Master thesis
The outperformance of sin stocks
A study on the abnormal returns of stocks in the global alcohol, tobacco, gambling and weapon industry.
By Nick de Keijser
Master of Finance
Faculty of Economics and Business University of Groningen
Abstract: In this thesis I investigate the returns of a self-constructed global sin stock portfolio of 160 publicly traded companies in the alcohol, tobacco, gam- bling and defense industry between 1996 – 2016. I find evidence for a significant risk-adjusted outperformance of sin stocks performing single and multi-factor time-series regression analysis. Additionally, I find even higher abnormal returns using the time period from 2007. Lastly, I conclude that sin stocks outperform a sample of comparables.
JEL classification: G15, G19, M14.
Keywords: Sin stocks, multi-regression analysis, socially responsible investing.
Author: Nick de Keijser Student number: s1686887 Supervisor: Dr. G.T.J. Zwart
Contents
1 Introduction 4
2 Literature review 6
2.1 Socially Responsible Investing . . . . 6
2.2 The history of sin . . . . 7
2.3 Sin stocks returns . . . . 8
2.4 Hypothesis . . . . 11
3 Data 12 3.1 Sin stock selection . . . . 12
3.2 Descriptive Statistics . . . . 14
4 Methodology 15 4.1 CAPM and Multi-Factor Models . . . . 15
4.2 Analysis . . . . 17
4.2.1 Weighting methods of returns . . . . 17
4.2.2 Different subsamples . . . . 18
5 Results 19 5.1 GlobalSin vs. Risk-free rate . . . . 19
5.2 GlobalSin vs. Comparables . . . . 20
5.3 Regression results from 2007 to 2016 . . . . 22
6 Robustness checks 24 6.1 GlobalSin vs. Region . . . . 24
6.2 GlobalSin vs. Industries . . . . 24
6.3 ViceX Fund vs. GlobalSin . . . . 25
6.4 Analysis based on equally-weighted returns . . . . 25
7 Conclusion 26
8 Discussion 27
9 References 28
Appendix A The complete sin companies list 30
Appendix B Country specific list of all the sin stocks 31
Appendix C GlobalSin vs. Region 32
Appendix D GlobalSin vs. Industries 33
Appendix E ViceX Fund vs. GlobalSin 34
Appendix F Regression results based on equally-weighted returns 35
1 Introduction
In June 2015 a remarkable article in the Dutch newspaper Volkskrant1 was released. It referred to the investigation of Eerlijke Verzekeringswijzer in collaboration with Volkskrant showing that as many as eight of the ten largest insurance companies in the Netherlands (Aegon, Delta Lloyd and Nationale Nederlanden to name a few) and several major Dutch pension funds, such as ABP, invest billions of euros in companies who are involved in arms production and defense. The same research shows that 80% of the Dutch population find this a reason to switch to another insurer.
Back in 2007, the TV program Zembla showed that many companies invested in stocks of companies in the weapon industry. These so-called conventional shares were being sold in result of the social pressure that arose. The idea is that social pressure is increasing making these shares becoming less socially acceptable, however, eight years later large companies are still investing in these sin stocks. Hence, large investors are very obstinate.
The stocks of companies involved with activities that are considered unethical are labeled as sin stocks. For that reason, according to Renneboog et al. (2008) and Hong and Kasperczyk (2009), sin sectors include alcohol, tobacco and gambling. Sin stocks are rejected by most of society because they play on human weaknesses and vice. Sin investing is the polar opposite of socially responsible investing (SRI), where in addition to achieving investment returns, social aspects also play a role. Therefore, sin stocks are often avoided by socially responsible investors.
Statman (2000) states that because large SRI do not buy sin stocks, the companies that own these stocks experience a higher cost of capital. For an individual investor, this means that the return of sin stocks is higher than unconventional stocks because the expected return is the cost of capital on an investor level. Hong and Kasperczyk (2009) cite other possible reasons for higher returns for sin stocks. The litigation risk is higher for sin companies and risk sharing has its limitations.
There are a number of reasons why my research is relevant and adds to the existing literature.
Although there has been research done into the outperformance of sin stocks, it is often done with data only dating to 2007. In this research, I used data from November 1996 to August 2016. In conjunction with the data from 1996-2016 I separately analyzed the period between 2007-2016. The financial crisis started in 2007, but also social pressure on large investors is growing (as mentioned before). Consequently, it is interesting whether or not sin stocks outperform during this specific period. I will investigate if my self-constructed value-weighted sin portfolio experience monthly abnormal returns net of the risk-free rate and to a set of comparable stocks .
Other studies on this same subject, have used a sample of only US stocks (Hong and Kasperczyk; 2009) or only European stocks (Salaber; 2007). I have created a global portfolio in which my sample contains companies from 47 different countries across six continents. I will perform single and multi-factor time-series regressions. The main research question is:
Do sin stocks outperform using a global portfolio, and do sin stock returns differ between regions and conventional industries?
The rest of the thesis is constructed as follows. In Section 2, I will discuss previous literature about SRI, the history of sin and sin stock returns. In section 3, I will explain the data used. Section 4 describes the methodology and in section 5, I will present the empirical results. Finally, the discussion and conclusion completes my thesis in section 6.
2 Literature review
2.1 Socially Responsible Investing
Socially responsible investing (SRI) is an investment behavior that considers environmental, social and corporate governance (ESG) criteria to examine both social and long-term financial goals. Hence, investments in companies that fit the social norms of the fund. Most of the SRI funds avoid investing in stocks of companies that are active in the alcohol, tobacco and gambling industry.
In the past 15 years, socially responsible investing (SRI) has become more and more a broad-based investment approach. A recent report of the US SIF Foundation2 from December 2015 shows that more than one out of every five dollars under professional management in the United States was invested in line socially responsible investing, $8.7 trillion in total. That is a 33% growth over the previous two years, and a 14-fold increase since 1995. This is in line with the findings of Ballestero et al. (2012), who state that mutual funds go further than just corporate social responsibility (CSR) behavior. When looking for investment opportunities, social and ethical interests play an increasingly important role in the investment decision.
