This paper concludes that companies that voluntarily disclose their greenhouse gas emissions and carbon reduction strategies see stock prices rise. Through reasearching CSR newswire releases and company stock prices, Griffin and Sun uncovered this interesting find. See the report in full to understand the process involved that led to these findings.
Read Yuan Sun's summary of the report in It Pays to Be Green.
To download the document and join the conversation,
Join Now or Log in
Electronic copy available at: http://ssrn.com/abstract=1995132
Going Green: Market Reaction to
CSR Newswire Releases
By
Paul A. Griffin1 and Yuan Sun2
1 University of California, Davis
2 University of California, Berkeley
January 29, 2012
Electronic copy available at: http://ssrn.com/abstract=1995132
ii
Going Green: Market Reaction to CSR Newswire Releases
Abstract
Voluntary disclosure theory predicts that an optimal disclosure decision should produce an overall
net benefit for shareholders, and that such net benefit should decrease in public information
availability. This study supports the predictions of disclosure theory in the context of climate
change. Using voluntary disclosures made through the CSR newswire service, we find that
managers’ disclosure decisions involving greenhouse gas emissions produce positive returns to
shareholders. This response varies negatively with company size and public information
availability. For small companies in a limited public information environment, we find that mean
adjusted share price increases significantly by 2.32 percent over days -2 to 2 around the CSR
newswire release date. Our sample of disclosing companies received an aggregate market value
boost from their CSR newswire releases of approximately ten billion dollars, independent of
differences in public information availability.
1
Going Green: Market Reaction to CSR Newswire Releases
1 Introduction
Growing awareness of climate change has led to increased demands from individuals,
advocacy groups, and government regulators for companies to provide additional information
about their climate change strategies and plans for managing and reducing carbon emissions.
These demands have resulted in companies allocating increased resources to communicating
information about their carbon footprint to interested parties. Although users will soon have access
to standardized carbon emissions data collected by the U.S. Environmental Protection Agency
(EPA 2009) and the California Air Resources Board (CARB 2010), companies currently report
most of their green information to interested parties voluntarily, either directly or through various
non-governmental channels. Voluntary disclosure theory suggests that companies should make
disclosure decisions in the best interests of shareholders, net of the costs of disclosure such as
agency costs, proprietary disadvantage (Verrecchia 1983, Diamond and Verrecchia 1991, Lambert
et al. 2007), and outsiders’ knowledge of information uncertainty (Dye 1985). Voluntary
disclosure theory also suggests that the net benefits of additional disclosure to shareholders should
depend on the firm’s information environment, in particular, the level of information available to
outside shareholders. For example, when outside shareholders’ information availability is low, the
theory predicts a new disclosure about company activities should elicit more price response as
traders look to benefit from the heretofore private information. On the other hand, a new
disclosure should yield less price response when outside information availability is high, as traders
have already priced into the stock their greater knowledge about company activities from
competitive sources. Yohn (1998) confirms this prediction in the context of earnings
announcements and finds that companies with less (more) public information availability yield
2
stronger (weaker) price reactions to earnings. Similarly, Freeman (1987) finds that smaller (larger)
companies’ shares respond more (less) around earnings announcements, reasoning that small-stock
traders have higher per share information costs than large-stock traders. This leads to reduced
public information availability for smaller companies, which increases the sensitivity of share price
to new information.
In this study, we test the preceding propositions in the context of climate change disclosures
and posit that shareholders should respond positively to voluntary climate change disclosures
consistent with the net benefits intended by the disclosure, and that such response should vary with
outside shareholders’ (public) information availability. We capture the overall net benefit to the
company as an increase in shareholder value, as shareholders’ expectations interpreted broadly
should embrace the business profit and the environmental and social consequences of managers’
disclosure decisions; and we proxy for public information availability using company size and
financial analyst variables. Our results are intended to contribute to a growing empirical literature
(reviewed in section 2) on shareholders’ recognition of and response to climate change disclosures.
Company managers should be informed about the overall effects of their disclosure decisions; and
shareholders, analysts, and investment managers should understand the shareholder valuation
implications of companies’ climate change disclosures.
Specifically, we examine shareholders’ response to a unique set of disclosures about climate
change made by U.S. companies through the Corporate Social Responsibility (CSR) newswire
service. We study the CSR newswire service for several reasons. First, companies’ CSR releases
meet the criteria of a voluntary disclosure. Companies decide on the content, timing, and
dissemination (CSR in this case) of a disclosure. Second, CSR newswire claims to be the global
leader in disseminating news about corporate social responsibility and sustainability, currently with
one-quarter of one million page views every month. This enables us to study a large sample of
disclosures over several years for a wide range of industries made through a single channel. To
3
further strengthen our tests, we restrict our analysis to CSR releases about greenhouse gas (GHG)
or carbon emissions, as recent evidence across Australia, Canada, and the United States (Chapple et
al. 2011, Griffin et al. 2011, Matsumura et al. 2011) concludes that shareholders find GHG
emissions as value relevant and price them as an off-balance sheet liability; whereas the literature
reports sparse evidence on responses to other categories of non-financial disclosure, possibly
because the content of such is more varied and the implications less clear. Eccles et al. (2011) also
support this view based on the number of user hits from Bloomberg. They document that, of the
many non-financial measures available to investors, information-intensive institutional investors
such as hedge fund managers and sell-side analysts have the highest demand for GHG emissions
amounts and environmental disclosure quality scores. Our results also contribute to a developing
literature on newswires in general, and how they aid in price discovery in capital markets (Li and
Ramesh 2010).
Third, CSR newswire conveys fresh information to users (see note 2), also a critical aspect of
our research design. This differs from recent studies that rely on survey data, such as from the
Carbon Disclosure Project (CDP). While CDP is another leading source of GHG data, investors
and others receive information on large companies’ emissions at best some nine to ten months after
year end, by which time other channels, including CSR newswire, may have preempted much of
the news content of that information. An alternative is to study companies’ 8-K filings, as in
Griffin et al. (2011). While 8-Ks reveal timely and potentially significant news because the filer
has a regulatory obligation to disclose promptly and to meet a materiality threshold, evidence of a
shareholder response to an 8-K does not imply that shareholders benefit from a voluntary
disclosure, which is the purpose of this analysis. In fact, in the case of emission-intensive
companies, the evidence suggests that 8-K climate change disclosures trigger a negative reaction
on the average, inconsistent with the intended stock market outcome of a voluntary disclosure.
4
To summarize, our study adds to the literature by examining whether shareholders benefit
when companies make decisions to disclose climate change information voluntarily, in our case,
through the CSR newswire service. We view fact-based analysis of companies’ voluntary
disclosure decisions as an essential element of the climate change disclosure debate, which thus far
has focused disproportionately on users’ demands for additional information absent evidence that
would buttress those demands (e.g., Coburn et al. 2011). To the best of our knowledge, the
following results relating to the market effects of voluntary disclosures about carbon emissions are
unique to the literature.
