3.1. Data Source
In this study, we use ESG performance scores from the MSCI ESG Ratings database, the largest provider of ESG data to investors. MSCI data are reliable insofar as they incorporate several sources: both the disclosures by firms analyzed and the information disclosed by external sources (in particular, governments, non-governmental organizations and the media) [
6].
The distinction needs to be made between exposure to ESG risks and management of ESG risks. MSCI considers exposure to risk to come from the specific industry, location of assets and income, and other measures such as raw material outsourcing. Risk management, on the other hand, represents the strategies and measures used by the firm to manage its risk level and opportunities. More important, MSCI ESG Ratings assess ESG performance through continuous measures (scores) of 37 ESG Key Issues capturing firms’ risk exposure and respective risk management practices. These issues then feed into the scores of the three pillars (environmental, social and governance) which make up the overall performance score. Therefore, contrasting with previous studies, such as [
17], we use a measure of ESG performance that contrasts firms’ ESG risk exposure to their ESG management practices. This is important because controversial firms (or sectors) with higher risk exposure need to compensate with more responsible practices to mitigate those risks. This approach has been recommended in more recent studies about ESG performance, see for example the study by [
25]. For instance, the results by [
22] also corroborate the idea that assessment and comparison of sin stocks’ ESG performance to less controversial sectors, should consider the level of risk exposure (concerns, not only the strengths). It should be noted that the following analysis relies on the scores before the adjustment for industry, since it is carried out only once, i.e., on the final score.
These two aspects are examined in turn within the framework of this study. More precisely, we anticipate that firms identified as sin stocks will have a greater exposure to risks than the control group, as well as better management of these risks and, consequently, an ESG performance similar to that of comparable firms. We thus formulate the following forecasts.
Due to the very nature of their business, we expect that sin stocks will have much greater exposure to ESG risks than the rest of the market. More specifically, the risks of controversy, environmental damage, scandals, boycotts, etc., are much higher than for firms operating in more traditional industries.
To mitigate the increased risks specific to their business and market expectations, we believe that sin stocks will exhibit better adapted risk management practices. All firms want to avoid scandals, but those more sensitive to controversies will have to put more effort into making this risk reasonable, compared to firms operating in more traditional sectors.
The final result, the ESG letters rating assigned by MSCI (which ranges from “best” (AAA) to “worst” (CCC)) [
6], is thus expected to be relatively similar between sin stocks and comparable firms operating in more traditional sectors, our control group.
3.2. Composition of the Treatment Group
In accordance with the literature discussed above and evidence documented in other studies, we selected three sectors to compose the subsample of sin stocks: casinos, alcohol and tobacco. Here are, briefly, the reasons which led us to exclude other sectors usually associated with sin stocks:
Nuclear energy: Some analysts consider that companies producing energy from nuclear sources are simply companies in the energy sector. For this reason, we prefer to exclude firms of this sector which are not unanimously seen as sin stocks.
Weaponry: It is tremendously difficult to draw the fine line between high tech companies and weaponry producers. The following three examples illustrate the difficulty of reaching a consensus: (1) Boeing, an aircraft manufacturer, derives about a third of its revenues from the military sector and the balance from the commercial sector. (2) Thales, a French multinational operating in aeronautics, derived 51% of its revenues from the military business in 2018 [
26]. (3) L3Technologies is a firm that develops aircraft avionics instruments (intelligence, surveillance and reconnaissance) and two-thirds of its revenues come from the military business [
27]. In short, it is very arbitrary to determine which of these firms are sin stocks, because it is indeed a continuum. This therefore led us to exclude the weapons industry from our sample of sin stocks, the treatment group.
Adult entertainment: This sector was excluded due to the very limited number of public firms. In addition, identification of firms operating in this sector is difficult since this sector is not an industry identified by MSCI dataset. This classification problem was also highlighted by [
20] as well as by [
21].
Therefore, our target sample for the treatment group—i.e., sin stocks—comprises all firms operating in the casinos, alcohol and tobacco sectors as identified by the MSCI ESG ratings database, regardless of the country. The most recent ESG scores available when this study was developed (December 2016) were used for the purposes of our empirical analysis.
3.3. Composition of the Control Group
In order to examine whether our sample firms of sin stocks have a significantly different ESG performance, their ESG score needs to be compared to that of a control group of similar firms operating in more traditional sectors. This pairing is a critical step, as it allows to control for several firm-level attributes that can explain firms’ ESG performance. For example, the control group must comprise firms of similar size and financial performance [
28]. Otherwise, the difference between sin stocks and the control group could be attributable simply to the size and financial constraints of the firms studied. More precisely, larger firms are assumed to exhibit better ESG performance since they have more resources, greater variety of stakeholders and more economies of scale for sustainable development programs. Further, this control group must represent a multitude of sectors. If a sector is over-represented in the control group, its ESG performance may not be comparable to that of sin stocks.
Therefore, following a similar approach to that of the study by [
24,
28], a matched sampling approach is adopted to compose the control group. In this sense, each firm designated as a sin stock was paired with a firm from a more traditional sector whose characteristics are as similar as possible, considering the following five attributes:
Location: The country where firms operate can greatly influence their exposure to ESG risks as well as their response (risk management). For instance, some countries adopt policies in terms of diversity, environmental protection or the fight against corruption that reduce the exposure of firms to these issues.