Several studies investigate the performance of social investing. Many of them conclude that SRI stocks indices do not outperform conventional stocks. Bauer et al. (2005) claim that the level of social responsibility does not have significant influence on stock performance.
They compare two different portfolios, one with conventional stocks and one with SRI funds, using the Carhart (1997) four-factor model and found no evidence for abnormal returns.
Schröder (2005) focuses on socially responsible investing indices instead of SRI funds.
With CAPM and the Fama and French (1993) three-factor and four-factor model he test the performance. As results, he found no evidence for an outperformance of SRI indices.
2http://www.ussif.org/content.asp?contentid=40
2.2 The history of sin
‘Mede’ made of honey is thought to be the oldest alcoholic drink that we know of. It was already consumed 8,000 years ago. It was discovered by accident from honey and water colliding with yeast in a warm place. Beer had already been known for 5400 years BC. In Egypt, papyrus scrolls were found in 3500 BC with the recipe for wine. There are indications that the Chinese where already familiar with distillation at that time along with the Greeks and Romans who knew how to create strong liquor from wine, however, the invention of distillation is attributed to the Arabs.
Germany had mastered the brewing process creating the consumption to quickly rise due to a constant lack of clean drinking water. Beer and wine was much safer than water until the 18th century . In fact, alcohol consumption was considered a basic need because of this.
That changed from the moment clean drinking water became more available. The need for drinking alcohol was less in Asia and the Middle East supporting Islamic beliefs in successfully prohibiting alcohol.
In the West around 1800 drinking alcohol was no longer seen as a basic need. In particular, distilled liquor was considered to be dangerous and unhealthy. This realization led to increased measures against alcohol.The United States was drained between 1920 and 1933. The conse- quence of the prohibition was that when liquor was illegally fired, it resulted in the death of 35,000 people. In 1930, one million people were arrested for violating the liquor ban. After the Great Depression, the prohibition was unsustainable. The money from the excise tax on alcohol was needed to build up the economy.
In the 20th century, the consideration of alcohol being a human vice became strengthened by the detection that consumption of alcohol has many negative health effects (Worman, 1995). Nowadays, an estimated 15 million Americans suffer from alcoholism and 40% of all car accident deaths in the US involve alcohol.
When Columbus discovered America in 1492, he saw the natives used tobacco to smoke.
In 1518, tobacco plants were brought from America to Europe. French ambassador Jean Nicot saw tobacco as a drug, the addictive substance in tobacco (nicotine) was named after him.
Smoking became fully established after the First World War and became a general habit, it was allowed everywhere. The US military was one of the most strongly influenced, 95% of the soldiers smoked.
The fifties and sixties were the heyday of the tobacco industry and despite the popularity, tmore and more stories about the health risks of smoking arose. In Europe around 1970, warnings about the harmful effects of smoking were required on all cigarette packs. Positive commercials advertising the use of tobacco became increasingly banned. The image of smoking has been changed from a drug into a harmful stimulant, causing the tobacco industry to be seen broadly as sinful.
For many decades, gambling has been experienced as sinful. The main reason for this being that it is addictive. In further research , gambling is often related to criminal activities.
Worldwide, the gambling sector is strongly regulated. In the last ten years, there is a pattern of deregulation and gambling continues to become more socially accepted, however, it is still often seen as a sinful business. In the 20th century there is a growing opposition to war and military intervention. The arms industry is therefore considered as sinful.
2.3 Sin stocks returns
According to the theoretical-framework of Merton (1987) there are two reasons for an outper- formance of sin stocks in comparison to “normal” stocks. Firstly, sin stocks being neglected by large norm-constrained investors (pension funds). He states that these neglected stocks are less often studied by professional analysts. Hence, the accessible information on certain stocks is poor. The decreased demand of sin stocks results in stock prices below their fundamental value. This means that the expected returns of sin stocks are higher than industry-comparable companies. Secondly, the study claims that a higher litigation risk of sin companies results in an increase of the expected returns.
Sin stocks experience more litigation risk than other stocks. Litigation risk is the assessment of the possibility that legal actions may be taken against the company. The risk department of a company will estimate the chance of a litigation. Besides that, the possible litigation risk of a single transaction may be considered. Therefore, a sin industry needs more support from legal experts, which logically involves extra costs. These costs are especially higher for for the tobacco and alcohol-industries. In short, it is likely that litigation is related to a certain risk premium, a form of compensation that sin-investors take for tolerating the extra risk.
Hong and Kacperczyk (2009) used time-series regressions to investigate the returns of their sin portfolio in comparison to other (non-sin) stocks for the period of 1965-2006. They created a zero-investment portfolio with long US sin stocks and short in their comparables and found a premium of 0.26% per month by using the Fama and French (1993) three-factor model and Carhart’s (1997) additional momentum factor. After controlling for several firm characteristics and using cross-sectional regressions, they conclude that sin stocks significantly outperform their comparables by a monthly 0.29%. Hong and Kacperczyk (2009) also checked the significance of the results excluding the tobacco sector, since the high returns may be caused by the litigation risk of the tobacco sector. However, sin stocks, even without the tobacco stocks, still significantly outperform their comparables with a monthly 0.21%. The main explanation for the findings of Hong and Kacperczyk is that the controversial stocks are neglected by a large group of investors under social pressure. The limited risk sharing between sin investors results in sin stocks prices below their fundamental value. Durand, Koh and Limkriangkrai (2013) confirm the results of Hong and Kacperczyk using a sample of US sin stocks between 1990 and 2008 founding a four-factor alpha of 0.31%.
