Disclosure of Environmental Violations and the Stock Market in the Republic of Korea

For almost 20 years, the Ministry of Environment of the Republic of Korea has published on a monthly basis a list of enterprises that fail to comply with national environmental laws and regulations. In this paper, the authors examine the reaction of investors to the publication of these lists and show that enterprises appearing on these lists have experienced a significant decline in their market valuation. Firms in developing countries are often said to have no incentives to invest in pollution control because they typically face weak monitoring and enforcement of environmental regulations. The findings of the authors, however, indicate that the inability of formal institutions to control pollution through fines and penalties may not be as serious an impediment to pollution control as is generally argued. Environmental regulators in developing countries could harness market forces by introducing structured programs to release firm-specific information about environmental performance. This paper - a product of the Infrastructure and Environment Team, Development Research Group - is part of the group's work on industrial pollution control.


I. Introduction
It is often said that firms in developing countries do not have incentives to invest in pollution control effort because of weak implementation of environmental regulations: the cost of complying with the regulation exceed expected benefits resulting from a reduction in expected penalty. However, this argument assumes that the environmental regulator is the only agent that can effectively penalize firms lacking compliance. Recent research indicates that local communities may exercise considerable leverage to pressure firms to improve their environmental performance. 1 The argument also ignores that capital markets may react negatively to the announcement of adverse environmental incidents (such as violation of permits, spills, court actions, complaints, etc.) or positively to the announcement of superior environmental performance. Hence, when accounting solely for regulators' fines and penalties and ignoring the pressure that communities and markets may bear, the expected costs associated with a poor environmental performance may be significantly under-estimated. The inability of formal institutions especially in developing countries to provide incentives for pollution control effort via the traditional channel of fines and penalties may not be as serious an impediment to pollution control as is generally argued: Communities and capital markets, if properly informed, may in specific circumstances provide appropriate incentives.
A limited number of papers have analyzed the reaction of capital markets to environmental news in Canada and the United States. These studies have generally shown that firms suffer from a decline in market values following the announcement of 1 See Afsah et al. (1996), Blackman et al. (1998), and Pargal and Wheeler (1996).
adverse environmental news. 2 The impact of firm-specific environmental news on market value may work its way through various channels: a high level of pollution intensity may signal to investors the inefficiency of the firm's production process; it may invite stricter scrutiny by environmental groups and/or facility neighbours; it may result in the loss of reputation, goodwill, etc. On the other hand, the announcement of a good environmental performance or of the investment in cleaner technologies may have the opposite effect: lesser scrutiny by regulators and communities (including the financial community), and greater access to international markets among other benefits. 3 Studies of this nature in developing countries have been very limited in numbers.
In a recent paper, Dasgupta et al. (2001) have shown that capital markets in Argentina, Chile, Mexico, and the Philippines do react negatively (decrease in firms' value) to citizens' complaints targeted at specific firms, and positively to the announcement of rewards and recognition of superior environmental performance. These results suggest that environmental regulators in developing countries may explicitly harness those market forces by introducing structured programs of information release on firms' environmental performance, and empower communities and stakeholders through environmental education programs.
Numerous countries, both developing and developed have in fact implemented such programs. An increasing number of environmental regulators around the world have 2 In the United States, these studies include analyses of the reaction of capital markets to releases of the Toxics Release Inventory (Hamilton (1995), and Konar and Cohen (1997)). Lanoie and Laplante (1994) analyze the reaction of capital markets to environmental news in Canada. For a survey of these studies, see Lanoie, Laplante and Roy (1998). 3 See Porter and Van Linde (1995), and Konar and Cohen (1997) for a more detailed discussion. While this may not be as well-known, the Republic of Korea (henceforth Korea) has developed its own extensive experience with the public disclosure of environmental performance of regulated facilities. Since the mid 1980s, the Ministry of Environment of Korea has published on a monthly basis a list of facilities in violation with existing Korean environmental laws and regulations. Over the sole period of 1993 to 2002, over 7,000 violations have been reported on those lists, involving more than 3,400 facilities.
As such, the Korean experience with a structured public disclosure program may very well be one of the most extensive experiences of this nature in the world. 5 In this paper, building upon the existing, albeit limited literature on this topic, we examine whether or not capital markets in Korea have reacted to the information contained in these monthly violation lists.
In the next section, we provide a brief description of the Korean public disclosure program. In Section 3, the event-study methodology is briefly described. We present the dataset and our results in Section 4, and conclude in Section 5.

