Master Thesis Finance Author: B.R.W. Kimman Mail: baskimman@gmail.com Phone: +31610643314 Student number: S1879235
Place and date: Groningen, 13-06-2016 Supervisor: Dr. L. Dam
Does the US market value corporate toxic waste?
Abstract
1
1. Introduction
There is an increased realization of rising costs either regulated by US policies to control environmental protection, or fuelled by society through reputational damage and decrease in goodwill. Investment managers that are willing to focus their portfolio composition on socially responsible investments have gained in numbers over the past few years. US SIF, the Forum for Sustainable and Responsible Investment, address in their annual US SRI trends report, that American located assets under management driven by SRI strategies expanded from $3,71 trillion in 2012 to $6.57 trillion in 2014. An increase of 76 percent meaning that every one dollar out of six under professional management is invested with a SRI strategy. The definition of when a firm or investment is “socially responsible” varies. According to Hamilton et al. (1993) SRI investors are usually interested in positive and negative criteria on environmental issues, social relations and pollution control management. SRI investors thus try to encourage proper social and environmental corporate behavior (positive screening), while avoiding companies that produce unhealthy or unethical products. Neither should firms exploit working environments in either developed or undeveloped countries (negative screening). In general, SRI investors expect firms to take all stakeholders in consideration in addition to maximizing value. Integrating these screens into a portfolio choice is often called the third generation of screens (Renneboog et al. 2008). The fourth and most recent generation focuses on combining the sustainable investment choice with shareholder activism. Investors now actively try to influence corporate social behaviour by executing voting rights or deviate their actions through direct dialogue. Renneboog et al. (2008) state that Corporate Social Responsibility (CSR), is the combination of a solid corporate governance, good stakeholder relations and an efficient environmental performance. Corporate Governance relates to the conflict between an agent (manager) and a principal (investor). Stakeholder relations can address both primary stakeholders such as employees, consumers, governments and suppliers. It can also include a so called ‘social issue participation’ which refers to getting involved in sin industries like gambling, drugs and weapons (Hillman & Keim, 2001). Summing up, the three main pillars of CSR leave room for interpretation. The increased interest in SRI might imply that investors care about corporate ethical behaviour as a whole. However, it is not clear if this implies that investors actually react towards released environmental statistics such as toxic waste emission. Neither do we know if it is embedded in perceived firm value.
2
market system. Signalling the existence of firms with high/low toxic disposal based on TRI data might alter the perception or preference of consumers and investors.
To research whether the release of this dataset had impact on the US stock market, Hamilton (2003) performed an event study finding statistically significant negative abnormal returns. Khanna et al. (1998), repeated this study finding that the abnormal returns are only observable if pollution levels differ from the expectations created by analysts based on the records of the year before. Since the research of both Hamilton and Khanna et al., environmental regulations have been tightened, firms have been dedicating more attention towards emission in their annual reports and the availability of information as a whole has grown exponentially. Therefore it is doubtful whether the annual release of TRI data is still perceived as news. Simultaneously, literature finds that CSR standards are positively correlated with long-term financial performance indicators. Pollution data can offer investors information about related costs to be suspected. These costs can influence the profitability of a firm and therefore affect firm value.
Based on the observations above, the following research questions will be addressed in this paper:
What is the short-term market impact of pollution data provided by the TRI? Is there a long-term relationship between TRI pollution data and firm value?
The research questions are addressed separately. To analyse the short term market reaction, I perform an event study that measures whether there is an abnormal return on the annual release date of the TRI dataset in 2011, 2012 and 2013. The sample consists of 295 US listed firms traded on NASDAQ and NYSE. For the long-term relationship I use the same sample. I perform a panel regression using Tobin’s Q as dependent variable and TRI pollution data as the most important dependent variable.
3
2. Literature review
Environmental and financial performance
Academic research has been studying various relationships between Corporate Social Responsibility and the performance of firms. General opinions towards environmental regulations are that they impose significant costs, reduce productivity and hinder competitiveness of firms. (Jaffe et al., 1995). It can also lead to fines, liabilities and administrative or legal action against polluters. However, some argue that these legislations create value instead of destroying it. The cost savings generated through not applying environmental investments may be exaggerated or are made up in some cases. Furthermore, firms can reduce pollution levels by making changes in the production process without the requirement of incurring costs of investments and there are possible benefits due to satisfied employees, which might generate more productiveness (Dowell et al., 2000).
To explain the importance of social responsibility perceived by investors and its relationship with financial performance, McGuire et. al (1988) illustrate three arguments. First of all, although environmental investments might be large, they can simultaneously decrease costs and increase revenues. An increase of for instance carbon dioxide emissions and offsite transfers can affect the judgement of projected costs for analysts. Examples of these costs are clean-up costs dealing with hazardous waste sites, transaction costs arising from litigation over pollution liability, pollution control and abatement expenditures and penalties and fines from enforcement actions. It might also lead to higher operating costs since a firm could be watched more carefully by regulators and environmentalists. The second argument is that costs incurred by environmental management can be limited and might turn out to be beneficial in other ways. They can for instance improve productiveness and employee morale. Finally McGuire et al. (1988) state that firms that do not influence the trade-off between environmental and financial performance, have to invest eventually in order to improve their environmental performance and are likely to suffer financial drawback. Another argument that social responsibility influences financial performance is mentioned by Hamilton (1995). He argues that high pollution levels can lead to a decrease in reputation, value of the brand name and goodwill. Investing in sustainable production therefore might prevent these costs while enhancing a firm’s reputation. These examples indicate several reasons to explain why there is reason to believe that investors use TRI data in firm valuation.