The sin stocks portfolio of Fabozzi, Ma and Oliphant (2008) outperformed the market, receiving a monthly abnormal return of 1.64% (19% annually). They tested a sample of 267 companies in 21 countries from six different industries between 1970 and 2007. Besides the known industries (alcohol, tobacco and gambling) they studied the defense, biotech and adult-entertainment industry. There are two main reasons for the outperformance according Fabozzi et al.(2008). Firstly, the initial price at launch (IPO) is lower than their fundamental value because of the bad image of these companies, and secondly, there are implicit and
explicit costs associated to satisfy certain social norms and sin stocks do not face such costs.
In comparison to other studies, the abnormal returns they found are very high. This could possibly be explained by their different interpretation of sin industries (they include the biotech and the sex/adult entertainment industries) and sample period.
The study of Salaber (2013) is based on a portfolio of 158 European sin stocks between 1975 and 2006. She confirms the outperformance (monthly abnormal return of 0.63%) of sin stocks, which is in line with Hong and Kacperczyk (2009). Moreover, there are cultural and tax-based explanations on sin stock outperformance. The outperformance of sin stocks in Protestant countries is higher than in Catholic countries. Additionally with it being higher when a company is located in a country with higher excise-tax.
Kim and Venkatachalam (2011) are looking for explanations of the higher excess returns of sin stocks. In contrast to other studies, they found that sin stocks have a high qualitatively financial reporting. Besides that, (non-financial) social norms may be an influence in the creation of higher returns of sin stocks.
A few analysts have researched the performance of the ViceX Fund, a fund that only invests in stocks of sin industries like alcohol, tobacco, gambling, defense and aerospace. The theory behind the fund is that these companies are less influenced by market conditions, and therefore can perform better in uncertain economic times. In section 6.3, as additional analysis, I compare my self-constructed sin portfolio with the ViceX Fund.
Chong et al. (2006) studied the performance of the Vice Fund in comparison to the DSE Fund (a SRI fund). They performed a single-factor regression using daily stock return data between 2002-2005 concluding a monthly outperformance of the ViceX fund of 0.72%.
Although, there is also criticism about this research. According to Hoepner and Zeume (2009), the time period that is used is too short. Besides that, there is a strong tilt towards tobacco and small cap stocks. After controlling for these factors, there is no observable significant outperformance.
2.4 Hypothesis
In line with the research of Hong and Kacperzcyk (2009) and several other related studies, I have compiled my first hypothesis:
H1: sin stocks show an outperformance using a global sample between 1996 - 2016.
I will test this hypothesis by performing time-series regression analyses based on single- factor capital asset pricing model (CAPM), the Fama and French (1993) three-factor model and the Carhart (1997) four-factor model. Using value-weighted portfolio returns (between November 1996 and August 2016) net over the risk-free rate and a portfolio long in sin stocks and short in a set of comparable companies to, I will investigate if they show a outperformance.
Although there have been numerous studies on sin stock outperformance, many of them have only used data samples until 2007. It is also interesting to see whether the subsequent period (2007-2016) affect the possible outperformance of sin stocks. Secondly, August 2007 is considered as the beginning of the financial crisis. This has had a huge impact on the global equity markets. In my view, it is valuable to investigate whether sin stocks outperform in the years since the financial crisis. Finally, given the growth of socially responsible investing in recent years, investors have become more aware of the origin (conventional or unconventional) of the shares they buy. Therefore, my second hypothesis is:
H1: sin stocks show an outperformance using a global sample between 2007 - 2016.
Finally, I perform the same regressions for different subsamples. The total sin portfolio has been divided into different smaller parts, based on region and sin industry.
3 Data
3.1 Sin stock selection
The selection process of the sin portfolio starts with the Standard Industry Code (SIC) as used in Fama and French (1997) that contains 48 different industries. The alcohol industry belongs to group 4 and includes SIC codes 2080 to 2085. Group 5 is the tobacco industry having SIC codes 2100 – 2199. For the weapon and defense industry SIC codes 3480-3489, 3760-3769 and 3795 are applicable. The gambling industry does not have unique SIC codes. Therefore, I will follow the study of Hong and Kasperczyk (2009) and use NAICS codes 7132, 71312, 713210, 71329, 713290, 72112, and 721120.
The ORBIS database is used to compose the company list. The complete company list can be found in Appendix A. I have chosen to use a twenty-year timespan between November 1996 and August 2016, since the financial company data (mainly in Asia) before 1996 is limited.
For the multi factor regressions, the data from Kenneth French’s website is used. Important in defining the final portfolio, a strong tilt towards a certain region or microcap (companies with a market capitalization of less than $300 million) should be avoided. I created a global sin portfolio, which can be divided into several regions (Table 1). These regions are in line with the available data from Kenneth French’s website3. A country specific list of the sin stocks per region can be found in Appendix B.
Table 1: Division of sin stocks across regions
Region Number of sin stocks
North America 29
Europe 44
Asia Pacific ex. Japan 60
Japan 7
Rest of the world 20 Global portfolio 160
Different industries
Figure 1: Number of companies in the different sin-industries
This global portfolio of 160 sin stocks contains 107 companies from the alcohol industry, 33 tobacco-related stocks, 14 from the weapon and defense industry and six from the gambling industry, as shown in Figure 1. 160 is the total number of companies that have been used in my study. Some companies did not exist 20 years ago, others may have gone bankrupt or being acquired during these 20 years. Hence, the average number of companies each year is 148, with a minimum of 130. The mean market capitalization of companies in the sample is
$2.1 billion. The largest company has a market value of $43 billion. As mentioned before, a strong tilt towards micro-cap must be prevented. Therefore, have I set the lower limit at $225 million, which is only 25% under the small-cap barrier.
I will follow the study of Salaber (2013) for the calculation of the returns using the monthly Thomson Reuters DataStream Total Return Index (RI). The end-of-the-month RI includes the re-investment of dividends. As follows, the returns can be calculated as natural logarithm using:
Ri,t =LN(RIt/RIt−1).
The risk-free rate for the global portfolio that will be used is the US the three-months Treasury bill rate, which is in line with other studies (e.g. Lobe and Walkshäusl, 2011).
Section 4.2.2 explains the different risk-free rates that are used when the sample is divided into different regions.
