The effects of socially responsible investment on sin stocks : a multiple event study on pension fund regulation

(1)

University of Amsterdam, Amsterdam Business School

The effects of socially responsible

investment on sin stocks

A multiple event study on pension fund regulation

Abstract

This thesis focuses on the effects of social responsible investment on sin stocks in Europe and the United States. Sin stocks belong to companies that produce harmful products according to societal norms. The performed analyses are based on four key events in the development stage of the IORP II directive, a new regulation that enforces European pension funds to divest from sin stocks. By computing the average abnormal returns using the market model and testing their significance, this study searches for evidence to support a negative effect on the market for sin stocks. However, there is not enough statistical evidence found to support a significant permanent negative effect on the stock markets. Furthermore, the effect of the events is compared between Europe and the United States to find significant differences in the reaction of the stock markets. This study finds sufficient statistical evidence to prove that the effects of this regulation are equal in Europe and the United States.

Student: Koen Groenendijk (11953543) Email: koengroenendijk@gmail.com MSc Finance, Corporate Finance track

Thesis supervisor: Mr. S. Kucinskas August, 2018

(2)

2

Statement of Originality

This document is written by Koen Groenendijk, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. Introduction

Nowadays, society increasingly demands for more Socially Responsible Investing (SRI) by institutions. As one of the most recent examples, the second largest European pension fund, the Dutch ABP (Kennedy, 2017), announced that it will completely stop investing in tobacco and nuclear weapons (Tooms, 2018). However, this is only one of the latest expressions of a current investing trend. Over the last couple of years, SRI has become increasingly popular and more and more investors are reconsidering their investments based on social responsibility. Not only has SRI been increasing in popularity, but it has evolved into a worldwide investment philosophy that is currently adopted by a large number of institutions. In other words, it has matured into a mainstream investment policy (Sparkes & Cowton, 2004). Consequently, the demand for stocks of the firms behind ‘sinful’ products is decreasing. Sin stocks are considered to be stocks from companies that produce harmful or socially unaccepted products. The most common examples are the alcohol, tobacco, weapons and gaming industries (Fabozzi et al., 2008; Ghoul et al., 2011; Hong & Kacperczyk, 2009; Kim & Venkatachalam, 2011; Salaber, 2007). If investors become less motivated, or even prohibited, to invest in sin stocks, this might have serious consequences for the firms’ market performance in these industries.

Currently, pension funds and the EU are in discussion about additional regulation concerning disclosure of their investment portfolios. Although just recently new regulations have passed with regards to SRI, the European Union is already looking into expanding the new legislation. On the one hand, the EU wants more transparency about the pension funds’ portfolios and limitations to the type of businesses they are allowed to invest in. On the other hand, the pension funds ask for more research into the effects of the current laws, since there is little known about the impact of such regulation. They argue that implementation of laws of this nature are too expensive or that the definition of sin stocks is too subjective. As such, it would be impossible to implement universal legislation in the entire European Union (Eatock, 2017; Rust, 2018). Also, the pension funds are worried about their returns, since sin stocks tend to yield high returns compared to comparable industries. Moreover, there have been several studies that provided evidence for underperformance of socially responsible portfolios (Geczy et al., 2003; Statman, 2000). In addition, there have been several studies that focus on the effects of SRI on stock prices and the cost of capital of a company, for both the cost of equity and cost of debt (Ghoul et al., 2011; Goss & Roberts, 2011). Furthermore, researchers have tried to determine factors that can explain the higher returns on sin stocks,

(5)

5

relative to ‘non-sinful’ similar stocks (Fabozzi et al., 2008; Hong & Kacperczyk, 2009; Kim & Venkatachalam, 2011; Salaber, 2007, 2009). However, the possible disruptive consequences of SRI on the stock market are not very clear yet. This is confirmed by the arguments of the European pension funds themselves (Eatock, 2017). Since there is much unclear about the effects of SRI on the market for sin stocks and pension fund regulations in particular, this study may contribute to the current discussion. This research may add to the discussion about SRI effects, through analyzing the effects on the market of new laws that enforce SRI. Therefore, the research question of this paper is formulated as follows:

What is the short-term effect of Social Responsible Investment on firm performance in the sin-stock market?

In order to analyze this question, two hypotheses are formed in the next chapter. The analysis will be performed by conducting four different event studies on sin stocks in Europe and the United States. This study concerns both regions, because the European pension funds have their largest (sin stock) investments in both these regions. Hence, there should not be a difference between the impact of European regulation on stocks within these different regions. The event studies will make use of the market model to estimate the stock returns of sin stocks during the event periods. In addition, the abnormal returns will be calculated by computing the difference between the expected returns and the actual returns. The actual returns are collected from different databases, depending on the companies’ countries of origin. Datastream will be used to gather data on European firms, while data applicable to firms from the United States will be obtained from CRSP. Subsequently, the significance of the abnormal returns during the event periods will be tested. Additionally, the effects on the American and European stock markets will be compared in order to test for a different influence from European regulation. The results do not provide enough evidence to support a large negative influence on the sin stock markets. However, there is enough evidence to assume that European pension fund regulations have similar effects on the European and American stock markets.

This paper continues as follows. First, the literature review will cover the institutional and academic backgrounds of the research. Secondly, the methodology for the analyses is described. Thirdly, the construction of the dataset is described, followed by descriptive statistics of this dataset. Fourthly, the results are presented and reflected upon. Fifthly, a description of the robustness tests is provided, as well as the results of these tests. And

(6)

6

finally, the last section of this paper concludes on the research performed and a summary of the relevant implications is provided.

2. Literature review

In order to investigate the effect of SRI, this thesis will analyze the impact of changes in investing by European pension funds. These pension funds are accountable for approximately €7 trillion (Kennedy, 2017) and therefore could have a significant impact on stock markets. Sethi (2005) discusses the influence of pension funds in the United States on stock markets. According to Sethi, the pension funds in the United States hold approximately $1 trillion in assets combined. Any divestment by the pension funds has to be conducted gradually in order to prevent large disruptions in the market. Following these findings, European pension funds should also be able to make a significant impact on stock markets. Therefore, a change of investing behavior by these funds, due to new regulation, may be likely to affect the performance of several stocks (Sethi, 2005; Schwert, 1981). In order to investigate this effect, this paper focuses on such new regulation by analyzing the effects of the IORP II directive in the European Union.

