The reduction in the likelihood of a suit occurs because, in addition to lower expected recoveries, federal courts may no longer have jurisdiction. For example, in Plumbers’ Union Local No. 12 Pension Fund v. Swiss Reinsurance Co., a US Pension fund filed an action under 10(b) against a Swiss Corporation alleging that the corporation and two of its senior officers misled investors about the company’s business fundamentals. The claim was dismissed under Morrison because the Swiss Reinsurance Co shares were purchased on a foreign exchange (Council of Institutional Investors, 2012).
1
lead to a decrease in disclosure by mitigating lawsuits related to insufficient disclosure. On the
other hand, a reduction in expected private litigation cost may lead to an increase in disclosure
by mitigating lawsuits related to (implicit) forecasts that do not materialize. These conflicting
forces highlight the importance of an exogenous setting to deriving any conclusions about the
association between litigation risk and voluntary disclosure.
We take advantage of the exogenous shock to expected private litigation costs
attributable to Morrison to provide clean insight into the relation between private litigation and
voluntary disclosure. We use two empirical strategies to identify this relation. First, we compare
foreign cross-listed firms (who were affected by Morrison) with a matched sample of US firms
(who were not affected by Morrison). Second, we take advantage of the fact that foreign cross-
listed firms experienced differential effects based on country-level attributes (e.g., the country’s
legal institutions) and firm-level attributes (e.g., the change in the firm’s expected private
litigation costs). These factors allow us to compare the relative change in disclosure behavior of
the cross-listed firms that are most affected by Morrison to those that are least affected.
We use the likelihood and frequency of management guidance as direct tests of voluntary
disclosure behavior and properties of analyst forecasts as indirect tests of voluntary disclosure
behavior (similar to Rogers and Van Buskirk, 2009). We find that there was a reduction in both
the likelihood of management guidance as well as the frequency of guidance for foreign cross-
listed firms relative to US firms following Morrison. We also find that there was a reduction in
analyst coverage for foreign cross-listed firms relative to US firms following Morrison. These
results indicate that firms responded to the reduction in expected private litigation costs by
reducing voluntary disclosure.
2
We further examine the effect of private securities litigation on voluntary disclosure by
conducting cross-sectional tests using a series of country-level institutional variables. We
conduct these tests because Morrison transfers private litigation rights of a firm’s shareholders
who purchased shares outside the US from the US regulatory system to the home country’s
regulatory system. Therefore, to the extent that the foreign cross-listed firm resides in a country
with a weak regulatory structure, there should be a larger impact associated with Morrison. Our
results are consistent with this expectation. For each country-level measure, we find that several
of our disclosure proxies experience larger changes when the foreign cross-listed firm is from a
country with weak institutions.
We then conduct additional cross-sectional tests using variables that capture the impact of
Morrison on the firm’s expected private litigation costs. More specifically, we use the volume of
shares traded on foreign exchanges relative to US exchanges. Since Morrison eliminated the
ability of shareholders who purchase shares on non-US exchanges to seek compensation in US
courts, this variable captures the proportional reduction in expected private litigation costs
conditional on a settlement. In addition, we use an alternate approach that is the product of this
conditional reduction in cost and the ex-ante likelihood of a lawsuit (using the litigation risk
model in Kim and Skinner (2012)) to capture the overall reduction in expected private litigation
cost. Under both approaches, our results are again consistent with the conclusion that firms
reduced voluntary disclosure in response to a reduction in private litigation costs. Firms with
more trading on non-US exchanges experienced the largest reductions in expected private
litigation costs as a result of Morrison, and these firms responded by curtailing voluntary
disclosure to a greater extent.
3
Overall, the findings from our country and firm level cross-sectional tests support the
hypothesis that the effects of the Morrison ruling on voluntary disclosure behavior are the
strongest for firms where the ruling should have the greatest impact. While a number of tests are
statistically insignificant and thus inconclusive, all significant coefficients support the conclusion
that a reduction in expected private litigation costs leads to a deterioration in voluntary
disclosure. In addition, we obtain similar results when we match firms based on the level of
voluntary disclosure in the pre-period, when we separately analyze Canadian and non-Canadian
firms, and when we consider short-window market based measures of a firm’s information
environment. In addition, our results do not seem to be driven by other concurrent events, and in
particular, the Dodd-Frank Act. Therefore, we have a reasonable degree of confidence in the
overall results even if several individual tests are inconclusive.
We make several contributions to the literature. Our primary contribution is to the
literature on the relation between private securities litigation and voluntary disclosure. Morrison
allows us to cleanly identify the causal effect of private securities litigation on disclosure
behavior. The exogenous shock to the litigation environment not only provides an ideal setting to
study the causal relation between litigation and disclosure, but it also keeps public enforcement
constant allowing us to focus specifically on private litigation. We find that firms responded to
the reduction in expected private litigation costs by reducing their voluntary disclosure.
Our findings also contribute to the literature on the costs and benefits of cross-listing.
Extant research documents that cross-listed firms have better information environments, which
are associated with higher market valuations (Lang, Lins and Miller, 2003; Doidge, 2004;
Doidge, Karolyi, and Stulz, 2004, 2009). The sources of these benefits have been largely
attributed to the firm’s voluntary bonding to a more stringent reporting regime (Coffee, 2002). In
4
addition, recent work by Cheng, Srinavasan, and Yu (2013) and Gande and Miller (2012) finds
that private securities litigation against cross-listed firms in US courts is relatively common,
which suggests that private litigation in US courts provides an important enforcement role that
supports the bonding mechanism. Our evidence adds to this literature by showing that two
aspects of the bonding mechanism (i.e. disclosure regime and enforcement) are related, and in
particular that the private securities litigation rights of a firm’s shareholders affects a cross-listed
firm’s voluntary disclosure behavior.
We also add to the literature on Morrison by identifying a specific firm response to the
ruling. The literature on Morrison has focused exclusively on the stock market response. For
example, Licht, Li, and Siegel (2012) investigate the overall wealth effects of private securities
litigation using a short window stock market event study. They find inconsistent results, and
conclude that private securities litigation does not increase firm value. Gagnon and Karolyi
(2012) investigate the stock market response of the US listed relative to the non-US listed shares
of the same firm. They find a positive relative effect, which they suggest is because investors
who acquired US listed shares can still collect damages through litigation, but only bear a portion
of the overall litigation costs, which fall on the firm as a whole. Neither of these studies
examines the firm’s response to Morrison, and in particular, whether and how Morrison
influences the firm’s voluntary disclosure behavior.
This paper proceeds as follows. In Section 2, we summarize the legal setting that we
exploit as a natural experiment. We then outline the existing literature and present our
hypotheses in Section 3. We present our data collection in Section 4, followed by our research
design in Section 5. The results are presented in Section 6 and Section 7, and we conclude in
Section 8.
5
2. Institutional Setting
The Morrison decision provides an ideal setting to study the effects of private litigation
on voluntary disclosure because expected private litigation costs were unexpectedly and
materially altered for certain types of foreign firms. This section discusses the regulatory
landscape for shareholder lawsuits under Section 10(b) of the Securities and Exchange Act to
establish that Morrison was unexpected and material, and to identify how it affected expected
private litigation costs for certain firms. We will rely on this discussion for our empirical
strategy, which is outlined in Section 4 and 5.
Section 10(b) and Rule 10b-5 give shareholders the right to recover damages from the
firm for any act or omission resulting in fraud or deceit in connection with the purchase or sale of
the firm’s shares. Prior to Morrison, the Court of Appeals had established two tests for applying
Section 10(b) to cases with foreign elements: the “effects test” and the “conduct test.” Under the
effects test, Section 10(b) applied to fraudulent conduct that directly affected US investors or
markets. In practice, the effects test was satisfied when either the securities were traded on a US
exchange or the securities were purchased by a US investor. Under the conduct test, Section
10(b) applied to fraudulent conduct in the US that caused losses abroad, even if those losses were
incurred by foreign investors. As a result, it was possible for a foreign issuer to be sued in US
courts by foreign plaintiffs who bought their securities on a foreign exchange—what was
popularly referred to as an “F-cubed” securities class action. In such cases, the conduct test was
typically satisfied if the fraudulent conduct occurred in the US.
The applicability of the effects and conduct tests are summarized in Panel A of Figure 1.
The two-by-two matrix identifies the applicable test for a foreign issuer based on the location of
the exchange and the nationality of investor. The effects test was used when either a US investor
6
or US exchange was involved. The conduct test was used for F-cubed cases. US courts have
heard a significant number of F-cubed cases and admitted a large portion of those cases. In the
15-year period between the Private Securities Litigation Reform Act of 1995 and the Morrison
decision, there were approximately 277 class action lawsuits against foreign companies filed in
US Courts, which represented approximately 10% of all class action lawsuits. Of these filings,
107 included F-cubed investors in the initial class, and there were 41 cases where the court
decided on F-cubed subject matter jurisdiction. In 27 of these 41 cases (66%), courts found that
subject matter jurisdiction existed over F-cubed investors’ trades (Buckberg and Gulker, 2011).
The Morrison case considered by the Supreme Court was an F-cubed securities class
action. Australian investors who purchased shares on a foreign exchange sued National Australia
Bank (“NAB”), a foreign issuer. The specific fraudulent act involved the overstatement of the
servicing rights asset of a Florida-based NAB subsidiary which was in the business of servicing
mortgages. The Court of Appeals ruled that there was no subject matter jurisdiction using the
conduct test, because the act of overstating the value of the servicing rights asset was undertaken
by executives located in Australia. The fact that the specific asset that was overstated was located
in the US was insufficient to create jurisdiction. The court concluded that any actions in the US
were, at most, a link in a securities fraud that was conducted abroad.