Our study first confirms the theoretical proposition that companies disclose voluntary climate
change information in the interests of shareholders by documenting a significantly positive
response to CSR newswire releases involving GHG emissions in the three day interval around the
disclosure date. Second, our study finds that this response varies negatively with company size and
public information availability. This result supports our second proposition that voluntary
disclosure of GHG emissions data benefits companies more in environments with low public
information availability. These results are also unchanged when we assess shareholder response in
different ways, namely, stock return in excess of a market index, stock return relative to a non-
event period, stock return relative to a matched sample of non-discloser companies, and stock
return from a Fama and French (1993) four-factor model that controls for market return in excess
of the risk free rate, company size, growth, and stock price momentum. We also challenge our
results by testing whether a portfolio that is long (short) in companies with low (high) public
information availability might earn a significant hedge portfolio return from possible mispricing.
We find no evidence to suggest that investors exploit differences in companies’ information
environment to earn positive hedge fund returns, which is another way of understanding that
investors recognize the benefits of climate change disclosures efficiently with an appropriate
adjustment in shareholder value based on public information availability, although measurement
5
error could also explain this result. Finally, our results send a statistically reliable “it pays to be
green” message to managers about the compatibility of green disclosure and shareholder value
maximization. The discloser companies in our study received an aggregate boost in market value
from their CSR newswire releases of approximately ten billion dollars, independent of differences
in public information availability calculated as the sum over all CSR releases of the market value
of stock i at CSR release day t-1 times the excess stock return on day 0.
Our study proceeds as follows. Section 2 discusses the literature. Section 3 describes the
sample and data. Section 4 summarizes the results of testing for a market response, and section 5
concludes.
2 Related literature
We first discuss disclosure theory, as this helps us understand the objectives and expected
consequences of voluntary disclosure for company shareholders. Verrecchia (1983) and Diamond
and Verrecchia (1991) inter alia predict that disclosure decisions by managers should be value-
enhancing for shareholders and part of a strategy that optimally trades off the benefits to
shareholders against the costs to the company, where those costs include out-of-pocket costs,
agency costs, litigation risk, and proprietary disadvantage. Others argue that a beneficial disclosure
strategy should be value-enhancing through a reduction in the cost of capital (Lambert et al. 2007)
or information asymmetry (Mensah et al. 2003), although others (e.g., Akins et al. 2011,
Armstrong et al. 2011) contend that a reduction in the information component of cost of capital has
limited impact on shares traded in competitive stock markets such as the NYSE and NASDAQ.
Disclosure theory also considers the effects of adverse selection (Grossman and Hart 1980) and the
prediction that shareholders in a voluntary disclosure setting who know that managers have private
information may interpret the absence of a disclosure negatively compared to an otherwise similar
company disclosure that reveals positive news. While it is beyond the scope of our study to
examine otherwise identical non-disclosers, our design checks for possible non-discloser effects by
6
analyzing the returns of a matched sample of non-discloser companies around the disclosure dates
of otherwise similar discloser companies.
Disclosure theory should also apply in the context of CSR newswire disclosures about climate
change, because even though the objectives of green disclosure relate to business profit and
broader considerations such as environmental and social goals, the intended consequences of
meeting those goals are ultimately the same, namely, improved shareholder expectations about
company performance and future cash flows. However, given the newness of the context and
investors’ changing views on corporate environmental and social responsibility, how climate
change information links to shareholder value may be more difficult to detect relative to traditional
financial statement measures such as earnings, although this mostly poses a design challenge.
Overall, this literature predicts that companies will disclose voluntarily only when it is worthwhile
to do so net of disclosure costs. This suggests a positive impact on shareholders (or a non-negative
effect at a minimum), which is our first proposition.
Disclosure theory further posits that managers condition their disclosure decisions and
strategies on attributes of the information environment. In capital markets with high public
information availability (and low information asymmetry) investors update their beliefs about
companies’ prospects frequently and promptly, based on multiple channels of information, so that
current prices reflect relatively well-formed expectations about future returns. In this setting, news
elicits a limited price response, as much is known already through the other channels. In contrast,
prices will be more responsive in an information environment with less frequent disclosure from
fewer sources, assuming investor sophistication and information quality unchanged. Voluntary
disclosure in such setting should therefore elicit a greater price response as investors update their
uncertain expectations in response to the new information.
Empirical studies by Freeman (1987) and Yohn (1998) confirm the notion that stock price
response differs on the basis of public information availability, although this is primarily in the
7
context of earnings announcements. Freeman (1987) proxies for public information availability
using company size, and Yohn (1998) uses analyst following and size as the proxies for public
information availability. This relation also occurs in the context of mandated disclosures about
climate change in regulatory filings, wherein Griffin et al. (2011) find that more emission-intensive
companies elicit less investor response than less-emission intensive companies around the 8-K
filing date because the former group’s shares trade in a richer information environment (as proxied
by company size and the CDP disclosure quality score). Overall these studies support our second
testable proposition – that shareholders’ response to a voluntary CSR newswire disclosure should
vary negatively with proxies for the level of public information availability.
Our study also builds upon a second stream of literature that relates green information to
shareholder value. By showing an association between stock prices and an attribute of a green
disclosure that varies across companies, several studies extrapolate this result to infer that
shareholders use this information to set prices. By documenting significant associations between
shareholder value or cost of capital and voluntary environmental or climate change disclosures,
these studies support the inference that share prices adjust appropriately to green information
conditional on an attribute of that information; such as the type (Hughes 2000), quality (Plumlee et
al. 2008), and timing (Dhaliwal et al. 2010) of the disclosure.
More recently, Chapple et al. (2011), Griffin et al. (2011), and Matsumura et al. (2011)
document a negative association between GHG emissions reported to the CDP and shareholder
value, and conclude that investors price carbon emissions as an undisclosed off-balance sheet
liability. Cormier and Magnan (1997) also document that investors price implicit environmental
liabilities through reports on water pollution. Griffin et al. (2011) further contend and find that
companies that do not report emissions to the CDP also show a negative relation between estimated
GHG emissions and stock price, consistent with the fact that the CDP is only one of many channels
of information about companies’ carbon emissions. These studies, however, mostly document
8
cross-sectional associations between stock price or cost of capital and an environmental or climate
change disclosure variable rather than a response that flows directly from the disclosure event per
se.1
A related set of studies examines the stock price response to environmental or climate change
disclosures using a design similar to this paper. The focus of these studies, however, differs from
the present paper, since the earlier papers address investors’ response to company disclosures about
one-off events. These include the Bhopal disaster (Blacconiere and Patten 1994), the passage of
Superfund legislation (Blacconiere and Northcutt 1997), the Exxon oil spill (Patten and Nance
1998), the Placer Dome accident (Magness 2010), and the first-time issuance of a corporate
sustainability report (Guidry and Patten 2010). Other studies examine environmental disclosures in
financial reports (Murray et al. 2006, Jones et al. 2007).
With the possible exception of Guidry and Patten (2010), who find no market reaction to the
issuance of an initial sustainability report by Australian companies, none of the events and/or
disclosures studied so far would seem purely voluntary by the company. This absence of evidence
about the consequences of specific, company-initiated, voluntary disclosures about climate change
is surprising given that most advocates of increased green disclosure premise their argument on the
expectation rather than the fact of a shareholder or company benefit from such disclosure (e.g.,
Coburn et al. 2011). The present study addresses this knowledge gap by using a unique data set of
green disclosures made voluntarily by U.S. companies through the CSR newswire service.