Description of the Korean public disclosure program 6
In the course of the rapid economic expansion of the 70's and 80's, the in their news coverage of the MVR.  Newspapers appear to be particularly interested by violations pertaining to the failure of pollution abatement equipment (Table 3). While this type of violation represents only 18.0% of the total number of events, it represents more than 25% of the events covered by newspapers. On the other hand, while the failure to report and failure of the monitoring system represent 9.2% of the total number of violations, these two types of violation represent only 5.5% of the events covered by the newspapers. In terms of government actions, orders and warnings appears to receive less interest from the newspapers than their weight as a percentage of the total number of violation events (Table 4). However, while prosecutions represent only 9.9% of the total violation events, they represent almost 16% of the violation events reported in the newspapers. Similarly, shutdowns (temporary or complete) and bans receive more attention in newspapers (7.5% of all events in the newspapers) than their overall importance in the monthly violation lists (11.8% of all violation events).

Event-study methodology
The event-study methodology is used here to see the extent to which investors react to environmental news (also called events). 11 The key assumption of the methodology is that capital markets are sufficiently efficient to evaluate the impact of new information (events) on expected future profits of the firms.
The methodology involves the following steps: (1) identification of the events of interest and definition of the event window; 12 (2) selection of the sample set of firms to 11 For more details, see MacKinlay (1997). 12 The event window consists of the day where the event occurred (day 0) and some days before and after the event.
include in the analysis; 13 (3) prediction of a "normal" return during the event window in the absence of the event; (4) estimation of the abnormal return within the event window, where the abnormal return is defined as the difference between the actual and predicted returns; and (5) testing whether the abnormal return is statistically different from zero.
The market model is of interest here to estimate abnormal returns.
The market model assumes a linear relationship between the return of any security to the return of the market portfolio: (1) , ,..., stands for security, R and R it mt are the returns on security i and the market portfolio respectively during period t , and e it is the error term for security i.
Equation (1) is generally estimated over a period which runs between 120 and 210 days prior to the event up to some days prior to the event. The event window is defined as the period from some days prior to the event to some days after the event. The size of the event window is really an empirical matter. With the estimates of α β i i and from equation (1), one can predict a "normal" return during the days covered by the event window. The prediction error (the difference between the actual return and the predicted 13 Firms may be excluded if simultaneous events are occurring within the event window. normal return), commonly referred to as the abnormal return (AR) for a single security i at a given time t, is then calculated as: Under the null hypothesis, the abnormal returns will be jointly normally determined with a zero conditional mean and conditional varianceσ 2 ( ) AR it : where L is the estimation period length (i.e. number of days used for estimation) and Hence, for any given subset of N events (or securities), the subset average abnormal returns ( AAR t ) at each instant t within the event window is computed as To test for the significance of AAR t a Z (or t ) test can be derived.
In order to test for the persistence of the impact of the event during a period , the abnormal return for a given security i can also be added to obtain the cumulated abnormal returns ( event window, and T a and T b are the lower and upper limits of the event window, respectively. 14 Asymptotically (as L increases) the variance of the cumulative abnormal return for security i is An aggregation of interest can also be performed across both time and events. In that scenario, the average cumulative abnormal return for a subset of N events between two dates T 1 and T 2 is defined as: Under the null hypotheses that the abnormal returns are zero, As pointed out by MacKinlay (1997), this distributional result is asymptotic with respect to the number of securities N and the length of estimation window L . Moreover, the validity of cross-sectional (or pooled) aggregation of abnormal returns rests on the assumption that the event windows do not overlap. If they do then the distributional results presented above are no longer valid since covariances across securities are no longer zero, particularly in the case of complete clustering. There are two solutions to the problem of clustering. The first one is to aggregate abnormal returns into portfolios. The second one is basically to leave abnormal returns unaggregated; that is, one has to deal with abnormal returns security by security (for details, see MacKinlay, 1997, 27).
In the next section, we present results obtained from using the single-index model (constant mean return model). 15

