Controversial Industries: does it pay to ignore social norms?

(1)

1

Controversial Industries: does it pay to

ignore social norms?

Implication of international sample

M.R.T.M. van Nuenen S2964414 (UoG)

T-235 (UU)

August 16th_{, 2018}

Supervisor: prof. dr. L.J.R. Scholtens Co-Assessor: prof. dr. W. Bessler

MSc International Financial Management Faculty of Economics and Business

MSc Business and Economics Department of Business Studies

Abstract

This paper investigates the impact of social norms on the performance and valuation of “controversial stocks”- publicly traded companies involved in the production of Adult Entertainment, Alcohol, Gambling, Nuclear Energy, Tobacco, Uranium, and Weapons. Their performance and valuation is directly compared with compare non-controversial stocks. The paper consider an international sample of 941 controversial stocks. Employing a multi-factor performance measure, seven countries provide a significant outperformance of controversial stocks across all relevant control factors. The valuation analysis, however, provide mixed results on a country level, but on the global market-to-book ratio provide a significant overvaluation of controversial stocks compared to comparable non-controversial stocks, which contradicts the prediction of an undervaluation.

JEL classification:

(2)

2

1 Introduction

Social norms are based on the beliefs of community members that operate as a driving force for individual behaviour (Kubler, 2001; Akerlof, 1980). In some occasions, the force for social norms influence economic behaviour and take precedence over the profit motive of individual investors and corporations. An early example of such a finding is the discrimination model of Becker (1957), which reported that agents (e.g. employees) with discriminatory beliefs, inherited from norms and standards within their communities, bear financial costs from avoiding interaction with particular types of people. A more recent and less extreme reflection of social norms is the rise of Socially Responsible Investment (SRI) strategies, which is a social norm based strategy that persuades investors to align ethical and financial concern to improve a corporations environmental, social and governance (ESG) performance (Renneboog, Ter Horst, & Zhang, 2008). SRI is one of the fastest growing investment strategies of the last twenty fives year being applied on 26 percent of all professionally managed assets in 2016 (Global Sustainable Investment Alliance, 2017, sd). The scope of SRI varies from investing in morally and ethically sounds companies to excluding stocks from unethical and controversial companies that produce and market goods perceived as unethical and controversial (e.g. Adult Entertainment, Alcohol, Gambling, Nuclear Energy, Uranium, and Weapons) (Kim & Venkatachalam, 2011). Social, environmental and ethical issues are not only of great concern for individual investors it also increase the pressure on institutional investors with public information on stock holdings (e.g. pension funds, universities, religious organisations, banks and insurance companies) to avoid controversial industries and to minimize the potential of public scrutiny (Hong & Kacperczyk, 2009).

There are also investment strategies that ignore social norms and include stocks from companies operating in industries that are excluded by responsible investors, institutional investors1

and SRI funds. Such a strategy is often referred to as “vice investing” or “sin investing” but as Trinks and Scholtens (2017) argue, referring to “sin stocks” is questionable and instead should be referred to as controversial.2_{The affection of holding neglected stocks could be evolved from Merton’s (1987)}

“neglected stock” theory, which states that the stock price of firms with a smaller investor base are depressed relative to their fundamental value as it limits information and risk sharing (Hong &

1_{Hong and Kacperczyk (2009) show that sin stocks have less institutional ownership (23%) than comparable} industries (28%) and less analyst coverage (1.3 and 1.7 analyst respectively)

(4)

4

Kacperczyk, 2009). In return, investors expect a higher risk adjusted return to compensate for the limited risk and information sharing, which makes it interesting for investors and corporations that do not consider social norms in their investment decisions. An alternative explanation is given by Heinkel, Kraus, and Zechner (2001), which created a model in the spirit of Merton (1987) and considers the price implications on the stock price of firms neglected by ethical investors because of the polluting business operations of the firm. The model highlights that exclusionary ethical investing result in a smaller investor pool for polluting firms as “green investors” eschew stocks of polluting firms. As a result, risk sharing is limited as less investors are willing to hold stocks of polluting firms, which depresses the price and increases the expected return to cover for the limitations on risk sharing. An example of such a strategy is the creation of the Vice Fund in 2001, which solely focuses on investing and promoting controversial industries.3_{In accordance with the “neglected stock” theory, the Vice}

Fund has since it inception an average excess return of two percent over the S&P500 (Liston, 2016). Nonetheless, examples like the Vice Fund are rare, which is intriguing given the extant of literature that determine sin stocks as stable, recession-resistant, profitable, and with superior market returns (Chong, Her, and Phillips (2006); Fabozzi, Ma, and Oliphant (2008); Hong and Kacperczyk (2009); Kim & Venkatachalam (2011) and Trinks and Scholtens (2017)).

The impact of investing in controversial industries and the affect on the financial performance of an investor, across different countries and social norms, is in its early stages, which leads to the following two research questions to remain open:

(1) Do controversial stocks outperform comparable non-controversial stocks? (2) Do controversial stocks have lower valuations than non-controversial stocks?

The first research question is rather general but remains open given the contrasting results when, in addition to the US, additional countries are included in the sample (see Table 1). The second research questions relates to the valuation of controversial stocks, and given the contrasting results by Hong and Kacperczyk (2009), and Durand et al., (2013b), show a interesting and relevant area to analyse and event extend to new geographical area’s. The financial performance analysis considers a systematic approach and consists of the following measures; Jensen’s Alpha (CAPM), Fama-French Three-Factor Model, Carhart Four Factor Model, Bet Against Beta factor by Frazzini and Pedersen (2014), and the more recent Fama-French Five-Factor Model. The application of the aforementioned models is in line

(5)

5

with the most recent papers by Blitz and Fabozzi (2017), and Richey (2017), (see Table 1). The systematic approach allows the paper to highlight the most applicable performance factors for controversial industries. This allows investors and companies to identify relevant and important factors when considering investments in these industries. Moreover, four country factors (e.g. (1) the investment freedom index, (2) government integrity index, (3) number of individuals using the Internet (% of the population), and (4) the business extent of disclosure index)4_{are used to analyse whether}

country specific characteristics influence the performance and valuation of controversial stocks. The investment freedom index and government integrity index are provided by the economic freedom index and based on Cumming and Johan (2007), which argues that different regions and countries have different legal and social standards, which impact the likelihood of an investor to apply an investment strategy that is based on social conscious. The last two variables percentage (%) of population with access to internet and business extent disclosure index are based on a recent discussion by The Economist (2017) on the rapid rise of SRI among millennials. They argue that the extensive access to Internet and an increase in the quantity of information provided by firms (partially due to government regulations) result in a rapid rise of SRI, especially among millennials. The first two variables are extracted from the economic freedom index5_{while the last two variables are extracted}

from the Worldbank database.

The results shows that controversial stocks in Belgium, Finland, Italy, Ireland, Singapore, the United Kingdom, and the United States, outperform comparable non-controversial stocks, even after adjusting for the relevant factor models. In addition, the CAPM analysis provide an outperformance of controversial stocks in Argentina, Hungary, the Isle of Man, Namibia, Pakistan, and Turkey, and a underperformance of controversial stocks in Israel. Moreover, country variable analysis show that, besides countries with low level of business disclosure, all country characteristics provide statistically significant outperformance of controversial stocks compared to comparable non-controversial stocks. In particular countries with low levels of government integrity or internet accessibility.