1 Another stream of literature investigates relations between environmental performance and environmental
disclosure, arguing that environmental performance should influence the amount and quality of
environmental disclosure. It is unclear from this literature, though, whether better environmental performers
are better voluntary disclosers or worse voluntary disclosers. For example, Clarkson et al. (2011a) find a
positive relation between environmental performance and disclosure based on a sample of Canadian
companies. On the other hand, Clarkson et al. (2011b) find that pollution-intensive companies in Australia
disclose more and better (more objective) information than their less pollution-intensive counterparts. Also,
9
3 Sample and data
We select the CSR newswire sample by searching all archived releases in www.csrwire.com
as of December 31, 2010 that contain the phrases “greenhouse gas emissions”, “carbon emissions”,
or “CO2 emissions”, hereafter, collectively, GHG releases. This initial search identifies 575 GHG
releases from a total of 18,588 CSR releases over 2000 through 2010. We then eliminate those that
do not relate to a specific company (48 releases), do not have a Compustat or CRSP identifier (312
releases), and have missing total assets at the end of the current fiscal year from Compustat and/or
missing stock return data around the disclosure event day from CRSP (43 releases). This results in
a final CSR sample of 172 GHG releases by 84 companies over 2000-2010. Panels A and B of
table 1 summarize the sample selection process. Panel B also shows the distribution of dates across
months and years. While the observations do not cluster monthly, more disclosures occur over
2007 to 2010, consistent with companies’ response to increased public interest in green information
and possibly through the growth efforts of CSR itself.2
Next, we select a matched control sample following a four-step process. The first step creates
a merged data set from Compustat and IBES that includes all company-year observations in years
2000 through 2010 with data for total assets (at from Compustat) and the number of analysts
following the company as of the end of each year (numest from IBES). In the second step, we
select a matching company from the merged Compustat/IBES data set for each company in the
CSR sample such that the company (a) is in the same sector as the CSR company (gsector from
Compustat) and (b) minimizes the sum of squares of the absolute difference in log of total assets
and number of analysts following in the year of the first CSR press release, that is, we choose a
2 We also checked to verify that CSR newswire conveys fresh information to users. For each CSR release,
we first examined the company’s website to determine whether the CSR release in our sample is new
information for investors. In all but two cases, the company’s website (often in the news section) disclosed
the CSR release on the same date. In two cases, the companies disclosed the same news on their websites
one day before CSR newswire release date. Second, we searched Direct Edgar for company press releases
about a CSR disclosure as an attachment to an 8-K filing. We found only one company that disclosed the
same news in the 8-K. However, the 8-K filing date followed the day of CSR newswire release.
10
matching company that minimizes (Log atmatched – Log atCSR)
2 + (numestmatched - numestCSR)
2 from
among all possible candidates in the merged Compustat/IBES data set. Third, we require that the
matched company should not be in the CSR sample and not serve as a control for more than one
CSR company. Fourth, we assign to the matched company a pseudo release date that corresponds
to the release date of the CSR company.
To illustrate the four-step process, consider a CSR sample company “A” with a December 31
fiscal year end and GHG releases through CSR on Feb 1, 2005, July 1 2005, Mar 1 2006, and Oct 1
2009 (four company-year filings). We start by finding a matching company in the same gsector
with the smallest log at and numest difference in fiscal 2005. We then tag the 2005
Compustat/IBES company observation, and create two matched company-year release dates in
fiscal 2005 (one is for Feb 1, 2005, the other is for July 1, 2005). We then tag the two
Compustat/IBES company observations with the smallest Log at and numest differences in 2006
and 2009, and assign the pseudo release dates Mar 1 2006 and Oct 1, 2009, respectively, to those
two observations. In this way, we match four company-year observations in the CSR sample with
four company-year observations in the Compustat/IBES data set. This procedure produces a final
sample of 172 company-year GHG releases for 84 CSR companies and 172 matched company-year
filings for 84 matched companies. We use the matched sample to control for the shareholder
returns of an otherwise similar sample with the closest asset and analyst forecast attributes. The
use of a matched set of pseudo release dates also checks for possible information transfer effects,
which could be positive if the CSR disclosure conveys favorable industry (gsector) information or
negative, for example, due to a possible adverse selection effect under the strict assumption that
investors view the matched sample companies as those than could have made a GHG emission
disclosure on the same day as the CSR company but chose not to do so.
Panel C of table 1 shows the distribution of industries and the mean asset size of each industry.
As the panel shows, each industry matches well on size, and our sample is reasonably diverse
11
across the sectors, although the financial and energy sectors represent the most frequent number of
observations, and industrials and materials sectors represent the least frequent. In the later analysis,
we assess whether differences in these observation frequencies have an impact on our statistical
tests.
Table 2 reports summary statistics for the CSR and matched samples, where we select the
Compustat variables for each CSR release that correspond with the year of a CSR disclosure as per
panel B of table 1. As panels A and B of table 2 indicate (and also panel C of table 1), the
matching procedure produces two samples that are statistically indistinguishable in size (mean log
at = 23.648 for CSR versus 23.365 for matched) and analyst following (mean numest = 15.264 for
CSR versus 14.972 for matched); and a two-sample t-test of difference in mean size or number of
analysts is not significant. The two samples are statistically similar on the other descriptor
variables as well, based also on a two-sample test of difference in the means. On the other hand,
panel C shows that the CSR and matched samples are larger, more profitable, and have greater
analyst following than the Compustat population in general (of 81,587 company/year observations
in 2000-2010), which means that our results based on the CSR and matched samples are not
representative of the larger population.
4 Results
Table 3 summarizes our initial results that test whether shareholders respond favorably to a
CSR newswire release around the announcement date (days -1 to 1) (proposition 1). Panel A
reports tests of whether mean market-adjusted return for the CSR sample (a) exceeds zero, (b)
exceeds the mean market-adjusted return for the matched sample, and (c) exceeds the mean
market-adjusted return for the CSR sample over a non-announcement interval (days -20 to 20,
excluding days -1 to 1). We define market-adjusted return for company i as the common stock
return on day t less the return on the CRSP value-weighted market index. As predicted, we find
that shareholders on average respond positively to a CSR newswire release, by 0.468 percent over
12
the three day window. This amount is significantly different from zero based on a one-tailed t test
(p-value = 0.042). We also find that the mean excess return from days -1 to 1 is significantly
higher than the mean market-adjusted return over a non-announcement period (p-value = 0.027).
Furthermore, the market-adjusted return for the CSR sample over the event period is significantly
higher than the return for the matched sample (p-value = 0.098). We detect no significant response
for the matched sample around the pseudo release date that corresponds to the release date of the
CSR company. Hence, independent of company size or public information availability, the tests in
panel A document a small but reliably positive shareholder response to a CSR newswire release.
Panel B of table 3 documents a significantly positive response using a multi-factor approach.