The Supreme Court affirmed, but in so doing, ruled that the main fraud-related provisions
of US securities laws apply only to transactions in securities that take place in the United States
or to transactions in securities listed on a US securities exchange. While the decision to find a
lack of subject matter jurisdiction was not surprising, the application of a presumption against
extraterritoriality was completely unexpected as it reversed more than 40 years of established
legal precedent. The new requirement set forth by the court is referred to as the transactional test,
7
and it represents a narrower version of the effects test. While the effects test is satisfied when
either the investor or the exchange is based in the US, the transactional test can only be satisfied
if the securities in question were sold on an US exchange. In addition to reducing the scope of
the effects test, the Morrison ruling eliminated the conduct test entirely. The applicability of the
transactional test is summarized in Panel B of Figure 1. For both US and foreign investors,
Section 10(b) can only be applied to fraudulent conduct that has caused losses on securities
traded in the US. Comparing Panel A and Panel B, the Morrison decision only impacted firms in
the bottom row of the two-by-two matrix.
The Morrison decision applied a presumption against extraterritoriality for all securities
actions, both public and private. As a result, Morrison also prevented the SEC from pursuing
actions against foreign companies under Section 10(b). However, on July 21, 2010, only four
weeks after Morrison, Dodd-Frank Act was signed into law. This Act restores federal district
courts’ jurisdiction over SEC actions charging entities with violating federal securities law
antifraud provisions if misconduct in the US has significantly furthered those violations, even if
the securities were traded outside the US and the transactions only involved foreign traders, or if
misconduct that happened abroad had a foreseeable substantial effect within the US. This new
legislation restored the SEC’s extraterritorial reach to its pre-Morrison level, with the result that
only private enforcement was affected by Morrison.
The provisions added to the Dodd-Frank Act in response to Morrison also tasked the SEC
with studying the need for further legislation to reinstate extraterritoriality for private securities
litigation. The SEC issued a study on April 11, 2012 (US Securities and Exchange Commission,
2012), which did not express an opinion regarding the desirability of reinstatement, despite
institutional investors urging them to do so (e.g., Council of Institutional Investors, 2012).
8
Commissioner Luis Aguilar issued a dissenting statement on April 11, 2012 in which he argued
that the private litigation rights removed by Morrison will have a profound negative effect on
shareholders.3 Despite its material impact on investor rights, no legislation has been passed since
Morrison to expand the extraterritorial reach of US courts for private securities litigation.
Therefore, the changes introduced by Morrison with respect to individual investors have not been
changed, providing us with a clean regulatory break that we exploit for our empirical design,
which is discussed in Section 4 and 5.
3. Literature Review and Hypothesis Development
Empirical findings on the relation between expected private litigation costs and voluntary
disclosure have been mixed. Skinner (1994) shows that firms are more likely to preemptively
disclose earnings information in bad news quarters relative to good news quarters. This suggests
that firms disclose more when the threat of litigation is perceived to be higher. This result
contrasts with Francis, Philbrick and Schipper (1994) who find that more than half of their
litigation sample is sued based on an earnings forecast or a preemptive earnings disclosure, not
an earnings announcement. Skinner (1997) similarly finds that firms that are eventually sued
have higher levels of voluntary disclosure, but suggests that this is due to the endogeneity of the
disclosure behavior. As the threat of litigation increases, firms increase their disclosure to reduce
expected litigation costs.
3 Consider the opening paragraph to Commissioner Luis Aguilar’s statement concerning the SEC Study “Today the Commission has authorized that a Study be sent to Congress expressing the views of the Staff on the cross-border scope of the private right of action under Section 10(b) of the Securities Exchange Act of 1934. However, my conscience compels me to write separately to record my views on the Study. I write to convey my strong disappointment that the Study fails to satisfactorily answer the Congressional request, contains no specific recommendations, and does not portray a complete picture of the immense and irreparable investor harm that has resulted, and will continue to result, due to Morrison v. National Australia Bank, Ltd.”
9
The main challenge in these studies is the endogenous nature of voluntary disclosure and
litigation. Subsequent research has tried to address this in different ways. Baginski, Hassell and
Kimbrough (2002) investigate the management guidance behavior of US and Canadian firms and
find that Canadian firms are more likely to provide guidance. However, in contrast to US firms,
Canadian firms are not more likely to voluntarily disclose in bad news quarters. Overall, their
findings suggest that the threat of litigation discourages voluntary disclosure, but encourages the
early disclosure of bad news. Using a simultaneous equations approach, Field, Lowry, and Shu
(2005) find no evidence that voluntary disclosure triggers litigation, and some evidence that
disclosure deters certain types of litigation. In contrast, Rogers and Van Buskirk (2009) find that
firms reduce their voluntary disclosure following lawsuits. They interpret this finding as
suggesting that firms act as if they believe voluntary disclosure increases expected litigation
costs. This suggests that an increase in expected litigation costs would lead to a reduction in
voluntary disclosure. Lowry (2009) challenges this interpretation and suggests that the
experiences of firms that have been sued may be not representative of the overall population.
In contrast to most of this prior work, we focus on foreign firms that are cross-listed in
the US. Recent work by Cheng, Srinavasan, and Yu (2013) and Gande and Miller (2012) finds
that private securities litigation against cross-listed firms in US courts, while not as common as
that against US firms, is also relatively common. In addition, the damages that arise under these
shareholder suits can be significant. Moreover, the allocation of damages to non-US investors
can represent a large proportion of the overall settlement. As noted already, for the 2003 suit
against Royal Ahold which was settled for $1 billion in 2006, only 2.4 per cent of trading during
the class period was in the US. Because the Morrison decision was unexpected, and because the
consequence of the Morrison decision is a material reduction in private litigation costs, we can
10
use the Morrison ruling as natural experiment to reexamine the relation between litigation and
disclosure. This setting allows us to examine the same firm under two regulatory regimes and
gives us a natural comparison group in the US firms which were not affected by Morrison. Given
the conflicting economic forces in the literature we state our first hypothesis in the null form:
H1: Foreign cross-listed firms do not differentially alter their voluntary disclosure
behavior after Morrison.
Under the alternative hypothesis that expected private litigation costs increase (decrease)
voluntary disclosure, we expect that foreign cross-listed firms increase (decrease) their voluntary
disclosure relative to US firms following Morrison.
The effect of a decline in expected private litigation costs likely differs across firms based
on both country-level and firm-level attributes. Morrison transfers private litigation rights of a
firm’s shareholders who purchased shares outside the US from the US regulatory system to the
home country’s regulatory system. Therefore, to the extent that the cross-listed firm resides in a
country with a weak regulatory structure, there should be a stronger impact associated with
Morrison. This leads to our second hypothesis, which we state in the alternative:
H2: The change in voluntary disclosure behavior following Morrison is stronger for
foreign cross-listed firms with weaker home country institutions.
Firms may also respond differently to Morrison based on the firm specific measure of the
change in expected private litigation costs due to Morrison. In particular, firms that have a
greater proportion of shares traded outside the US and firms with a higher ex ante litigation risk
should experience larger changes in expected private litigation costs. Therefore, if the drop in
expected private litigation costs reduces voluntary disclosure, we would expect the reduction to
be greatest for firms with the greatest decline in expected private litigation costs. In contrast, if
11
the drop in expected private litigation costs increases voluntary disclosure, we would expect the
increase to be greatest for firms with greatest decline in expected private litigation costs. We
therefore state our third hypothesis in the alternative:
H3: The change in voluntary disclosure behavior following Morrison is stronger for
foreign cross-listed firms with a greater decline in expected private litigation costs.
4. Data and Sample
This section proceeds as follows. First, we explain our choice of event window. We then
summarize the data collection for the treatment firms (i.e. those firms affected by Morrison) and
then the control firms (i.e. those firms not affected by Morrison).
4.1 Event Window
The event window is summarized in Figure 2. We use equal-length two year periods
before and after Morrison for the pre- and post-periods, respectively. The pre-Morrison period
commences on January 1, 2008 and ends on December 31, 2009. This period starts after the
elimination of the 20-F reconciliation requirement for IFRS and ends before the oral arguments
for Morrison, which occurred on March 29, 2010. The post-Morrison period commences on
January 1, 2011 and ends on December 31, 2012. This period is after the Morrison decision is
published and the Dodd-Frank Act is adopted, the latter of which occurred on July 21, 2010. We
use calendar year periods for both the pre- and post-periods to ensure that our data is comparable
over time.
4.2 Selection of Treatment Firms
The Morrison ruling affects shareholders of foreign firms with shares traded on a non-US
exchange. For these firms, shareholders who purchase shares outside the US can no longer use
12
US courts to initiate shareholder actions. Following prior research on Morrison (Gagnon et al.,
2012; Licht et al., 2012), we select as our treatment firms the subset of foreign firms that are
cross-listed in the US, and thus have shares that trade both on a US and non-US exchange. We
categorize a firm as foreign if it is incorporated and headquartered outside the US based on
information collected from Compustat.4 In addition, we exclude financial firms as the Dodd-
Frank Act, which was passed shortly after Morrison, resulted in a significant change in the
regulatory framework for financial institutions.5
For each firm, we obtain financial and stock return data from Compustat and CRSP, and
analyst forecast and management guidance from I/B/E/S. We exclude firms with insufficient data
for our tests. In addition, we collect information on the US dollar equivalent trading volume on
US and non-US exchanges on which the firm is traded using Bloomberg terminal. We exclude
firms whose proportion of trading volume on US exchanges exceeds 98 percent from the
treatment group because for those firms Morrison only affects a very small fraction of the firm’s
shareholders. Finally, we balance our dataset by requiring that the foreign cross-listed firms have
at least one observation in both the pre- and post-Morrison period.