Size: Firm market capitalization is used insofar as larger firms are expected to have more stakeholders creating an incentive to adopt more responsible practices. In addition, they have more resources that allow for a more complete and fleshed out disclosure of sustainable practices.
Financial leverage: The valuation by the ratio of Total Assets/Equity aims to control the capital structure of the firm. We believe that the source of funds (debt or equity) can impact business incentives. For example, debt financing creates fewer expectations of responsible practices than equity financing. The Total Assets/Equity ratio of the firms was compared.
Return on Assets: We believe that a firm’s financial performance creates different incentives for adopting responsible practices. Among other things, a firm generating losses will have more pressing concerns than sustainable development. The average return for the past five years was compared, computed as Net Income/Total Assets.
Sector exclusions: In the control group, the financial sector has been excluded. As documented in previous studies, financial institutions face very different regulation and incentives and it is therefore preferable to exclude them in a context of matched sampling.
3.4. Sampling Strategy
In order to build the two subsamples described above (treatment and control groups), the starting point of our sampling strategy was all firms to which MSCI had assigned an ESG score on 1 December 2016; 11,289 firms met this criterion. A second database (WRDS, Wharton Research Data Services) was combined with the MSCI database in order to generate the firms’ NAICS and SIC codes. A criterion was then applied to retain the firms whose industry is either Casinos and Gaming, Tobacco or Alcohol. For the Alcohol sector, firms were retained if their NAICS code was 312120 (Breweries), or if their SIC code was 2084 (Wines, Brandy and Brandy Spirits) or 2085 (Distilled and Blended Liquors). The following firms and their subsidiaries were added manually to the sin stocks subsample, since they are well-known brewers: Anheuser-Busch, Diageo, SABMiller and Molson Coors. As a result, 123 sin stocks were identified with ESG scores for the year of 2016 in the MSCI database.
We then paired each sin stock with a comparable firm operating in a more traditional business sector. The paired control group needed to present similarity for the four attributes explained above that are as close as possible to that of the sin stocks: the firm’s country, market capitalization (size), financial leverage and average return on assets over the past five years. This approach is similar to the pairing methodology used in previous literature [
28,
29].
The Worldscope dataset was used to collect information on the financial fundamentals used to measure the four criteria applied in the paired matching sample procedure. Of the 123 stocks identified in the MSCI dataset, only one did not have the identifier code (the ISIN code) and could not be matched in the Worldscope. For the 122 remaining sin stocks, 115 entries contained financial information. Worldscope was able to identify the home country of all of these 115 firms, but was unable to provide market capitalization, leverage or return on assets for some firms. It was deemed necessary for firms in the control group to have at least the firm’s country and two of the financial attributes. Therefore, the final subsample of sin stocks is made up of 81 firms.
The list of firms that can be included in the control group has been established according to the following criteria: (1) the firm must have an ESG score in the MSCI database, (2) it must not be included in the 123 sin stocks identified previously and (3) it must not belong to the financial sector. More precisely, eight industries were excluded according to MSCI industry classification: Banks, Life & Health Insurance, Asset Management & Custody Banks, Consumer Finance, Property & Casualty Insurance, Diversified Financials, Multi-Line Insurance & Brokerage, and Investment Banking & Brokerage). There was a total of 5295 firms meeting these criteria. The next step was to match each of the 81 sin stocks with a comparable firm (control group). As much as possible, the paired firm came from the same country and had the three closest attributes (according to a Z-score calculation, allowing us to calculate which firm is most similar to the sin stock in terms of standard deviation from the average of the attribute). Two of the 81 sin stocks had to be excluded as they had very specific attributes that prevented them from being matched with any firm in the control group. These firms are Multi Soft II, Inc. with an unknown market capitalization, leverage of −0.07 and return on assets of −956.76%; and Multi Solutions II, Inc. whose market capitalization was 379,915 US$, leverage of −0.08 and return on assets of −913.23%.
The final sample is therefore made up of 158 firms, that is 79 sin stocks individually matched with 79 comparable firms belonging to more traditional sectors.
Table 1 presents the distribution of the industries of the two subsamples. As expected, the control group is very diversified in terms of industries and not overweighted towards any single industry. In fact, a total of 36 different industries are represented in our control group. For reasons of space, in
Table 1 we grouped 31 industries in “Other Industries” because they have three and less firms included in our control group. In
Table 1, details are omitted to conserve space but, “Other Industries” comprises 50 firms belonging to 31 different industries as follows: Utilities (3), Wireless Telecommunication Services (3), Food Products (3), Paper & Forest Products (3), Professional Services (3), Steel (3), Oil & Gas Refining & Marketing (2), Commodity Chemicals (2), Energy Equipment & Services (2), Integrated Telecommunication Services (2), Metals and Mining—Precious Metals (2), Retail—Food & Staples (2), Road & Rail Transport (2), Aerospace & Defense (1), Air Freight & Logistics (1), Broadcasting, Cable & Satellite (1), Building Products (1), Commercial Services & Supplies (1), Construction & Engineering (1), Construction Materials (1), Containers & Packaging (1), Electrical Equipment (1), Electronic Equipment, Instruments & Components (1), Hotels & Travel (1), Marine Transport (1), Media (1), Metals and Mining (1), Specialty Chemicals (1), Technology Hardware, Storage & Peripherals (1), Textiles, Apparel & Luxury Goods (1), and Trading Companies & Distributors (1) As expected,
Table 2 (high
p-values) shows that the two subsamples are statistically similar.