The impact of CSR is often measured through the effect it has on valuation of assets or profitability. The results and views differentiate when it comes to the effect CSR has on firm performance. Some academic research argues that high social standards and environmental behaviour result in excessive operating costs, which will eventually have a negative influence on the profitability of a firm. Others state that strong environmental management adds value in reputation and decreases or avoids for instance clean-up costs, which will strengthen the performance of a firm. Like the research in this paper, empirical studies that measure the relationship between CSR and firm performance or value can be roughly separated in short-term responses and long-short-term relationships.
Short-term responses through environmental event studies
4
R&D investments by technological firms. It is argued that investors only focus on firms with short-term profits and firms that are undervalued with long-term investments, such as R&D. These investments are perceived to create strategic opportunities for a firm, which lead towards a competitive advantage (Porter, 2008). They find a positive reaction towards increased expenditures while finding a negative reaction when R&D projects are announced to be discontinued. These studies show that investors react significantly towards announcements of investing and adding new information to the market.
It might be obvious that R&D expenditures are perceived as relevant to shareholders since investments can weaken or strengthen the competitive advantage of a firm. However there are also several other aspects that can trigger abnormal returns. One of these aspects is considered to be environmental performance. According to Epstein (1991) shareholders reflect that one of the top three priorities of a firm should be to clean up the environment, using capital expenditures.
Several studies find significant abnormal returns around environmental events. Klassen et al. (1996) and Filbeck et al. (2004), find a significant positive relation between firms announced to win environmental awards and financial performance. Klassen et al., create a theoretical model, which categorizes cost savings into market share gains, environmental product or process certification and higher product contribution margins. These savings can be achieved through establishing industry wide standards, prevention of spills and environmental liabilities and reduced energy consumption. To test the model, Klassen et al. perform an event study by looking for significant abnormal returns around environmental award announcements. To test negative response, events around firms being involved in environmental crises and penalties are taken into consideration. Based on this study, both environmental awards and crises seem to generate significant impact and reasons to develop a sound environmental strategy. The magnitude of crises events were largest, resulting in an average market valuation loss of 390 million US dollars.
Another way to measure short term market reaction towards environmental information, as is used in this research, is by measuring the effect of the annual release of the Toxic Release Inventory (TRI). Both Hamilton (1993) and Khanna et al. (1998) find significant negative abnormal returns on the day that toxic releases of firms are publicly announced.
5
Khanna et al. (1998) examine a similar TRI data study on firms from the chemical industry. They state that since there is no possibility for investors to continuously receive information about environmental performance, it is likely that TRI data will lead to significant abnormal returns. Because there is a lag of approximately one year between the release of TRI data and actual generated waste, investors or analysts formulate their own beliefs or expectations. Therefore they claim that it is not the information itself that generates abnormal returns, but the deviation of predictions based on for instance the previous year’s TRI report. The study shows that if these firms are known for their large amount of pollution, the release of the data will not directly lead to statistically significant negative returns. However, analysing the data more frequently over the years allows the investor to compare pollution against competitors. Repeating the study therefore does lead to significant abnormal negative returns. Khanna et al. (1998) also find stronger reactions with a lag of one day, than on the day of the TRI data release. According to them, it might take investors some time to evaluate and analyse new information before they act.
Although the speed of the absorption of information is researched extensively, there is no clear consensus to be found. Many economists support the existence of under-reaction when information is added to the market regarding stock returns. With information mostly signalling and planned news items are used. Michaely et al. (1995) find under-reaction in dividend initiations and omissions that are caused by signalling. Bernard and Thomas (1990) conclude that a drift exists in earnings after an announcement that covers scheduled news. Womack (1996) finds an asymmetric delay in response that is based on the alteration of recommendations by analysts. Furthermore, investors have a lag in reaction regarding a change in capital structure. According to Loughran et. al (1995) this can be found in seasoned equity offerings. Ikenberry and Vermaelen (1995) find the lag after researching tender offers. There is also literature that shows investors over-react to information and therefore trade more than what should be desirable. The variance of stock returns for instance is found to be greater if the stock market is open compared to a closed market. The results are robust even if the amount of information added is the same in both situations (French and Roll, 1986). While looking at the R-squared regarding CAPM and APT factors, part of the variance in daily and monthly stock returns cannot be clarified (Roll, 1988). Finally Mitchell and Mulherin (1994) find that the relationship between news and stock markets exists, but it is not very strong. Since literature suggests that the market can show a lagged response or under- and over-reactions, I present all days within the event window to capture these issues.
6
Long-term relationship between CSR and financial performance
Literature attempts to evaluate the long-term relationship between Corporate Social Responsibility and financial performance several ways. Theories so far have been inconclusive whether CSR affects performance in a positive or negative way.