Table 1 provides summary statistics for the treatment firms. There are 453 unique firms
and 1,548 firm-year observations from 37 countries. Canadian firms comprise the largest
proportion, with about one-third of the firms headquartered in Canada, followed by Israel, the
4 It is technically possible for a firm that is incorporated and headquartered overseas to not qualify as a private foreign issuer, and hence be subject to the same regulatory regime as US listed firms. The SEC’s definition of foreign private issuer, which is provided in Exchange Act Rule 3b-3, excludes firms that are technically foreign (i.e. incorporated outside the US) but are essentially American (i.e. majority of voting rights are held by US investors AND either the majority of top management is American, the majority of assets are held in the US, or the business is managed in the US). To the extent that any such firms are included in our sample of treatment firms, it would bias against finding any results since these firms would not be affected by Morrison. 5 We also conduct robustness tests to provide assurance that the results we document are not because the foreign and US firms in our sample were differentially affected by other provisions of the Dodd-Frank Act. These tests are described in Section 7.
13
United Kingdom, and Brazil. Table 1 also presents the country institutional variables we use in
the cross-sectional analyses.
4.3 Selection of Control Firms
The control firms in our setting are US firms that are not affected by Morrison. An
alternative control group would be to use foreign firms that are not cross-listed. We do not use
these foreign non-cross-listed firms for two reasons. First, foreign non-cross-listed firms may
have been impacted by Morrison. This is because the pre-Morrison rules permitted suits against
foreign firms that had operations in the US under the conduct test, regardless of whether the firm
was cross-listed or not. In addition, even if the foreign firm did not have operations, it was still
possible to file a claim under the effects test if the plaintiffs were US investors. Second, the
predictions associated with tests that use foreign non-cross-listed firms as controls are ambiguous
because the relative decline in litigation risk for foreign non-cross-listed firms could be greater or
less than the decline for foreign cross-listed firms. Foreign non-cross-listed firms likely had
lower expected private litigation costs when compared with foreign cross-listed firms prior to
Morrison. Therefore, the overall decline in expected private litigation costs for the foreign non-
cross-listed firms should be lower. However, the proportional reduction is greater for these firms
since their exposure to US private securities litigation is likely eliminated entirely following
Morrison.
Foreign non-cross-listed firms that did not have any US operations or US investors would
provide a useful control sample, as these firms did not experience any change in expected private
litigation costs due to Morrison. However, these firms would likely be very different than the
foreign cross-listed firms in our sample. In addition, there are some more practical concerns.
First, identifying firms without US operations or US investors is not a straight forward task.
14
Second, even if these firms could be identified, we would need to be able to collect information
on management guidance and analyst activity to conduct our empirical tests. Currently,
management guidance data for foreign firms is only obtainable for a subset of our sample period
as that information on Capital IQ is only updated through 2011.6 Therefore, we use US firms as
our control sample.
Rather than including all US firms as control firms in our empirical tests, we generate a
matched sample of US firms using a propensity score model. We use a propensity score model in
lieu of matching on a small set of specific characteristics, such as firm size and industry
classification, because it places less restrictive assumptions on the functional form for the
relation between the control variables and the outcome variable (Rosenbaum and Rubin, 1983).
Typically, propensity score matching is used to allow the researcher to approximate an
experimental setting such that causal inferences may be made. In our setting, this would entail
identifying a set of firms that were equally likely to be treated (i.e. to be subject to Morrison), but
did not receive the treatment. That is not the reason we employ propensity score matching, as our
setting provides an exogenous shock that allows us to make casual statements. Rather, we
employ propensity score matching because it reduces noise in our estimation that would
otherwise be present due to fact that the average cross-listed firm is not easily comparable to the
average US firm (Karolyi, 2006).
We populate the propensity score model using the standard determinants of voluntary
disclosure that originated with Lang and Lundholm (1993). The matching variables are firm size
(Total Assets), growth opportunities (Book-to-Market), performance (Stock Return, Return on
Assets, and Earnings Surprise), and firm risk (Return Variability). Each variable is defined in
6 See the Appendix in Li and Yang (2014) for a description on how to identify management guidance data for foreign firms using Capital IQ.
15
Table 2. All variables are winsorized at the 1st and 99th percentile. We do not include the returns-
earnings correlation, because it requires a long time-series of data, thereby unduly restricting the
sample size. We match firms based on the last available year within the pre-period. Once a
treatment firm is matched to a control firm, the firms remain matched for the entire sample
period.
Table 3 provides the propensity score model for the management guidance tests.7 The
propensity score model we use is the same for our analyst tests. However, for that sample we add
an additional restriction that each firm has at least one analyst in the pre-period before matching.
This ensures that the pair is not dropped in subsequent analyses if one of two firms had no
coverage. We do not make this requirement in the post-period, as doing so would be sampling on
the outcome we are trying to document. Table 3 Panel A reveals that the average cross-listed
firm is indeed different from the average US firm. It shows that cross-listed firms are generally
larger, have greater growth opportunities and lower volatility than the average US firm. The
pseudo R2 is 13% suggesting a reasonable fit. We use this model to match treatment (i.e. foreign
cross-listed) and control (i.e. US) firms by pairing firms with the closest propensity score in the
same year and industry. We sample the control firms with replacement as this allows for the
closest possible match and best control. We cluster standard errors by firm to correct for the
possibility that a single control firm may have multiple occurrences. Panel B provides the
covariate balance between the matched treatment and control firms for the management guidance
tests. Overall, the sample seems well-matched on most dimensions included in the propensity
score model. Two of the pairwise differences are statistically significant. To address problems
7 Management guidance is considered mandatory disclosure in Japan and Taiwan, thus we exclude foreign cross-listed firms from these two countries in our management guidance tests.
16
with imperfect covariate balance, we include the matching variables as control variables in our
regressions.
5. Research Design
We analyze the impact of private litigation on voluntary disclosure using the following
specification:
Vol_Discj,t = β0 + β1Postt + β2Postt * Foreignj + Controlsj + IndustryFE
+ CountryFE + εj,t (1)
We use five different variables to proxy for voluntary disclosure behavior. We use Management
Guidance and Guidance Frequency as direct tests of voluntary disclosure behavior. We use
Analyst Following, Forecast Accuracy and Forecast Dispersion as indirect tests of voluntary
disclosure behavior (similar to Rogers and Van Buskirk, 2009). Each of these voluntary
disclosure variables is defined in Table 2. Post is an indicator variable that takes on the value of
‘1’ for fiscal years that begin after January 1, 2011 and end before December 31, 2012; and takes
on the value of ‘0’ if the fiscal year begins after January 1, 2008 and ends before December 31,
2009. Foreign is an indicator variable that takes the value of ‘1’ if the firm is a foreign cross-
listed firm and ‘0’ otherwise. We do not include Foreign as a non-interacted variable since we
already include country fixed effects in all our regressions. The control variables are the
explanatory variables from the propensity score model, which are firm size (Total Assets),
growth opportunities (Book-to-Market), performance (Stock Return, Return on Assets, and
Earnings Surprise), and firm risk (Return Variability). We also include a control variable for the
change in GDP (GDP Growth), which controls for differential economic opportunities across the
countries in our sample. Each of these control variables is defined in Table 2. Industry fixed
17
effects are derived from Fama French industry classification codes based on 17 industry
portfolios.
The coefficient of interest in this specification is β2, which measures the differential
change in voluntary disclosure behavior for firms affected by Morrison relative to firms
unaffected by Morrison. Our first hypothesis, stated in null form, predicts that β2 = 0. To the
extent that a decline in expected private litigation costs leads to a deterioration (improvement) in
voluntary disclosure, this implies that β2 < 0 (β2 > 0).
We further examine the effect of private securities litigation on voluntary disclosure by
conducting cross-sectional tests using a series of country-level institutional variables. We
conduct these tests using the following specification.
Vol_Discj,t = β0 + β1Postt + β2Postt * Foreignj + β3Postt * Foreignj * Weak Home
Institutionj + Controlsj + IndustryFE + CountryFE + εj,t (2)
Vol_Disc, Post and Foreign are defined as in equation (1). We use three different variables to
proxy for the strength of the home country institutions. We use a country’s Legal Origin as a
proxy for its investor protection as the rights of minority shareholders are arguably better
protected in common law countries than in code law countries (e.g., La Porta et al., 1998). We
use the Aggregate Earnings Management score from Leuz et al. (2003) as earnings management
is generally associated with weak shareholder protections. Finally, we use the Public
Enforcement Resources variable from Jackson and Roe (2009) because shareholders in countries
with weak public enforcement are likely to rely more on the US regulatory structure. Each
country level variable is defined in Table 1. The control and fixed effects variables mirror those
used in equation (1).
18
The coefficients of interest in this specification are β2 and β3. The differential change in
voluntary disclosure behavior for foreign cross-listed firms with strong home country institutions
relative to US firms is measured by β2. The differential change in voluntary disclosure behavior
for cross-listed firms with weak home country institutions relative to cross-listed firms with
strong home country institutions is measured by β3. Our second hypothesis implies that the
change in voluntary disclosure behavior following Morrison is stronger for firms with weak
home country institutions. Therefore, to the extent that a decline in expected private litigation
costs leads to a deterioration in voluntary disclosure behavior, then our specification implies that
both β2 < 0 and β3 < 0.
We then examine the effect of private securities litigation on voluntary disclosure by
conducting cross-sectional tests using firm-level variables. More specifically, we split our
sample based on the change in the individual firm’s expected private litigation costs following
Morrison. The specification we employ is as follows:
Vol_Discj,t = β0 + β1Postt + β2Postt * Foreignj + β3Foreign * High Morrison Impactj +
β4Postt * Foreignj * High Morrison Impactj + Controlsj + IndustryFE +
CountryFE + εj,t (3)
Vol_Disc, Post and Foreign are defined as in equation (1). We use two different
approaches to estimate the change in expected private litigation costs and hence the firm-level
impact of Morrison. First, we use the proportion of the volume of shares traded on foreign
exchanges relative to US exchanges. Since Morrison eliminated the ability of shareholders who
purchase shares on non-US exchanges to seek compensation in US courts, this variable captures
the proportional reduction in expected private litigation costs of the firm conditional on a
settlement. We refer to this variable as High Foreign Trading, and it is defined as an indicator
19
variable that equals ‘1’ if the foreign firm’s Foreign Trading is above the foreign firm sample
median.