2.1 Institutional background

The directive on the activities and supervision of institutions for occupational retirement provision (IORPs) II, or simply the IORP II directive, encompasses new regulation with regards to the investment behavior of European pension funds based on SRI principles. It entered into force on January 12th, 2017 (Eatock, 2017). The three main pillars of the new directive are risk assessment, improved governance and disclosure to both members and the public (EP, 2016). The improved risk assessment prescribes internal audits, written risk policies and risk assessment reports. Furthermore, these risks must be communicated to the members of the pension funds and realistic forecasts of the returns must be presented. However, the part on which this study will focus is concentrated on the new investment rules. All pension funds have to take Environmental, Social and Governance (ESG) factors into account. The ESG factors are quite similar to Socially Responsible Investment. The new directive thus prohibits investments in any companies that are harmful for the environment or not socially accepted. A more distinct explanation of these stocks will follow further on in this paper. One of the most important and relevant factors of the IORP II directive is the ‘prudent person rule’. The prudent person rule implies that the introduced ESG factors are to be considered to have the same status as a person would have. In other words, failing to

(7)

7

adhere to the ESG factors were as if the pension fund would not be operating in the interest of one of its members. This results in direct conflict with another investment rule that is included in the directive, namely to invest the pension fund’s assets in a way that achieves the best long-term results for the members and beneficiaries of the fund. In order to ensure adherence to the new IORP II directive, European member states have the right and are required to impose adequate administrative sanctions to pension funds. If a pension fund does not operate in the interests of its members or does not adhere to regulations in general, the supervisory system may freeze its assets. Moreover, if there is a serious violation, member states may redirect the power of the persons that lead the pension fund to someone who they deem more capable of executing this function (EP, 2016).

All in all, there are quite a few implications for the investment behavior of European pension funds. However, the most important implication of the IORP II directive is the implementation of the new investment rules. For many pension funds, this means that operations have to be adjusted and divestments have to be made (Willis Towers Watson, 2016). In this light, the new regulation provides a possibility to investigate the influence of SRI on sin stocks. Since regulations do not suddenly come into being, but rather take years to develop, it does not suffice to only research the finalizing date of the directive. As the stock markets in Europe and the United States are regarded as efficient markets, all news surrounding the directive should be directly incorporated into the stock prices (Fama, 1970). Therefore, several event periods surrounding dates on which important news about the directive was presented or leaked are investigated in this study. These event periods will be elaborated on in the chapter about methodology.

2.2 Academic background

In academic literature, sin stocks are considered to be stocks of companies with sinful operations, as the name suggests. On the one hand, one could think of products that may harm people, such as guns and tobacco. On the other hand, sin stocks could belong to companies whose operations are not necessarily harmful, but are not socially accepted. Since “socially accepted” is a quality that differs across countries and cultures, it creates uncertainty about which stocks should and should not be included. Moreover, the qualification of sin stocks would create a dichotomy between ‘sinners’ and ‘saints’, which demonizes certain companies solely based on the type of product that they make (Putnam, 2004). Therefore, the focus would not be on the way in which a company conducts business, but purely on the

(8)

8

industry in which it operates (De Colle & York, 2009). The mistake in the screening of these companies is explained by De Colle and York (2009) in an example of beer brewers and wine makers. Although alcohol is considered harmful nowadays, as it is a large cause of traffic accidents, suicides and several types of cancer (WHO, 2014), beer brewing and wine making can also be considered as a craftsmanship. Especially in Europe, several countries have a rich tradition in either beer brewing or wine making. The Trappist monasteries originally brew their beer because it was healthier than the water at that time. Moreover, wine is part of the tradition of drinking “the blood of Jesus” in the Catholic Church. Consequently, one could argue that one industry may be more ‘sinful’ than the other. Therefore, this so-called “social screening” is a reflection of the social, religious or political beliefs of investors (Schwartz, 2003). Based upon the beliefs of investors, the definition of being ‘sinful’ may differ across markets. Fortunately, sinful industries have been researched in several papers, which mostly agree upon which industries are to be classified as sinful. For example, sin stocks are defined by Hong and Kacperczyk (2009) as companies that operate in the tobacco, alcohol, gambling, military, sex and nuclear industries. However, they mainly focus on the tobacco, alcohol and gaming industries. This is due to uncertainty about the “sin-status” of the weapon industry in the United States and a lack of data on behalf of the sex and nuclear industries. Salaber (2007) agrees with their study and also focuses on the tobacco, alcohol and gaming industries, although her study was situated around the European sin stocks. Furthermore, Fabozzi et al. (2008) also handle a broader definition of sin stocks. They argue that the weapons and biotech alterations industries should be included as well. Kim and Venkatachalam (2011) extend the definition even more by describing sin stocks more generally as part of companies whose operations are involved in producing and marketing unethical products. As some of these industries facilitate human vices, it is quite easy to classify them as sinful. Furthermore, products of these industries may be harmful for people. For example, these companies could be harmful by either supplying weapons for military conflicts or by selling addictive substances. Also, there may be long-term health issues involved, such as the higher risk of cancer due to alcohol, tobacco or being exposed to radiation from nuclear power plants.

The main motivation for interested parties to invest in these sin-stocks are the relatively high stock returns, when compared to similar non-sinful stocks (Fabozzi et al., 2008; Fauver & McDonald, 2014; Hong & Kacperczyk, 2009; Kim & Venkatachalam, 2011). Scholars have found evidence to support multiple theories that explain these structural higher returns.

(9)

9

Firstly, the sin-industries profit from conservative accounting rules due to the addictive nature of the products of the companies. Therefore, they bear more attention from regulators, which enhances the scrutiny with which their financial data is monitored (Kim & Venkatachalam, 2011; Salaber, 2009). This results in a lower cost of capital in two ways. The quality of accounting information has an effect on both the perception of future cash flows by investors and indirectly on the decisions made by the company, which affect the stock price and cost of capital (Lambert et al., 2007). Secondly, several types of sin stocks, most of all the tobacco industry, have a high litigation risk compared to non-sin stocks. However, this type of risk does not seem to be supported by sufficient statistical evidence (Hong & Kacperczyk, 2009). Thirdly, Hong and Kacperczyk (2009) find that sin stocks are undervalued because of a trend in SRI. Since socially responsible investing has become a common way of investing one’s money (Sparkes & Cowton, 2004), the decreased demand causes sin stocks to be priced below their fair price. As a consequence, these stocks yield higher returns than stocks of an equal value. Among others, this effect is addressed by Merton (1987), who argues that prices of stocks fall below their fundamental value if an important group of investors, specifically institutions such as pension funds, are neglecting them. In addition, he finds that institutions are allowed and able to have large enough stakes in a company to seriously influence the stock price. The theory behind his findings is called the neglect effect hypothesis. In general, so-called neglected stocks seem to be undervalued. Consequently, they yield a higher return than comparable firms that have more media attention and analyst coverage (Arbel et al., 1983; Carvell & Strebel, 1987; Hong & Kacperczyk, 2009). One of the main reasons for the abnormal returns of neglected stocks is the amount of information that is available. Since these stocks usually have less analyst coverage, investors have access to less information about these stocks. This increases the risk and therefore the possibility of false pricing. Subsequently, it reduces the attractiveness of these stocks for investors. As a consequence, neglected stocks are found to have positive abnormal returns due to the increased volatility (Arbel et al., 1983; Carvell & Strebel, 1987). In contrast, Adamsson and Hoepner (2015) argue that there is a size bias in the methodology of Hong and Kacperczyk (2009). According to them, the abnormal returns of sin stocks compared to other stocks disappear when the variation in size is controlled for. However, Arbel et al. (1983) found strong evidence to support that the neglected firm effect does not only apply to companies with a small market capitalization. In their findings, the effect is also present for middle sized and even large firms. These results have been confirmed in a later study by Carvell and Strebel (1987), who find strong statistical evidence that the neglected

(10)

10

firm effect is stronger and more robust than the size effect, one of the Fama and French (1993) factors. Moreover, the neglected firm effect is one of the main reasons for sin stocks to earn a positive abnormal return, compared to the total market (Hong & Kacperczyk, 2009; Salaber, 2009). Nonetheless, this study will take this discussion into consideration and account for both equally weighted and value weighted portfolios.