3.2 Descriptive Statistics
The descriptive statistics are shown in Table 2, with the mean, median and standard deviation of the global sample and the different regions. The sin stocks for all sample countries together earn on average a monthly return of 1.02%. When taking a look at the sin stock regions separately, almost all regions show an average monthly return above 1.0%.
Table 2: Descriptive Statistics: monthly sin stock returns
#Months #Comp #Obs Mean Median St. Dev
Global 236 160 37760 0.0102 0.0091 0.1227
North America 236 29 6844 0.0103 0.0087 0.1259
Asia Pacific ex. Japan 236 60 14160 0.0089 0.0117 0.0492
Europe 236 44 10384 0.0102 0.0087 0.1256
Japan 236 7 1652 0.0104 0.0096 0.0824
4 Methodology
4.1 CAPM and Multi-Factor Models
The Capital Asset Pricing Model (CAPM) is a financial investment theory to determine return requirements. It is composed out of a so-called risk-free rate and a risk premium which contains the market risk. The CAPM is based on the modern portfolio theory (MPT) of Markowitz (1952). The MPT assumes that there is - under ideal market conditions - a link between the expected return and level of risk that is taken. Hence, higher returns can only be achieved by accepting greater risks. William Sharpe (1964), along with Lintner (1965), adjusted the model of Markowitz in order to make a more practical applicable model. Their CAPM states that there are two kind of risks, systematic and specific (non-systematic) risk.
Systematic risk is unavoidable for an investment opportunity, but specific risk can be eliminated by diversification. The CAPM is represented as:
Rp,t−Rf,t= α+ βm(Rm,t − Rf,t)+ εp
where Rp,t−Rf,t stands for the excess return of the portfolio for time period t, α stands for Jensen’s alpha, βmstands for the market beta for portfolio p, (Rm,t− Rf,t) stands for the market premium of a value-weighted portfolio at time t and εp stands for the error term for portfolio p. The alpha-coefficient is the measure for out- or underperformance for a portfolio or single stock. The beta of the model βm used to analyzerelative risk, is a measure of the market risk exposure of an index. The underlying beta of the market is always one. For a beta higher than one, the index has a higher exposure to risk compared to the benchmark. A beta coefficient is lower than one indicates a lower risk in comparison to the benchmark. Although the Capital Asset Pricing Model is a commonly used method, it has its limitations. Various studies have shown the lack of accuracy given the differences between the actual realized returns and the CAPM predictions. A famous model that is often used nowadays is the three-factor model of Fama and French (1993). The three factors correspond to the market index, the firm size and the ratio book-to-market value. Empirical studies have shown that stocks with high ratios of book-to-market value and stocks of relatively small firms generate a higher return than predicted by the CAPM. According to Bodie et al. (2008) this means that the CAPM does not
account for these influences.
According to Banz (1981), small stocks earn higher risk-adjusted returns than large stocks.
This phenomenon, called the size effect, was later confirmed by several other studies. Fama and French measure the firm size factor by the difference in returns between the one-third smallest firms and the one-third largest largest firms (based on market capitalization) for every period and is general known as the SMB factor ("Small-Minus-Big"). The HML factor ("High- Minus-Low"), often called the value premium, accounts for the difference between the returns of value stocks (stocks of companies with high book-to-market values) and growth stocks (low book-to-market values). The Fama-French three factor model is represented as:
Rp,t−Rf,t = α+ βm(Rm,t− Rf,t)+ βS M BRS M B,t+ βH M LRH M L,t + εp
In addition to the earlier explained parameters, the α stands now for the three-factor alpha, βp,SM Bstands for the beta with regard to size for stock i, RS M B,tstands for the return difference between a small cap portfolio and a large cap portfolio at time t, βp,H M L stands for the return difference between a value and growth portfolio at time t, RH M L,t is the value premium.
Jegadeesh and Titman (1993) discovered the momentum abnormality. They found out that abnormal returns can be earned by looking at the previous 12 months performance of stocks.
Hence, taking short positions in bad performing stocks or long positions in well performing stocks results in an outperformance. Carhart (1997) created the momentum factor, which shows the trends exposure of a stock. When this factor is positive, it indicates the outperformance of stocks with high prior returns. The momentum factor is explained as the portfolio composed of 30 percent of the stocks with the highest return over the last year minus the return of the 30 percent of stocks with the lowest previous 12 months return. The Carhart four-factor model is represented as:
Rp,t−Rf,t= α+ βm(Rm,t − Rf,t)+ βS M BRS M B,t + βH M LRH M L,t+ βMO MRMO M,t + εi
where the additional βi,MOM stands for the momentum beta, RMO M,t stands for the mo- mentum premium. In the Carhart (1997) four-factor model, the alpha coefficient gives that part of the index returns that cannot be explained by the market performance, size, value or
4.2 Analysis
In this research, analysis will be performed with the Ordinary Least Squares (OLS) estimator for obtaining the parameters. Furthermore, the Newey-West (1987) approach will be used for correcting the residuals for autocorrelation and heteroscedasticity by adjusting the standard errors. This is in line with related studies, like Hong and Kasperczyk (2009) and Salaber (2013).
4.2.1 Weighting methods of returns
There are two possible methods for weighting the portfolio returns, namely equally-weighted and value-weighted. An equally-weighted portfolio weights each sin stock return equally regardless of its market capitalization. The method gives a more balanced insight of the expected returns for a sin investor who does not prefer to determine the asset allocation to the size of a specific (sin) company. In addition, the calculations are much easier. The equally- weighted method is often used in several studies (e.g. Hong and Kasperczyk, 2009; Fabozzi et al., 2008).
On the other hand, value-weighted portfolio returns are valued based on their market capitalization relative to the overall market capitalization of the sin portfolio. For a (sin) investor, the value-weighted average returns method is more precisely in showing the total wealth effects (Fama, 1998). A portfolio composed using value-weighting show a more realistic picture of investing in sin stocks. Investors are generally more inclined to invest in large-caps instead of small caps and moreover biased by investing more in well-known stocks.