Second, we create a composite expected litigation cost measure by multiplying the
proportion of foreign trading by the ex-ante likelihood of being sued (using the litigation risk
model from Kim and Skinner (2012)). This provides a slightly stronger measure of the change in
expected private litigation costs since it places additional weight on the proportion of foreign
trading for firms with higher overall levels of ex ante litigation risk. We refer to this variable as
Litigation Cost. In our analyses we will use High Litigation Cost, an indicator variable that
equals ‘1’ if the foreign firm’s Litigation Cost is above the foreign firm sample median. We do
not include Foreign, High Foreign Trading, or High Litigation Cost as non-interacted variables
due to the inclusion of country fixed effects in all our regressions and the fact that foreign trading
is zero for US firms. We include the same control variables as with equation (1).
The coefficients of interest in this specification are β2 and β4. The differential change in
voluntary disclosure behavior for foreign cross-listed firms with a relatively small change in
expected litigation cost relative to US firms is measured by β2. The differential change in
voluntary disclosure behavior for cross-listed firms with large changes in expected private
litigation costs relative to cross-listed firms with small changes in expected private litigation
costs is measured by β4. Therefore, to the extent that a decline in expected private litigation costs
leads to a deterioration in voluntary disclosure behavior, our specification implies that both β2 <
0 and β4 < 0.
20
6. Results
The results of equation (1), provided in Table 4, indicate that expected private litigation
costs have a positive effect on voluntary disclosure. Following an exogenous reduction in
expected private litigation costs, we find a statistically significant relative deterioration in the
likelihood of management guidance, management guidance frequency, and analyst coverage for
cross-listed firms relative to US firms. However, we do not identify any change in forecast
accuracy or forecast dispersion. In other words, we find that both of our direct measures of
voluntary disclosure (i.e. management guidance and guidance frequency) and one out of the
three indirect measures of voluntary disclosure (i.e. analyst coverage), are positively associated
with private litigation. Across all five specifications, the coefficients on the control variables are
consistent with the literature. Firm size is positively associated with our proxies for voluntary
disclosure, and book-to-market, earnings surprise and return variability are generally negatively
related to our proxies for voluntary disclosure. Overall, the results in Table 4 document a decline
in voluntary disclosure in response to a reduction in expected private litigation costs.
We further investigate the effect of expected litigation costs on voluntary disclosure using
cross-sectional tests involving country-level institutional variables. We examine the differential
effects using the specification in equation (2), the results of which are presented in Table 5.
Within each panel, columns 1-5 present tests using each of our voluntary disclosure proxies from
Table 4, and Panels A-C present these regressions based on each of the three different country
institutional measures (i.e. Legal Origin, the Aggregate Earnings Management score from Leuz
et al. (2003), and the Public Enforcement Resources variable from Jackson and Roe (2009)).
The coefficient on the interaction, Post*Foreign, measures the differential change in
voluntary disclosure behavior for cross-listed firms with strong home country institutions relative
21
to US firms. For each of our three specifications shown in Panels A-C, the coefficients on these
variables are consistent with the main results in Table 4. We find a statistically significant
relative deterioration in the likelihood of management guidance, management guidance
frequency, and analyst coverage for cross-listed firms relative to US firms, but we do not identify
any change in forecast accuracy or forecast dispersion.
In Panel A, the coefficient on the triple interaction, Post*Foreign*Code Law, measures
the differential change in voluntary disclosure behavior for cross-listed firms with weak investor
protections relative to cross-listed firms with strong investor protections. The coefficient on this
term is negative and significant for guidance frequency and analyst coverage. In addition, the
coefficient on this term in the management guidance specification is weakly significant. Overall,
these results provide corroborating evidence supporting our second hypothesis. In other words,
we find that within our sample of foreign firms, those firms with weaker home country investor
protections experience a more significant decline in voluntary disclosure in response to the
exogenous reduction in expected private litigation costs.
Panel B provides the results using the aggregate earnings management score as a proxy
for the strength of the home country shareholder protections. Once again, we find statistically
significant results on the Post*Foreign interaction term for three of the five measures we
examine, with each result consistent with a deterioration in voluntary disclosure behavior. In
addition, we find that three of the five coefficients on the triple interaction, Post*Foreign*High
Aggregate Earnings Management, are consistent with a decline in voluntary disclosure. The
relative decrease in management forecast guidance frequency and analyst following for cross-
listed firms is significantly larger for firms located in countries with weaker shareholder
protections. In addition, while on average we do not find a change in forecast dispersion, we do
22
find that firms with weaker home-country institutions experienced a significantly larger increase
in analyst forecast dispersion post Morrison. Consistent with Panel A, these results suggest that
firms with weaker home country institutions experienced a more significant decline in voluntary
disclosure in response to Morrison.
In Panel C, we report results using a measure of the public enforcement of securities law
based on regulatory budgets. We expect firms located in countries with fewer resources devoted
to public enforcement to react more strongly to Morrison, as these firms are less likely to have
local regulatory oversight. The results mirror those found in Panel B. The coefficients on the
Post*Foreign interaction term are consistent with a decline in voluntary disclosure for three out
of the five measures we tabulated. In addition, the coefficients on the Post*Foreign*Low Public
Enforcement interaction term are also consistent with a decline in voluntary disclosure for three
of the five measures. There is a relative decrease in management forecast guidance frequency
and analyst following, and a relative increase in forecast dispersion, for cross-listed with lower
levels of public enforcement.
In Table 6, we examine the differential impact of Morrison based on measures of the
change in the individual firm’s expected private litigation costs. Panel A reports the results using
a binary variable that is derived from the volume of shares traded on the foreign exchange
relative to the US exchange. This variable captures the effect of Morrison because shares traded
on the foreign exchange can no longer file a claim under US securities law. Therefore, firms with
high levels of foreign trading will experience a greater decline in expected private litigation costs
in response to Morrison.
The coefficient on Post*Foreign measures the differential change in voluntary disclosure
behavior for cross-listed firms with low foreign trading relative to US firms. The results are
23
consistent with Table 4. We find a statistically significant relative deterioration in the likelihood
of management guidance, management guidance frequency, and analyst coverage for cross-listed
firms relative to US firms, but we do not identify any change in forecast accuracy or forecast
dispersion.
The coefficient on Post*Foreign*High Foreign Trading measures the differential impact
of Morrison for cross-listed firms with more trading volume outside the US relative to cross-
listed firms with less trading volume outside the US. Consistent with our third hypothesis, the
coefficient on Post*Foreign*High Foreign Trading is negative and significant for management
guidance frequency and analyst following. However, we do not find statistically significant
results for our other specifications. Overall, these results suggest that the reduction in voluntary
disclosure is stronger for foreign cross-listed firms with higher expected private litigation costs.
The results in Panel B measure the change in expected private litigation costs by
multiplying the proportion of foreign trading volume by a measure of litigation risk developed by
Kim and Skinner (2012). This provides a slightly stronger measure of the impact of Morrison on
the change in the firm’s expected private litigation costs than Panel A since it places additional
weight on the proportion of foreign trading for firms with higher overall levels of ex ante
litigation risk. The results in Panel B are similar to, but slightly stronger than those in Panel A.
The coefficient on Post*Foreign*High Litigation Cost is negative and significant for
management guidance frequency and analyst following, similar to Panel A. In addition, the triple
interaction is now significant for the management guidance variable.
We do not report the results using the Kim and Skinner (2012) measure by itself because
it is derived from a prediction model of the total litigation risk for US firms. Therefore, it likely
provides a noisy estimate of the impact of Morrison relative to the foreign trading measure.
24
Nonetheless, for robustness, we conduct tests using the Kim and Skinner (2012) measure. In
untabulated results, we find results that are directionally similar to those in Table 6, but not
statistically significant. Thus, while the ex-ante litigation risk measure contributes to the joint
significance of the litigation cost variable, it is not statistically significant by itself.
Overall, the findings from our cross-sectional tests in Table 5 and Table 6 support the
hypothesis that the effects of the Morrison ruling on voluntary disclosure behavior are the
strongest for firms where the ruling should have the greatest impact. While a number of tests are
statistically insignificant and thus inconclusive, all significant coefficients support the conclusion
that a reduction in expected private litigation costs leads to a deterioration in voluntary
disclosure. Out of the twenty five specifications we run in Tables 5 and 6, thirteen of the
coefficients on the cross-sectional interaction terms are in the predicted direction and statistically
significant at the 10% level or better. Thus, we have a reasonable degree of confidence in the
overall results even if several individual tests are inconclusive.
7. Robustness Tests
7.1 Alternative Matching Procedure
We use an alternative matching approach where we match based on the level of the
dependent variable in the pre-period conditional on year and industry. When there are multiple
matches, we select the firm with the closest propensity score. We use this alternative approach
because even though the foreign cross-listed firms and US firms may be economically similar in
our primary analysis, this does not imply that they are similar for each of our measures of
voluntary disclosure. This is only an issue to the extent that these differences are not adequately
controlled for by the inclusion of the control variables, as well as industry and country fixed
25
effects. For example, if there were one less analyst for each firm domiciled in Germany, this
difference would be controlled for by the country fixed effect.
Similar to our primary matching approach, we sample the control firms with replacement
as this allows for the closest possible match and best control. We cluster standard errors by firm
to correct our standard errors for the possibility that a single control firm may have multiple
occurrences. The results presented in Table 7 mirror our base specifications in Table 4. We find a
statistically significant relative deterioration in the likelihood of management guidance,
management guidance frequency, and analyst coverage for cross-listed firms relative to US
firms. Once again, this is consistent with a decline in voluntary disclosure in response to a
reduction in expected private litigation costs.