The valuation analysis, however, contradicts Merton’s neglected stock theory, as the global market-to-book ratio is statistically significant and 26.08% higher than comparable non-controversial

(6)

6

stocks, which resembles an overvaluation instead of an undervaluation. In addition, nine countries (e.g. Australia, Belgium, the Cayman Islands, China, Finland, Ireland, Israel, Malaysia, the Philippines, and the Republic of Korea (south)) provide evidence of an overvaluation of controversial stocks, while six other countries (e.g. Germany, Greece, Italy, Spain, Sri Lanka, and Sweden) provide evidence of an undervaluation of controversial stocks, which was in line with the prediction. Moreover, the country variable analysis show a Market-to-Book ratio analysis that yield seven positive Cdum coefficients at a 1% significance level, which indicate that controversial stocks have significantly higher MTB ratios than comparable non-controversial stocks. Nonetheless, the overvaluation of MTB tend to be stronger in countries with low levels of government integrity, investment freedom, internet accessibility, and high level of business disclosure. The Price-Earning ratio analysis provide a statistically significant overvaluation for controversial stocks with either low government integrity or investment freedom while the PEBITDA ratio provide a statistically significant overvaluation for controversial stocks in with low levels of government integrity or internet accessibility.

Overall, the aim of this research paper is to extent the geographical area of the existing literature on the impact of social norms on the performance and valuation of controversial stocks. In addition, it extends the common “triumvirate of Sin” approach and use seven industries (e.g. Adult entertainment, Alcohol, gambling, Nuclear Energy, Uranium, Tobacco, and Weapons) collectively on a risk adjusted basis. Moreover, a systematic factor model approach create an overview useful for both corporate and individual investors to consider the impact of each individual factor on the performance of the controversial industries, which is useful when they consider investing in one of these industries. Finally, four country variables operate as social norm indicators and analyse whether country specific characteristics influence the performance and valuation of controversial stocks.

This paper relates the fields of international financial management, economics, and business through the use of multiple country level characteristics across a large international sample of 102 countries. Moreover, the analysis consist of both traditional (e.g. CAPM, Fama-French, and Carhart Momentum factor) and the more recent (e.g. Bet Against the Beta factor and Fama-French 5 factor) portfolio management models.

(7)

7

2 Literature Review and Hypotheses

The aim of the literature review is to introduce relevant existing literature on the performance of controversial stocks on which this research can build and make further contributions. Table 1 provides an overview and separate the performance of the VICE fund, portfolio’s, and the valuation of controversial stocks separated in panel a, b, and c, respectively.

Chong, Her, and Phillips (2006) introduced a seminal paper on the comparison between a socially responsible fund (Domini Social Equity Fund) and socially irresponsible fund (Vice Fund) due to the lack of research on socially irresponsible investing. The paper considered traditional performance measures followed by a generalised autoregressive conditional heteroscedasticity (GARCH (1,1)) model, which allow the paper to evaluate both funds over a three-year period from 2002-2005. The results show that the Vice Fund outperformed the Domini Social Equity Fun during the sample period, which indicate that the socially irresponsible outperformed the socially responsible fund.

Fabozzi, Ma, and Oliphant (2008), consider the analysis of sin stocks from and individual perspective and was the first paper to include both U.S. and Non-U.S. based firms while combining the classical Triumvirate of Sin portfolio with additional industries (e.g. defence, biotech (compromising animal testing, genetic engineering, and ordinary stem cells), and adult entertainment). The empirical evidence shows that the sin stocks outperform the market by an annual average excess return over the market of 11.15%. Moreover, the sin portfolio outperformed the market indexes 35 out of the 37 years analysed, which means that the sin portfolio not only outperform the market index on magnitude but also on frequency. The paper identifies not conforming to or upholding implicit or explicitly costly social standards as the main reason why sin stock outperform the market.

(8)

8

remain to be undervalued by 15-20%. As a result, sin firms tend to rely more on private debt as it is more expensive to finance their operations in equity, which could be considered as an advantage as debt market are less transparent than equity markets. Finally, the paper conducted a robustness check by expanding the study with an international sample that consists out of Canada and seven large markets in the European Union, hereafter referred to as EU, between 1985 and 2004. The international sample shows a similar result as sin stocks outperform the benchmark by up to 21 basis points on a daily basis, which is smaller but still a significant result. Overall, the results show that social norms have a significant impact on the investing environment of sin stocks and highlight the cost associated with norm-constrained investing.

Salaber (2009a) analyse the performance of the Triumvirate of Sin stocks relative to industry-comparable stocks on the US market between 1926 and 2005. Results show that US sin stocks outperform the market but it disappears when the control group consists of industry-comparable stocks with similar defensive characteristics. Moreover, sin stocks outperform the market during recessions but underperform during economic growth, which indicate that the earnings of sin stocks are not as sensitive to economic conditions due to the addictive properties of sin industry products. Overall, the paper provides evidence that mitigates the neglect effect on US sin stock time series because it is not valid when economic conditions change and it suggest that SRI investors pay a financial cost when they avoid these stocks because of ethical and social reasons (Salaber, 2009a).

Salaber (2009b) analyses whether country specific factors (e.g. religion, level of excise taxation, and the degree of litigation risk) has an influence on time-series variation of the average sin stock returns within Europe between 1975 and 2006. Salaber (2009b) considered Europe because the majority of the population is Christian, which has two main denominations; (1) Catholics and (2) Protestants, and despite the European Union, individual countries have different cultural and legislation environment. The results show that sin stocks within Protestant countries have a higher return due to “sin aversion”. Moreover, sin stocks in countries with either a high level of excise taxation or a high level of litigation risks have higher risk adjusted returns than other stocks (Salaber, 2009b). As such, Salaber (2009b) provide evidence that religion, litigation risk, and excise taxation has an influence on the time-series variations on average sin stock returns.

Moreover, Kim and Venkatachalam (2011) consider the “neglected stock” theory by Merton (1987) and analyse whether the higher returns of US controversial stocks6_{can be explained by greater}

levels of information risk arising from poor financial reporting quality (precision of estimation). Merton

(9)

9

(1987) highlighted that less awareness and analyst coverage can lead to a reduction in the quantity of information provided (e.g. less frequent). This reduction led to an increase in information risk, which lead to a decrease in the stock price and a higher expected return as result. Merton (1987), however, does not consider the potential impact of social neglect and business risk on the quality of financial reporting (precision of estimations) and assume the quality is the same for all securities in his analysis. Kim and Venkatachalam (2011), in contrast, mainly focus on the quality of the provided information as a cause of higher expected returns. The results, however, show a superior quality of financial reporting quality for these firms, in particular, on timely loss recognition and predictability of future cash flow earnings, which show that a lack of quality cannot explain the high-risk adjusted returns of controversial stocks. Moreover, Kim and Venkatachalam (2011), argue that investors are willing to bear a financial cost by neglecting sin stocks to comply with societal norms, which provide evidence of non-financial factors to affect investor portfolio decisions.