Lastly, the market portfolio parameter I use is also value-weighted. Therefore, I commit to the value-weighted average returns method for my analysis. As a robustness check, I will also perform an analysis based on equally-weighted returns of my global sin portfolio. The regression results using this equally-weighted method are presented in section 6.4.
4.2.2 Different subsamples
Besides looking only at the performance of the global portfolio, I have made several subsamples.
The global portfolio is divided into the different regions and can also be split into different sin industries. Furthermore, for testing my second hypothesis I have used a time period from 2007 to 2016. The results are shown in section 5.3.
The global portfolio can be divided into different regions. The regions that I use are North America, Europe, Asia and Pacific without Japan and a subsample with only Japanese sin companies. This categorization is in line with the regional distribution of Kenneth French.
Therefore, I can use precise factor values for my four-factor regression. For each region, I have used different risk-free rates. For example, using an US three-month T-bill rate for a portfolio with only Japanese sin companies would not give proper results. Hence, the risk-free rates that are used for North America, Europe, Asia Pacific and Japan are respectively the US one-month T-bill4, the three-months Euribor5, Hong Kong three-month interbank rate and the Japanese Yen LIBOR three months interbank rate. The four-factor regression results of the different regions are discussed in section 6.1.
Finally, I made a zero-investment portfolio with long the monthly ViceX Fund returns and short my GlobalSin portfolio. This investigates whether or not the fund has an added value, since investors must pay a management fee. The ViceX fund was established and tradeble since August 2002. Hence, I used a 14 year time span (between August 2002 and August 2016).The result of the four-factor regression is shown in section 6.3.
5 Results
5.1 GlobalSin vs. Risk-free rate
Table 3 shows the results of the CAPM, three-factor model and four-factor model of the time-series return regression of the value-weighted global sin portfolio of 160 companies (GlobalSin) net of the risk-free rate. It is clear that all three models earn significantly higher returns. To start with the CAPM results, the value-weighted global sin portfolio earns a 0.54%
higher monthly return than predicted by the CAPM between November 1996 - August 2016 (significance level 1%). This alpha is in line with the study of Salaber (2013), but lower than the sin portfolio of Fabizzo et al. (2008). A beta of 0.5150 means that the GlobalSin portfolio has a lower sensitivity towards the volatility of the market portfolio. This beta below one corresponds to the findings and calculations of Hong and Kacperczyk (2009). They found for several sin industries betas below one.
The results of the GlobalSin regression of the three-factor model give similar outcomes after controlling for size (SMB) and book-to-market (SMB). The value-weighted portfolios earn a monthly 0.42% higher return between November 1996 - August 2016, at a significance level of one percent. The SMB and HML factor are both significant, respectively at a one and a five percent level.
On the other hand, the four-factor alpha shows an outperformance of 0,38%, at a significance level of five percent . The explanatory power of the additional momentum is not that large, since it is not statistically significant. However, I will use the four-factor model as main model in order to compare between different region and industry subsamples, which is in line with related studies.
Table 3: Time-series return regressions (net of risk-free rate) between 1996-2016
GlobalSin vs. Risk-free α βM KT βS M B βH M L βMO M
GlobalSin - Rf ,t 0.0054*** 0.5150***
(0.0016) (0.0351)
GlobalSin - Rf ,t 0.0042*** 0.5396*** -0.1418** 0.3018***
(0.0015) (0.0333) (0.0717) (0.0628)
GlobalSin - Rf ,t 0.0038** 0.5569*** -0.1603** 0.3260*** 0.0573 (0.0016) (0.0351) (0.0726) (0.0646) (0.0376) NOTE:This table shows the time-series regression results obtained from regressing the excess
monthly returns of the value-weighted global sin stocks (alcohol, tobacco, defense and gambling) portfolio between 1996-2016 by using OLS estimation on the CAPM, Fama and French (1993) three factors and Carhart (1997) four factors. The risk-free rate is the US three-months T-bill. The intercept (alpha) shows the outperformance. βM KT is the parameter on the market risk premium. βS M B is the return of a portfolio long small stocks and short large stocks, βH M Lis the return of a portfolio long high book-to-market stocks and short low book-to-market stocks and βMO M is the return of a portfolio long past 12-month return winners and short past 12-month return losers. In brackets are shown the standard errors which are adjusted for serial correlation using the Newey-West (1987) correction.
*** significance level of 1% . ** significance level of 5%. * significance level of 10%.
5.2 GlobalSin vs. Comparables
Most of the studies, like Fabrizzo et al. (2007), related to this specific subject calculated the abnormal (out)performances of a sin portfolio against the risk-free rate or market as bench- mark. Contrary to these studies, Hong and Kacperczyk (2009) calculated the performance benchmarked against a set of comparable stocks. They used stocks that belong to the Fama and French (1997) industry groups 2 (food), 3 (soda), 7 (fun) and 43 (meals and hotels).
I created a similar portfolio of comparables for my analysis with data from November 1996 - August 2016. I tried to create a portfolio as comparable as possible formed from the aforementioned industries. My comparables sample consists of 150 companies and the mean, minimum and maximum market capitalization are respectively $2.4 billion, $235 million and
$39 billion. The interesting thing about this method is that it can be checked whether or not the selected risk-free rate that is incorrect (too low). When the GlobalSin - Comparables portfolio would not outperform, the risk-free rate that is chosen for the main analysis might be too low.
Table 4 shows the CAPM, three-factor and four-factor regression results of the so-called zero-investment portfolio, which is long in my self-constructed GlobalSin (the monthly return
and short in Comparables (the monthly return for a value-weighted portfolio of comparable stocks). The monthly value-weighted alpha of the CAPM, three-factor and four-factor model is respectively 0.27%, 0.33% and 0.28%, all significant at a five percent level. Looking at the results of Hong and Kacperczyk (2009), there are many similarities. They experience alphas of 0.25% (CAPM), 0.26% (three-factor) and 0.26% (four-factor) between 1965 - 2006.