One potential concern with this alternative matching procedure is that the US firms in the
control sample do not follow the disclosure practices of the typical US firm, and therefore any
change in disclosure practice over our sample period might also be unusual. For example, when
the dependent variable is management guidance, we match the foreign cross-listed firm with a
US firm based on whether both firms provide management guidance in the pre-period. Because
many foreign firms do not provide management guidance, this means that the control firms are
US firms that also do not provide guidance. The issue that this creates is that the US firm may
converge to normal US disclosure practices over our sample period, and hence institute
management guidance in the post-Morrison period. This would bias our results toward finding
that foreign cross-listed firms reduce voluntary disclosure when compared with the control firms.
This potential bias is evident by looking at the coefficient on Post in Table 7. This coefficient
implies that there is a 30 percentage point increase in US firms that provide management
26
guidance between the pre and post periods. This is much higher than in our primary tests and
does not seem plausible.
7.2 Sample Selection
We split our sample into two groups—Canadian and non-Canadian firms. We do this
because Canadian firms make up approximately one-third of the sample of foreign cross-listed
firms, and we want to ensure that the results are attributable to foreign firms generally, and not
just to Canadian firms. The results presented in Table 8 suggest that our results appear in both
subsamples, but are generally stronger in the sample of non-Canadian firms. The coefficients on
the Post*Foreign interaction term is significant for two of the five specifications for the Canadian
firms in Panel A, but three of the five specifications for the non-Canadian firms in Panel B. In
addition, the coefficient magnitudes and statistical significance levels are higher in Panel B. This
finding is also consistent with the notion that Canadian firms have stronger home-country
institutions than most other cross-listed firms. These results provide evidence that our results are
not driven by Canadian firms.
7.3 Alternative Proxies for Voluntary Disclosure
We conduct a series of alternative proxies for voluntary disclosure. First, we consider
market based measures of a firm’s information environment which provide indirect measures of
disclosure behavior. Following Rogers and Van Buskirk (2009), we use absolute size adjusted
returns, measured over the three day window surrounding earnings announcements. To the extent
that firms reduce voluntary disclosure throughout the year, we expect that the proportion of new
information provided during earnings announcements will increase, and therefore that there will
be a higher absolute price reaction. Following a similar logic we also examine share turnover and
bid-ask spreads around the earnings announcement. Table 9 presents the results for these market-
27
based measures. We find statistically significant results for the price and volume specifications,
but not the spread specification. These results provide some additional support for the conclusion
that there is a positive association between private litigation risk and voluntary disclosure.
Second, we consider finer definitions of management forecasts. In particular, we create
three management guidance variables that reflect the percentage of management forecasts that
are annual, point, and short-term forecasts, respectively. A short-term forecast is a forecast that
pertains to the current quarter (Johnson, Kasznik, and Nelson, 2001). In untabulated results, we
find results that are directionally consistent, but not significant at conventional levels.
Third, we perform tests to capture the differential response of bad versus good news.
Because we find that voluntary disclosure is reduced in response to a reduction in private
litigation risk, we expect that the voluntary disclosure of bad news will be reduced more
following Morrison (Skinner, 1994). We test this prediction by using a specification that mirrors
equation (3), except that High Expected Private Litigation Costs is replaced by Bad News. We
use two different measures of Bad News. The first measure defines Bad News as ‘1’ if earnings
decrease from the previous year, and zero otherwise. The second measure defines Bad News as
‘1’ if the realized earnings are less than the analyst consensus at the start of the year, and zero
otherwise.
The untabulated results are generally consistent with our expectation, but many of the
specifications are statistically insignificant. Under the first definition of Bad News, the
coefficient on the Post*Foreign*Bad News interaction term is negative and significant for the
analyst number variable (t-statistic of 1.83). Under the second definition of Bad News, the
coefficient on the Post*Foreign*Bad News interaction term is negative and significant for the
management guidance variable (t-statistic of 1.69). None of the other specifications have
28
significant coefficients. Both significant results suggest that firms are more likely to withhold
bad news following Morrison. This provides some additional support for our finding that there is
a positive relation between private litigation cost and voluntary disclosure. However, the support
is relatively weak given that most of our results are inconclusive.
We also investigate whether timely loss recognition changes differentially for those firms
subject to a reduction in expected private litigation costs. Because we find a reduction in
voluntary disclosure in response to Morrison, we predict that firms affected by Morrison are
likely to experience a reduction in timely loss recognition. We use two measures of timely loss
recognition, both of which are taken from Ball, Robin and Wu (2003). The results for each
specification are not significant at conventional levels and therefore inconclusive.
7.4 Concurrent Events
We investigate whether our results could be attributable to concurrent events. In
particular, our event window coincided with the passage of the Dodd-Frank Act. To the extent
that this legislation changed the disclosure requirements for US firms relative foreign cross-listed
firms, our results could be attributable to Dodd-Frank rather than Morrison. As noted in Section
4, our analysis excludes financial firms, who were the primary targets of the reforms contained in
Dodd-Frank. However, it is still possible that other provisions of Dodd-Frank may affect our
results.
Our investigation did not uncover any financial statement disclosure provisions included
in the Dodd-Frank that applied differentially to US versus foreign cross-listed firms. For
example, new disclosure rules relating to conflict minerals were adopted for mining and natural
29
resource companies.8 However, those rules applied to both US and foreign issuers.9 In addition,
Dodd-Frank included several provisions related to executive compensation. Some of these
provisions, such as the “say-on-pay” advisory vote requirement, are applicable to proxy
statements issued in advance of annual meetings. Because foreign issuers are typically exempt
from SEC proxy rules, this could lead to differences in proxy requirements for foreign issuers
relative to US firms following the implementation of Dodd-Frank.
We do not expect that differences in proxy rules could explain changes in voluntary
disclosure. Nonetheless, to see if this difference could partially explain our results, we replicate
our results using a subsample of 195 foreign cross-listed firms that are subject to the proxy rule
changes by virtue of their listing on the NYSE. This occurs because NYSE-listed companies are
required to solicit proxies for all meetings of shareholders, with limited exceptions.10 We find a
positive relation between voluntary disclosure and expected litigation costs for this subsample,
consistent with our full sample results. This supports our expectation that the effect we document
is not due to differences in proxy rules across US and foreign cross-listed firms.
8. Conclusion
We use a natural experiment to examine whether and how expected private litigation
costs affect voluntary disclosure behavior. The Morrison ruling unexpectedly and substantially
reduced the expected litigation costs of foreign cross-listed firms. We find that these firms
respond to an exogenous decrease in expected private litigation costs by reducing voluntary
8 The final rules adopted by the SEC to implement these sections of Dodd-Frank can be viewed at https://www.sec.gov/spotlight/dodd-frank/speccorpdisclosure.shtml. 9 There are 79 firms in our sample that are part of the mining and natural resource industry. Excluding these firms from our analysis does not change any of the results. 10 See NYSE Exchange Rule 402.04 for more details.
30
disclosure relative to a control sample of matched US firms. We document a reduction in both
the likelihood of management guidance as well as the frequency of guidance for foreign cross-
listed firms relative to US firms. We also find a reduction in analyst coverage and, in some tests,
an increase in forecast dispersion for foreign cross-listed firms relative to US firms. Our
interpretation of the data is supported by cross sectional tests, which show that firms with weak
home country institutions and cross-listed firms with greater reductions in expected private
litigation costs have stronger reductions in measures of voluntary disclosure in response to
Morrison.
31
References
Baginski, S. P., J. M. Hassell, and M. D. Kimbrough. 2002. The effect of legal environment on voluntary disclosure: Evidence from management earnings forecasts issued in US and Canadian markets. The Accounting Review 77 (1): 25–50.
Ball, R., A. Robin, and J. Wu. 2003. Incentives versus standards: Properties of accounting income in four East Asia countries and implications for acceptance of IAS. Journal of Accounting and Economics 36: 235-270.
Buckberg, E., and M. Gulker. 2011. Cross-border shareholder class actions before and after Morrison. Working paper, NERA Economic Consulting.
Cheng, B., S. Srinivasan, and G. Yu. 2013. Securities litigation risk for foreign companies listed in the US. Working paper, Harvard Business School.
Coffee, J. 2002. Racing towards the top? The impact of cross-listings and stock market competition on international corporate governance. Columbia Law Review 102: 1757-1831.
Council of Institutional Investors, 2012. Morrison v. National Australia Bank: The impact on institutional investors. White paper.
Doidge, C., 2004. US cross-listings and the private benefits of control: evidence from dual-class firms. Journal of Financial Economics 72: 519–553.
Doidge, C., Karolyi, G.A., Stulz, R., 2004. Why are foreign firms listed in the US worth more? Journal of Financial Economics 71: 205–238.
Doidge, C., Karolyi, G.A., Stulz, R., 2009. Has New York become less competitive than London in global markets? Evaluating foreign listing choices over time. Journal of Financial Economics 91: 253–277.
Field, L., M. Lowry, and S. Shu. 2005. Does disclosure deter or trigger litigation? Journal of Accounting and Economics 39 (3): 487–507.
Francis, J., D. Philbrick, and K. Schipper. 1994. Shareholder litigation and corporate disclosures. Journal of Accounting Research 32 (2): 137–164.
Gande, A., and D. Miller. 2012. Why do US securities laws matter to non-US Firms? Evidence from private class-action lawsuits. Working Paper, Southern Methodist Univeristy.
Gagnon, L. J., and G. A. Karolyi. 2012. The economic consequences of the US Supreme Court’s Morrison v. National Australia Bank decision for foreign stocks cross-listed in US markets. Working Paper, Queen’s University and Cornell University.