There are a few questions that arise to oppose these theories. First of all, the presence of arbitrageurs in the market for neglected stocks, which should eventually bring the stock price back to its fair value. However, multiple studies find that the arbitrageurs, such as mutual funds, do not bring enough capital to the market to cancel out the effects of the neglected frim effect (Merton, 1987; Shleifer & Vishny, 1997). Moreover, Fauver and McDonald (2015) conclude that after controlling for arbitrage opportunities, sin stock still yield annual excess returns of 1-2% in countries included in the G20. Secondly, there is contradictory evidence to support that there are no excess returns. Although sin stocks may outperform the total market, they do not outperform stocks from comparable industries (Salaber, 2009). However, Salaber (2009) also finds that it is less risky to hold sin stocks than stocks from comparable industries. In the same study, the excess returns are investigated during economic expansion and recession. Although there are no significant excess returns during economic prosperity, there is enough evidence to support that sin stocks outperform their comparables when the economy is not doing well. As a consequence, it is advisable for investors to include sin stocks in their portfolio. This also means that there is a significant cost to investors who invest according to SRI and neglect sin stocks (Salaber, 2009). The IORP II directive solves a paradox that has already been discussed in academic literature. Since there is quite some statistical evidence to support pension funds’ investments in sin stocks, it must be rewarding to invest in these stocks. However, if these stocks are superior, pension funds should be investing mainly in sin stocks in order to achieve the highest returns for their members and beneficiaries. In spite of higher returns, these stocks represent products that are known for their negative effects on the well-being of people. On the one hand, acting in the interest of the members and beneficiaries would mean to achieve high returns on investments. On the other hand, if these same companies are responsible for health issues, it would indirectly harm the members if these companies are supported (Hoepner & Zeume, 2014).

In the case of European pension funds, the IORP II directive requires them to divest in sin stocks. As discussed by aforementioned studies, these stocks are a profitable investment opportunity for these pension funds. Moreover, there is statistical evidence that avoiding

(11)

11

these stocks inflicts a cost to investors (Salaber, 2009). The presence of sin stocks in pension funds’ investment portfolios before the IORP II directive is confirmed by news articles such as Tooms (2018). He describes that European funds are starting to divest in the tobacco and weapons industries. This is also confirmed by Willis Towers Watson (2016), as they argue that significant changes have to be made by pension funds in Europe. These should be recognized by the stock markets and therefore be observable in stock prices and returns. Following the neglected firm theory, the sin stock prices should therefore decline as a reaction to news about the implementation of ESG rules in the IORP II regulation. Consequently, the first hypothesis of this paper is as follows:

H1: The implementation of the IORP II law shall lead to a decline in stock returns in the

sin-stock market.

Furthermore, there is hardly any literature on the influence of European pension funds on the United States stock markets. The available literature discusses the effects of the investing behavior of pension funds on the stock market. Also, the literature discusses the effects of pension funds on sin stocks in particular. However, there are hardly any examples of studies that investigate the pension funds’ investing behavior overseas and its effects. Nonetheless, European pension funds invest in a large mix of international stocks, from which most are either European or American. For example, a quick look at the top 100 investments of the Dutch pension fund ABP learns that the proportion of stocks in the portfolio from either region is quite similar1. As such, the influence of investments of European pension funds in stocks from the United States may be as large as the effect on stocks from their own region. Sethi (2005) argues that United States pension funds have the capability to influence stock prices through the size of their investments. Since European pension funds have combined assets that are seven times larger than their American counterparts (Kennedy, 2017), they should be capable of affecting American stocks as well. Because of the free flow of capital and similarly efficient markets, the European and American stock markets should react within a similar timespan to news about new investment regulations (Fama, 1970). Therefore, the second hypothesis is as follows:

H2: The implementation of the IORP II law shall have a similar effect in both the European

and American stock markets.

(12)

12

3. Methodology

As mentioned before, this thesis revolves around a number of event studies in order to test if the regulatory change for European pension funds influences the market of sin stocks. The event periods will include 5 days before and after the event date in order to include early news leaking and possible delay due to weekends, for instance. To do so, normal returns will be calculated using the market model, which is prescribed by Schwert (1981):

𝐸(𝑅𝑖𝑡) = 𝛼𝑖 + 𝛽𝑖𝑅𝑚𝑡+ 𝛽𝑗𝐿𝑂𝐺𝑆𝐼𝑍𝐸 + 𝜀𝑖𝑡

Here the expected return of a security at point t in time is depicted by Rit, whereas Rmt

describes the market return. The variables 𝛼_𝑖 and 𝛽_𝑖 reflect the constant and the coefficient, respectively. In addition, a control variable is added for the size of the companies. There are significant differences in the size of the companies, as will be explained further in the chapter about the data gathering and summary statistics. The coefficient of the control variable

LOGSIZE is depicted by 𝛽_𝑗. Furthermore, the estimation error is explained by the variable 𝜀_𝑖𝑡

in the formula and has an expected value of 𝐸(𝜀_𝑖𝑡) = 0. In the market model, the daily stock prices of all firms in the sample over the year 2012 are used, since it is advisable to use at least 120 trading days between the event window and the estimation window (MacKinlay, 1997). By taking the day to day differences between the logarithmic values of both the stock prices and index prices, the daily stock returns and market returns can be computed. However, CRSP does not extend data on the market index values. The market returns that are used instead for the regression of stock listed in the United States are therefore obtained from the S&P 500 (MacKinlay, 1997). These market returns are gathered directly from the database, since CRSP does supply the S&P 500 return. Furthermore, the stock returns are regressed on the market returns for each single firm in order to calculate each constant and coefficient. Of course, the European firm returns are regressed on the European market returns and the American firm returns on the American market returns. These regression outcomes, in turn, are used to make an estimation of the expected stock returns in the event windows, based on their correlation with the market return i.e. the market return coefficient 𝛽_𝑖. As soon as the expected returns are determined, the abnormal returns will be computed. The abnormal returns will be calculated as described by MacKinlay (1997), by using the formula:

(13)

13

The abnormal returns consist of the difference between the expected return that was calculated using the market model and the actual returns. Here, 𝐴𝑅_𝑖𝜏 denotes the abnormal return of security i at time τ. The expected return is explained by 𝐸(𝑅_𝑖𝜏) given the available information 𝑋𝜏. In order to use this variable to calculate the test statistic, the average abnormal return (AAR) is needed. The AAR is calculated by taking the daily average of these abnormal returns:

𝐴𝐴𝑅_𝑡 = 1

𝑁∑ 𝐴𝑅𝑖𝑡 𝑁

𝑖=1

Subsequently, the first test statistic can be calculated by using the following formula:

𝑇𝑆1 =

√𝑁 ∗𝐴𝐴𝑅_𝜎 𝑁

Where N is the number of stocks in the sample, AAR is the average abnormal return and σ denotes the standard deviation of the average abnormal returns. This test statistic is calculated per day and independently for sin stocks and the control group. It calculates the probability of having abnormal returns that are statistically different from zero for every day. If the result of the test statistic is negative, it shows that there were negative abnormal returns. In other words, the stocks would have performed worse than expected. With regards to the first hypothesis, a significant negative result would support the idea that the IORP II directive has a negative influence on the stock prices. In order to control for a size bias, the test statistic is calculated for two portfolios. The first portfolio is equally weighted, whereas the second portfolio is value weighted (Schwert, 1981). Moreover, by calculating the test statistics for these portfolios rather than just the individual stocks, the correlation between these individual stocks is accounted for (King, 1966; Jensen et al., 1972; Schwert, 1981). The weight of the second portfolio is determined by taking into account the variable LOGSIZE, which is the logarithmic value of the market capitalization of each firm. After calculating the first test statistic, the second test statistic will calculate the significance of the cumulative average abnormal returns (CAAR) over the entire event window. The CAAR is calculated according to the following formula:

𝐶𝐴𝐴𝑅_𝑡 = ∑ 𝐴𝐴𝑅_𝑡 𝑡

(14)

14

Consequently, the formula for the second test statistic will be as follows: 𝑇𝑆₂ = √𝑁 ∗𝐶𝐴𝐴𝑅

𝜎

Where the new variable CAAR reflects the cumulative average abnormal return. That is, the average abnormal return over the entire event period. N is still the number of stocks in the sample and σ describes the standard deviation of the cumulative average abnormal return. The result of this test statistic is similar to the first test statistic. In order to support the hypothesis, the outcome should yield significant negative results. This test statistic is also calculated for both the equally weighted and value weighted portfolios. Furthermore, the test statistic is again calculated for both the sin stock group and the control group. The second test statistic calculates if the combined abnormal returns over the event window are statistically different from zero.

In order to test the second hypothesis, the same tests as before are performed, only this time for the individual regions. These tests may give an indication as to what extent the two subsamples are comparable. Subsequently, T-tests will be performed in order to analyze if there is a statistically significant difference between the average abnormal returns in Europe and the United States. First, the variance of the two subsamples is tested in order to see if a standard T-test is applicable. If the variances are unequal, an unequal variance T-test has to be conducted. The results of the T-tests have to be insignificant to meet the expectations. If there are significant changes to be found, it will mean that there are differences between the effects of IORP II on sin stocks in Europe and the United States. This could implicate several things, among which asymmetric information or different implications of the neglected firm theory in terms of foreign direct investment.

There are four different event periods that will be analyzed. These periods are formed around dates on which important information regarding the IORP II directive was first made public. First of all, the initial announcement of increasing the transparency for European pension funds by law. On its third annual conference in Frankfurt, the chairman and the executive director of the European Insurance and Occupational Pensions Authority (EIOPA) announced that the focus for the next IORP directive was on the improvement of governance, reporting requirements and transparency rather than solvency rules (Bernardino, 2013; Montalvo, 2013; Ottawa, 2013). Therefore, the first event period is based on the conference date: November 20th, 2013. Secondly, the next event period is formed around February 28th,

(15)

15

2014. On this date, the first draft version of the IORP II proposal was leaked. The main objectives of the proposal included focusing on governance, transparency to members of the pension funds and providing supervisors with the tools to enforce the new rules (Woolfe, 2014). Thirdly, the next event date to be analyzed was January 25th, 2016. On this date, the European Parliament voted and agreed on the first new draft since the leaked IORP II proposal in 2014. This draft would become the base for negotiations with the EU member states and their pension funds (Williams & Woolfe, 2016). Lastly, the final event date was set at June 20th, 2016. This date marks the leak of the final document, which was planned to be presented on the 27th of June (Williams, Rust & Kennedy, 2016). This finalized document describes accurately what is expected of pension funds with regards to responsible investing. Furthermore, it describes the supervisory role of member states and the means at their disposal to enforce the directive. Next to that, the prudent person rule is introduced in this document, which is one of the most important ways of ensuring responsible investment by the European pension funds (EMA, 2016).

Regarding the robustness of the tests that are performed, the test statistics will be computed multiple times with the omission of an industry. In this way, it can be ensured that the test results are not driven by a single industry. The robustness tests include an analysis without the largest sin industry in the sample and an analysis without the largest control industry. Accordingly, the comparable industries will also be left out. For example, the largest sin industry in the sample is the alcohol industry. Consequently, the soda industry will also be left out of the sample for the robustness test. These robustness tests are especially applicable to this study, since the number of firms per industry varies significantly (details on the composition of the sample will be provided in the data description chapter).

4. Data

Since European pension funds not only invest in sin stocks within the European Union, the stocks included in the sample consist of both European and American listed companies. Therefore, the necessary data has to be retrieved from multiple databases. Although pension funds are able to invest in other markets as well, this analysis is confined to firms that are listed in either Europe or the United States, because of the quality and availability of data in these markets. The data on European firms is collected from Datastream, whereas the data on firms listed in the United States is retrieved from the Center for Research in Security Prices (CRSP).

(16)

16

From Datastream, the collected data includes the daily stock prices, company name, the market value of the firm, daily index returns and the ISIN and SIC codes for identification purposes. Similarly, the variables that are obtained from CRSP are daily stock prices, number of outstanding shares, daily returns from the S&P 500, share code, company name, permno, SIC codes and eight-digit CUSIP codes. The data is collected over the complete year 2012 and four event periods, which all signify important moments in the process of forming the IORP II directive. In order to be included in the sample, firms must have available data during all of these periods. This implies that all companies that do have missing values for either the market value, stock return or index return, are not included in the sample. Furthermore, all firms must have a four-digit SIC code and either a valid ISIN or CUSIP code in order to be identifiable. The ISIN codes are translated to eight-digit CUSIP and checked for duplicates in order to streamline the data for analysis. Furthermore, the eight-digit CUSIP is used to merge the datasets into one. The string variable REGION is then created in order to keep the European and American firms distinguishable. The variable takes the straightforward values of either “EU” or “US”. In addition, financial firms were not included in the sample, since these companies are subject to strict regulation and are not as relevant to this study. Next to that, only firms with a share code of 10 or 11 will be included in the study, following the example set in other studies. The result is a sample of 405 companies, from which 126 firms are considered sin stocks and 279 firms are in the control group. In addition, 282 firms in the sample are from the European Union, while 123 firms are listed in the United States.