Table 4: Time-series return regressions (net of comparables-portfolio) between 1996-2016
GlobalSin vs. Comparables α βM KT βS M B βH M L βMO M
GlobalSin - Comparables 0.0027** -0.3766***
(0.0027) (0.0587)
GlobalSin - Comparables 0.0033** -0.3674*** -0.7258*** 0.0274 (0.0025) (0.0552) (0.1187) (0.1039)
GlobalSin - Comparables 0.0028** -0.2741*** -0.8264*** 0.1044 0.3109***
(0.0024) (0.0552) (0.1140) (0.1016) (0.0591) NOTE:This table shows the measurement results obtained from the time-series regressions of a global sin
portfolio (GlobalSin - Comparables) that is long in GlobalSin (the monthly return for value-weighted global portfolio of sin stocks— alcohol, tobacco, gabling and defense) and short in Comparables (the monthly return for an value-weighted portfolio of comparable stocks) between 1996-2016 using OLS estimation on the CAPM, Fama and French three factors (1993) and Carhart (1997) four factors. The intercept (alpha) shows the
outperformance. βM KT is the parameter on the market risk premium. βS M Bis the return of a portfolio long small stocks and short large stocks, βH M Lis the return of a portfolio long high book-to-market stocks and short low book-to-market stocks and βMO Mis the return of a portfolio long past 12-month return winners and short past 12-month return losers. In brackets are shown the standard errors which are adjusted for serial correlation using the Newey-West (1987) correction.
*** significance level of 1% . ** significance level of 5%. * significance level of 10%.
5.3 Regression results from 2007 to 2016
Besides my main analysis, there are a number of reasons for using a research period from 2007. As mentioned in section 2.4, there has been very little research into sin stocks returns with financial data from 2007. In consideration that it was the year when the financial crisis began. The idea of a fund like ViceX is that companies in sin industries perform better in economically uncertain times. Finally, the last decade socially responsible investing (SRI) has also become more important.
Table 5 shows the time-series return regression of the value-weighted global sin portfolio of 160 companies (GlobalSin) between August 2007 and August 2016 based on the CAPM, three-factor model and four-factor model. The first estimation, as shown in Panel A, of the CAPM shows the excess return of the sin stock portfolio net of risk-free rate of monthly 0.61%
(significance level 1%). The results of the three and four-factor models show similar high outperformances (0.57% and 0.55%). These results are significantly higher than the results from section 5.1. A shorter and more recent time period (only the last nine years) seems to have a positive influence on the performance of sin stocks.
Panel B shows the excess return of the portfolio long in global sin stocks and short the comparables (GlobalSin – Comparables) between August 2007 and August 2016. The CAPM result shows 0.27% monthly excess return over market return and is statistically significant at 5%. Looking at the three and four-factor results, I observed a monthly 0.33% and 0.28%
outperformance (both 5% significant). The results of Panel B are in line with the time period (1996-2016) used in previous sections.
Table 5: Time-series return regressions (net of risk-free rate and comparables) between 2007- 2016
A: GlobalSin vs. Risk-free α βM KT βS M B βH M L βMO M
GlobalSin - Rf ,t 0.0061*** 0.5861***
(0.0019) (0.0384)
GlobalSin - Rf ,t 0.0057*** 0.6133*** -0.2492** -0.2703***
(0.0018) (0.0387) (0.1243) (0.1136)
GlobalSin - Rf ,t 0.0055*** 0.6236*** -0.2425** -0.2363** 0.0444
(0.0019) (0.0407) (0.1248) (0.1214) (0.0554)
B: GlobalSin vs. Comparables
GlobalSin - Comparables 0.0027** -0.3766***
(0.0027) (0.0587)
GlobalSin - Comparables 0.0033** -0.3674*** -0.7258*** 0.0274 (0.0025) (0.0552) (0.1187) (0.1039)
GlobalSin - Comparables 0.0028** -0.2741*** -0.8264*** 0.1044 0.3109***
(0.0024) (0.0552) (0.1140) (0.1016) (0.0591) NOTE:Panel A shows the time-series regression results obtained from regressing the value-weighted returns (in
excess of the risk-free rate) of the global sin stocks (alcohol, tobacco, defense and gambling) portfolio between 2007-2016 by using OLS estimation on the CAPM, Fama and French three factors (1993) and Carhart (1997) four factors. Panel B shows the measurement results obtained from the time-series regressions of a global sin portfolio (GlobalSin - Comparables) that is long in GlobalSin (the monthly return for value-weighted global portfolio of sin stocks— alcohol, tobacco, gabling and defense) and short in Comparables (the monthly return for an value-weighted portfolio of comparable stocks) between 2007-2016 by using OLS estimation on the CAPM, Fama and French three factors (1993) and Carhart (1997) four factors. The intercept (alpha) shows the outperformance. βM KT is the parameter on the market risk premium. βS M Bis the return of a portfolio long small stocks and short large stocks, βH M Lis the return of a portfolio long high book-to-market stocks and short low book-to-market stocks and βMO Mis the return of a portfolio long past 12-month return winners and short past 12-month return losers. In brackets are shown the standard errors which are adjusted for serial correlation using the Newey-West (1987) correction.
*** significance level of 1% . ** significance level of 5%. * significance level of 10%.
6 Robustness checks
6.1 GlobalSin vs. Region
The four-factor regression results of the samples per region are shown in Appendix C. The subsample North America, comprising a total of 29 companies, shows a clear positive four- factor alpha. The alpha is a monthly 0.72%, and is also significant at a one percent level. Only the HML-factor is a little significant, at a 10 percent level, the other factors are not significant.
The European sin portfolio contains 40 companies and shows an even bigger outperfor- mance, namely monthly 0.80% (at one percent significance) between November 1996 – August 2016. This result is in line with the European study of Salaber (2013). She found a monthly outperformance of 0.63%.