32
Jackson, H. E., and M. J. Roe. 2009. Public and private enforcement of securities laws: Resource-based evidence. Journal of Financial Economics 93 (2): 207–238.
Karolyi, G.A. 2006. The world of cross-listings and cross-listings of the world: Challenging conventional wisdom. Review of Finance 10(1): 73-115.
Johnson, M., R. Kasznik, and K. Nelson. 2001. The impact of securities litigation reform on the disclosure of forward-looking information by high technology firms. Journal of Accounting Research 39 (2): 297-328.
Kim, I., and D. J. Skinner. 2012. Measuring securities litigation risk. Journal of Accounting and Economics 53 (1): 290–310.
La Porta, R., F. López de Silanes, A. Shleifer, and R. Vishny. 1998. Law and finance. Journal of Political Economy 106: 1113–1155.
Lang, M., and R. Lundholm. 1993. Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research 31 (2): 246–271.
Lang, M., Lins, K., Miller, D. 2003. ADRs, analysts, and accuracy: Does cross listing in the US improve a firm’s information environment and increase market value? Journal of Accounting Research 42: 589-623.
Leuz, C., D. Nanda, and P. D. Wysocki. 2003. Earnings management and investor protection: an international comparison. Journal of Financial Economics 69 (3): 505–527.
Li, X., and H. Yang. 2014. Mandatory Financial Reporting and Voluntary Disclosure: Evidence from Mandatory IFRS Adoption. Working Paper, Temple University and University of Pennsylvania.
Licht, A., X. Li, and J. Siegel. 2012. What makes the bonding stick? A natural experiment involving the Supreme Court and cross-listed firms. Working Paper, Harvard Business School.
Lowry, M. 2009. Discussion of “Shareholder litigation and changes in disclosure behavior”. Journal of Accounting and Economics 47 (1): 157–159.
Rogers, J. L., and A. Van Buskirk. 2009. Shareholder litigation and changes in disclosure behavior. Journal of Accounting and Economics 47 (1): 136–156.
Rosenbaum P.R., Rubin D.B. 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70:41–55.
Skinner, D. J. 1994. Why firms voluntarily disclose bad news. Journal of Accounting Research 32 (1): 38–60.
33
Skinner, D. J. 1997. Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics 23 (3): 249–282.
US Securities and Exchange Commission (SEC). 2012. Study on the cross-border scope of the private right of action under Section 10(b) of the Securities Exchange Act of 1934. Washington, D.C.
34
Figure 1: Subject Matter Jurisdiction for Investors in Foreign Firm
Panel A: Pre-Morrison Investor Nationality
US Foreign
Location of exchange
US Admitted via Effects Test
Admitted via Effects Test
Foreign Admitted via Effects Test
Admitted via Conduct Test
Panel B: Post-Morrison Investor Nationality
US Foreign
Location of Exchange
US Admitted via Transactional
Test
Admitted via Transactional
Test
Foreign Fails
Transactional Test
Fails Transactional
Test
35
Figure 2: Timeline
Pre Period (1/1/2008 – 12/31/2009)
1/1/2005 Mandatory
Adoption of IFRS in the EU
11/15/2007 Elimination of
20-F Reconciliation Requirement for
IFRS Users
3/29/2010 Morrison v.
NAB Oral Arguments
6/24/2010 Morrison v.
NAB Ruling
6/25/2010 Conference Committee
Approved the Final Versions of Dodd-Frank Act
Post Period (1/1/2011 – 12/31/2012)
7/21/2010 Dodd-Frank Act Signed into Law
36
Table 1: Sample Composition by Country for Foreign Cross-listed Firms
Country Unique Firms
Firm- Years
Legal Origin
(1 = Code Law; 0 = Common Law)
Aggregate Earnings Management
(1 = Less Transparent Earnings)
Public Enforcement
(1 = Lower Securities Enforcement Budget)
Argentina 5 18 1 NA
15,994 (1) Australia 10 22 0 4.8 (0) 89,217 (0) Belgium 2 8 1 19.5 (1) 27,276 (1) Bermuda 5 18 0 NA NA Brazil 22 85 1 NA 31,729 (1) Canada 159 558 0 5.3 (0) 82,706 (0) Chile 7 24 1 NA 66,093 (0) China 12 43 1 NA NA Colombia 1 4 1 NA 42,660 (1) Denmark 2 8 1 16.0 (0) 25,940 (1) Finland 4 14 1 12.0 (0) 45,937 (1) France 15 59 1 13.5 (0) 28,851 (1) Germany 13 46 1 21.5 (1) 12,903 (1) Greece 3 10 1 28.3 (1) 60,111 (0) Hong Kong 10 34 0 19.5 (1) 320,531 (0) India 7 15 0 19.1 (1) NA Indonesia 2 6 1 18.3 (1) 5,576 (1) Ireland 7 21 0 5.10 (0) 72,639 (0) Israel 40 136 0 NA
145,673 (0)
Italy 5 18 1 24.8 (1) 61,239 (0) Japan 16 34 1 20.5 (1) 15,754 (1) Luxembourg 3 12 1 NA 473,894 (0) Mexico 17 63 1 NA 49,864 (0) Netherlands 12 45 1 16.5 (0) 131,285 (0) Norway 3 12 1 5.8 (0) 25,109 (1) Peru 1 4 1 NA
108,353 (0)
Philippines 1 4 1 8.8 (0) 65,848 (0) Portugal 2 7 1 25.1 (1) 75,562 (0) Russia 4 16 NA NA NA South Africa 8 24 0 5.6 (0) 49,291 (1) South Korea 5 20 1 26.8 (1) 80,192 (0) Spain 1 4 1 18.6 (1) 29,873 (1) Sweden 6 23 1 6.8 (0) 21,988 (1) Switzerland 8 24 1 22.0 (1) 29,340 (1) Taiwan 6 24 1 22.5 (1) 44,090 (1) Turkey 1 4 1 NA
58,893 (0)
United Kingdom 28 81 0 7.0 (0) 80,902 (0) Total 453 1,548
(continued)
37
Table 1 (continued) The foreign cross-listed firm sample comprises a maximum of 453 unique firms and 1,548 firm-year observations from 37 countries, for which we have sufficient data to estimate our propensity score model (see Table 3). We classify a firm as foreign cross-listed if it is incorporated and headquartered outside of the US and is listed on an US exchange. We eliminate firms with over 98 percent of the trading volume on the US exchanges. We require the foreign cross-listed firms to have at least one observation pre and post Morrison. The table also lists the institutional variables we use in the cross-sectional analyses: (1) We distinguish between countries of code law Legal Origin (= ‘1’) and countries of common law legal origin (e.g., La Porta et al. 1998). (2) The Aggregate Earnings Management score from Leuz et al. (2003). Higher values represent countries with more earnings management. (3) Public Enforcement resources measured as the securities regulators’ budget per billion US$ of GDP based on the extended sample in Jackson and Roe (2009). Variables (2) and (3) are transformed into binary indicators (in parentheses) by splitting the sample at the median.
38
Table 2: Descriptive Statistics for Variables Used in the Regression Analyses Panel A: Foreign Cross-listed Firms
Variable N Mean Std. Dev. Min P25 Median P75 Max
Dependent Variables:
Management Guidance 1,490 0.17 0.37 - - - - -
Guidance Frequency 1,490 0.73 2.23 0.00 0.00 0.00 0.00 19.00
Analyst Following 925 6.21 5.32 1.00 2.00 4.42 8.83 28.58
Forecast Accuracy 883 -0.06 0.16 -1.91 -0.04 -0.02 -0.01 -0.00
Forecast Dispersion 552 0.02 0.03 0.00 0.00 0.01 0.02 0.28
Propensity Score Matching and Other Variables:
Log (Total Assets) 1,490 7.92 2.54 -0.34 6.01 8.27 9.86 12.2
Book-to-Market 1,490 0.71 0.70 0.02 0.33 0.56 0.89 8.09
Earnings Surprise 1,490 0.10 0.21 0.00 0.02 0.04 0.10 3.45
Return Variability 1,490 0.16 0.09 0.01 0.10 0.14 0.19 0.47
Stock Return 1,490 0.64 0.91 -0.89 0.08 0.53 1.01 2.69
Return on Assets 1,490 -0.01 0.21 -2.34 -0.03 0.03 0.07 0.62
GDP Growth 1,490 0.02 0.05 -0.13 -0.02 0.03 0.05 0.20
Foreign Trading 1,384 0.67 0.31 0.00 0.41 0.75 0.96 1.00
Litigation Cost 1,160 0.26 0.15 0.00 0.14 0.25 0.36 0.89 Panel B: Matched US Firms
Variable N Mean Std. Dev. Min P25 Median P75 Max
Dependent Variables:
Management Guidance 1,490 0.79 0.41 - - - - -
Guidance Frequency 1,490 6.66 6.49 0.00 1.00 5.00 10.00 23.00
Analyst Following 925 13.81 6.79 1.00 8.83 13.33 18.08 28.58
Forecast Accuracy 883 -0.05 0.20 -1.91 -0.03 -0.01 -0.00 -0.00
Forecast Dispersion 552 0.01 0.03 0.00 0.00 0.00 0.01 0.29
Propensity Score Matching and Other Variables:
Log (Total Assets) 1,490 7.94 2.14 1.49 6.60 8.09 9.41 12.20
Book-to-Market 1,490 0.57 0.42 0.02 0.29 0.5 0.71 2.64
Earnings Surprise 1,490 0.11 0.24 0.00 0.01 0.03 0.11 2.37
Return Variability 1,490 0.17 0.10 0.03 0.09 0.14 0.21 0.47
Stock Return 1,490 0.78 0.86 -0.80 0.17 0.48 1.19 2.69
Return on Assets 1,490 0.00 0.17 -2.79 -0.02 0.03 0.06 0.41
GDP Growth 1,490 0.01 0.04 -0.03 -0.03 -0.03 0.05 0.05 (continued)
39
Table 2 (continued) The table presents descriptive statistics for the variables used in the regression analyses for the foreign firm sample (Panel A) and matched US firm sample (Panel B). The table presents descriptive statistics for the management guidance matching sample; descriptive statistics for the analyst forecast matching samples are largely similar. We use five dependent variables in our primary analyses: (1) Management Guidance is a binary variable equals to one if the firm issue any management guidance for the fiscal year. (2) Forecast Frequency is the number of issuances of management guidance for the fiscal year. (3) Analyst Following is the monthly number of analysts providing an EPS forecast averaged over the fiscal year. (4) Forecast Accuracy is defined as the monthly absolute difference between the consensus EPS forecast and actual EPS averaged over the fiscal year, scaled by end of fiscal year stock price. We multiply by negative one such that larger values represent more accurate forecast. (5) Forecast Dispersion is the monthly standard deviation of analyst EPS forecasts averaged over the fiscal year, scaled by end of fiscal year stock price. We use the following variables in the propensity score matching model: (1) Total Assets are denominated in US$ millions. (2) Book-to-Market is the ratio of book value of equity divided by market value of equity. (3) Earnings Surprise is the absolute change in earnings before extraordinary items scaled by beginning total assets. (4) Return Variability is the annual standard deviation of monthly stock returns over a firm’s fiscal year. (5) Stock Return is the annual buy-and-hold return including dividends for the fiscal year. (6) Return on Assets is the ratio of earnings before extraordinary items divided by total assets. We control for GDP Growth defined as the two-year growth rate over the contemporaneous period in all our regression analyses. We use two additional variables for our cross-sectional predictions: (1) Foreign Trading equals the proportion of trading volume outside of the US. For US firms, this variable is set to zero. (2) Litigation Cost is measured based on the model in Kim and Skinner (2012), normalized between zero and one, and multiplied by Foreign Trading. Accounting data and market values are measured as of the fiscal year end. Except for variables with natural lower or upper bounds, we winsorize all variables at the first and 99th percentile, and we use the natural log of the raw values where indicated.