The identification of sin stocks is based upon the Fama and French (1997) classification of stocks. They identify stocks by their SIC codes and divide them into 48 different industries. Since this study only focuses on the alcohol, tobacco and weapons industries as sin stocks, respectively Fama and French groups 4, 5 and 26, there are only a few SIC codes that classify as such. In order to be considered a sin stock, the firm’s SIC code must be between 2080 and 2085 for alcohol, between 2100 and 2199 for tobacco and either be between 3480 and 3489, between 3760 and 3769 or exactly 3795 or 3812 in order to be classified as weapons industry. The classification of firms in the control group is done similarly. Fama and French groups 2, 3 and 24 consist of natural comparable firms for the groups of sin stocks in the dataset (Hong & Kacperczyk, 2009). These groups consist of firms in the food industry, soda industry and aircraft industry, respectively. Firstly, firms are included in the food control group if SIC codes are in between 2000 and 2046, between 2050

(17)

17

and 2063, between 2070 and 2079, between 2090 and 2095 and, finally, either 2098 or 2099. Secondly, to be qualified as active in the soda industry, the SIC code needs to fall between 2064 and 2068 or be either 2086, 2087, 2096 or 2097. Thirdly, firms in the aircraft industry must have SIC codes that fall between 3720 and 3729. With regards to conducting the analysis, the sin stock group and control group are distinguished by creating the dummy variable SINID, which takes on a value of 1 if a company is considered a sin stock. As a result of this categorization, 81 firms are in the alcohol industry, 27 in the weapons industry and 18 in the tobacco industry. Furthermore, 183 firms are categorized into the food industry, whereas 49 firms are in the aerospace industry and 47 firms are in the soda industry.

Table 1 shows the summary statistics for the full sample. N stands for the total number of observations in that particular sample, which should not be confused with the number of firms. The first variable that is depicted in the table is the logarithm of the daily closing price of the stocks in the sample, which is LOGPRICE. Secondly, LOGINDEX refers to the logarithm of the daily market index. Thirdly, the daily returns of both the stocks and the market are explained by RETURN and MRETURN, respectively. Fourthly, LOGSIZE reflects the logarithm of the market capitalization of the firms in the sample. Finally, the systematic risk of the firms in the sample is shown by the coefficient between the stock returns and the market returns, which is BETA. Furthermore, each part of the table illustrates a different time period. The first part focuses on the estimation period, which covers the entire year of 2012. The other parts consist of the event periods, which have a span of 5 days before until 5 days after the event date. The aforementioned event dates were November 20th, 2013; February 28th, 2014; January 25th, 2016; and June 20th, 2016.

As can be seen in Table 1, the values for all variables are quite consistent in the different time periods. As could be expected, the mean of the daily market returns is zero. Since this should be a most diversified portfolio, the daily average returns should compensate each other completely (Hillier et al., 2012). In case of the portfolio of sin stocks and stock from the control group, the mean return is 6% during the estimation period. Furthermore, the average returns on the sample portfolio are 6% in event period 1, 4% in event period 2, 3% in the third event period and 2% in the last event period. Also, the market return remains consistent over all periods considered in this research. Although not much can be derived from solely the total market average, these numbers do not directly support market-wide price shocks during these periods. Furthermore, the prices of the stocks seem to be increasing over time. This means that, on average, the prices of the stocks in the sample are increasing,

(18)

18

Table 1. Summary Statistics

This table reports the summary statistics for the variables used in the 5 different time periods. The first table describes the summary statistics for the estimation window, which is the entire year of 2012. The second table shows the statistics for the event period around November 20th, 2013. The third table reflects the event period around February 28th, 2014, whereas the fourth table describes the event period around January 25th, 2016. The final part of the table describes the event period around June 20th, 2016. N is the total number of observations. LOGPRICE is the logarithm of the daily closing price of the firms in the sample. LOGINDEX is the logarithm of the daily market index. RETURN is the daily stock return for the firms in the sample. MRETURN is the daily market return. LOGSIZE is the logarithm of the market capitalization of the firms in the sample. BETA is the company beta of the firms in the sample, which reflects the systematic risk.

Estimation window

Variable N Mean Median SD Min Max

LOGPRICE 124,963 2.05 2.46 2.25 -4.61 10.53 LOGINDEX 90,841 5.02 4.97 1.39 -0.11 11.91 RETURN(%) 123,906 0.04 0.00 0.36 -0.73 2.72 MRETURN(%) 123,906 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 123,906 13.18 13.29 2.54 4.25 19.02 BETA 123,906 0.59 1.00 0.72 -1.32 1.61 Date 1

LOGPRICE 2,845 2.30 2.79 2.29 -4.61 10.53 LOGINDEX 2,845 5.22 5.27 1.42 0.00 11.91 RETURN(%) 2,845 0.06 0.00 0.48 -0.73 2.72 MRETURN(%) 2,845 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 2,845 13.67 13.86 2.42 6.61 19.00 Date 2

LOGPRICE 3,164 2.50 2.90 2.10 -4.61 10.65 LOGINDEX 3,164 5.35 5.33 1.37 0.26 11.74 RETURN(%) 3,164 0.04 0.00 0.34 -0.73 2.72 MRETURN(%) 3,164 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 3,164 13.83 13.98 2.30 7.17 18.99 Date 3

LOGPRICE 2,701 2.55 3.00 2.26 -4.61 11.05

LOGINDEX 2,701 5.46 5.50 1.54 -1.61 11.25

RETURN(%) 2,701 0.03 0.00 0.28 -0.73 2.64

MRETURN(%) 2,701 0.01 0.01 0.02 -0.07 0.08

(19)

19

Date 4

LOGPRICE 2,692 2.61 3.11 2.26 -4.61 11.07

LOGINDEX 2,692 5.51 5.57 1.54 -2.30 11.40

RETURN(%) 2,692 0.02 0.00 0.29 -0.73 2.64

MRETURN(%) 2,692 0.00 0.00 0.02 -0.07 0.08

LOGSIZE 2,692 14.02 14.30 2.47 7.66 19.15

reflecting an aggregate increase in value. In accordance, LOGSIZE increases over time, since this variable is calculated by multiplying the stock price and the number of outstanding shares. Something else that one may notice is the size of the standard deviations compared to the means of the variables. In the cases of BETA, RETURN and MRETURN the standard deviations are consistently larger than the mean values. For LOGPRICE the standard deviation is more likely to be approximately equal to the mean. In addition, the number of observations differs within the estimation period. While most of the variables have approximately the same number of observations, LOGINDEX has significantly less. This is due to the different databases from which the data is gathered. Whereas Datastream provides the market index values as a variable, CRSP does not provide these numbers. Consequently, the market index data for firms based in the United States cannot be provided. However, CRSP provides the S&P 500 returns directly. The S&P 500 is considered to be the best equity benchmark and economic indicator of the United States market (Bullock & Platt, 2015); therefore these returns are used for the analysis of American stocks in the sample (MacKinlay, 1997). In contrast, Datastream provides values for the market index, which are used in order to calculate the daily market returns. Therefore, these are included in the summary statistics.