Some literature supports the theory that costs of social-environmental responsibility are higher than the benefits, and thus influence firm performance negatively. Jones et al. (2007) show that SRI strategies statistically underperform traditional funds by 1,52% annually in Australia. Chen and Metcalf (1980), find that the correlation between their pollution performance index and firm performance is negative. The same conclusion is stated by Jaggi and Freedman (1992), who therefore argue that social responsible behaviour and environmental management will not be rewarded.
Others like Climent and Soriano (2011) find no significant difference between traditional and environmental mutual funds. By using the CAPM model in the US, they study financial performance while using returns of equally weighted portfolios over time. Between 2001 and 2009 returns of environmental funds were not significantly different from the rest of SRI or conventional mutual funds. Dam and Scholtens (2015), provide a theoretical framework for the relationship between social and financial performance on firm level. Although SRI does not seem to generate positive significant abnormal stock returns, there are other economic performance measures, which indicate a positive relationship. By analysing more than 60 studies, positive connections can be found between social responsibility and market-to-book ratio and ROA. They conclude that although a firm may realize lower stock market returns after announcing social responsibility, this does not necessarily lead to a decrease in firm value.
7
3. Data and Methodology
3.1
Data
To gather all data, I use four different databases. EPA’s Toxic Release Inventory, CRPS data, Orbis and DataStream. The EPA requires all American facilities that operate within SIC codes 20-39, and who employ more than 10 persons to submit releases of various chemicals on the fist of July since 1986. Each facility is identified by name, SIC-code, parent company name and Dun & Bradstreet number. Once the EPA gathers the data, it is submitted online each year with a delay of approximately one year. Data gathered on the first of July in 2012 will therefore be submitted around July 2013. To compare TRI data used in this study with their stock market reaction, I only include publicly listed firms in the sample. Since all data is reported per facility, I first consolidate all facilities based on their parent company name. Besides facilities, I allocate and sum up all chemicals based on their type of emission. These types include air, water, land and underground (on-site release). Until 1991, firms only needed to report the number of total chemicals that were going to be stored, treated or disposed off-site. Currently firms are obliged to report recycling and energy recovery as well. Therefore, I calculate total waste as the sum of on-site release, off-site release, energy recovery, total recycling and total treatment (TRI basic data files guide).
Consolidating all 82108 facilities leaves 5334 firms based on their parent company name. I identify ticker symbols through the Orbis database in order to filter unlisted firms and gather CRPS data for the remaining listed firms. After removing firms that have insufficient TRI or daily stock data, 301 firms are left to form the sample for this study.
The release date of the TRI database differs annually. Therefore to collect the exact publication days for 2011, 2012 and 2013, the EPA was contacted. The datasets included are the three most recently released complete annual databases from the EPA. The distribution of chemicals based on the type of emission for all 295 firms within the TRI dataset, as compared to the sample firms are summed up for 2012 in Table AI and for 2013 in Table AII of the appendix. The database for 2011 is constructed differently by the EPA. Therefore it is not possible to present the data the same way. The tables illustrate that although only 6% of the firms are included in the sample, they account for 19% of total waste. Moreover, the tables also indicate that total production waste is increasing, while the number of companies included decrease. The TRI database does however not scale the emissions based on assets or other forms of size measurements.
Daily stock data is gathered from the CRPS database for all firms that remain in the sample. Hereafter, I calculate the daily stock returns for each individual firm. Since the studied firms are traded on three different stock markets, daily returns of the NYSE, NASDAQ and TSX are gathered to calculate the market return for each individual company.
8
3.2 Methodology
In the first section, similar to Hamilton (1993) and Khanna et. al (1998), I use event study methodology to measure the stock price reaction of firms on de day that the TRI dataset is published. This methodology is commonly used for academic research (MacKinley, 1997). In the second section, I perform a panel regression to detect whether there is a relationship between TRI pollution data and firm value as measured by Tobin’s Q.
3.2.1 Short-term reaction: Event study methodology
Market modelTo evaluate the market reaction for stocks on the day of the release of the TRI data, I use event study methodology based on similar study of MacKinley (1997). For each individual firm within the sample, 𝑖 = 1, 2, … . , 𝑁, the market model is composed:
𝑟𝑖𝑑 = 𝛼𝑖 + 𝛽𝑖 𝑟𝑚𝑑+ 𝜀𝑖𝑑
(1)
Where 𝑟𝑖𝑑 is the continuously compounded return of stock 𝑖 on day 𝑑, 𝑟𝑚𝑑 is the market
return of the specific stock market on which stock 𝑖 is traded on day 𝑑. Alpha 𝛼𝑖 is the
intercept (estimation of the constant daily return for stock 𝑖) and Beta 𝛽𝑖 is the slope (firm specific parameter of the market model). 𝜀𝑖𝑑 is the error term (unexplainable part of the return by market movements). The model assumes that if there is no unanticipated news or information to be expected, the relationship between the return of the firm and the market remains the same and the error term 𝜀𝑖𝑑 is zero. The parameters are estimated through
Ordinary Least Squares regression (OLS).