40
Table 3: Identification of a Matched US Firm Sample Panel A: Propensity-Score Estimation Using a Probit Regression
Dependent variable: Foreign Log (Total Assets) 0.373***
(15.27)
Book-to-Market -0.222**
(-2.49)
Earnings Surprise 0.039
(0.42)
Return Variability -1.860***
(-2.92)
Stock Return 0.731***
(10.25)
Return on Assets -0.076
(-0.51)
Pseudo R2 0.134 Log likelihood -1347.0 Observations 3,973
Panel B: Covariate Balance between Matched Pairs
Mean Foreign
Mean US
Median Foreign
Median US
t-test difference p-value
KS Bootstrap test difference
p-value Log (Total Assets) 7.827 7.809 8.011 8.122 0.915 0.031** Book-to-Market 0.583 0.733 0.497 0.564 0.000*** 0.006*** Earnings Surprise 0.116 0.100 0.032 0.041 0.347 0.152 Return Variability 0.170 0.162 0.141 0.141 0.239 0.355 Stock Return 0.728 0.565 0.447 0.450 0.007*** 0.003*** Return on Assets -0.003 -0.010 0.028 0.025 0.589 0.235 N (Firms) 431 431 431 431
The table reports the propensity score matching procedures for the management guidance sample; results for the analyst forecast samples are largely similar. Panel A presents logit coefficient estimates and (in parentheses) z-statistics based on robust standard errors clustered by firm. Foreign is an indicator variable set to ‘1’ for the sample of foreign firms. For details of the remaining variables see Table 2. Panel B presents the test statistics of covariate distributions for the foreign and matched US firms. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
41
Table 4: Changes in Voluntary Disclosure Behavior Following Changes in Private Litigation Risk
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.068* 0.850** 3.616*** 0.008 0.007
(1.83) (2.48) (8.06) (0.25) (0.88)
Post*Foreign -0.085*** -0.788** -2.072*** 0.016 -0.003
(-3.08) (-2.55) (-5.38) (0.58) (-0.72)
Log (Total Assets) 0.063*** 0.581*** 2.007*** -0.001 -0.001
(6.20) (4.95) (12.15) (-0.42) (-1.34)
Book-to-Market -0.028 -0.514** -1.982*** -0.090*** 0.013*
(-1.27) (-2.11) (-3.32) (-2.98) (1.97)
Earnings Surprise -0.115* -1.098* 1.172 -0.068 0.007
(-1.72) (-1.89) (0.94) (-0.90) (0.48)
Return Variability 0.071 -3.679 -3.983 -0.451** 0.090**
(0.23) (-1.31) (-0.96) (-2.45) (2.35)
Stock Return 0.016 -0.100 -0.323 -0.019** 0.003
(0.98) (-0.53) (-0.74) (-2.14) (1.26)
Return on Assets 0.078 0.141 -1.196 0.066 -0.034
(1.49) (0.22) (-0.54) (1.09) (-1.14)
GDP Growth -0.112 -0.565 -0.005 -0.069 -0.076 (-0.28) (-0.22) (-0.00) (-0.24) (-0.81)
Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4840 0.3615 0.6034 0.1172 0.1967 N (Observations) 2,980 2,980 1,850 1,766 1,104 N (Foreign-US Pairs) 431 431 284 273 169
The table reports changes in firms’ voluntary disclosure behavior following a change in private litigation risk. We report OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors clustered by firm. The Post variable takes on the value of ‘1’ for fiscal year starts after January 1, 2011 and ends before December 31, 2012; and takes on the value of ‘0’if the fiscal year starts after January 1, 2008 and ends before December 31, 2009. Foreign is an indicator variable set to ‘1’ for the sample of foreign firms. All other variables are defined in Table 2. We include industry- and country-fixed effects in the regression, but do not report the coefficients. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
42
Table 5: Cross-sectional Analyses using Foreign Country Institutions Panel A: Legal Origin (1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.079** 0.908** 3.802*** 0.008 0.007
(2.14) (2.57) (8.70) (0.23) (0.84)
Post*Foreign -0.071** -0.641** -1.727*** 0.018 -0.005
(-2.49) (-2.00) (-3.92) (0.64) (-0.89)
Post*Foreign*Code Law -0.046 -0.432** -0.905*** -0.003 0.004
(-1.60) (-2.24) (-2.63) (-0.15) (0.82)
Control Variables Included Included Included Included Included Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4899 0.3647 0.6106 0.1170 0.1759 N (Observations) 2,948 2,948 1,826 1,746 1,084 N (Foreign-US Pairs) 427 427 281 270 166
Panel B: Aggregate Earnings Management
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.014 1.058** 3.503*** 0.006 -0.003
(0.27) (2.25) (6.82) (0.13) (-0.34)
Post*Foreign -0.071** -0.887** -1.805*** 0.022 -0.006
(-2.31) (-2.31) (-4.10) (0.60) (-0.98)
Post*Foreign*High -0.026 -0.576* -1.154*** 0.006 0.011** Aggregate Earnings Management (-0.62) (-1.76) (-3.28) (0.26) (2.39)
Control Variables Included Included Included Included Included Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.5038 0.3816 0.6207 0.1198 0.1598 N (Observations) 2,112 2,112 1,352 1,294 790 N (Foreign-US Pairs) 311 311 211 203 122
(continued)
43
Table 5 (continued) Panel C: Public Enforcement (1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.099** 1.182*** 3.852*** 0.009 -0.004
(2.43) (3.16) (8.12) (0.25) (-0.72)
Post*Foreign -0.087*** -0.742** -1.840*** 0.021 -0.004
(-3.08) (-2.29) (-4.62) (0.73) (-0.88)
Post*Foreign*Low -0.029 -0.419* -0.829** -0.016 0.009** Public Enforcement (-0.90) (-1.91) (-2.47) (-0.67) (2.04)
Control Variables Included Included Included Included Included Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4910 0.3648 0.6150 0.1182 0.1705 N (Observations) 2,796 2,796 1,730 1,662 1,036 N (Foreign-US Pairs) 403 403 266 257 158
The table presents country-level cross-sectional analyses of changes in firms’ voluntary disclosure behavior following a change in private litigation risk. The Post variable takes on the value of ‘1’ for fiscal year starts after January 1, 2011 and ends before December 31, 2012; and takes on the value of ‘0’if the fiscal year starts after January 1, 2008 and ends before December 31, 2009. Foreign is an indicator variable set to ‘1’ for the sample of foreign firms. Panels A – C report results for the following foreign country institutions: (1) Legal Origin, (2) Aggregate Earnings Management and (3) Public Enforcement. All variables are defined in Tables 1 and 2. The panels only report the OLS coefficient estimates and (in parentheses) t-statistics of the main variables of interest, but include the full set of controls and fixed effects (see Table 4). ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
44
Table 6: Cross-sectional Analyses using Firm-Level Measures of Change in Expected Private Litigation Costs Panel A: Proportion of Foreign Trading
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.066 0.854** 3.587*** 0.007 0.007
(1.64) (2.32) (7.96) (0.21) (0.90)
Post*Foreign -0.073** -0.583 -1.711*** 0.027 -0.002
(-2.02) (-1.61) (-3.82) (0.90) (-0.46)
Foreign*High Foreign Trading -0.177*** -0.503* -3.022*** 0.025 -0.002 (-4.51) (-1.78) (-4.89) (1.50) (-0.32) Post*Foreign*High Foreign Trading -0.015 -0.419** -0.669* -0.023 -0.002
(-0.48) (-2.04) (-1.92) (-1.15) (-0.36)
Log (Total Assets) 0.063*** 0.603*** 1.949*** -0.001 -0.001
(5.55) (4.69) (11.90) (-0.37) (-1.39)
Book-to-Market -0.019 -0.411 -1.905*** -0.090*** 0.013**
(-0.80) (-1.63) (-3.38) (-3.01) (2.11)
Earnings Surprise -0.102 -1.129* 0.797 -0.066 0.007
(-1.56) (-1.77) (0.65) (-0.89) (0.48)
Return Variability 0.121 -3.532 -4.677 -0.448** 0.088**
(0.36) (-1.12) (-1.11) (-2.41) (2.30)
Stock Return 0.015 -0.088 -0.317 -0.019** 0.003
(0.92) (-0.43) (-0.73) (-2.14) (1.34)
Return on Assets 0.080 0.096 -1.438 0.067 -0.034
(1.42) (0.14) (-0.67) (1.09) (-1.13)
GDP Growth -0.139 -0.525 0.476 -0.052 -0.077 (-0.31) (-0.19) (0.12) (-0.18) (-0.83)
Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4926 0.3618 0.6187 0.1168 0.1952 N (Observations) 2,764 2,764 1,842 1,758 1,096 N (Foreign-US Pairs) 397 397 283 272 168
(continued)
45
Table 6 (continued) Panel B: Ex-ante Litigation Cost
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.058 0.841** 3.771*** 0.009 0.008
(1.29) (2.05) (7.61) (0.25) (0.82)
Post*Foreign -0.044 -0.330 -1.436*** 0.028 -0.005