Durand, Koh, and Limkriangkrai (2013a) build on Hong and Kacperczyk (2009) and analysed the impact of social norms on both “sinners” and “saints” in the US from 1990 till 2008. The main research question is whether social norms operate not only as a constraint for sinful activities but also as incentive to pursue activities that are perceived to be virtuous. The paper identifies “sinners” in accordance with the standards of Hong and Kacperczyk (2009) while the “saints” are firms listed on the MSCI KLD400 Social Index, which is the leading index of stocks considered to be appropriate for socially responsible investors. The evidence confirm the findings of Hong and Kacperczyk (2009) and show a positive risk-adjusted performance of “sinners”. Nonetheless, the results do not show a correspondingly negative risk-adjusted performance for “Saints”. In addition, similar as Hong and Kacperczyk (2009), “sinners” have less institutional ownership, less fund interest, less coverage and as a result are cheaper than “saints”. Furthermore, social norms force “sinners” to rely more on debt instead of equity, and as result force them to make financial policies that could be sub-optimal (Durand, Koh, & Limkriangkrai, 2013a). In contrast, “saints” rely more on the equity market and are less leveraged than “sinners”, which allow them to distribute more of their earnings to shareholders via dividends and share repurchases. As such, “saints” tend to adopt corporate financial policies that allow them to reduce the agency cost of management (Durand, Koh, & Limkriangkrai, 2013a).

(10)

10

shareholders hold less sin stocks. This conclusion, however, is debatable given the significant governmental holdings of East-Asian countries in their domestic market. It could, instead, highlight the strength of governmental signals on the social norms of a society. In particular, japan shows a strong affection on holding significant amount of controversial shares in controversial. Moreover, the price performance and valuation analysis contradicts Hong and Kacperczyk (2009), by providing an underperformance (e.g. significant negative risk adjusted returns) of controversial stocks in Japan, Malaysia, the Republic of Korea (South), and Singapore, while the valuation performance provide and overvalued MTB Ratios for six out of the seven countries. In fact, investors from these countries are willing to pay more for sin stocks than comparable non-sin stocks. The results, however, also contradicts Hong and Kacperczyk (2009), by providing significant negative risk adjusted returns for controversial stocks in four countries (e.g. Japan, Malaysia, The Republic of Korea (South), and Singapore) while the remaining countries provide insignificant negative values. Moreover, the valuation analysis provide overvalued MTB ratios for controversial firms in six out of the seven countries (e.g. Australia, India, Japan, Malaysia, the Republic of Korea (South), and Singapore), which contradicts the predicted undervaluation by Hong and Kacperczyk (2009). In fact, investors from these countries are willing to pay more for sin stocks than comparable non-sin stocks. Overall, the paper argue that the more individualistic countries like the US, and, to some extent, Australia and New Zealand, tend to avoid sin stocks as investors in these societies take responsibility for their actions and believe their decisions could change the world and the believe of others (Durand, Koh, & Tan, 2013b). Moreover, it shows the potential strength of governmental signal and strong difference in social norms between two types of cultures, which influence the likelihood of holding and the price of controversial stocks. Finally, it shows that United States results, as provided by Hong and Kacperczyk (2009), are not universal and differ across countries.

Lobe and Walkshäusl (2016) conduct an empirical test among the “Sextet of Sin” (adult entertainment, alcohol, gambling, nuclear power, tobacco and weapons) to show whether a portfolio of socially irresponsible firms trade at a discount. This paper is among the first to extend the widely used “Triumvirate of Sin” by considering additional industries that are often excluded from SRI funds. The paper created three separate sin indexes (e.g. global, sector, and domestic (regional or national)) and compares the stock market performance of these indexes with a set of important international socially responsible investment indexes. The results provide no evidence whether ethical and/or unethical screens create a significant difference among the financial performance of both portfolios (Lobe & Walkshäusl, 2016)

(11)

11

referred to as FF3, while Richey (2016) adds the Carhart Four Factor model, hereafter referred to as C4. Both papers show that vice stocks have significant and positive alpha throughout the sample period and outperform the overall market. The most recent paper by Richey (2017) introduces the Fama-French Five-Factor Model, henceforth FF5, which extends the FF3 by adding “profitability” and “investment” as the fourth and fifth factor. The adaptation of both variables, however, dissolve alpha’s significance in the vice stocks returns (Richey, 2017).

Moreover, a recent paper by Blitz and Fabozzi (2017) extended the existing literature on controversial stocks performance by introducing both the FF5 model and the BAB factor on an individual (e.g. Japan and United States), regions (e.g. Europe) and a global sample. Blitz and Fabozzi (2017) highlight the limitations on data availability, which resulted in the creation of three samples as Kenneth French data set provides information on the US between 1963 and 2016, the Thomson Reuters DataStream consider the US between 1973 and 2016 while the European, Japanese and global sample are between 1990-2016 (Blitz & Fabozzi, 2017). The results show that, besides Japan, all countries and regions provide a significant and positive one-factor Alpha, which is in line with Durand et al., (2013b). In fact, Japan doesn't provide any significant Alpha within the paper. The Alpha for the United States, Europe, and Global sample, remains positive and significant after introduction of FF3 and the Momentum factor. However, the control for the 4th_{and 5}th_{Fama-French factor (e.g.}

“profitability” and “investment”), dissolves the significance of alpha, which is in line with the results of Richey (2017). Moreover, Frazzini and Pedersen’s BAB factor with the C4 improves the explanatory power of the model but, besides a significant Alpha for the US 1963-2016 sample, dissolves the significance of the remaining alphas. A similar result is provided by the combination of the FF5 and BAB factor, which does not provide any significant alpha. Overall the paper shows that sin stocks outperform the market when controlling for traditional factors. However, this premium disappears when the paper controls for newly developed factors like the BAB, profitability (RMW) and investment (CMA). A significant limitation of the paper is, however, the use of regional factors (e.g. European level indicators), as Griffin (2002) highlights; Fama and French factors are highly country specific and not regional specific as multiple variations occur among countries.

Finally, Trinks and Scholtens (2017), investigate the impact of negative screening among fourteen (potential) controversial industries.7_{The paper conduct a comparative mean–variance}

analysis based on 1634 controversial stocks, which in the majority of the industries (exception of adult

(12)

12

entertainment and stem cells) increase the risk adjusted returns, which is in line with the majority of the existing literature on controversial industries. Moreover, they paper highlight that a screened market portfolio (S&P500) significantly underperform compared to unscreened market portfolio. Overall, Trinks and Scholtens (2017) argue there is an opportunity cost of reframing from investing in controversial industries and the height of the cost depend the type of industry screen applied.

The final row of Panel B and Panel C represent this paper, which extend the geographical focus to 102 developed and emerging markets while using the same models as the most recent articles by Blitz and Fabozzi (2017), and Richey (2017) (see Table 1). The reason to include the relatively new Fama French 5 factor model and Bet Against Beta model by Frazzini and Pedersen (2013) is to create an extensive but systematic adaption of each model to highlight and specify what factors have the most contribution and/or impact on the performance of controversial industries versus non-controversial industries. For example, Blitz and Fabozzi (2017) show that controversial stocks tend to have a lower beta than average in the United States and the BAB factor controls for the return difference between low- and high-beta stocks.

Overall, based on the “neglected stock” theory by Merton (1987) and the tendency of the existing literature to provide a higher risk adjusted return for controversial “sin” stocks, the following hypothesis is created.