We use the Fama and French (1993) four-factor model for this purpose and conduct the following
regression:
Rt – rft = α + β1RMRFt + β2SMBt + β3HMLt + β4MOMt + et, (1)
where t refers to days -1 to 1, Rt is the stock return on day t, rft is the risk-free rate (measured as the
one month Treasury bill rate); RMRFt (excess market factor) is the return of the value-weighted
stock market portfolio minus the return on the one month Treasury bill; SMBt HMLt and MOMt are
portfolio returns on zero-investment, factor-mimicking portfolios for size, book-to-market equity,
and stock return momentum, respectively; and e is the residual error. The intercept or alpha
coefficient in the regression tests for an announcement effect after controlling for market return in
excess of the risk free rate, company size, growth, and a momentum factor. Panel B reports a
positive and significant alpha coefficient for this regression for the CSR sample (coef. = 0.00153,
p-value = 0.075, robust standard error). Because of differences in industry composition (table 1,
panel C), we also test for positive alpha where the regression standard errors adjust for return
associations (clusters) within an industry. The results are more significant under this assumption
(p-value = 0.034). Furthermore, we show that the alpha coefficient for the CSR sample is
significantly higher than that for the matched sample under the clustered standard error. For the
13
matched sample, on the other hand, we find no significant and positive alpha coefficient. Finally,
as a check on our results, we estimate model 1 where t refers to days -20 to 20, excluding days -1 to
1for both the CSR sample and the matched sample. We should not observe a significant alpha
under this alternative, and we do not (e.g., p-value = 0.533 and 0.471 under the clustered standard
error for the CSR sample and the matched sample respectively). Overall, the results of table 3
confirm our first proposition from voluntary disclosure theory, namely, that CSR disclosures
should be in the best interests of shareholders. Contrariwise, our results show no evidence of an
overall detriment to shareholders following voluntary disclosures about green activities.
Next, to strengthen this initial result, we examine whether shareholders’ response in the CSR
announcement interval varies with an attribute of the information environment. This makes it more
likely that the response relates to the climate change release and not an unrelated information item.
In proposition 2, we predict a more positive response for smaller companies and companies that
trade in an environment with less public information availability. We proxy for public information
availability in three ways, namely, the number of analysts following the stock, absolute value of
analysts’ earnings forecasts errors, and standard deviation in analysts’ earnings forecast (Brennan
and Subrahmanyam 1995, Chung et al. 1995, Yohn 1998, Hong et al. 2000, Frankel and Li 2004,
Zhang 2006). Because size can also be a systematic risk factor priced by investors to reflect return
uncertainty, we concentrate on excess stock returns using the Fama and French (1993) model
before partitioning on analyst forecast variables hypothesized to reflect public information
availability, although we present results based on market-adjusted returns as well.
Table 4 present results partitioned by large and small company size and high and low
information availability. We summarize the key results as follows. First, large companies do not
show mean excess return over days -1 to 1 that is significantly greater than zero, the mean excess
return over days -20 to 20 excluding days -1 to 1, or the mean excess return for the matched sample.
Small companies, on the other hand, show significantly greater mean excess return over days -1 to
14
1 for all three significance benchmarks. The mean excess return difference between the large and
small companies is also significant at less than 5 percent (p-value= 0.031).
Second, for companies with greater analyst coverage, and therefore high public information
availability, we show that the mean excess return over days -1 to 1 is marginally significant for two
significance tests (i.e., p-value = 0.091 when we test the mean excess return against zero, and p-
value = 0.062 when we test the mean excess return against the return over the non-event period).
The mean excess return over days -1 to 1, however, is not significantly different from the return for
the matched sample. For companies with less available public information, we find that the mean
excess return over days -1 to 1 is significant for all three significance tests. However, the return
difference between the high and low information availability companies is not statistically
significant.
Third, when we compare the mean excess return for the extreme groups of small companies
with low public information availability and large companies with high public information
availability, we show that the former experience a strong positive stock reaction over event days -1
to 1. Overall, these results support our second proposition that voluntary disclosure benefits
companies more in environments with low public information availability.
Panel B of table 4 summarizes the alpha coefficients from a Fama and French (1993) four-
factor model estimated over CSR release days -1 to 1, partitioned on company size and number of
analysts following. First, similar to the results in panel A, in most tests, we find no significant
alpha coefficients for large or high information availability companies. But we observe positive
and significant alpha for both small and low information availability companies. Second, for the
two extreme groups, we observe a positive and significant alpha for companies with low
information availability/low size, and a positive but insignificant alpha for companies with high
information availability/high size. Third, the table shows insignificant t-statistics for excess stock
15
returns over days -20 to 20, excluding days -1 to 1, thus indicating essentially zero returns over
these days.
Table 4 uses the number of analysts following the stock as the proxy for public information
availability. We also generate qualitatively unchanged results when we use standard deviation of
analysts’ forecasts as an alternative proxy for information availability. For example, untabulated
results show that the mean excess return over days -1 to 1 is significantly greater than zero (p-value
= 0.0618), or the mean excess return over the non-event period (p-value = 0.0358) for companies
with greater than median standard deviation of analyst forecasts than companies with less than
median standard deviation of analyst forecasts. The alpha coefficient is also significantly positive
for companies with large standard deviation of analyst forecasts (coefficient = 0.00205, p-value =
0.060), and insignificant for companies with small standard deviation of analyst forecasts
(coefficient = 0.00053, p-val = 0.553). The results are weaker when we use the absolute value of
analyst forecast error as a third analyst-based proxy for public information availability.
Panel A of figure 1 plots the cumulative mean excess returns from day -10 to 10 for group 1
and compares this group with the matched group 1 sample of non-CSR release companies. This
plot shows a positive response for the CSR group 1 (significantly different from zero over days -1
to 1 as per table 4), whereas the matched non-CSR group tends to show negative returns around the
pseudo release date (although the mean is not significantly different from zero over days -1 to 1).
Over days -2 to 2, the CSR group increases by 2.32 percent whereas the matched group decreases
by 1.63 percent (and both these five-day excess returns are significantly different from zero). Also,
beyond day 2, we observe no sign of a price reversal for the CSR sample. Overall, the evidence in
table 4 confirms our second proposition that shareholders respond to CSR disclosures conditional
on public information availability.
Second, we partition the sample into size and information availability terciles to check for a
monotonic relation between shareholder response and size or information availability. Table 5
16
summarizes the results of estimating a Fama and French (1993) four-factor model for each of the
six partitions. This analysis shows that the alpha coefficient increases monotonically as size
decreases (panel A) or information availability decreases (panel C). These results, however, while
significant across the information availability terciles are less pronounced across the size terciles.
This occurs possibly because each regression already includes a size factor (SMB) to capture return
uncertainty from that factor. Panels B and D of table 5 also show no relation between shareholder
response and the partitioning variables for the non-CSR release matched sample over the same
event days as the CSR sample. Panel B of figure 1 plots the alpha coefficients for the CSR sample
from panels A and C. The plot clearly shows a monotonic relation between shareholders’ response
and size or public information availability.
Third, we add size (Log at) and/or number of analysts (numest) to the Fama and French (1993)
four-factor model and report the results in table 6. This table shows that both additional factors
vary negatively with shareholders’ response (regressions 1 and 2). Regressions 1 and 2 also
document positive and significant alphas, consistent with an overall significant response after
controlling for size and public information availability (similar to panel B of table 3). Both factors
together (regression 3) also significantly explain shareholders’ response (based on the incremental
R2 from the two variables combined), but none is significant individually in that regression. In
addition, regressions 4-6 show that none of the additional factors is significant for the matched
sample of pseudo release dates.