In order to predict the normal return of the stock, I first calculate the expected return on day 0 of the event. The abnormal return for any firm 𝑖 within the sample on day 𝑑 is then calculated as the actual return on that specific day minus the expected return. These steps will be further explained below. An abnormal return is achieved when unanticipated information is added, which alters the return of the firm 𝑟𝑖𝑑 while the market return 𝑟𝑚𝑑 remains the same. 𝜀𝑖𝑑 is
therefore the prediction error generated by the TRI data.
This model is used on the three most recent years 2011, 2012 and 2013. Since the release date of the data is different each year, the event day 𝑑 is different for all years. All individual release dates can however be equally regarded as event day 0.
To estimate the market model, I use stock price data from 100 trading days before the event window. An estimation period of 40 trading days is necessary to make the model statistically valid (Hendricks & Singhal, 2003). The windows itself consists of 10 days before and 10 days after the release of the TRI data. Therefore the market model is based on 𝑑: -110 till 𝑑:-10. Over the 100 trading days intercept 𝛼𝑖 and slope 𝛽𝑖 are identified for each specific firm.
To calculate the expected return of the individual firms within the sample for 𝑑:-10 until 𝑑:10, I use the following equation:
9
To estimate the abnormal return on 𝑑:-10 till 𝑑:10, I subtract the expected return in equation (2) from the actual stock return on day 𝑑 for each firm 𝑖.
𝐴𝑅𝑖𝑑 = 𝑟𝑖𝑑− 𝐸(𝑟𝑖𝑑) (3)
Where 𝐴𝑅𝑖𝑑 is the abnormal return for firm 𝑖 on day 𝑑 and 𝐸(𝑟𝑖𝑑) is the expected return for
firm 𝑖 on day 𝑑.
The abnormal returns are averaged for all firms on any day 𝑑 within the sample. This average abnormal return (𝐴𝑅̅̅̅̅𝑑) represents the reaction of shareholders on day 𝑑. The 𝐴𝑅̅̅̅̅𝑑 for day 𝑑 with 𝑁 firms is calculated as follows:
𝐴𝑅 ̅̅̅̅𝑑 = 1
𝑁 ∑ 𝐴𝑅𝑖𝑑
𝑁
𝑑=1 (4)
The variance of the market model is calculated in order to measure the significance of 𝐴𝑅̅̅̅̅𝑑 on day 𝑑. 𝑣𝑎𝑟(𝐴𝑅̅̅̅̅𝑑) = 1 𝑁2∑ 𝜎𝜀 2 𝑁 𝑖=𝑖 (5)
Finally, the z-value for all average abnormal returns (𝐴𝑅̅̅̅̅𝑑) within the sample on day 𝑑 is calculated as the average abnormal return 𝐴𝑅̅̅̅̅𝑑 divided by its standard deviation 𝑆(𝐴𝑅̅̅̅̅𝑑) :
𝑇𝑆𝑑 = 1 𝑁 ∑ 𝐴𝑅𝑖𝑑 𝑁 𝑑=1 √1 𝑁2∑ 𝜎𝜀 2 𝑁 𝑖=𝑖 (6)
Where 𝑇𝑆𝑑 is the z-test-statistic value for day 𝑑. Since I am interested in the shareholder reaction regarding the TRI announcement, I analyse the average abnormal return and the z-statistic values on day 0. However, since Khanna et al. (1998) argue that it takes time for analysts to analyse the data, I will regard event day 1 as well. There are no further theoretical reasons to assume information being leaked prior to the event day 0. I remove non-trading days and other days in which the stock market did not operate from the event window. There is no reason to expect that the reaction will differ between years. So, since there are no time-specific effects, the results can be pooled to increase reliability. For robustness all individual event studies will be available in the appendix.
Mean model
10
3.2.2 Long term relationship: Panel regression using Tobin’s Q
In order to measure whether the toxic emissions are embedded in perceived firm value, I perform a panel regression. The outcome of the regression will reflect the impact of the TRI emission data on the Tobin’s Q ratio corresponding to the firms within the sample.
Tobin’s Q will therefore be the dependent variable. To calculate Tobin’s Q, I use the methodology used by Chung & Pruitt (1994). This model assumes there is equilibrium when the market value of a firm is equal to the replacement value of all its assets, such as a plant, equipment and inventory. If this is found to be true, Tobin’s Q ratio will be equal to 1. A Q-ratio below 1 can imply that a firm does not have profitable investment opportunities. Alternatively, a Q-ratio above 1 indicates expected profitable investments or perceived valuable intangible assets.
Tobin’s Q is calculated for each individual firm as follows:
𝑞
𝑖=
𝑀𝑉𝐸+𝑃𝑆+𝐷𝐸𝐵𝑇𝑇𝐴 (7)
Where 𝑀𝑉𝐸 is the share price of firm 𝑖 accumulated with the number of outstanding common stocks. 𝑃𝑆 is defined as the liquidating value of a firm’s outstanding preferred stock. 𝐷𝐸𝐵𝑇 indicates the short-term liabilities excluding short-term assets plus the book value of long-term debt. Finally 𝑇𝐴 represents the book value of a firm’s total assets. The key independent variable is the amount of total waste distributed by each firm within the sample represented by TRI. To control for size, I divide total emission by the total assets of a firm. For robustness I also divide TRI data by net sales, both representing size measures. By doing so we can compare the pollution per asset or sale across firms instead of total waste amounts.