H1: controversial stocks have higher risk-adjusted returns than comparable non-controversial industries

(13)

13

Durand, Koh, and Tan (2013b) replicate the US-based study of Hong and Kacperczyk (2009) in seven Pacific-Basin Markets: Australia, India, Japan, South Korea Malaysia, New Zealand, and Singapore. The results show that substantial shareholders in Australia and New Zealand are less likely to hold sin stocks while their co-shareholders in Japan and South Korea are actually more likely to hold sin stocks. They identify the cultural closeness of Australia and New Zealand towards the US as the main reason why substantial shareholders hold less sin stocks. This conclusion, however, is debatable given the impact significant governmental holdings of East-Asian countries in their domestic market. It could, instead, highlight the strength of governmental signals on the social norms of a society. In particular, japan shows a strong affection on holding significant amount of controversial shares in controversial. Overall, the results show a positive and significant market to book ratio for controversial stocks in six out of the seven countries. This means that investors are actually willing to pay more for sin stocks, which is in direct contrast with the findings of Hong and Kacperczyk (2009). This, however, shows how important investor perspective and government holdings are on the valuation of controversial stocks and that country differences do exist.

Based on the “neglected stock” theory by Merton (1987) and the findings of Hong and Kacperczyk (2009), the following hypothesis is developed:

(14)

(15)

(16)

16

A visual representation of the hypothesis is presented below. The first hypothesis (e.g. H1) considers the impact of social norms on the performance of controversial stocks while the second hypothesis (H2) represent the relation between social norms and the valuation of stocks in controversial industries. Based on existing literature (see table 1), the expected impact of social norms are positive on the performance of controversial stocks and a negative impact on the valuation of firms operating in controversial industries. In addition, similar to Salaber (2009b), Durand et al., (2013b), and Fauver and McDonald IV (2014), four additional country factors (e.g. investment freedom index, government integrity index, % of population with access to internet, and Business extent disclosure index) are applied to consider and evaluate the impact of country specific characteristics on the performance and valuation of controversial stocks. These four factors, however, are not represented by individual hypothesis but provide additional information on the country specific characteristics that allow for potential differences in social norms and eventually the performance and evaluation of controversial stocks.

Social norms

Peformance of

Controversial

industries

Valuation of

Controversial

industries

+

H2

-H1

Fig. 1. Structure of the model

3 Methodology, Data, and Sample Description

This section describes the models that are used to answer the research questions and hypotheses. All variables are introduced and justified in the applicable model description. Moreover, the data and sample selection explain the information sources and the reasoning behind sample collection.

3.1 The Model

(17)

17

3.1.1 Single and multi-factor performance measurement frameworks 3.1.1.1 Capital Asset Pricing Model

The first model within the analysis is the Capital Asset Pricing Model (CAPM), which is developed inter alia by Sharpe (1964) and Lintner (1965). The intercept of the CAPM, Jensen’s alpha (1968) is interpreted as a measure of out- or underperformance relative to the benchmark.

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽(𝑟𝑚𝑡−𝑟𝑓𝑡)+ 𝜀𝑡 (1)

The dependent variable is EXCOMP, which denotes the excess return of controversial industries minus the excess return of non-controversial industries. The excess returns for both industries represent the return in excess of the risk free rate, which is the continuously compounded daily return of 30-day US treasury bills. rmt represent the continuously compounded daily return of the

appropriate national market index8_{. Subtracting r}

ft from rmt creates the market risk premium, the

expected return when investing in the market portfolio. β represent assets systematic risk of being exposed to the return of the market portfolio while ε represent the error term a well-behaved random disturbance with mean zero. The CAPM implies that the level of market risk exposure, as measured by β, should be compensated by the market risk premium. The main interest of the CAPM relates to the value of Alpha (αp), which measures the performance of a portfolio earned in excess of (positive alpha)

or below (negative alpha) the market portfolio.

3.1.1.2 Fama-French Three-Factor Model (FF-3)

The validity of the CAPM, however, diminished after number of empirical studies shows anomalies left unexplained by systematic risk (Fama & French, 1992). For example, Black, Jensen and Scholes (1972) reported that portfolios with low beta stocks earn higher returns than implied by the CAPM and show a positive alpha while portfolios that consist of high beta stocks earn lower returns than implied by CAPM and show a negative alpha. As such, the risk-rewarded relation is not as steep as the CAPM intend, which is often referred to as the beta anomaly (Barrosso & Maio, 2017). Fama and French (1993) extended these results and converted the CAPM into a multi-factor model by introducing two additional factors beyond systematic rick that control for abnormal returns from investing in small stocks (SMB: Small-minus-big capitalization portfolio) and from firms with high book-to-market ratios (HML: High-minus-low book-book-to-market portfolio). The additional factors improve the explanatory power of evaluating stock performance by incorporating both firm size and firm value within the original CAPM.

(18)

18

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽(𝑟𝑚𝑡−𝑟𝑓𝑡)+ 𝛾𝑝𝑆𝑀𝐵𝑡+ 𝛿𝑝𝐻𝑀𝐿𝑡+ 𝜀𝑡 (2)

The firm size factor (SMB) is computed in equation 2a while the value factor (HML) is computed in equation 2b. Fama and French (1992, 1993) create two categories based on the market capitalization (50-50) into Small (S) and Big (B), which is divided into three valuation groups based on the book-to-market ratios. “Growth” (G), represent the lowest 30%, “Neutral” (N) represent the middle 40% while the remaining 30% are the “Value” (V) category, which represent high the book-to-market ratios (Richey G. , 2017)

𝑆𝑀𝐵 = 1/3(𝑆𝑚𝑎𝑙𝑙 𝑉𝑎𝑙𝑢𝑒 + 𝑆𝑚𝑎𝑙𝑙 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝑆𝑚𝑎𝑙𝑙 𝐺𝑟𝑜𝑤𝑡ℎ) − 1/3(𝐵𝑖𝑔 𝑉𝑎𝑙𝑢𝑒 + 𝐵𝑖𝑔 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝐵𝑖𝑔 𝐺𝑟𝑜𝑤𝑡ℎ)

(2a)

𝐻𝑀𝐿 = 1/2(𝑆𝑚𝑎𝑙𝑙 𝑉𝑎𝑙𝑢𝑒 + 𝐵𝑖𝑔 𝑉𝑎𝑙𝑢𝑒) − 1/2(𝑆𝑚𝑎𝑙𝑙 𝐺𝑟𝑜𝑤𝑡ℎ + 𝐵𝑖𝑔 𝐺𝑟𝑜𝑤𝑡ℎ) (2b)

3.1.1.3 Carhart four factor Model (C4)

The FF3 significantly improved the explanatory power of the original CAPM in evaluating stock performance. Nonetheless, the three factor model was not able to explain the one-year momentum anomaly as detected by Jegadeesh and Timan (1993). Carhart (1997) considered the anomaly by Jegadeesh and Titman (1993) and presented an expansion of the three-factor model by introducing a “momentum” factor, or “hot hand” factor. The momentum factor (MOM) control for the ability to achieve a superior performance by buying stocks that performs well (winner) and sells stocks that perform poorly (losers) within the last 3 to 12 months (Richey, 2017). The momentum factor (MOM) considers the return of two high prior return portfolios minus the average return of two low prior return portfolios (See equation 3a) (Richey, 2017). The extension to four factors is often referred to as Carhart’s four-factor model and is applied in this study as presented by equation 3.