As a final analysis, we compute the zero investment hedge portfolio return by the calculating
the three-month compounded stock return before CSR release day 0 and after CSR release day 0
for portfolios based on size quintiles and three information availability terciles (number of analysts,
standard deviation of analysts’ forecasts, standard deviation of analysts’ forecast error). We then
test for a non-zero return on a portfolio that is long in size quintile 1 (smallest) and short in size
quintile 5 (largest) and long in information availability tercile 3 (least) and short in information
17
uncertainly tercile 3 (most) in the three months before and after CSR release date. In unreported
results, we find that seven of the eight hedge portfolios (four metrics times two time periods)
generate returns that are virtually indistinguishable from zero. This supports the view that stock
prices adjust appropriately around the CSR release dates for differences in public information
availability related to size or analyst forecast attributes, since otherwise the hedge portfolio returns
would be significantly different from zero, and potentially a source of arbitrage trading profits.
5 Conclusions
This study adds to the literature on the relevance to capital market participants of green
information by documenting shareholders’ response to companies’ voluntary disclosures about
GHG emissions made through the CSR newswire service (an organization that claims to be the
global leader in climate change disclosure). Our evidence is important because managers should
understand the risks and consequences of their green disclosure decisions to outside parties, whose
demands have risen substantially in recent years due to heightened concerns about climate change.
A more fundamental contribution of our study relates to tests of voluntary disclosure theory in the
context of climate change, where the theory predicts that an optimal disclosure decision should (a)
produce an overall net benefit for shareholders and (b) that such net benefit should decrease in
public information availability. To date, the empirical literature yields little to support the
predictions of voluntary disclosure theory in the context of climate change, possibly because
managers face a myriad of additional challenges and uncertainties in making optimal climate
change disclosures. For instance, too much climate change disclosure could increase proprietary
disadvantage and litigation and insurance costs; whereas too little could trigger adverse selection
problems, increase information asymmetry, and raise the cost of capital. Changing and uncertain
views about the appropriate balance of companies’ profit and environmental and social goals
further complicate a voluntary green disclosure decision. So it is unclear that we would observe
18
empirical results for voluntary climate change disclosures consistent with the theory (and with
findings in other settings).
Our study generates two new results regarding companies’ voluntary disclosures about GHG
emissions. First, our evidence shows that managers’ voluntary green disclosure decisions produce
positive returns to shareholders. This means that we can reasonably infer that managers’ green
disclosure decisions reflect an appropriate balance of the costs and benefits of disclosure, since the
theory also posits that too much or too little disclosure can be harmful to the company. Second, we
find that shareholders of smaller companies with limited public information availability benefit the
most from voluntary green disclosure, since in this setting investors have fewer other channels and
less access to competing information. For instance, for small companies in a limited public
information environment, mean adjusted share price increases significantly, by 2.32 percent over
days -2 to 2 around the CSR release date. Shareholders of large companies also benefit from
voluntary green disclosure but less significantly, as their larger size attracts more attention from
financial analysts and others whose analysis preempts some of the value of the additional
disclosure.
As part of our research design, we also check whether an investor could exploit differences in
shareholder response from differences in public information availability to earn an abnormal hedge
portfolio return. We construct a portfolio that is long (short) in companies with low (high)
information availability and test for a significant hedge portfolio abnormal return over three months
before and/or three months after the CSR disclosure. We find no evidence of an abnormal hedge
portfolio return over these intervals, which is another way to conclude that investors recognize the
benefits of climate change disclosures efficiently and without bias, although measurement error
could explain this result.
Also, as part of our research design, we compare the adjusted share price response of CSR
releases with the response of a matched sample of non-CSR companies on the same day. We find
19
no similar response for the matched sample as a whole, although the mean adjusted price of the
matched sample of small companies with low public information availability decreases around
CSR release date, which could be an information transfer effect, although we leave a thorough
investigation of such effects in the context of climate change disclosures to future research.
As a closing thought, we find our affirmative “it pays to be green” result interesting and
encouraging, as shareholders’ overall positive response to companies’ voluntary green disclosures
bodes well for the future, since shareholders, regulators, and others will surely place even greater
demands on companies for additional disclosure and transparency regarding climate change
impacts and strategies. So far, management appears to be making optimal voluntary disclosure
decisions in this regard.
20
References
Akins, B., J. Ng, and R. Verdi, 2011. Investor competition over information and the pricing of
information asymmetry. Working paper, MIT Sloan School of Management, February.
Armstrong, C., J. Core, D. Taylor, and R. Verrecchia, 2011. When does information asymmetry
affect the cost of capital? Journal of Accounting Research 49(1) 1-40.
Blacconiere, W., and D. Northcut, 1997. Environmental information and market reactions to
environmental legislation. Journal of Accounting, Auditing & Finance 12(2), 149-178.
Blacconiere, W., and D. Patten, 1994. Environmental disclosures, regulatory costs, and changes in
firm value. Journal of Accounting & Economics 18(3), 357-377.
Brennan, M., and A. Subrahmanyam, 1995. Investment analysis and price formation in securities
markets. Journal of Financial Economics 38(3), 361–81.
California Air Resources Board (CARB), 2010. California cap on greenhouse gas emissions and
market-based compliance mechanisms. Article 5, Sections 95800-96022, Title 17, California
Code of Regulations, also known as AB 32, December.
Chapple, L., P. Clarkson, and D. Gold, 2011. The cost of carbon: Capital market effects of the
proposed emission trading scheme (ETS). Abacus, forthcoming.
Chung, K., T. McInish, R. Wood, and D. Wyhowski, 1995. Production of information, information
asymmetry, and the bid-ask spread: Empirical evidence from analysts’ forecasts. Journal of
Banking and Finance 19(6), 1025-1046.
Clarkson, P., Y. Li, G. Richardson, and F. Vasvari, 2011a. Does it really pay to be green?
Determinants and consequences of proactive environmental strategies. Journal of Accounting
and Public Policy 30(2), 122-144.
Clarkson, P., M. Overell, and L. Chapple, 2011b. Environmental reporting and its relation to
corporate environmental performance, Abacus 47(1), 27-60.
Coburn, J, S. Donahue, S. Jayanti, 2011. Disclosing Climate risks and opportunities in SEC filings,
CERES, Boston: Mass.
Cormier, D., and M. Magnan, 1997. Investors' assessment of implicit environmental liabilities: An
empirical investigation. Journal of Accounting and Public Policy 16(2), 215-241.
Dhaliwal, D., O. Li, A. Tsang, and Y. Yang, 2010. Voluntary nonfinancial disclosure and the cost
of equity capital: The initiation of corporate social responsibility reporting, Working paper,
University of Arizona, October.
Diamond, D., and R. Verrecchia, 1991. Disclosure, liquidity, and the cost of capital. The Journal of
Finance 46(4), 1325-1359.
Dye, R., 1985. Disclosure of Nonproprietary Information. Journal of Accounting Research 23(1),
123-145.
Eccles, R., G. Serafeim, and M. Krzus, 2011. Market interest in nonfinancial information. Journal
of Applied Corporate Finance, 23(4), 113-127.
Environmental Protection Agency (EPA). 2009. Mandatory reporting of greenhouse gases. 40
CFR Part 98, Docket ID No. EPA-HQ-OAR-2008-0508-2278. Washington DC: EPA. Effective
date, December 29.