Table 2 also describes the summary statistics, although by industry. As mentioned before, N denotes the total number of observations, not the number of firms per category. On the one hand, the sin industry with the largest number of publicly listed firms in the sample is the alcoholic beverages industry with 81 firms. The second largest sin industry is the weapons industry, with 27 firms. The smallest sin industry is the tobacco industry, with 18 firms that meet the conditions to be included in the sample. On the other hand, the largest industry in the control group is the food industry, consisting out of 183 firms. Second is the aerospace industry, which includes 49 firms. The smallest industry in the control group is the non-alcoholic beverages, or soda, industry with 47 firms. The small number of sin stocks

(20)

20

compared to the control group, is not surprising. Hong and Kacperczyk (2009) found that there were only 63 publicly listed sin stocks in 2004. In addition, their sample included 193 distinct sin companies. However, half of their sample consisted of companies in the gaming industry and their sample consists of firms over a period of almost 80 years. Since these gaming stocks are not included in the Fama and French (1997) classification of industries, Hong and Kacperczyk constructed a way in which these stocks could be classified. Unfortunately, this method is not applicable to European stocks, which explains why these stocks are not included in this thesis. Moreover, Salaber (2007) constructs a sample of 158 European sin stocks. However, her sample includes stocks over a period of 32 years, in which a number of these firms has either been delisted, acquired by another company or gone bankrupt.

Looking at Table 2, the industries seem quite comparable in terms of the mentioned variables. One of the larger differences is in the LOGSIZE of the tobacco industry. The average tobacco company in the sample is larger than the average company in one of the other industries. Moreover, the smallest tobacco company in the sample is almost comparable with the average size of the other industries. This may imply that the market is mostly formed by large companies, for example because it does not add value to list a smaller tobacco company or the smaller companies are being acquired by the larger companies. However, this difference in size may also be caused by a general lack of data for smaller listed tobacco companies. Nonetheless, the former may be more likely, as Hong and Kacperczyk (2009) could only find 36 distinct publicly listed tobacco companies over a period of almost 80 years. In addition, some of these companies have been acquired by or merged with other companies in the tobacco industry over time. For example, the Altria Group holds several companies such as Philip Morris, John Middleton and Nat Sherman2.

Furthermore, the average return during the estimation period for the alcohol industry is 4%. The returns for the tobacco and weapons industries are 2% and 8%, respectively. For the control group, the returns are 12% for the soda industry, 2% for the food industry and 4% for the aerospace industry. Also, there is quite a large variation in the stock returns for each industry, which is explained by the standard deviation. For each industry the standard deviation is far larger than the mean return. This shows that there are significant differences

(21)

21

Table 2. Summary Statistics by Industry

Table 2 reports the summary statistics for the variables used by industry. The summary statistics are calculated over the estimation window, i.e. the year 2012. N is the total number of observations. LOGPRICE is the logarithm of the daily closing price of the firms in the sample. LOGINDEX is the logarithm of the daily market index. RETURN is the daily stock return for the firms in the sample. MRETURN is the daily market return. LOGSIZE is the logarithm of the market capitalization of the firms in the sample. BETA is the company beta of the firms in the sample, which reflects the systematic risk.

Alcohol

LOGPRICE 23,252 2.14 2.80 2.62 -4.61 7.67 LOGINDEX 18,913 5.01 5.11 1.49 0.00 8.86 RETURN(%) 23,252 0.04 0.00 0.33 -0.73 2.72 MRETURN(%) 23,252 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 23,252 12.47 11.91 2.81 6.61 18.94 BETA 23,252 0.78 1.00 0.58 -1.33 1.61 Tobacco

LOGPRICE 4,978 2.92 3.42 2.17 -3.51 6.20 LOGINDEX 3,200 6.02 6.32 0.91 3.85 8.02 RETURN(%) 4,978 0.02 0.00 0.23 -0.73 2.72 MRETURN(%) 4,978 0.00 0.00 0.01 -0.07 0.08 LOGSIZE 4,978 15.96 16.58 2.01 12.28 18.89 BETA 4,978 0.65 1.00 0.65 -0.34 1.61 Weapons

LOGPRICE 8,930 2.24 2.35 1.75 -3.91 4.85 LOGINDEX 5,045 4.63 4.66 1.16 0.26 11.91 RETURN(%) 8,930 0.08 0.00 0.51 -0.73 2.72 MRETURN(%) 8,930 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 8,930 13.26 13.58 2.34 8.67 18.14 BETA 8,930 0.51 0.99 0.85 -1.32 1.61 Soda

LOGPRICE 13,021 2.59 2.63 2.11 -3.22 10.53 LOGINDEX 8,783 5.30 5.27 1.31 2.12 9.61 RETURN(%) 13,021 0.12 0.00 0.62 -0.73 2.72 MRETURN(%) 13,021 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 13,021 13.51 13.42 2.51 8.01 19.03 BETA 13,021 0.46 0.97 0.79 -1.32 1.61

(22)

22

Food

LOGPRICE 58,910 1.59 2.03 2.19 -4.61 7.95 LOGINDEX 44,803 4.83 4.82 1.26 -0.11 8.89 RETURN(%) 58,910 0.02 0.00 0.27 -0.73 2.72 MRETURN(%) 58,910 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 58,910 12.91 13.01 2.28 6.99 18.14 BETA 58,910 0.60 1.00 0.74 -1.32 1.61 Aerospace

LOGPRICE 14,815 2.76 3.05 1.73 -4.61 6.80 LOGINDEX 9,046 5.65 5.30 1.68 1.13 11.61 RETURN(%) 14,815 0.04 0.00 0.35 -0.73 2.72 MRETURN(%) 14,815 0.00 0.00 0.02 -0.07 0.08 LOGSIZE 14,815 14.06 14.40 2.48 4.25 18.44 BETA 14,815 0.43 0.27 0.70 -1.32 1.61

in the returns of the companies within these industries. The total risk on stock returns, which is captured by the standard deviation, does not differ significantly. The average daily stock return for the sin stocks during the estimation window is 5%, whereas it is 4% for the control group. The standard deviation is often used to assess the level of risk of a security. The higher the variability of a stock, and thus its standard deviation, the higher the risk. The average standard deviations of the stock returns for the sin industries and the control group are 0.37 and 0.36, respectively. According to these values, the sin stocks and their comparables are quite similar in terms of risk and return. This is in accordance with Salaber (2009), who argues that there are no significant excess returns between sin stocks and their comparables during economic prosperity and these stocks behave quite similarly. Regarding the sin stocks, the most risky industry is the weapons industry with a standard deviation on stock returns of 0.51. Second is the alcohol industry with a standard deviation of 0.33 and the least risky sin stocks on average are from the tobacco industry, with a standard deviation of 0.23. The industries in the control group have quite similar values for the stock return standard deviation. The industry with the highest value is the soda industry with a standard deviation of 0.62. Next are the aerospace industry with a value of 0.35 and, lastly, the food industry with a standard deviation of 0.27. These values, however, only reflect the level of risk of the stock returns.