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽𝑝(𝑟𝑚𝑡− 𝑟𝑓𝑡) + 𝛾𝑝𝑆𝑀𝐵𝑡+ 𝛿𝑝𝐻𝑀𝐿𝑡+ 𝜇𝑝𝑀𝑂𝑀𝑡+ 𝜀𝑡 (3)

Similar to the first two models, the dependent variable is EXCOMP, which is the return of a controversial stock over the risk free rate net over the return of comparable non-controversial stock over the risk free rate.

𝑀𝑂𝑀 = 1/2(𝑆𝑚𝑎𝑙𝑙 𝐻𝑖𝑔ℎ + 𝐵𝑖𝑔 𝐻𝑖𝑔ℎ) − 1/2(𝑆𝑚𝑎𝑙𝑙 𝐿𝑜𝑤 + 𝐵𝑖𝑔 𝐿𝑜𝑤) (3a)

3.1.1.4 Bet Against Beta factor

(19)

19

consistently provide negative (positive) risk adjusted returns (Barrosso & Maio, 2017). More recently, Frazzini and Pedersen (2014) considered the beta anomaly and that a “Bet Against the Beta” (BAB) strategy that consist of short high-beta stocks and long low-beta stocks, generate positive abnormal returns9_{relative to standard risk models (Bali, Brown, Murray, & Tang, 2014). In addition to the US}

equities, the BAB strategy also provide positive abnormal returns in twenty international equities, bonds, and currencies (Barrosso & Maio, 2017). Furthermore, Asness, Frazzini, and Pederesen (2014) found similar results for industry profiles, and show that, in contrast to other equity anomalies, the returns of exploiting the beta anomaly seems to be robust to transaction costs. Frazzini and Pedersen (2014) attribute the beta anomaly to leverage constrained investors (e.g. pension and mutual funds) that purchase and overweight (under-weight) high-beta (low-beta) stocks in their portfolios in order to try and boost their expected returns. As a result, the prices of high-beta (low-beta) stocks increase (decrease), which result in the low-beta stocks to generate positive risk-adjusted returns. These empirical results of the BAB strategy and the leverage constrained investors resulted in the creation of a BAB factor, which control for the excess return differential between low and high-beta stocks. The BAB factor is constructed to be market neutral with a weighted average beta of one. The BAB factor will be analysed as addition to Carhart four market model (equation 4) and Fama and French five-factor model (see equation 6), which is in accordance with Blitz and Fabozzi (2017).

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽𝑝(𝑟𝑚𝑡− 𝑟𝑓𝑡) + 𝛾𝑝𝑆𝑀𝐵𝑡+ 𝛿𝑝𝐻𝑀𝐿𝑡+ 𝜇𝑝𝑀𝑂𝑀𝑡+ 𝜈𝑝𝐵𝐴𝐵𝑡+ 𝜀𝑡 (4)

The BAB will be applied as provided by equation 4 and the equation of the BAB factors is provided by equation 4a.

𝐵𝐴𝐵(𝑅𝑡+1𝑠 ) = 1 𝛽𝑡𝐿 (𝑟𝑡+1𝐿 − 𝑟𝑓) − 1 𝛽𝑡𝐻 (𝑟𝑡+1𝐻 − 𝑟𝑓) (4a)

𝛽𝑡𝐿 and 𝛽𝑡𝐻 are the estimated beta based on low and high beta portfolio while the 𝑟𝑡+1𝐿 and the

𝑟𝑡+1𝐻 are the actually returns of these portfolios for the applicable period (Frazzini & Pedersen, 2014). 3.1.1.5 Fama-French Five-Factor Model

The latest development in asset pricing model is based on Titman, Wei, and Xie (2004) & Novy-Marx (2013), which shows that the existing Three-Factor model lack the variation in average returns on profitability and investment. As a result, Fama and French (2015) extended the explanatory power of the existing model by adding a profitability (RMW) and investment (CMA) factor. The profitability

9_{Frazzini and Pedersen (2014) provide a given a Sharpe ratio of 0.78 between 1926 and 2012, which almost} doubles that of the US equity market and is 40% higher than the momentum factor for the same period (Frazzini & Pedersen, 2014). The Sharpe ratio is a widely used performance measure that accounts for both systematic and unsystematic risk and is calculated as follows; SRi=

∆𝑟_𝑖

(20)

20

factor is computed in equation 5a while the investment factor is computed in equation 5b. The RMW consider the difference in weak and robust profitability of stock returns while CMA considers high investment firms as aggressive and low investment firms as conservative. The global size (SMB), value (HML), momentum (MOM), profitability (RMW) and investment (CMA) factors are obtained from the extended Frazzini and Pedersen dataset of the AQR Capital Management Database (2017).

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽(𝑟𝑚𝑡− 𝑟𝑓𝑡) + 𝛾𝑝𝑆𝑀𝐵𝑡+ 𝛿𝑝𝐻𝑀𝐿𝑡+ 𝜌𝑝𝑅𝑀𝑊𝑡+ 𝜏𝑝𝐶𝑀𝐴𝑡+ 𝜀𝑡 (5)

𝑅𝑀𝑊 = 1/2(𝑆𝑚𝑎𝑙𝑙 𝑅𝑜𝑏𝑢𝑠𝑡 + 𝐵𝑖𝑔 𝑅𝑜𝑏𝑢𝑠𝑡) − 1/2(𝑆𝑚𝑎𝑙𝑙 𝑊𝑒𝑎𝑘 + 𝐵𝑖𝑔 𝑊𝑒𝑎𝑘) (5a)

𝐶𝑀𝐴 = 1/2(𝑆𝑚𝑎𝑙𝑙 𝐶𝑜𝑛𝑠𝑒𝑟𝑎𝑡𝑖𝑣𝑒 + 𝐵𝑖𝑔 𝐶𝑜𝑛𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑣𝑒) − 1/2(𝑆𝑚𝑎𝑙𝑙 𝐴𝑔𝑔𝑟𝑒𝑠𝑠𝑖𝑣𝑒 + 𝐵𝑖𝑔 𝐴𝑔𝑔𝑟𝑒𝑠𝑠𝑖𝑣𝑒)

(5b)

Finally, equation 6 is a combination of the BAB factor and Fama French five-factor model and is in accordance with Blitz and Fabozzi (2017), which shows an insignificant alpha but a sharp increase in explanatory power after the inclusion of the BAB factor.

𝐸𝑋𝐶𝑂𝑀𝑃𝑡= 𝛼𝑝+ 𝛽(𝑟𝑚𝑡− 𝑟𝑓𝑡) + 𝛾𝑝𝑆𝑀𝐵𝑡+ 𝛿𝑝𝐻𝑀𝐿𝑡+ 𝜈𝑝𝐵𝐴𝐵𝑡+ 𝜌𝑝𝑅𝑀𝑊𝑡+ 𝜏𝑝𝐶𝑀𝐴𝑡+ 𝜀𝑡 (6)

Overall, hypotheses null states that the intercept is zero and there is no difference between controversial and non-controversial stocks. Hypothesis 1, in contrast, expects a positive alpha that indicates after a systematic adaption of each model to highlight and specify what factors have the most contribution and/or impact on the performance of controversial industries versus non-controversial industries. A positive alpha indicates that controversial stocks outperform comparable non-controversial industries. In addition, the null Hypothesis expect the alpha of non-controversial stocks to be equal among countries with higher investment freedom, government integrity, % of population using the internet, and business disclosure extent. The alternative Hypotheses 1a, b, c, and d, expects a higher alpha for controversial stocks in countries with high levels (top 33% of all countries) of investment freedom, government integrity, % of population using the internet, and business disclosure extent.