21
Fama, E., and K. French, 1993. Common risk factors in the returns on stocks and bonds. Journal of
Financial Economics 33(1), 3-56.
Frankel, R., and X. Li, 2004.Characteristics of a firm’s information environment and the
information asymmetry between insiders and outsiders. Journal of Accounting and Economics
37(2), 229-259.
Griffin, P., D. Lont, and Y. Sun, 2011. The relevance to investors of greenhouse gas emission
disclosures. UC Davis Graduate School of Management Research Paper No. 01-11. Available at
SSRN: http://ssrn.com/abstract=1735555.
Grossman, S., and O. Hart, 1980. Disclosure laws and takeover bids. The Journal of Finance, 35(2),
323-334.
Guidry, R., and D. Patten, 2010. Market reactions to the first-time issuance of corporate
sustainability reports. Sustainability Accounting, Management and Policy Journal 1(1), 33-50.
Hong, H., T. Lim, and J. Stein, 2000. Bad news travels slowly: size, analyst coverage, and the
profitability of momentum strategies. The Journal of Finance 55(1), 265-295.
Hughes, K., 2000. The value relevance of nonfinancial measures of air pollution in the electric
utility industry. The Accounting Review 75(2), 209-228.
Hughes, S., A. Anderson, and S. Golden, 2001. Corporate environmental disclosures: Are they
useful in determining environmental performance? Journal of Accounting and Public Policy
20(3), 217-240.
Jones S., G. Frost, J. Loftus, and S.Van der Laan, 2007. An empirical examination of the market
returns and financial performance of entities engaged in sustainability reporting. Australian
Accounting Review 17(41), 78-87.
Lambert, R., C. Leuz, and R.Verrecchia, 2007. Accounting information, disclosure, and the cost of
capital. Journal of Accounting Research 45(2), 385-420.
Leung, S., and S. Yu, 1996. On the choice between sample selection and two-part models. Journal
of Econometrics 72 (1/2), 197-229.
Leuz, C., and R. Verrecchia, 2000. The economic consequences of increased disclosure. Journal of
Accounting Research 8(Suppl.), 91-124.
Li, E., and K. Ramesh, 2009. Market reaction surrounding the filing of periodic SEC reports. The
Accounting Review 84(4), 1171-1208.
Magness, V., 2010. Environmental disclosure in the mining industry: A signaling paradox. In
Freedman, M., and Jaggi, B., eds., Sustainability, Environmental Performance and Disclosures,
Advances in Environmental Accounting and Management 4, 55-82.
Matsumura, E., R. Prakash, and S. Vera-Muñoz, 2011. Carbon emissions and firm value. Working
paper, University of Wisconsin, August.
Mensah, Y., X. Song, and S. Ho, 2003. The long-term payoff from increased corporate disclosures.
Journal of Accounting and Public Policy 22, 107–150.
Murray, A., D. Sinclair, D. Power, and R. Gray, 2006. Do financial markets care about social and
environmental disclosure? Further evidence from the UK. Accounting, Auditing &
Accountability Journal 19(2), 228-255.
Patten, D. M. and J. R. Nance, 1998. Regulatory effects in a good news environment: The
intraindustry reaction to the Alaskan oil Spill. Journal of Accounting and Public Policy 17(4/5),
22
409-429.
Plumlee, M., D. Brown, and S. Marshall, 2008. The impact of voluntary environmental disclosure
quality on firm value, Working paper, University of Utah, May.
Yohn, T., 1998. Information asymmetry around earnings announcements. Review of Quantitative
Finance and Accounting 11(2), 165-182.
Zhang, F., 2006. Information uncertainty and stock return. The Journal of Finance 61(1), 105-136.
23
Figure 1: Shareholder response to CSR release
Panel A: Cumulative mean excess return for days -10 to 10: CSR versus matched sample1
Panel B: Fama and French (1993) alphas for days -1 to 1: CSR sample by size and public
information availability2
1Panel A plots the cumulative mean market-adjusted excess return for the CSR smallest size/least public
information availability portfolio and the equivalent matched sample portfolio, cumulated over days -10 to 10
(table 4 tests for a significant response over days -1 to 1).
2Panel B graphs the Fama and French (1993) four-factor model alpha coefficients from table 5, panel A (CSR
sample by size) and panel C (CSR sample by public information availability). The intercept coefficients
estimate the announcement effect over days -1 to 1 for each partition after controlling for market return in
excess of the risk free rate (RMRF) and the portfolio returns on zero-investment, factor-mimicking portfolios
for company size (SMB), growth (HML), and stock momentum (MOM).
-2.00%
-1.50%
-1.00%
-0.50%
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0 1 2 3 4 5 6 7 8 9 10
M
ar
ke
t-a
dj
us
te
d
ex
ce
ss
r
et
ur
n
Smaller size/
Lower info.
availability
Matched sample
0.0003
0.0013
0.0031
0.0007
0.0026
0.0050
0
0.001
0.002
0.003
0.004
0.005
0.006
Larger size/More info.
Availability
Middle companies Smaller size/Lower
info. Availability
A
lph
a
co
eff
ic
ie
nt
Size
Public Info.
Availability
24
Table 1: Sample characteristics
Panel A: Sample selection
CSR release
18,588
CSR "greenhouse gas emissions" releases
575
Non-company disclosure
48
Non-Compustat (GVKEY) or CRSP (PERM) identifier 312
Non-Compustat at or CRSP daily return
43
All
172
Panel B: Calendar month/year distribution
Month % of obs. Year % of obs.
Jan 6.25 2000 2.33
Feb 3.98 2001 0.58
Mar 9.09 2002 0.58
Apr 12.50 2003 1.74
May 8.52 2004 1.74
Jun 11.36 2005 2.91
Jul 7.39 2006 8.14
Aug 5.11 2007 17.44
Sep 6.82 2008 19.77
Oct 14.20 2009 23.84
Nov 6.25 2010 20.93
Dec 8.52
All 100.00
100.00
Panel C: Mean Log at by industry
gsector
% of
obs.
CSR
sample
% of
obs.
Matched
sample
1 Utilities 4.60 23.870 4.60 23.342
2 Consumer discretionary 13.22 24.350 13.22 23.497
3 Consumer staples 16.67 22.180 16.67 21.796
4 Energy 17.24 23.615 17.24 23.646
5 Financials 18.97 24.260 18.97 23.753
6 Health care 9.20 24.400 9.20 24.414
7 Industrials 2.30 23.580 2.30 23.679
8 Information technology 9.20 24.260 9.20 23.828
9 Materials 3.45 26.310 3.45 26.133
10 Telecommunication 5.17 24.260 5.17 24.057
This table summarizes the sample selection process and the distribution of the sample across months in a
calendar year, year of CSR release, and Compustat gsector.