There are other firm specific characteristics that are expected to have a relationship with the value of a firm. These control variables include net sales, total assets and debt-to-equity ratio. Large firms, represented by both net sales and total assets, can influence firm value due to scale of economy, and their visibility to investors. Both are measured by collecting data at the end of the fiscal year for 2011, 2012 and 2013. To control for skewness, I use the natural logarithm. If the debt-to-equity ratio of a firm is large, it might cause investors to assess that the firm is not favorable. It can signal a higher risk profile and therefore influence firm value. I calculate the debt-to-equity ratio by dividing the book value of debt by the sum of the book value of debt and the book value of equity. R&D expenditures might also influence Tobin’s Q. Firms investing in R&D are likely to have profitable projects or investment opportunities which can positively influence perceived firm value. However, there is not sufficient data available for the firms within my dataset to include this control variable in my regression. Since I have multiple years and firms within the sample, I perform a panel regression. It is not possible to perform a pooled regression since I would deny the heterogeneity or individuality that may exist among the firms. If I do not take this firm effect into account it will result in a specification error. After performing a Hausman test, I conclude that the model has to be performed under fixed-effect panel regression. To account for the annual fluctuations of Tobin’s Q, I include year dummy’s. Therefore, I will perform the following regression:
11
Where the dependent variable is the firm specific Tobin’s q for year 𝑡 𝑞𝑖𝑡. The independent
variable is the total waste of a firm divided by total assets 𝑇𝑅𝐼𝑎𝑖𝑡 generated by firm 𝑖 for year 𝑡. The control variables are the natural logarithm of total assets 𝑆𝐴𝐿𝐸𝑆𝑖𝑡 for year 𝑡 and the debt-to-equity ratio 𝐷𝐸𝑖𝑡 for year 𝑡
.
𝛼𝑖 represents the individual firm effect and 𝛿𝑡 represents the year dummy’s. The error term 𝜀𝑖𝑡 is the measurement error in the dependent variable and adjusts for unobserved explanatory variables which are uncorrelated with the independent variables, firm- and year effects. To add robustness, I add a panel regression with sales instead of total assets to control for size effects:𝑞𝑖𝑡 =𝛼𝑖+ 𝛽𝑖 𝑇𝑅𝐼𝑠𝑖𝑡+ 𝛽𝑖𝑆𝐴𝐿𝐸𝑆𝑖𝑡+ 𝛽𝑖𝐷𝐸𝑖𝑡+ 𝛿𝑡+ 𝜀𝑖𝑡
(9)
Where the dependent variable is the firm specific Tobin’s q for year 𝑡 𝑞𝑖𝑡. The independent
variable is the total waste of the firm divided by sales 𝑇𝑅𝐼𝑠𝑖𝑡 generated by firm 𝑖 for year 𝑡. The control variables are the natural logarithm of net sales 𝑆𝐴𝐿𝐸𝑆𝑖𝑡 for year 𝑡 and the debt-to-equity ratio 𝐷𝐸𝑖𝑡 for year 𝑡.
3.2.3. Descriptive statistics
Table I represents the descriptive statistics of both the event study and panel regression.
Both abnormal returns on the event day and the day after the release of the TRI data, show average positive returns, ranging from -0,60% to 15,11%. This indicates that the short-term response will probably not be large. On average firms within the sample emit 13.5 million pounds of toxic waste each year individually. The market value of firms within the sample is on average higher than the replacement value of its assets. This might indicate that on average the firms within the sample are expected to have valuable intangible assets. However, a ratio
Table I Descriptive statistics
Variable Mean Standard
Deviation Minimum Maximum
Abnormal return % (day 0) 0.02 0.47 -0.60 13.50
Abnormal return % (day 1) 0.01 0.52 -0.63 15.11
TRI (Mlb) 13.51 49.36 0.00 569.21 Q-Ratio 1.74 0.73 0.49 6.61 D-E ratio 1.49 1.62 -14.09 14.81 Total Assets ($M) 12.62 42.71 0.05 685.33 Sales ($M) 8.78 18.05 0.02 171.60 Firms (N) 295
12
that is very high such as the maximum descriptive of 6.61, can also be interpreted as a firm being overvalued by the market. A mean debt-to-equity ratio above 1 indicates that the majority of assets within the firms included in the sample is financed through debt.
Minimum and maximum values of both total assets as sales indicate that there are small and large firms within the sample. Therefore it is indeed necessary to take the natural logarithm as well as scaling total waste to size. Table II represents the correlation coefficients of the control variables in the panel regression. Since total assets and net sales are separately regressed, there is no reason to suspect multicollinearity among independent variables.
4. Results
In this part I will first present the event study results. After analysing the event output, I will evaluate the results of the panel regression on Tobin’s Q.
4.1 Event study results
The event study results, combining 2011, 2012 and 2013 indicate that there is no significant abnormal return on any given day in the event window. Table III, Figure 1(a) and Figure 1(b) present the results surrounding event day 0. Where figure 1(a) represents the Market Model and figure 1(b) represents the Mean Model. As mentioned before, there is a possibility that the information is leaked to the market before the event day, therefore I present all days within the event window before event day 0. As argued by Khanna et. al (1998), it is also possible that it takes time to analyse the data and therefore the impact of the information might be lagged. To capture this, I present all days beyond event day 0 are as well. Figures 1(a) and 1(b) also provide 5% confidence bounds for the entire event window.