(23)

23

In contrast, the industry betas explain the systematic risk in the industries of interest. The average beta values for the sin industries are 0.78 for the alcohol industry, 0.65 for the tobacco industry and 0.51 for the weapons industry. Looking at the industries in the control group, first of all, the soda industry has a beta of 0.46. Secondly, the food industry has a beta of 0.60. And, thirdly, the aircraft industry has a beta of 0.43. These results imply that, on average, all industries have a positive relationship with movements in the market. Furthermore, the betas are consistent with the claim that sin stocks do not strongly depend on financial markets (Salaber, 2009). On average, the stock prices in all of the industries move less strong than the market. However, the sin industries have a higher overall beta than the industries in the control group, or at least their comparables. Since the industry beta is the measurement of systematic risk, it seems that the sin industries are overall riskier investments than the industries in the control group. Subsequently, one could say that the alcohol industry is perceived as the most risky and the aircraft industry as the least risky one. If compared with Hong and Kacperczyk (2009), the beta values are overall lower than in their study, except for the alcohol industry. However, these betas are solely focused on stocks that are listed in the United States. If the betas are compared with the beta calculations for Europe by Damodaran (2018), they seem more correct. Overall, the values for beta lie in between the values in the United States and Europe, as calculated by these other studies. However, the weapons and aerospace industry betas are still low compared to these other studies, which can be compared in Appendix A.

5. Findings

As mentioned in the methodology section, the first analysis seeks to find the level of abnormal returns due to signals to the market that are based on information about the IORP II directive. Table 3 shows the results for the abnormal returns and the first test statistic in the four different event periods. The columns on the left side display the values for the control group, whereas the columns on the right side of the table show the values for the sin stocks. For both groups, the test statistics for both the equally weighted and value weighted portfolios have been included. Since the AARs do not differ between these portfolios on a two-decimal scale, only one set of AAR values is included for each event period. In order to support the first hypothesis, the test statistics for the control group should be insignificant, since news about the IORP II directive does not concern these industries. Therefore, the returns should be quite similar to their expected values. The results for the sin stocks should

(24)

24

be negative and statistically significant in order to fully support the hypothesis, since there is an expected negative effect from news concerning divestment in sin stocks by European pension funds.

First of all, the announcement date on which the general plans for the IORP II directive were made public is analyzed. The first event period shows fluctuating abnormal returns for sin stocks over the first event period. The average daily abnormal return for sin stocks during the event period is 0.17%. In contrast, the control group has a negative average daily abnormal return of 0.95% over the entire period. As table 3 shows, the event caused no significant excess returns for the control group. This supports the expectation that the announcement of the IORP II directive did not have any significant effect on the control group. With regards to the sin stocks, the average abnormal returns fluctuate quite a lot on a day to day basis. On the first two days of the event period, -4 and -3, there are positive excess returns of respectively 0.38% and 0.29%. These are significant at the 10% level. However, one day before the announcement speech, there are relatively large negative abnormal returns of 3.05% on average. These negative results are highly significant, namely at the 1% level.

Remarkably, there are no significant abnormal returns on the day of the event itself. This could mean that the event had little effect or that the information may have leaked before the announcement. Moreover, all days after the event date yield positive abnormal returns at either the 5% or 1% level or yield no significant abnormal returns at all. This indicates that the market reaction is rather positive than negative on the days after the announcement. Furthermore, the value weighted portfolio generally yields more significant results than the equally weighted portfolio. Although the test statistics may differ at some points, both portfolios yield similar test results with regards to the sign of the returns. The positive results after the event could originate from a positive surprise rather than negative. This could be due to not presenting any concrete measures yet regarding the ESG factors and their means of enforcement. Also, the market reaction could be due to a high level of uncertainty about the new directive on the day before the announcement. As the plans for the new directive may still be far away or milder than expected, the market readjusts for the negative returns before the announcement date.

Secondly, the second event date marks the day on which the first draft of the IORP II directive was leaked. This draft included more concrete intentions of focusing on ESG factors and the initial means of enforcing the plans in the directive. The second event period shows

(25)

25

Table 3. Average Abnormal Returns and the First Test Statistic

This table shows the average abnormal returns (AARs) and corresponding test statistics for the industries in the sample. SD represents the standard deviation. TS1 is the first test statistic. In the left set of columns, the values for the control group are presented. On the right side, the values for the group of sin industries are presented. Both the equally weighted and value weighted portfolios have been included. Since the AARs did not differ for the equally weighted and value weighted portfolios (on a two-decimal scale), the AARs of only one of the portfolios is provided. The results are significant at the 10 percent (*), 5 percent (**) and 1 percent levels (***).

Date 1

Control Sin

Event date AAR (%) SD

Equally Weighted TS1 Value Weighted TS1 AAR (%) SD Equally Weighted TS1 Value Weighted TS1 -4 -0.26 0.05 -1.08 -1.07 0.38 0.05 1.43 1.66* -3 -0.14 0.05 -0.56 -0.68 0.29 0.06 1.02 1.81* -2 -0.17 0.05 -0.63 -0.47 -0.04 0.06 -0.16 -0.28 -1 -0.36 0.05 -1.41 -1.54 -0.37 0.02 -3.05*** -2.56** 0 0.02 0.06 0.06 -0.01 0.06 0.06 0.20 0.05 1 0.17 0.05 0.62 0.79 0.35 0.06 1.27 2.04** 2 -0.16 0.06 -0.55 -0.62 0.58 0.04 2.76*** 2.83*** 3 0.15 0.05 0.61 0.49 -0.09 0.02 -1.01 -0.47 4 -0.20 0.04 -0.92 -1.19 0.36 0.04 1.74* 2.10** Date 2 Control Sin

Equally Weighted TS1 Value Weighted TS1 AAR (%) SD Equally Weighted TS1 Value Weighted TS1 -4 -0.28 0.05 -1.10 -1.19 -0.02 0.02 -0.20 0.00 -3 0.04 0.05 0.16 0.27 0.29 0.04 1.37 1.81* -2 -0.04 0.05 -0.18 -0.26 -0.24 0.02 -2.99*** -3.13*** -1 -0.18 0.05 -0.76 -0.83 0.72 0.04 3.34*** 3.61*** 0 -0.05 0.05 -0.21 -0.16 -0.15 0.01 -2.92*** -2.75*** 1 -0.09 0.06 -0.28 -0.40 0.81 0.05 3.56*** 4.00*** 2 0.11 0.06 0.38 0.60 0.66 0.04 3.42*** 3.77*** 3 -0.09 0.06 -0.31 -0.49 0.71 0.03 4.40*** 3.65*** 4 0.33 0.04 1.74* 1.80* 0.44 0.04 2.03** 2.14**

(26)