3.1.2 Valuation analysis

(21)

21

stocks) as presented by Durand et al., (2013b). In accordance with Hong and Kacperczyk (2009) and Hong, Kubik, and Stein (2008), the dependent variable (Valuation) represent three different valuation ratios namely; the natural logarithm of market-to-book (LNMB), price-to-earnings ratio (LNPE), and the price-to-earnings before taxes, interest, deprecation, and amortization ratio (LNPEBITDA). In contrast to the aforementioned studies this paper makes a direct comparison between the valuation of controversial and non-controversial industries instead of comparing to the overall market. Similar to the previous model a direct comparison allows this paper to give an improved indication on the impact of social norms on investors. The valuation values of controversial stocks are compared with the same non-controversial industry as applied in the first time-series regression for hypothesis 1 (See section 4.2 for further explanation on the pairing of controversial and non-controversial industries). Similar to equation 7, the independent variables is CDUM and equals one when the firm operate in a controversial industry and zero otherwise. The control variables for Eq. (8) are in accordance with Hong and Kacperczyk (2009) and Durand et al., (2013b), and represent firm characteristics and consist of Return On Equity (ROE), the next three years of ROEs (FROE, F2ROE, and F3ROE), expenditures in research and development (R&D) as fraction of sales (RDSALES), and a dummy variable when the information on R&D is missing (RDMISS). Finally ε represent the error term.

𝑉𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑖𝑡= 𝛽1+ 𝛽2𝐶𝐷𝑈𝑀𝑖,𝑡+ 𝛽3𝑅𝑂𝐸𝑖,𝑡+ 𝛽4𝐹𝑅𝑂𝐸𝑖,𝑡+ 𝛽5𝐹2𝑅𝑂𝐸𝑖,𝑡+ 𝛽6𝐹3𝑅𝑂𝐸𝑖,𝑡 + 𝛽7𝑅𝐷𝑆𝐴𝐿𝐸𝑆𝑖,𝑡+ 𝛽8𝑅𝐷𝑀𝐼𝑆𝑆𝑖,𝑡+ +𝜀𝑖𝑡

(8)

The null hypothesis expects the values of LNMB, LNPE, and LNPEBITDA for controversial industries to be equal to comparable non-controversial industries. Hypothesis 2, in contrast, expects the three coefficients to be smaller for CDUM portfolio, which indicate a lower valuation of controversial sin stocks. In addition, the null Hypothesis expect the alpha of controversial stocks to be equal across all countries. Hypothesis 2a, b, c, and d, expects the valuation of controversial stocks to be higher in countries with high levels of investment freedom index, government integrity, % of population with access to internet, and business extent discourse.

3.2 Data and Sample Selection

3.2.1 Selection of controversial stocks

(22)

22

Venkatachalam, 2011). The “Triumvirate of Sin”, however, is rather limited as other issues could be considered as (potentially) controversial by responsible investors too (Trinks and Scholtens, 2017). Lobe and Walkshäusl (2016) identifies a similar problem as they could not find a widely accepted definition of “vice” stocks, which resulted in an analyse of the exclusion criteria’s of 32 SRI funds across nine regions (e.g. Global, US, Eurozone, Europe, United Kingdom, Australia, Japan, Canada, and Africa). They apply a disapproval voting strategy10_{to highlight the most common exclusion criteria’s, which}

identified adult entertainment, alcohol, defence, gambling, nuclear energy (incl. uranium) and tobacco as the most common exclusion criteria items among SRI funds (Lobe & Walkshäusl, 2016). This combination is often referred to as the “Sextet of Sin”. This paper follows Lobe and Walkshäusl (2016) and will analyse the “Sextet of Sin” industries but with a separation of the uranium and nuclear energy industry as both industries have different industry comparable. The main reason to follow Lobe and Walkshäul (2016) is based on the global exclusion criteria analysis as it covers nine different regions all represented in the sample of this paper.

The author of this paper is aware that the consideration of (potential) controversial issues can be debated both, as the products produced in each industry serve a certain purpose. For instance, nuclear energy can be of vital importance for the provision of electrical power in a society. Moreover, alcoholic beverages play an important and profitable role in the hospitality industry. However, the aim of this paper is not to individually judge each industry but instead to provide evidence whether industries that are often shunned by socially responsible investors, outperform comparable non-controversial industries.

3.2.2 Data

The study consists of an international sample for the period of 1965-2016 and uses the screening methodology of Thomson DataStream by Ince and Porter (2006)11_{. Monthly financial data}

from January 1965 to December 2016 is collected from Thomson Reuters DataStream. All information is denoted in US Dollars and is based on Thomson Reuters DataStream exchange rates. The monthly returns are measured as the natural logarithm of a stocks Total Return index t=0 divided by the index on t = -1. The Total return index represents a stocks theoretical growth in value starting at 100. The Total return index assumes that dividends are reinvested. This is in line with Lobe and Walkshäul (2016) and Trinks and Scholtens (2017). Moreover, this paper circumvents survivorship bias12_{by including all}

delisted stocks until they disappear in the sample. Firms operating in controversial industries are selected through the use of four industry classification systems namely; Thomson DataStream Industry

10_{Disapproval voting strategy is a reverse of the approval voting technique by Brams and Fishburns (1978)} 11_{Returns are, in accordance with Ince and Porter (2006) winsdorized at 0.5% and 99.5% level}

(23)

23

Classification Benchmark (ICB), US Standard Industrial Classification (SIC), Thomson Reuters Business Classification Code (TRBC) and the North American Industry Classification System (NAICS). The first selection considers the ICB and SIC codes that create an initial sample of firms operating in controversial industries with clear industry codes (e.g. Alcohol industry consists of “Brewers” (3535) and “Distillers and Vintners” (3533), “Tobacco” (3785), “Defence” (2717), and “Gambling” (5752)).

Unfortunately, the ICB and SIC system does not have a sector code for adult entertainment, nuclear energy, and uranium mining/extraction, which lead to additional approaches to identify firms operating in one of the industries. At first, both the NAICS and TRBC were identified as industry code systems that do have separate codes for the Nuclear energy and uranium industry.13_{The NAICS codes}

were used to obtain an initial sample from Orbis, which is cross-checked with three different indexes consisting of uranium and nuclear energy oriented companies (DAX World nuclear power index, DAX global Nuclear energy index, and S&P global nuclear energy index (lifted November 2016)). Afterwards, the industrial securities identification numbers (ISIN) provided by Orbis are implemented in Thomson DataStream and allow for the extraction of TRBC codes, to implement an additional justification to identify potential misfit. Finally, firms with irregularities among the NAICS and TRBC codes, are individually checked to justify the consistency and quality of the sample.

Finally the adult entertainment industry, which unfortunately does not have any industry code among the four industry classification systems, is selected in accordance with Lobe and Walkshäusl (2016), through a selection of adult and sexual related content searches on Orbis. This resulted in a list that is cross-checked with two published books on “sin” stock investment strategy by Ahrens (2004) and Waxler (2004), to confirm the accuracy of the final list.