Table II Correlation coefficients
1 2 3 4
1. TRI 1
2. D/E Ratio 0.023 1
3. Total Assets 0.000 -0.042 1
4. Net Sales -0.017 -0.037 0.931*** 1
13
Figure I(a) Market Model
-4,00 -3,00 -2,00 -1,00 0,00 1,00 2,00 3,00 4,00 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Da ily a v er a g e a bn o rm a l re turns ( %)
Time (event day)
Average Abnormal Return Upper bound test statistic Lower bound test statistic Table III Average Abnormal Returns 2011-2013
Market Model Mean Model
Day AR Z-value AR Z-value
-10 0.41% 0.27 0.33% 0.22 -9 0.11% 0.07 0.39% 0.26 -8 0.48% 0.33 0.76% 0.51 -7 -0.31% -0.21 -0.14% -0.09 -6 0.25% 0.17 -0.23% -0.16 -5 0.33% 0.22 -0.15% -0.10 -4 -0.33% -0.22 -0.75% -0.50 -3 -0.03% -0.02 0.02% 0.01 -2 -0.32% -0.21 0.48% 0.33 -1 0.22% 0.15 1.13% 0.77 0 0.02% 0.01 -0.08% -0.05 1 -0.35% -0.24 -0.52% -0.35 2 0.29% 0.20 0.14% 0.10 3 0.06% 0.04 -0.19% -0.13 4 0.99% 0.67 1.96% 1.33 5 -0.12% -0.08 0.07% 0.04 6 0.22% 0.15 0.51% 0.34 7 -0.58% -0.39 -0.61% -0.41 8 -0.27% -0.18 -0.17% -0.11 9 0.01% 0.01 0.11% 0.08 10 0.12% 0.08 0.01% 0.01 Source: DataStream, CRPS * Significant at the 10% level
14
Figure I(b) Mean Model
As presented in table III, the average abnormal return on event day 0 is 0,02% according to the market model and -0,08% according to the mean model. The results indicate that according to the market model, the return on event day 0 is 0,02% higher than we would expect based on historical stock and market movements of the firms within our sample. Both models support that the abnormal returns are not significantly different on the day that the TRI data is published. Looking at other days within the event window, there is no reason to believe that the information was leaked into the market or lagged due to analysing the dataset. Figure I(a) and Figure II(b) visualize that there is a difference between the market and the mean model results. When looking at the appendix where individual event studies for each year are presented for robustness, this difference is spotted in the 2011 event study results presented in Table AIII, Figure AI(a) and AI(b). This can be explained by large market movements on several days within the event window of 2011. These market movements are included in the Market Model, where the Mean Model only takes the average company return of hundred days before the event into account. Individual event studies presented in the appendix show more fluctuations throughout the entire event window, but it cannot be concluded that the release of the TRI data creates a statistically significant response on the stock market. The individual event studies of 2012 and 2013 presented in the appendix, show more fluctuating abnormal returns. When looking at the entire event window, it cannot conclude that the release of TRI data triggers abnormal returns on event day 0 or event day 1. The results differ from previous studies performed by Hamilton (1993) and Khanna et al. (1998) who both find significant negative abnormal returns on the day that the EPA releases TRI data. However as mentioned in the literature section, the impact of the reaction relies on the extent to which the dataset is perceived as unanticipated news. In other words, the more investors and analysts know about pollution level of firms, the less impact the release of the dataset is expected to have. Since the articles of both Hamilton and Khanna et al., the availability of such information has grown. Therefore I do not conclude it is against economic theory that TRI data no longer triggers significant abnormal returns on the publishing day.
-4,00 -3,00 -2,00 -1,00 0,00 1,00 2,00 3,00 4,00 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Da ily a v er a g e a bn o rm a l re turns ( %)
Time (event day)
15
4.2 Panel regression results
Although event study results suggests that the release of TRI information is no longer perceived as news to investors, it does not indicate that the information is not included in firm value. The relationship between Tobin’s Q, TRI data and the control variables are presented in table IV. Model A contains the fixed effects panel regression of equation (8), where TRI data is divided by total assets and firm size is calculated by the natural logarithm of total assets. Model B contains the fixed effects panel regression of equation (9).
Table IV Regression results
Model A: Assets Model B: Net sales Explanatory variables Coefficient T-Statistic Coefficient T-Statistic
Constant 5.850 4.365*** -2.533 -1.874* TRI (assets) -0.002 -0.335 TRI (sales) 0.008 2.065** Total Assets -0.274 -3.058** Sales 0.287 3.161*** D/E ratio 0.046 2.047** 0.033 1.481 Year 2011 -0.271 -10.230*** -0.220 -8.661*** Year 2012 0.008 0.305 0.035 1.397 Explanatory value (R2) 0.902 (90,2%) 0.903 (90,3%) Source: DataStream, TRI Data
* Significant at 10% level ** Significant at 5% level *** Significant at 1% level
The explanatory value (R2) of both Model A and B are 90,2% and 90,3% respectively. The value is very high which indicates that the explanatory variables explain around 90% of the dependent variable. In model A total waste indicated by TRI data does not seem to have a statistically significant relationship with the explanatory variable Tobin’s Q. A higher amount of toxic waste per asset will thus not have consequences for the market value or replacement value of the assets of a firm. In Model B however, TRI shows a positive statistically significant relationship with firm value of 0.008 (p<0,05). The result indicates that when a firm increases its toxic release with one pound per dollar in net sales and all other variables held constant, the Q-ratio of a firm will increase with 0.008. Although small, this result is against economic theory that more toxic waste will impose significant costs, which will decrease profitability and thus firm value. It is also against the findings of Konar and Cohen (2001) that bad environmental performance is negatively correlated with the intangible asset value of a firm. Regarding both results in models A and B, there is a very weak positive relationship between the TRI emission data and firm value measured by Tobin’s Q based on the panel regression.