Also the Asia and Pacific portfolio excluding Japan experiences a monthly outperformance net of the risk-free rate of 0.67%, which is significant at a level of 5%. The Japan sin portfolio shows a small abnormal return, since this is not statistically significant it cannot be assumed that it differs from zero. A possible explanation is the fact the JapanSin sample contains only seven Japanese companies.
6.2 GlobalSin vs. Industries
In Appendix D, the four-factor regression results of the different industry samples are pre- sented. I created subsamples of the four sin industries (alcohol, tobacco, gambling and defense/weapons). By adding a fifth subsample namely GlobalSin without using companies of the tobacco industry I’m able to check whether previous results are heavily boosted by the tobacco stocks. As mentioned earlier, the litigation risk is higher in the tobacco industry which may result in higher returns.
The subsample Alcohol contains 107 companies and shows monthly abnormal returns of 0.18% using the four-factor model (significant at 5%. A beta of 0.4825 means that the Alcohol portfolio has a lower sensitivity towards the volatility of the market portfolio. The Tobacco
significant 1%. The Gambling sample (six companies) does not show a statistically significant outperformance. This also applies to the Defense and weapons sample (14 companies), an outperformance of 0.06%, but not significant.
Finally, the GlobalSin without Tobacco sample, comprising a total of 127 companies, experiences a monthly outperformance net of the risk-free rate of 0.16%, which is significant at a level of 5%.
6.3 ViceX Fund vs. GlobalSin
As far as I know, ViceX is the only fund that focuses completely on companies in sin industries.
I made a zero-investment portfolio with the long monthly ViceX Fund returns and short my GlobalSin portfolio returns. This investigates whether or not the fund has an added value.
The four-factor results are shown in Appendix E. The ViceX Fund outperform my GlobalSin portfolio with a monthly 0.11% (significant at 5%, which is annually 1,32%. However, the annual management fee is 1.48%6.
6.4 Analysis based on equally-weighted returns
As mentioned in section 4.2.1, there are two possible methods for weighting the portfolio returns, namely equally-weighted and value-weighted. For my main analysis, I used value- weighted portfolio returns. In addition, I also performed the analysis based on equally-weighted portfolio returns. The regression results are shown in Appendix F.
The four-factor alpha of my GlobalSin portfolio net of the risk-free rate (Panel A) shows an outperformance of 0,42% (significant at 1%), which is in line with my value-weighted results. Also the subsamples have many similarities . The four-factor alpha of the GlobalSin portfolio net of a set of comparables earn monthly 0.18% (5% significant) which is lower than the outcome of section 5.2 (0.28%).
Finally, Panel C shows the outperformance of the GlobalSin portfolio net of the risk-free rate from 2007 until 2016, which is lower at a monthly value of 0.19% (5% significant).
6http://www.morningstar.com/funds/XNAS/VICEX/quote.html
7 Conclusion
In this thesis, I provide evidence for the outperformance of sin stocks. The sample includes sin stocks from 47 countries across the world. In conclusion, I have tested two hypotheses.
Between November 1996 and August 2016, the value-weighted global sin stock portfolio reported net of the risk-free a significant monthly abnormal return of 0.54%, using the one-factor model (CAPM). This is significantly lower in comparison to the results of Fabozzi et al. (2008), who earned a monthly 0.96%.
After correcting our sample for size, value and momentum using the three-factor and
four-factor models, the significant monthly outperformance is respectively 0.42% and 0.38%.
Benchmarked against a set of comparables my self-constructed global sin portfolio as a result shows outperformances of 0.27%, 0.33% and 0.28% using one, three and four-factor models.
Herewith, I followed the research of Hong and the results are significantly similar, however, contrary to my research they have a sample used only by US firms.
In the introduction, I explained that there are several reasons to also look at the period August 2007 – August 2016. Looking at the results, the four-factor alpha of the global portfolio net of the risk-free rate is 0.55% (significant of 1%), which is remarkably higher than the initial time period. A portfolio with long the global sin portfolio and short a set of comparable stocks results in a significant monthly abnormal return of 0.28%, which is in line with the initial time period.
In conclusion, it can be said that there is a significant outperformance for sin stocks.
Although, several studies in the last 15 years confirm this outcome, it is still not corrected by the market. In fact, it only seems to keep growing.
8 Discussion
In this research, I self-constructed a global portfolio of sin stocks of 160 companies. Since I used Kenneth French’s data for the multi-factor regressions and a strong tilt towards a certain region or microcap should be avoided, the sample size is limited. Future research may focus on self-creating the SMB, HML and momentum factors to get a larger sample of sin stocks.
As shown in my study, I found for my global sample and for the several subsamples betas below one. When a beta coefficient is lower than one it indicates a lower risk in comparison to the benchmark. The under lying theory is that companies are less influenced by market conditions, and therefore can perform better economically uncertain times. Hence, it paves a way for further research to check whether or not sin stocks outperform during the financial crisis (2007-2012). The presumption that sin stocks show higher abnormal returns during recessionary periods could be checked.
In this study, I found out that sin stocks still show significant outperformances. Using different regional or industrial subsamples, I experience similar results. But as I said, the results should be interpreted with caution. Finally, in my opinion the performance of ViceX Fund deserves further investigation, especially during economic uncertain times.
9 References
Bauer, R., Koedijk, K., Otten, R. (2005). International Evidence on Ethical Mutual Fund Performance and Investment Style.Journal of Banking and Finance, 29, 1751–1767.
Bodie, Z., Kane, A., and Marcus, A.J. (2008). Investments. New York: McGraw-Hill/Irwin
Carhart, C. (1997). On persistence in mutual fund performance, Journal of Finance, 45(5), 57-82.
Chong, J., Her, M., and Phillips, G. (2006). To sin or not to sin? Now that’s the question, Journal of Asset Management, 6(6), 406-417.
Durand, R.B., Koh, S., Limkriangkrai, M., (2013). Saints versus Sinners. Does morality matter? Journal of International Financial Markets, Institutions and Money, 24(4),166-183.