The stocks are compared through performance factors and cross sectional regression with comparable “non-controversial industries”. Non-controversial industries are, in accordance with Hong and Kacperczyk (2009), selected on similarities among industry classification codes (ICB). The selection of non-controversial industries for the nuclear energy, uranium and adult entertainment industry is based on existing literature. Overall, it resulted in seven combinations of controversial and non-controversial industries.

1. The Adult entertainment (no ICB code) is compared with Media (5550), which include Broadcasting & entertainment (5533), Media Agencies (5555) and Publishing (5557). Media is

(24)

24

considered as the majority of listed firms in adult entertainment focus on the production and distribution of pornographic images and videos.

2. The alcohol industry (3533 and 3535) is compared to Soft Drinks (3537)

3. Tobacco industry (3780) is linked to Food Producers (3570), which include Farming & Fishing (3573) and Food Products (3577).

4. The Weapon industry is represented by the Defence industry (2717) and compared to the aerospace industry (2713). All firms denoted as aerospace are individually checked for business segment earnings to consider the percentage of revenue earned, if applicable, in the defence industry. Firms earning less than 20 percent of their revenue from the defence industry and have an ICB code of 2713 remain relevant for the comparison and are considered as aerospace company. Firms with an ICB code of 2713 but earn more than 20 percent but less than 50 percent are still considered as Aerospace companies but excluded from the comparison due to high interdependency among both industries. Firms denoted as aerospace but earn more than 50 percent in the defence industry are manually changed to 2717 and are included in the defence industry. Finally firms with no available information on business segment, nor give indications of highly defence oriented industries, remain in the 2713 aerospace industry and applicable for the comparison

5. Gambling (5752) is linked through Travel & Leisure (5750), which considers Airlines (5751), Recreational services (5755), Restaurants and Bars (5757) and Travel & Tourism (5759). Hotels (5753) are also part of Travel &leisure sector but excluded for comparison as many listed hotels also exploit casinos.

6. Nuclear energy (no ICB code) is compared to conventional electricity (7535). Nuclear energy is considered to be part of conventional electricity (7535) and similar to the Aerospace industry, all firms are cross-checked to make sure they do not operate in nuclear energy.

7. Uranium (no ICB code) is compared to Industrial Metal and Mining (1750), which considers Aluminium (1753), Nonferrous Metals (1755), and Iron & Steel (1757). Uranium is considered to be part of the nonferrous metal industry, which again, is individually crossed checked to select companies that operate in Uranium the firms that don’t. After the application of the TRBC, multiple relevant industries were considered in the Mining industry (1770). As such, a cross-check for firms operating in either Aluminium, nonferrous metals, and Iron & steel are included in the existing sample.

(25)

25

one additional cluster from existing literature separately, namely; “Triumvirate of Sin” (e.g. Alcohol, Tobacco, and Gambling) as applied by Hong and Kacperczyk (2009) and the majority of the existing literature on controversial industries (See Table 1).

Overall, the sample period differs among models due to data availability limitations and the use of country specific control variables instead of global and regional because Griffin (2002) highlights that Fama and French factors are highly country specific and not regional specific as multiple variations occur among countries. The CAPM is based on the national market index or the S&P500 in case Thomson Reuters DataStream does not provide information on a national market index. Overall, 34 countries use the S&P500 as market return and 68 countries apply a national market index (see Appendix B for an overview). An additional CAPM (e.g. CAPM2) is calculated for countries with pre-market index information (e.g. stocks with return information before a national pre-market index is introduced). In total 22 countries have an additional CAPM with S&P500 as market index that allows for a complete coverage of the available information. Moreover, Appendix C provides an overview of the models and time period covered by each country and show that out of 102 countries, 2 countries (e.g. United States and Japan), cover the full range models (e.g. CAPM, FF3, C4, BAB, and FF5), 24 countries (incl. the United States and Japan) provide CAPM, FF3, C4, and BAB, and 75 countries provide an actual CAPM analysis between controversial and non-controversial industries. The remaining 27 countries do not provide a country specific analysis but contribute to the country specific variable analysis, which uses the entire sample to select the top, mid, and lowest 33%. The main reason for the differences in sample period and available factors is due to data limitations. FF3 factor, Carhart Momentum Factor and BAB factor by Frazzini and Pedersen (2014), is limited to 23, mainly developed, countries with different sample periods. Nonetheless, the full model (e.g. CAPM, FF3, C4, BAB, and FF5) is applied within the United States from January 1973 till December 2016 and Japan from September 1990 till December 2016, which allows for direct comparison with Blitz and Fabozzi (2017) whom cover the same period.

3.2.3 Country Variables

(26)

26

on the investors. As such, this paper selected four investor related country factors to analyse the impact of country characteristics on the performance of controversial stocks. The additional factors are; Investment Freedom index, the Government integrity index, % of population with access to Internet, and the business disclosure index. The first two variables are extracted from the economic freedom index14_{while the last two variables are extracted from the Worldbank database.}

The investment freedom index considers the number of restrictions individuals and company’s encounter when moving capital both abroad and domestic. The government integrity index operates as an indication on the level of systematic corruption applied in government institutions. In some countries it may reflect a traditional interaction but by allowing people to gain governmental benefits interfere with the principles of an economically free society (The Heritage Foundation, 2017). The Investment Freedom Index and Government Integrity index are based on Cumming and Johan (2007), which argue that different regions and countries have different legal standards and social norms in regards to SRI policies, which impact the likelihood of an investor to apply such a strategy. In particular, Asian and less-developed countries with low levels of government integrity tend to have less affection towards SRI and have weaker standards for SRI than European and North American countries (Dunning, 2002; Doh, Rodriguez, Uhlenbruck, & Eden, 2003). This lack of affectation towards SRI may explain the findings of Durand et al., (2013b) as to why investors from the Pacific-Basin are willing to pay a premium for controversial stocks and why Hong and Kacperczyk (2009) present undervalued controversial stocks in the United States. In addition, it shows the impact of governmental signs on the social norms of the society and investor behavior because Durand et al., (2013b) show significant government holdings among East Asian countries in domestic controversial industries while Dunning (2002) show they have a weaker standard towards SRI than other developed nations and regions. Both studies imply that governmental signs have impact on the social norm and the behaviour of investors. Moreover, both government integrity and investment freedom are based on governmental decisions, which justify the selection for both country variables

The last two variables, % of population with access to internet and business extent disclosure are based on a recent article by the Economist (2017) that discussed the increase in socially responsible investing activities. They highlights the relevancy of Internet on the shaping of social norms and investment behavior of the millennial generation born in the 1980s and 1990s. This millennial generation grew up in a digital age that made them more aware of global issues and increase the likelihood of using electronic investment tools. The combination of these factors resulted in a new

(27)

27

generation of investors that believe they can improve the world by investing in industries and stocks that meet their environmental, social and governance criteria’s.15_{As a result a new SRI generation is}

born that made SRI a mainstream investment strategy (Belghitar, Clark, & Deshmukh, 2017). Moreover, the Economist (2017) argue that technology development made SRI available for the ordinary investors as improvements in the computing power simplify the assessment of companies actions, as more company data becomes publically available due to both voluntarily and statutory reporting initiatives. In fact, the European Union and 12 stock exchanges require listed firms and pension funds to disclosure ESG information. In addition, Kim and Venkatachalam (2011) and Merton’s neglected stock theory highlight the importance of the quantity of information provided, not necessary the quality of the estimations.