16
Tobin’s Q while measuring the impact of CSR scores in Asian emerging markets. As expected in economic theory, net sales in Model B has a positive significant relationship with Tobin’s Q. Furthermore, the debt-to-equity ratio has a small but positive statistically significant relationship with the Q-ratio of a firm. An increase of the debt-to-equity ratio by 1 will increase the Q-ratio by 0.046 (p<0,001). Lang et al. (1996) and Hennessy (2004) argue that leverage is negatively related to growth and Tobin’s Q. However this only accounts for firms with a low Q-ratio. Since the mean Tobin’s Q in our sample is relatively high, this might imply that higher leverage can optimize the benefits of debt in its capital structure and therefore increase firm value. Year dummy’s for 2011 and 2012 are presented in table IV as Year 2011 and Year 2012 respectively. Only 2011 differs statistically significant from 2013 in both models. In 2011 the Q-ratio of firms is on average 0.271 (Model A) and 0.22 (Model B) lower as compared to reference year 2013.
5. Conclusion
The Toxic Release Inventory provided annually by US EPA does not result in a short-term stock market reaction. Performing an event study on 295 US listed firms between 2011 and 2013 shows no statistically significant abnormal returns on the release date of the TRI report. There is no reason to suspect earlier leakage of information or a lag for analysing data. Although these results are not in line with previous literature by Hamilton (1993) and Khanna et al. (1998), there is reason to believe that analysts are more frequently informed about the toxic emissions of listed firms than in the past. Listed firms are obliged to mention pollution levels in their annual reports, US EPA reports indications at several periods throughout the year and information is more easily accessible. Therefore investors and analysts can predict actual pollution levels and their consequences more accurately. There is however a small positive statistically significant long-term relationship between the TRI measures and firm value as measured by Tobin’s Q. By performing a fixed effects panel regression on the same 295 listed US firms between 2011 and 2013, I find that an increase of one pound per dollar in net sales, results in an increase of a firm’s Q-ratio by 0.008. This result contradicts the theory that increasing toxic waste imposes significant costs, which will reduce profitability and firm value. It is also not in line with results by Konar and Cohen (2001) who find a statistically significant negative relationship between toxic release and the value of intangible assets. A fixed effects panel regression using total assets instead of net sales to control for firm size does not result in statistically significant results. Therefore the long-term relationship found in this paper between the Toxic Release Inventory and firm value is weak. Overall the annual release of TRI does not trigger a short-term market reaction. It neither has a strong direct long-term relationship with financial performance measured by Tobin’s Q. Based on the results of this study, the market does not react strongly to the annual release of the Toxic Release Inventory provided by US EPA, neither do I find results that the information is embedded in perceived firm value.
5.1 Limitations and further research
17
18
6.
Literature
Bharadwaj, A. S., Bharadwaj, S. G., & Konsynski, B. R. (1999). Information technology effects on firm performance as measured by Tobin's q. Management science, 45(7), 1008-1024.
Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of
finance, 52(1), 57-82.
Chan, S. H., Martin, J. D., & Kensinger, J. W. (1990). Corporate research and development expenditures and share value. Journal of Financial Economics, 26(2), 255-276.
Chen, K. H., & Metcalf, R. W. (1980). The relationship between pollution control record and financial indicators revisited. The Accounting Review,55(1), 168-177.
Cheung, Y. L., Tan, W., Ahn, H. J., & Zhang, Z. (2010). Does corporate social responsibility matter in Asian emerging markets?. Journal of Business Ethics, 92(3), 401-413.
Chung, K. H., & Pruitt, S. W. (1994). A simple approximation of Tobin's q. Financial
management, 70-74.
Dowell, G., Hart, S., & Yeung, B. (2000). Do corporate global environmental standards create or destroy market value?. Management science, 46(8), 1059-1074.
Epstein, M. J. (1991). What shareholders really want. New York Times, 28.
Fama, E. F. (1991). Efficient capital markets: II. The journal of finance, 46(5), 1575-1617. Filbeck, G., & Gorman, R. F. (2004). The relationship between the environmental and
financial performance of public utilities. Environmental and Resource Economics, 29(2), 137-157.
Geczy, C., Stambaugh, R. F., & Levin, D. (2005). Investing in socially responsible mutual funds. Available at SSRN 416380.