25
Table 2: Sample descriptive statistics
Panel A: CSR sample
No. of obs. Mean 25% 50% 75%
Log at 172 23.648 22.780 23.930 24.640
Log revt 172 23.464 22.715 23.652 24.490
Gross margin (1-cosg/revt) 172 0.390 0.241 0.360 0.565
Leverage (dltt/at) 172 0.217 0.123 0.211 0.296
ROA (ib/at) 172 0.060 0.030 0.063 0.092
Capital expenditures (capx/at) 172 0.053 0.030 0.046 0.063
Earnings to price (epspx/prcc) 172 0.046 0.042 0.057 0.072
Book to market (ceq/prcc.csho) 172 0.425 0.218 0.365 0.542
IBES number of analysts (numest) 144 15.264 10 15 19
Panel B: Matched sample
No. of obs. Mean 25% 50% 75%
Log at 172 23.365 22.649 23.646 24.185
Log revt 172 23.094 22.468 23.316 24.014
Gross margin (1-cosg/revt) 172 0.371 0.234 0.343 0.469
Leverage (dltt/at) 172 0.257 0.136 0.260 0.339
ROA (ib/at) 172 0.040 0.016 0.043 0.072
Capital expenditures (capx/at) 172 0.055 0.021 0.043 0.080
Earnings to price (epspx/prcc) 172 0.039 0.035 0.056 0.074
Book to market (ceq/prcc.csho) 172 0.526 0.296 0.434 0.671
IBES number of analysts (numest) 144 14.972 10 15 19
Panel C: Compustat/IBES population
no. of obs, mean 25% 50% 75%
Log at 81,587 19.225 17.471 19.313 21.024
Log revt 81,587 18.575 16.948 18.704 20.505
Gross margin (1-cosg/revt) 81,587 0.397 0.223 0.381 0.577
Leverage (dltt/at) 81,587 0.170 - 0.078 0.267
ROA (ib/at) 81,587 (0.074) (0.076) 0.010 0.053
Capital expenditures (capx/at) 81,587 0.050 0.007 0.026 0.062
Earnings to price (epspx/prcc) 81,587 (0.083) (0.094) 0.026 0.065
Book to market (ceq/prcc.csho) 81,587 0.684 0.264 0.519 0.879
IBES number of analysts (numest) 81,587 6.066 2 4 8
This table summarizes the CSR, the matched sample, and the Compustat population of companies by key
financial statement and stock market characteristics. The italicized variable descriptors refer to the terms in
Compustat or IBES.
26
Table 3: Shareholder response to CSR newswire release
Model CSR sample Signif.
Matched
sample Signif. Diff. Signif.
No. of
obs.
Panel A: Market-adjusted excess return1
Day -1 0.00202
-0.00075
0.00277
172
Day 0 0.00103
-0.00010
0.00113
172
Day 1 0.00163
0.00015
0.00148
172
Days -1 to 1 0.00468
-0.00070
0.00538 0.098 * 172
Signif. versus 0 0.042 ** 0.581 ns
Signif. versus days -20 to 20, excl. -1 to 1 0.027 ** 0.528 ns
Model CSR sample Signif.
Matched
sample Signif. Diff.
Signif.
robust
Signif.
cluster
No. of
obs.
Panel B: Fama-French 4-factor model2
Days -1 to 1 alpha 0.00153
-0.00057 ns 0.00210 0.125 ns
0.016** 516
Signif. versus 0, robust std. err. 0.075 * 0.593 ns
516
Signif. versus 0, sector-clustered std. err. 0.034 ** 0.450 ns
516
Days -20 to 20, excl. -1 to 1, alpha -.000155
-0.00018
0.00003 0.940 ns 0.906 ns 6,260
Signif. versus 0, robust std. err. 0.521 ns 0.558 ns
6,260
Signif. versus 0, sector-clustered std. err. 0.533 ns 0.471 ns
6,260
1Panel A summarizes the mean market-adjusted excess return for the CSR and the matched sample over days -1 to 1 relative to day 0, the day of CSR
release. The significance tests show the one-sided t statistic probability of a type 1 error under the null hypothesis that the mean excess return over days -1
to 1 is greater than zero, or greater than the mean excess return over days -20 to 20 excluding days -1 to 1. The significance test for the difference between
the CSR sample and the matched sample tests the null hypothesis that the mean excess return over days -1 to 1 for the CSR sample is different from the
mean excess return for the matched sample over the same period. * indicates significant at less than 10%, ** indicates significant at less than 5%, and “ns”
indicates not significant at less than 10%.
2Panel B summarizes the alpha coefficient from a Fama and French (1993) four-factor model estimated over each sample as a cross-sectional regression
over CSR release days -1 to 1. The intercept coefficient or alpha in the regression tests for an announcement effect after controlling for market return in
excess of the risk free rate (RMRF) and portfolio returns on zero-investment, factor-mimicking portfolios for company size (SMB), growth (HML) and
stock momentum (MOM). The significance tests show the two-sided t statistic probability of a type 1 error under the null hypothesis that the alpha
coefficient equals zero. The significance test for the difference between the CSR sample and the matched sample tests the null hypothesis that the alpha
for the CSR sample is different from the alpha for the matched sample. * indicates significant at less than 10%, ** indicates significant at less than 5%,
and “ns” indicates not significant at less than 10%.
27
Table 4: Shareholder response to CSR newswire release: By company size and public information availability
Sample CSR
CSR
CSR
CSR
CSR
CSR
Size or public information availability Large Sig. 3 Small Sig.
High
Public
Info. Sig.
Low
Public
Info. Sig.
Large/
High
Public
Info. Sig.
Small/
Low
Public
Info. Sig.
Panel A: Market-adjusted returns1
Day -1 0.00092
0.00332
0.00075
0.00530
0.00222
0.00950
Day 0 -0.00022
0.00254
0.00077
0.00185
0.00101
0.00414
Day 1 -0.00062
0.00427
0.00289
0.00262
-0.00037
0.00435
Days -1 to 1 0.00008
0.01013
0.00441
0.00976
0.00285
0.01799
Signif. versus 0 0.487 ns 0.025 ** 0.091 * 0.022 ** 0.216 ns 0.007 ***
Signif. versus days -20 to 20, excl. -1 to 1 0.376 ns 0.018 ** 0.062 * 0.010 *** 0.143 ns 0.002 ***
Signif. versus matched sample, days -1 to 1 0.478 ns 0.050 ** 0.175 ns 0.085 * 0.068 * 0.003 ***
Signif. versus large or/and high public info.
0.031 **
0.172 ns
0.022 **
Panel B: Fama and French (1993) model2
Fama-French 4 factor alpha (α) over days -1 to 1 -0.00005
0.00329
0.00137
0.00330
0.00078
0.00595
Signif. α versus 0, robust standard error 0.950 ns 0.050 ** 0.186 ns 0.037 ** 0.484 ns 0.013 **
Signif. α versus 0, sector clustered std. err. 0.945 ns 0.019 ** 0.114 ns 0.060 * 0.429 ns 0.038 **
Signif. small v. large; high v. low, robust std. err.
0.069 *
0.302 ns
0.048 **
Signif. small v. large; high v. low, sector-clustered
std. err.
0.042 **
0.339 ns
0.093 *
Fama-French 4 factor alpha (α) over days -20 to
20, excl. -1 to 1 -0.00026
-0.00005
-0.00015
0.00003
-0.00028
-0.00010
Signif. α versus 0, robust std. err. 0.279 ns 0.908 ns 0.609 ns 0.927 ns 0.361 ns 0.855 ns
Signif. α versus 0, sector-clustered std. err. 0.083 * 0.946 ns 0.376 ns 0.823 ns 0.089 * 0.735 ns
1Panel A summarizes the mean market-adjusted excess return over days -1 to 1 relative to day 0, the day of CSR release, partitioned on company size and public information
availability (number of IBES analysts following the stock). The significance tests show the one-sided t statistic probability of a type 1 error under the null hypotheses that the mean
excess return for the CSR sample over days -1 to 1 is greater than zero, greater than the mean excess return over days -20 to 20 excluding days -1 to 1, or greater than the mean
return for the matched sample. The significance test also tests the null hypothesis that the mean excess return for small or/and low public information companies is higher than
large or high public information companies. * indicates significant at less than 10%, ** indicates significant at less than 5%, *** indicates significant at less than 1%, and “ns”
indicates not significant at less than 10%.