Overall the Economist (2017) indicate that the rise of internet and stricter business disclosure lead to more investors to adopt a SRI investment strategy and less investors and institutions to consider controversial industries.

3.3 Descriptive statistics

This section of the paper shows the descriptive statistic for the entire sample including both controversial and non-controversial stocks across 102 countries.

(28)

28

Table 3 considers the descriptive statistics for the time-series return regressions for the period 1965-2016. An interesting feature is the difference between EXCOMP and EXCONP, which indicates that the average monthly return net over the risk free rate for a portfolio of controversial stocks is negative while the average monthly return of controversial stocks net over comparable portfolio of non-controversial stocks is positive. Moreover, the negative MKTPREM indicates a market discount for the period 1965-2016. These results are significantly lower than the US-based study of Hong and Kacperczyk (2009), which is the only study that provides descriptive statisitcs. In addition, the SMB, HML, MOM, are also lower than the average for Hong and Kacperczyk (2009). The reamining control factors (e.g. RMW, CMA, and BAB) are not provided by Hong and Kacperczyk (2009), and therefore not able to compare with the existing literature.

Variables Mean Std. Dev. Median IQR (75%-25%) min max

EXCOMP (%) 0.0009 0.1942 0.0018 0.1287 -10.5588 8.9864 EXCONP (%) -0.0019 0.1968 0.0010 0.1213 -5.8374 4.0698 MKTPREM (%) -0.0005 0.0667 0.0017 0.0617 -0.8329 1.3209 SMB (%) 0.0008 0.0323 0.0004 0.0380 -0.2563 0.3210 HML (%) 0.0044 0.0328 0.0040 0.0332 -0.2363 0.3170 MOM (%) 0.0074 0.0477 0.0091 0.0425 -0.5680 0.3400 RMW (%) 0.0057 0.0432 0.0070 0.0397 -0.3439 0.1836 CMA (%) 0.0027 0.0218 0.0016 0.0250 -0.2484 0.1360 BAB (%) 0.0090 0.0427 0.0092 0.0408 -0.4460 0.5130

This table reports all the descriptive statistics of the dependent and independent variables used in the price factor analyses. EXCOMP denotes the excess montly return net of the risk free rate for an equal-weighted portfolio of controversial stocks net over non-controversial stocks. EXCONP is the excess montly return net

of the risk free rate of controversial stocks. MKTPREM is the excess montly return of the market index net over the risk free rate. SMB is the monthly return of a portfolio long small stocks and short high stocks.

HML is the montly return of a portfolio long high book-to-market stocks and short low book-to-market

stocks. MOM is the montly return of a portfolio long past 12-month return winners (well-performing) and short past 12-month return losers (poor-performing). RMW is the monthly return of a portfolio long robust (high) operating profitability stocks and short weak (low) operating profitability stocks. CMA is the monthly

return of diversified portfolios of conserative (low) and aggressive (high) investment firms. BAB is the montly return of a portfolio long low-beta stocks and short high-beta stocks. Variables that are denoted as

ratios are divided by 100 to provide decimals results.

Time-series return regressions: 1965-2016

(29)

29

Table 4 presents the descriptive statistics of the valuation regression. The results are

windsorized at a 0.5 and 99.5% and represent 99% of the sample. The main reason to windsorize the results are the extreme maximum and minimum ROE and RDSALES values (see Appendix D for the original descriptive statistics), which incidental factors such as insolvency. The time series average of LNMTB, LNPE, and LNPEBITDA, indicate that for the entire sample, a typical firm in a typical year has a negative Market to Book ratio of -0.89 but a positive Price Earnings ratio of 1.05 and a positive Price-to-earnings before interest, taxes, deprecation, and amortization of 0.44. These figures,

including the standard deviations, are in line with the US-based study of Hong and Kacperczyk (2009). Moreover, the average ROE is negative at -1.34% and the average firm spend 3.42% of their sales on research and development. Both are, however, rather low compared to Hong and Kacperczyk (2009), which may be the result of the inclusion of the additional countries and extending the sample period to 2016. In fact, the average ROE increased to 1.24% when considering 1965 to 2007.

In addition to the overall sample, industry specific are provided in Appendix E Noteworthy mentioning are the 1.3057 average LNMTB for the tobacco industry while the remaining industries vary between an average LNMTB of 0.1168 and 0.8654. This indicates that the tobacco industry is the only industry with a positive average Market-to-Book ratio of 0.267 during 1965 and 2016. Moreover, the average LNPEBITDA of the Uranium mining industry is 2.9856 while the other studies vary

between 1.0347 and 1.9965. This, however, may be the result of the limited amount of observations (84), which is significantly smaller compared to the other industries

4 Results

This section of the paper presents the analysis of this study on the performance and valuation of controversial stocks on a global scale between the period of 1965 and 2016. The stock performance valuation consist of a time series regression while the valuation analysis consist of an unbalanced panel data.

Variables Nr. of Obs. Nr. of Firms Mean Std. Dev. Median IQR (75%-25%) min max

LNMTB 1299996 8084 0.42 1.12 1.62 1.18 -2.53 3.44 LNPE 1052364 7087 2.86 1.09 2.81 1.07 -0.92 6.67 LNPEBITDA 1019856 7431 1.55 1.61 0.40 1.48 -3.82 7.02

ROE (%) 1236766 8086 -1.34 59.94 7.22 15.45 -553.57 149.49

RDSALES (%) 393955 3993 3.42 20.97 0.19 1.05 0.00 254.30

This table reports all the descriptive statistics of the dependent and independent variables used in the valuation analyses. LNMTB is the natural logarithm’s of a stock’s Market-to-Book Ratio. LNPE is the natural logarithm’s of a stock’s Price-Earnings Ratio. LNPEBITDA is the natural logarithm’s of a stock’s Price Earnings Before Taxes, Interest and after Deprecation ratio. ROE is the the Return of Equity. RDSALES is

the fraction of research and development expenditures to a firm over sales. Results are windsorized at 0.5% and 99.5%

Table 4

Descriptive statistics

(30)

30

4.1 Time Series Regression

This section test whether controversial stocks outperform comparable non-controversial stocks and whether specific country variables have an impact on the performance of both type of stocks. An overview of the control factors covered by each country is provided by Appendix C. In general, two countries (e.g. US and Japan) cover all seven control factors (e.g. MKT-RF, SMB, HML, MOM, RMW, CMA, and BAB), 24 countries (incl. Japan and US) cover five control factors (e.g. MKT-RF, SMB, HML, MOM, and BAB), and 75 countries16_{(incl. the previous 24 countries) control for the exposure to the}

market (e.g. MKT-RF), also refereed to as the CAPM analysis. The remaining 27 countries do not provide a country specific analysis but, instead, are considered for the country variable analysis. Tables included in the main text either cover the United States or a global perspective, while the tables of the remaining 74 countries are provided in the Appendix section.

4.1.1 Fama-French 5 Factor Model analysis

The only two countries that provide data on the final two factors of the Fama-French 5 factor model (e.g. RMW & CMA) are the United States and Japan.

Controversial Industries: does it pay to ignore social norms?