Fabozzi, F.J., Ma, K.C., and Oliphant, B.J. (2008). Sin Stock Returns, The Journal of Portfolio Management, 35(1), 82-94.
Fama, E.F. (1998). Market efficiency, long-term returns, and behavioral finance, Journal of Financial Economics, 49(3), 283-306.
Fama, E.F., French, K.R. (1993). Common risk factors in the returns on stocks and bonds, Journal of Financial Economics, 33(1), 3-56.
Fama, E.F., French, K.R., (1997). Industry costs of equity, Journal of Financial Economics, 43(2), 153-193.
Hong H., and Kacperczyk M. (2009). The Price of Sin: the Effects of Social Norms on Markets, Journal of Financial Economics, 93(1), 15-36.
Hoepner, Andreas G., and Zeume, S. (2009). The Dark Enemy of Responsible Mutual
Jegadeesh, N., Titman, S. (1993). Returns to Buying Winners and Selling Losers:
Implications for Stock Market Efficiency.Journal of Finance, 48(1), 65-91.
Merton, R.C. (1987). A Simple Model of Capital Market Equilibrium with Incomplete Information. The Journal of Finance, 42(3), 483-510.
Newey, W.K., West, K.D. (1987). A Simple Positive-Definite Heteroskedasticity and Autocorrelation - Consistent Covariance Matrix, Econometrica, 55, 703-708.
Renneboog, L., Ter Horst, J., Zhang, C., 2008a. Socially responsible investments:
Institutional aspects,performance, and investor behavior, Journal of Banking Finance, 32(9), 1723-1742.
Salaber, J. (2007). The determinants of sin stock returns: evidence on the European market.
Paris Finance International Meeting, 1-34.
Salaber, J. (2013). Religion and returns in Europe, European Journal of Political Economy, 32(1), 149-160.
Sharpe, W.F. (1964). Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk, Journal of Finance, 19(3), 425-442
Statman, M. (2000). Socially Responsible Mutual Funds. Financial Analyst Journal, 56, 30-39.
Worman, H. J. (1995). Signals and Structural Features Involved in Integral Membrane Protein Targeting to the Inner Nuclear Membrane. The Journal of Cell Biology, 130, 15-27.
Appendix A The complete sin companies list
30
Appendix B Country specific list of all the sin stocks
Appendix C GlobalSin vs. Region
Table 6: Regression results of different regions
Region α βM KT βS M B βH M L βMO M
NorthAmericaSin - Rf ,t 0.0072*** 0.4622 -0.0112 -0.0819* 1.2229
(0.0023) (0.0344) (0.0663) (0.0481) (3.0521)
EuropeSin - Rf ,t 0.0080*** 0.5123** -0.0709 -0.0202 0.0641*
(0.0025) (0.0341) (0.0816) (0.0559) (0.0333)
AsiaPacificSin - Rf ,t 0.0067** 0.4868* -0.0558 0.0504 0.0252
(0.0034) (0.0345) (0.0821) (0.0826) (0.0379)
JapanSin - Rf ,t 0.0019 0.5406*** -0.1127 0.2773*** 0.0847**
(0.0030) (0.0575) (0.0942) (0.0989) (0.0355)
NOTE:This table shows the time-series regression results obtained from regressing the monthly returns (excess of the risk-free rate) of the different regions of value-weighted sin stocks between 1996-2016 by using OLS estimation on the Carhart (1997) four factors. The risk-free rate that is used varies by region. The Rf ,t for North America, Europe, Asia Pacific and Japan are respectively the US one-month T-bill, the three-months Euribor, Hong Kong three-month interbank rate and the Japanese Yen LIBOR three months interbank rate. The intercept (alpha) shows the outperformance. βM KT is the parameter on the market risk premium. βS M Bis the return of a portfolio long small stocks and short large stocks, βH M Lis the return of a portfolio long high book-to-market stocks and short low book-to-market stocks and βMO Mis the return of a portfolio long past 12-month return winners and short past 12-month return losers. In brackets are shown the standard errors which are adjusted for serial correlation using the Newey-West (1987) correction.
*** significance level of 1% . ** significance level of 5%. * significance level of 10%.
Appendix D GlobalSin vs. Industries
Table 7: Regression results of different industries net of the risk-free rate
Industries α βM KT βS M B βH M L βMO M
Alcohol 0.0018** 0.4825*** -0.0073 0.2369*** 0.0358
(0.0014) (0.0315) (0.0651) (0.0579) (0.0337)
Tobacco 0.0076*** 0.4249*** -0.3121*** 0.3073*** 0.0564
(0.0023) (0.0510) (0.1053) (0.0938) (0.0546)
Gambling 0.0034 0.7442*** 0.0408 0.5169*** -0.1179
(0.0038) (0.0855) (0.1767) (0.1574) (0.0917)
Defense and weapons 0.0006 0.7982*** -0.0669 0.4029*** 0.0047
(0.0032) (0.0715) (0.1477) (0.1315) (0.0767)
GlobalSin ex. Tobacco 0.0016** 0.5559*** -0.0074 0.2790*** 0.0253
(0.0014) (0.0316) (0.0654) (0.0582) (0.0339)
NOTE:This table shows the time-series regression results obtained from regressing the monthly returns (excess of the risk-free rate) of the different regions of value-weighted sin stocks between 1996-2016 by using OLS estimation on the Carhart (1997) four factors. The risk-free rate that is used is the US 3-month T-bill. The intercept (alpha) shows the outperformance. βM KT is the parameter on the market risk premium. βS M Bis the return of a portfolio long small stocks and short large stocks, βH M Lis the return of a portfolio long high book-to-market stocks and short low book-to-market stocks and βMO Mis the return of a portfolio long past 12-month return winners and short past 12-month return losers. In brackets are shown the standard errors which are adjusted for serial correlation using the Newey-West (1987) correction.
*** significance level of 1% . ** significance level of 5%. * significance level of 10%.