26 Date 3

Control Sin

Equally Weighted TS1 Value Weighted TS1 AAR (%) SD Equally Weighted TS1 Value Weighted TS1 -4 0.28 0.06 0.95 1.05 0.91 0.05 3.59*** 3.29*** -3 -0.31 0.06 -1.12 -1.36 0.05 0.04 0.25 0.85 -2 -0.03 0.07 -0.09 -0.20 0.58 0.04 2.71*** 2.98*** -1 -0.23 0.06 -0.79 -0.74 0.45 0.02 4.15*** 3.78*** 0 -0.25 0.06 -0.91 -0.67 0.28 0.04 1.34 1.70* 1 0.17 0.07 0.53 0.01 0.46 0.04 2.09** 2.45** 2 -0.50 0.07 -1.50 -1.23 0.18 0.02 2.14** 2.22** 3 -0.18 0.06 -0.60 -0.78 0.23 0.04 1.09 1.52 4 0.06 0.06 0.20 0.31 0.06 0.04 0.30 0.97 Date 4 Control Sin

Equally Weighted TS1 Value Weighted TS1 AAR (%) SD Equally Weighted TS1 Value Weighted TS1 -4 0.37 0.07 1.02 0.60 -0.07 0.02 -0.88 -1.07 -3 -0.16 0.05 -0.66 -0.61 0.26 0.05 1.06 1.76* -2 0.15 0.04 0.74 0.75 0.30 0.04 1.43 2.06** -1 -0.18 0.06 -0.60 -0.54 0.11 0.02 1.45 0.78 0 -0.24 0.06 -0.76 -0.66 0.55 0.04 2.44** 2.88** 1 0.09 0.04 0.43 0.51 0.11 0.05 0.44 1.13 2 -0.49 0.06 -1.53 -1.35 0.00 0.01 -0.08 -0.13 3 -0.33 0.06 -1.16 -1.31 0.31 0.04 1.36 1.69* 4 0.23 0.05 0.84 1.11 0.67 0.05 2.89*** 3.10***

that again there are no significant abnormal returns for the control group and almost exclusively highly significant results for the sin industries. The average daily AAR for the control group is -0.03%, whereas the sin stocks yield a 0.31% positive AAR on daily average. Similar to the first event period, there is no evidence that the average abnormal returns for the control group are statistically different from zero. This supports the expectations that the control group is not affected by news about the new directive. The abnormal returns for sin stocks are fluctuating before the event date and turn positive after day 0. On day -2, the sin stocks make an average abnormal return of -0.24%, whereas the return one day later is 0.72%. On the event date itself, sin stocks make a return of -0.15% again, which supports the first hypothesis. However, the days after the event date, sin stocks yield average abnormal returns that are relatively high and significant at the 1% level (5% for day 4). For both the

(27)

27

equally weighted and value weighted portfolios, the results are almost equally significant. These results show that although the initial market reaction to the leaked draft was negative, the reaction of the market in the following days was not according to expectations. Following these results, the only evidence that supports the first hypothesis is found on the exact event date. In contrast, the surrounding days seem to support an exactly opposite reaction to the leaked draft. One of the reasons for this reaction may be found in the process of establishing the new directive. Since the plans in the draft would not be executed for at least a few years, the initial reaction to the news might be negative. However, as the example of ABP’s top 100 investments over 2017 proves, pension funds at least remained with their investments on the short-term. This implicates that the divestment in the sin industries may be too far away to affect the stock prices for a longer period.

Thirdly, the third event period was designed around the agreement upon a new draft version of the IORP II directive. This was the first draft after the initial leaked draft in 2014. The average daily AAR for the control group is -0.11%. However, the third analysis also does not yield any significant results for the control group. As mentioned before, this supports the expectations for the first hypothesis. The average daily AAR for the sin stock group is 0.36%. For the third event window, the sin stocks score positive abnormal returns on every day. The majority of these days is statistically significant. Furthermore, the average abnormal returns on day -2 and -1 are respectively 0.58% and 0.45%, both significant at the 1% level. Contrastingly, the day of agreement on the new draft yields no significant excess returns. Although this shows that there were no excess positive returns on the event day, it does provide any evidence to support the first hypothesis. Moreover, the excess returns of the two days after the event are 0.46% and 0.18%, which are statistically significant at the 5% level. Both the equally weighted and the value weighted portfolios yield fairly similar results for the third event period. Quite similar to the second event date, these results provide more evidence against the hypothesis rather than to supporting it.

Finally, the fourth event date has been determined to be the date on which the final document was leaked. This document described the exact rules with regards to the announced ESG factors. In addition, it described the options within the power of European member states to enforce the new rules, in case a pension fund would not adhere to the new investment rules. The most important news on this date was the prudent person rule. The average daily AAR for the control group is -0.06%. However, none of the results for the control group are statistically significant. Again, this is according to expectations. The

(28)

28

average daily AAR for the sin stocks is 0.25%. The most noticeable result here is the average abnormal return on day 0. The sin group has positive excess returns of 0.55% at a 5% significance level. This implies that the market reacted positively on the news of the leaked document. Besides day 0, the equally weighted portfolio only yields a significant result for day 4, where the sin stocks have an abnormal average return of 0.55%. The value weighted portfolio, however, also provides significant results for day -3 and -2. On day -3, the sin stocks have an abnormal return of 0.26% at a 10% significance level. Two days before leaking the final document, the sin stocks have an excess return of 0.30% at a 5% significance level. Also for the fourth event period, there seems to be no evidence to support the first hypothesis.

In order to analyze the abnormal returns for the entire periods, a second test statistic has to be calculated. Table 4 illustrates the cumulative average abnormal returns (CAAR). The CAAR is the sum of average abnormal returns in one period. The values for the equally weighted and value weighted portfolios are virtually the same. The CAAR for the first event period is 0.19%. For the second event period, the CAAR is 0.37%. For the third and fourth event periods, the CAAR is 0.40% and 0.26%, respectively. Since none of the test statistics in table 4 is statistically significant, the CAAR is not considered to be statistically different from zero. Based on these results, there is not sufficient evidence to reject the null hypothesis. Therefore, the first hypothesis cannot be accepted. However, there is a possibility that the results are driven by a single industry, since the difference in size of each industry in the

Table 4. Cumulative Average Abnormal Returns

This table shows the cumulative average abnormal returns (CAAR) of the sin stocks. The CAAR is the sum of all AARs during one period. N is the number of firms. SD explains the standard deviation. TS2 is the second test statistic, as described in the methodology. On the left side are the results for the equally weighted portfolio. On the right side are the results for the value weighted portfolio. The results are significant at the 10 percent (*), 5 percent (**) and 1 percent (***) levels.

Equally Weighted Value Weighted

Time

Period N CAAR (%) SD TS2 CAAR (%) SD TS2

Date 1 115 0.19 0.03 0.69 0.19 0.03 0.69

Date 2 100 0.37 0.02 1.55 0.37 0.02 1.56

Date 3 100 0.40 0.03 1.20 0.40 0.03 1.20

The effects of socially responsible investment on sin stocks : a multiple event study on pension fund regulation