2Panel B summarizes the alpha coefficients and tests of significance from a Fama and French (1993) four-factor model estimated over each sample as a cross-sectional regression
over CSR release days -1 to 1, partitioned on company size and public information availability (number of IBES analysts following the stock). The intercept coefficient or alpha in
the regression tests for an announcement effect for each partition after controlling for market return in excess of the risk free rate (RMRF) and portfolio returns on zero-investment,
factor-mimicking portfolios for company size (SMB), growth (HML) and stock momentum (MOM). The significance tests show the two-sided t statistic probability of a type 1
error under the null hypothesis that the alpha (α) coefficient equals zero. The significance test also tests the null hypothesis that the mean excess return for small or/and low public
information companies is higher than large or high public information companies. * indicates significant at less than 10%, ** indicates significant at less than 5%, *** indicates
significant at less than 1%, and “ns” indicates not significant at less than 10%.
28
Table 5: Fama and French (1993) four-factor regressions
Panel A: CSR sample by size
Larger size Middle size Smaller size
Coefficient Sig. Coefficient Sig. Coefficient Sig.
Intercept (α) 0.0003 0.3759 0.0013 0.2173 0.0031 0.1799
RMRF -0.0016 0.0850 0.0003 0.3676 -0.0022 0.1453
SMB 0.0010 0.3548 -0.0026 0.1868 0.0062 0.1822
HML -0.0025 0.2125 -0.0028 0.2912 0.0085 0.1024
MOM 0.0002 0.3939 -0.0007 0.3180 0.0037 0.0779
Adjusted R2 14.60%
0.00%
1.24%
No. of obs. 171
204
145
Panel B: Matched sample by size
Larger size Middle size Smaller size
Coefficient Sig. Coefficient Sig. Coefficient Sig.
Intercept (α) -0.0023 0.1454 -0.0002 0.3964 0.0007 0.3548
RMRF 0.0023 0.1373 -0.0006 0.3851 0.0050 0.0362
SMB 0.0037 0.2221 0.0084 0.1314 0.0001 0.3982
HML -0.0052 0.3071 -0.0015 0.3861 -0.0047 0.2196
MOM -0.0031 0.2749 -0.0018 0.3496 -0.0015 0.3304
Adjusted R2 5.15%
0.88%
6.66%
No. of obs. 183
177
162
Panel C: CSR sample by public information availability
More public information Middle public information Less public information
Coefficient Sig. Coefficient Sig. Coefficient Sig.
Intercept (α) 0.0007 0.3264 0.0026 0.1373 0.0050 0.0253**
RMRF -0.0011 0.1237 0.0011 0.3264 -0.0015 0.2340
SMB -0.0020 0.2630 0.0025 0.3026 -0.0028 0.2677
HML -0.0035 0.2125 0.0004 0.3955 -0.0023 0.3221
MOM -0.0010 0.2866 0.0027 0.0896 0.0016 0.2653
Adjusted R2 3.72%
0.62%
1.81%
No. of obs. 192
132
108
Panel D: Matched sample by public information availability
More public information Middle public information Less public information
Coefficient Sig. Coefficient Sig. Coefficient Sig.
Intercept (α) 0.0021 0.1294 -0.0030 0.2054 0.0001 0.3970
RMRF -0.0005 0.3422 0.0074 0.0062 -0.0054 0.0284
SMB -0.0011 0.3460 0.0112 0.1128 0.0165 0.0005
HML -0.0006 0.3905 -0.0092 0.2268 0.0017 0.3758
MOM -0.0013 0.2678 -0.0034 0.3049 -0.0006 0.3904
Adjusted R2 0.00%
16.72%
18.74%
No. of obs. 201
132
102
This table summarizes the coefficients and tests of significance from a Fama and French (1993) four-factor
model estimated over each sample as a cross-sectional regression over CSR release days or a matched sample
pseudo release days -1 to 1, partitioned on three groups of company size (larger, middle, smaller) and three
groups of public information availability based on number of IBES analysts following the stock (less, middle,
more). The intercept coefficients or alphas in the regressions test for an announcement effect for each
partition after controlling for market return in excess of the risk free rate (RMRF) and the portfolio returns on
zero-investment, factor-mimicking portfolios for company size (SMB), growth (HML), and stock momentum
(MOM). ** indicates significant at less than 5%.
29
Table 6: Fama and French multi-factor regressions with size and public information availability
Sample CSR Matched
Regression 1 2 3 4 5 6
Intercept (α) 0.03566 0.00660 0.04399 0.01951 0.00067 0.03024
p-value 0.06211 0.01065 0.08489 0.15175 0.38828 0.14139
Signif. * ** ns ns ns ns
RMRF -0.00061 -0.00066 -0.00065 0.00176 0.00160 0.00157
p-value 0.24416 0.21761 0.22240 0.12187 0.16900 0.17569
Signif. ns ns ns ns ns ns
SMB -0.00019 -0.00071 -0.00065 0.00420 0.00563 0.00555
p-value 0.39539 0.36024 0.36799 0.08942 0.04515 0.05105
Signif. ns ns ns ns ** ns
HML 0.00005 -0.00199 -0.00209 -0.00347 -0.00605 -0.00602
p-value 0.39857 0.21761 0.20571 0.25864 0.14548 0.14963
Signif. ns ns ns ns ns ns
MOM 0.00072 0.00048 0.00039 -0.00237 -0.00287 -0.00296
p-value 0.24174 0.32266 0.34637 0.20575 0.18476 0.17345
Signif. ns ns ns ns ns ns
Log at -0.00144
-0.00173 -0.00086
-0.00135
p-value 0.06577
0.11273 0.14550
0.16461
Signif. *
ns
ns
numest
-0.00029 -0.00006
-0.00005 0.00008
p-value
0.02795 0.36650
0.37877 0.37364
Signif.
** ns
ns ns
Adjusted R2 1.37% 3.03% 4.05% 2.99% 3.49% 3.66%
No. of obs 520 432 432 522 435 435
This table summarizes the coefficients and tests of significance from a Fama and French (1993) five- or six-
factor model estimated for each sample as a cross-sectional regression over CSR release days or matched
sample pseudo release days -1 to 1. The coefficients for Log at and numest test for a relation between stock
return and size and public information availability, respectively, on release days -1 to 1 after controlling for
market return in excess of the risk free rate (RMRF) and portfolio returns on zero-investment, factor-
mimicking portfolios for company size (SMB), growth (HML), and stock momentum (MOM). The intercept
coefficients or alphas in the regressions test for an announcement effect after controlling for RMRF, SMB,
HML, MOM, Log at, and numest. * indicates significant at less than 10%. ** indicates significant at less than
5%.