(-1.13) (-0.80) (-2.93) (0.87) (-0.83)
Foreign*High Litigation Cost -0.151*** -0.414 -1.443** -0.027 0.001 (-3.44) (-1.28) (-2.05) (-1.21) (0.21) Post*Foreign*High Litigation Cost -0.066* -0.689*** -1.032** -0.011 -0.000
(-1.78) (-2.72) (-2.58) (-0.45) (-0.05)
Log (Total Assets) 0.071*** 0.637*** 2.025*** -0.002 -0.001
(6.84) (4.90) (12.01) (-0.64) (-1.10)
Book-to-Market -0.040* -0.477** -1.930*** -0.085*** 0.011*
(-1.88) (-2.00) (-2.97) (-2.69) (1.71)
Earnings Surprise -0.050 -0.719 0.899 -0.135 0.011
(-0.93) (-1.11) (0.47) (-1.24) (0.51)
Return Variability 0.357 -3.233 -3.742 -0.586*** 0.101**
(1.40) (-1.18) (-0.83) (-2.83) (2.59)
Stock Return 0.034** 0.110 -0.395 -0.015* 0.003
(1.98) (0.50) (-0.74) (-1.68) (1.28)
Return on Assets 0.148*** 0.615 -1.687 0.042 -0.019
(2.64) (0.83) (-0.63) (0.70) (-0.63)
GDP Growth -0.070 -1.735 -1.134 -0.117 -0.083 (-0.13) (-0.53) (-0.25) (-0.36) (-0.77)
Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4745 0.3590 0.5906 0.1327 0.2202 N (Observations) 2,202 2,202 1,496 1,438 972 N (Foreign-US Pairs) 317 317 227 221 146
The table reports firm-level cross-sectional analyses of changes in foreign firms’ voluntary disclosure behavior following a change in private litigation risk. The Post variable takes on the value of ‘1’ for fiscal year starts after January 1, 2011 and ends before December 31, 2012; and takes on the value of ‘0’if the fiscal year starts after January 1, 2008 and ends before December 31, 2009. Panels A and B report results for the following proxies of firm-level expected litigation cost: (1) High Foreign Trading is an indicator variable equals to ‘1’ if the foreign firm’s Foreign Trading is above the foreign firm sample median. Foreign Trading equals the proportion of trading volume outside of the US. (2) High Litigation Cost is an indicator variable equals to ‘1’ if the foreign firm’s Litigation Cost is above the foreign firm sample median. All other variables are defined in Table 2. We report OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors clustered by firm. We include industry- and country-fixed effects in the regression, but do not report the coefficients. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
46
Table 7: Sensitivity Analyses using Alternative Matching Procedure
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.300*** 1.585*** 2.573*** -0.004 0.001
(4.91) (4.05) (5.42) (-0.17) (0.14)
Post*Foreign -0.309*** -1.349*** -1.089*** 0.005 0.001
(-5.55) (-3.77) (-2.75) (0.22) (0.49)
Log (Total Assets) 0.052*** 0.204*** 1.658*** 0.001 -0.003**
(4.57) (2.90) (10.72) (0.25) (-2.16)
Book-to-Market 0.011 0.086 -2.734*** -0.091*** 0.007
(0.47) (0.48) (-4.52) (-3.15) (1.28)
Earnings Surprise 0.006 -0.021 1.560 0.015 0.032**
(0.11) (-0.10) (1.52) (0.30) (2.36)
Return Variability 0.227 -0.050 -4.390 -0.359** 0.029
(0.78) (-0.03) (-1.13) (-2.35) (1.29)
Stock Return -0.010 -0.071 -0.258 -0.019*** 0.002
(-0.56) (-0.69) (-0.76) (-2.79) (1.16)
Return on Assets 0.025 0.072 -1.035 -0.005 -0.024
(0.77) (0.55) (-0.85) (-0.21) (-1.41)
GDP Growth -0.290 -2.376 2.447 -0.020 -0.055 (-0.69) (-1.01) (0.57) (-0.11) (-1.63) Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.1963 0.1305 0.4408 0.1244 0.1791 N (Observations) 2,850 2,750 1,688 1,636 1,032 N (Foreign-US Pairs) 419 402 262 251 159
The table reports additional sensitivity analyses of our base specifications (see Table 4) examining changes in firms’ voluntary disclosure behavior following a change in private litigation risk. We use an alternative matching procedure for the US control firms, where we match based on the level of the dependent variable in the pre-period conditional on year and industry. We select the firm with the closest propensity score in case of multiple matches. We report OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors clustered by firm. We include industry- and country-fixed effects in the regression, but do not report the coefficients. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
47
Table 8: Sensitivity Analyses using Alternative Sample Specifications Panel A: Canadian Firms
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.060* 0.560 3.517*** -0.039 0.005 (1.93) (1.11) (8.89) (-0.84) (0.63) Post*Foreign -0.070* -0.338 -1.143** 0.057 -0.012 (-1.86) (-0.64) (-2.11) (1.16) (-1.29)
Control Variables Included Included Included Included Included Industry Fixed Effects Included Included Included Included Included Adjusted R2 0.5047 0.3556 0.5832 0.1295 0.1905 N (Observations) 1,116 1,116 630 614 496 N (Canadian-US Pairs) 159 159 95 93 74
Panel B: Excluding Canadian Firms
(1) (2) (3) (4) (5)
Dependent variables: Management Guidance
Guidance Frequency
Analyst Following
Forecast Accuracy
Forecast Dispersion
Post 0.074* 1.079*** 3.820*** 0.029 0.002 (1.80) (2.83) (7.56) (1.03) (0.32) Post*Foreign -0.096*** -1.082*** -2.555*** -0.005 0.005 (-2.73) (-3.03) (-5.58) (-0.20) (1.58)
Control Variables Included Included Included Included Included Industry Fixed Effects Included Included Included Included Included Country Fixed Effects Included Included Included Included Included Adjusted R2 0.4928 0.3683 0.6340 0.1209 0.3596 N (Observations) 1,864 1,864 1,220 1,152 608 N (Foreign-US Pairs) 272 272 189 280 95
The table reports additional sensitivity analyses of our base specifications (see Table 4) examining changes in firms’ voluntary disclosure behavior following a change in private litigation risk. In Panel A, we include only Canadian-US firm pairs. In Panel B, we exclude Canadian-US firm pairs. The panels only report the OLS coefficient estimates and (in parentheses) t-statistics of the main variables of interest, but include the full set of controls and fixed effects. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
48
Table 9: Market-based Measures of Disclosure (1) (2) (3)
Dependent variables: Absolute Size Adjusted Returns
Share Turnover
Bid-Ask Spread
Post -0.010* -0.002 -0.002
(-1.85) (-1.32) (-1.49)
Post*Foreign 0.008** 0.002** -0.001
(2.09) (2.02) (-0.98)
Log (Total Assets) -0.002*** 0.002*** -0.002***
(-3.69) (5.78) (-6.81)
Book-to-Market 0.007** -0.002* 0.008***
(2.57) (-1.93) (4.44)
Earnings Surprise -0.007 0.004 -0.002
(-0.80) (0.91) (-0.60)
Return Variability 0.113*** 0.047*** 0.001
(7.39) (3.30) (0.12)
Stock Return 0.000 -0.000 -0.000
(0.29) (-0.35) (-0.37)
Return on Assets 0.007 0.008*** -0.010*** (1.06) (2.90) (-3.04) GDP Growth -0.118* -0.000 -0.006 (-1.96) (-0.03) (-0.33)
Industry Fixed Effects Included Included Included Country Fixed Effects Included Included Included Adjusted R2 0.2507 0.3454 0.3998 N (Observations) 1,644 1,684 1,630 N (Foreign-US Pairs) 240 245 236
The table reports changes in firms’ voluntary disclosure behavior following a change in private litigation risk using market-based measures as proxies for voluntary disclosure. We use the following four market-based dependent variables: (1) Absolute Size Adjusted Returns is the absolute value the three-day raw return minus the return of the corresponding size-decile index centered at the dates of the quarterly earnings announcement (“QEA”). (2) Share Turnover is the three-day average of the ratio of shares traded to total shares outstanding around QEA. (3) Bid-Ask Spread is the three-day average of the difference between the bid and ask price divided by the mid-point of the bid and ask price around QEA, multiplied by 100. All other variables are defined in Tables 2 and 4. We exclude foreign-US firm pairs where Foreign Trading is above 90% to ensure sufficient US trading. We report OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors clustered by firm. We include industry- and country-fixed effects in the regression, but do not report the coefficients. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed).
49