Hamilton, J. T. (1995). Pollution as news: media and stock market reactions to the toxics release inventory data. Journal of environmental economics and management, 28(1), 98-113. Hennessy, C. A. (2004). Tobin's Q, debt overhang, and investment. The Journal of
Finance, 59(4), 1717-1742.
Jaffe, A. B., Peterson, S. R., Portney, P. R., & Stavins, R. N. (1995). Environmental regulation and the competitiveness of US manufacturing: what does the evidence tell us?.
Journal of Economic literature, 132-163.
Jaggi, B., & Freedman, M. (1992). An examination of the impact of pollution performance on economic and market performance: pulp and paper firms.Journal of Business Finance &
19
Jensen, M. C. (1993). The modern industrial revolution, exit, and the failure of internal control systems. the Journal of Finance, 48(3), 831-880.
Jones, S., Van der Laan, S., Frost, G., & Loftus, J. (2008). The investment performance of socially responsible investment funds in Australia. Journal of Business Ethics, 80(2), 181-203.
Klassen, R. D., & McLaughlin, C. P. (1996). The impact of environmental management on firm performance. Management science, 42(8), 1199-1214.
Khanna, M., Quimio, W. R. H., & Bojilova, D. (1998). Toxics release information: a policy tool for environmental protection. Journal of environmental economics and
management, 36(3), 243-266.
Konar, S., & Cohen, M. A. (2001). Does the market value environmental performance?. Review of economics and statistics, 83(2), 281-289.
Lang, L., Ofek, E., & Stulz, R. (1996). Leverage, investment, and firm growth. Journal of
financial Economics, 40(1), 3-29.
McGuire, J. B., Sundgren, A., & Schneeweis, T. (1988). Corporate social responsibility and firm financial performance. Academy of management Journal, 31(4), 854-872.
MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of economic
literature, 35(1), 13-39.
Morck, R., Shleifer, A., & Vishny, R. W. (1988). Management ownership and market valuation: An empirical analysis. Journal of financial economics, 20, 293-315.
Porter, M. E. (2008). Competitive advantage: Creating and sustaining superior performance. Simon and Schuster.
Ramiah, V., Martin, B., & Moosa, I. (2013). How does the stock market react to the announcement of green policies?. Journal of Banking & Finance, 37(5), 1747-1758.
Renneboog, L., Ter Horst, J., & Zhang, C. (2007). The price of ethics: Evidence from socially responsible mutual funds. ECGI-Finance Working Paper, (168).
20
7.
Appendix
TABLE AI
The Distribution of Pollution 2012
All Firms Listed firms (sample) % of total firms Air (M lbs.) 726 165 23% Water (M lbs.) 206 33 16% Underground (M lbs.) 197 3 2% Land (M lbs.) 1945 402 21% Offsite (M lbs.) 405 64 16% Tot releases (M lbs.) 3479 667 19%
Tot energy recovery (M lbs.) 2805 393 14%
Tot recycling (M lbs.) 7317 1618 22%
Tot treatment (M lbs.) 8335 1279 15%
Tot production waste (M lbs.) 21941 3957 18%
Number of companies 5370 301 6%
TABLE AII
The Distribution of Pollution 2013
All Firms
Listed firms
(sample) % of total firms
Air (M lbs.) 729 169 23% Water (M lbs.) 201 34 17% Underground (M lbs.) 199 4 2% Land (M lbs.) 2418 460 19% Offsite (M lbs.) 407 64 16% Tot releases (M lbs.) 3955 731 18%
Tot energy recovery (M lbs.) 2783 395 14%
Tot recycling (M lbs.) 7506 1663 22%
Tot treatment (M lbs.) 8466 1512 18%
Tot production waste (M lbs.) 22523 4303 19%
21 -10,00 -8,00 -6,00 -4,00 -2,00 0,00 2,00 4,00 6,00 8,00 10,00 -1 0 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Market model 2011 Average Abnormal Return Upper bound test statistic Lower bound test statistic
Figure AI(a)
Figure A1(b)
Table AIII Event study results year 2011
Market Model Mean Model
Day 0 Day 1 Day 0 Day 1
Average abnormal return Mean AR 0.32% 1.02% 0.22% 0.38%
22
Table AV Event study results year 2012
Market Model Mean Model
Day 0 Day 1 Day 0 Day 1
Average abnormal return Mean AR -0.61% -0.43% -0.52% -0.42%
Z-value -1.85* -1.32* -1.59* -1.28* *** Significant at α=0.01 ** Significant at α=0.05 * Significant at α=0.1 Figure AII(a) Figure AII(b) -2,00 -1,50 -1,00 -0,50 0,00 0,50 1,00 1,50 2,00 -1 0 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Market model 2012 Average Abnormal Return Upper bound test statistic Lower bound test statistic
23
Figure AII(a)
Figure AII(b)
Table AVI Event study results year 2013
Market Model Mean Model
Day 0 Day 1 Day 0 Day 1
Average abnormal return Mean AR 0.20% -1.57% 0.08% -1.46%
Z-value 1.45* -11.14*** 0.53 -10.35*** *** Significant at α=0.01 ** Significant at α=0.05 * Significant at α=0.1 -2,00 -1,50 -1,00 -0,50 0,00 0,50 1,00 1,50 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Market model Average Abnormal Return Upper bound test statistic Lower bound test statistic