Dataset and Results
The initial dataset comprises 96 environmental news (details appear in Appendix 1). These events were examined to identify whether or not other events (positive or negative) were observed during the identified window. These confounding events could impact the results from the event-study methodology. As a result, 9 events were eliminated from the dataset. 16  We apply the event-study methodology to each of these 87 events. The study uses an estimation period of 210 days before the event window and an event window of 7 days (3 days prior the event, the day of the announcement, and 3 days after the event). 15 The single-index model is a particular case of the market model described above. Where market returns were available, we also obtained results using the market model. Results were similar to those presented here. In fact, Henderson (1990)  The latter size of the event window has been determined empirically through a search over a period of 10 days before and 10 days after the event.
Appendix 2 presents the list of 87 events and differentiates those events for which a statistically significant stock market reaction has been estimated from those for which no such reaction has been estimated. Detailed statistical results for those events with market reaction are presented in Appendix 3.
As shown in Appendix 2, 52 of the 87 events (60%) included in our dataset show a statistically significant market reaction. However, 5 of these exhibit a positive market reaction. 17 We were not able to identify whether or not these 5 events were plagued by the presence of simultaneous, positive, events. We cannot therefore offer a credible explanation for the unexpected market reaction for these 5 events. Of those events with negative market reactions, the average percentage reduction in market value has been calculated to be 9.7%. As shown in Table 5 below, this average reduction in market value is much higher than results obtained in Canada and the United States, but of a similar order of magnitude as results obtained in Argentina, Chile, Mexico, and the Philippines (Dasgupta et al. (2001)). This would tend to re-enforce the hypothesis that capital markets in developing countries may attach a greater premium to information which otherwise may generally not be as readily available as in more developed markets. If we include all events with market reaction, we note in Table 6 that 69% (36/52) of the events with market reaction are violation of emission standard, which is a slightly higher percentage than the percentage of those events in our dataset (65%). On the other hand, the failure to operate PCE properly represents only 15% of those events exhibiting market reaction, which is a slightly lower percentage than the share of those events in the dataset. This would appear to indicate, perhaps as may have been expected, that the market react slightly more frequently to the violation of emission standard than to the failure to operate PCE properly. However, the average percentage reduction of market value for those events pertaining to the violation of emission standard is 8.96%, while the average reduction is calculated to be 15.3% for those events pertaining to the failure of operating PCE properly. It should be noted however that we accept the null hypothesis at the 10% level of significance that there is no significant difference between negative market reaction from events of violation of emission standard and that of events of failure to operate PCE properly. 18  of those events without market reaction, the average number of news coverage is 1.77.
Though the difference may not be very large, it does support the assumption that the larger the number of newspapers coverage, the greater the likelihood of market reaction.
To this effect, it may be further noted that the 2 events which have been covered by 5 newspapers, and the (one) event that has been covered by 6 newspapers have all experienced market reaction. We have also grouped these events into 3 sub-groups: (1) a sub-group of events which have been covered by 1 or 2 newspapers; (2) a sub-group of events which have been covered by 3, 4, 5 and 6 newspapers; (3) a sub-group of events which have received coverage by 5 or 6 newspapers. Our results reveal an average percentage reduction of market value for each of these sub-groups of 4.46%, 16.1%, and 38.23% respectively. 20 It would thus appear that the larger or wider the coverage of the events by newspapers, which may itself be reflective of the nature and/or importance of the event, the larger the percentage reduction in market value.

Conclusion
Since the late 1980's, the government of Korea has actively implemented a public disclosure program to inform citizens of the fact that some large companies in Korea are not complying with Korean environmental laws and regulations. Perhaps contrary to expectations that capital markets in developing countries may not reach to such news, it was shown in this paper that investors on the Korean Stock Exchange do in fact strongly react to the disclosure of such news. The average reduction in market value was estimated to be much higher than the estimated changes in market value for similar events in Canada and the United States, and of a similar magnitude as observed changes in other developing countries (Argentina, Chile, Mexico, and Philippines). It was further shown that the larger the extent of coverage by newspapers, the larger the reduction in market value, reaching above 35% for those events covered by 5 or more newspapers.
While a number of papers have examined the reaction of stock markets to environmental news, it is not immediately clear whether or not such reactions then induced changes in the actual environmental performance of the involved facilities. This is subject to on-going research. 1 AR stands for abnormal return and CAR is the cumulative abnormal return. CAR is computed for Day -3 up to the specified day of interest. Within brackets is the value of z statistics. '*' denotes significance at the 10% level (two tailed test). If an event had no statistically significant AR for any of the days over the period of the even window (-3 to +3), then that event was not retained as statistically significant even if some of its CAR may have been statistically significant.

Appendix 1 Description of the dataset
On the other hand, if an event has at least one statistically significant AR during the event window, then all statistically significant results are reported for that event, even on those days where only the CAR is statistically significant.

Annex 5 Testing for differences in changes in market value: Violation of emission standard vs. Failure to operate PCE properly
The null hypothesis is 0 : , we accept the null hypothesis at the 10% level of significance that there is no significant difference between negative market reaction from events of violation of emission standards and that of events of failure to operate PCE.

Annex 6 Testing for differences in changes in market value: Changes over time
The null hypothesis is where the s ' µ are the respective population means. We have the comparison of more than two means. Here the use of the t test statistic is not appropriate (see Ott and Longnecker (2001)). We recourse to the analysis of variance to solve the problem of size distortion that may entail the use of t. Precisely, we build an F-test statistic: at the 10% level of significance, we reject the null hypothesis and conclude that the average % reduction in stock market is changing over time.