Corporate Social Responsibility and Value Creation: Evidence from
Mergers & Acquisitions
by
Remco Westerbeek
University of Groningen Faculty of Economics and Business
Table of Contents
Abstract ... 3
1. Introduction ... 4
2. Research framework ... 7
2.1 CSR and value creation ... 7
2.2 M&A and value creation ... 8
2.3 Relationship between CSR and M&A ... 10
3. Methodology and data ... 12
3.1 Methodology ... 12 3.1.1. Research method ... 12 3.1.2 Dependent variable ... 14 3.1.3 Independent variables ... 15 3.1.4 Control variables ... 16 3.1.5. Sensitivity analysis ... 17 3.2 Data collection ... 18 4. Empirical results ... 19 4.1 Sample characteristics ... 19 4.2 Univariate analysis ... 22
4.3 Cross-sectional regression analysis ... 24
4.3.1 Cumulative abnormal returns ... 24
4.3.2 Outperformance ... 25
4.4 Data comparisons ... 28
4.4.1 US-based sample ... 28
4.4.2. Split sample UK vs. Eurozone ... 28
4.5 Sensitivity analysis ... 29 5. Discussion ... 30 6. Conclusion ... 32 7. Bibliography ... 34 Appendix A: Correlations ... 40 Appendix B: Comparisons ... 41
Corporate Social Responsibility and Value Creation in Mergers &
Acquisitions
Abstract
Using a sample of 714 mergers and acquisitions (M&A) in Europe from the period 2010-2016, this paper examines whether corporate social responsibility (CSR) performance creates value for the shareholders of acquiring firms. It is hypothesized that CSR, measured by ESG scores of the ASSET4 database, indeed generates positive abnormal returns around M&A announcements. Findings suggest that low CSR acquirers may generate higher abnormal returns around
announcements than high CSR acquirers, converse to hypotheses. However, multivariate results are insignificant, sensitive to changes in measurement and differ for a US-based sample and for UK-based and Eurozone-based subsamples. Hence, results may suggest that there is no
substantial relationship between CSR and value creation. Another explanation may be that this relationship differs substantially per country or region. It remains unclear whether CSR is a process of shareholder value creation or not.
Keywords: Corporate Social Responsibility (CSR), Stakeholder Theory, Value Creation,
1.
Introduction
During the last few decades, corporate social responsibility (CSR) has garnered much attention from media and an extensive body of diversified academic literature has been written on the subject. Moreover, CSR and concurrent environmental, social and governmental (ESG) issues have become an integral part of the operations of firms around the world. One can conclude that there is an increasing pressure by stakeholders for businesses to be operated more socially responsible. Nowadays, many firms publish CSR reports that provide extensive information about their CSR policies and activities and/or devote a significant part of their annual report to CSR (Deng et al, 2013). Corporate social responsibility entails that managers consider trade-offs between shareholders and other stakeholders in the firm, e.g. employees, customers, and communities. As such, CSR is in contrast with the classic assumption that the only social responsibility of a business is to make profits and consequently create value for its shareholders (Friedman, 1970; Krüger, 2015). A related subject is socially responsible investing (SRI). According to US Social Investment Forum’s report (2017), the amount invested in SRI funds increased from $2.16 trillion in 2003 to $8.72 trillion in 2016 in the US alone, illustrating the growing attention on sustainability and responsibility among investors.
In light of the need for further research on value creation by CSR, this paper analyzes to what degree CSR can help to create value in the case of mergers & acquisitions (M&A). M&A is a subject that has an even larger slew of literature behind it than CSR, receiving more scrutiny on its implications as well. Certain is that M&A at least has the potential to create value for firms. Acquisitions have become a dominant mode of expanding the operations of a firm (Kogut & Singh, 1988). M&A allows firms to benefit from economies of scale, access scarce resources embedded in organizational cultures, teams and individuals, enhance firms’ revenue through market share gains, generate tax advantages and eliminate inefficiencies (Chakrabarti, Gupta-Mukherjee & Jayaraman, 2009). Thus, M&A can have large implications on shareholder value. Furthermore, the approval process of mergers & acquisitions tends to be subject to a range of challenges as well as support from stakeholders other than the shareholders, who have an impact on the eventual outcome of a merger and play a significant role in the integration process post-acquisition (Deng et al, 2013). Considering this, mergers & acquisitions are ought to be sufficiently important events to analyze the impact of CSR on value creation for shareholders.
The resulting research question is as follows:
RQ: How does CSR performance influence stock market-based value of acquirers during M&A events?
Considering mergers & acquisitions are largely unanticipated events, abnormal returns around announcements serve as a satisfactory proxy for value creation. Thus, the effects of CSR on firm value are examined on a stock market-based assessment, using the event study methodology. Here, value creation is based on the cumulative abnormal returns (CAR) on the stock of the acquirer in a period around the acquisition announcement (King et al, 2004). The potential value for outside investors is proxied by an outperformance measure, which compares abnormal returns before and after the announcement. A sample of 714 observations in the years 2010-2016 gathered from data of Bureau Van Dijk is analyzed in this paper. The focus is on acquirers based in Europe, considering how current literature has focused mostly on US acquirers (e.g. Deng et al, 2013). In doing so, this paper extends the existing analysis towards the European region. US data is used to make a comparison between the different regions by creating a sample of 585 observations, applying similar methods.
abnormal returns after the announcement versus before the announcement. I.e. it is expected that significant value is created for the acquiring firm’s shareholders.
This paper majorly contributes to existing literature, considering how only a small part of research has focused on the value implications of CSR in an M&A setting. Moreover, this paper is the first attempt to assess outperformance of abnormal returns post-announcement versus pre-announcement in order to analyze potential value creation for outside investors. Most existing literature has focused purely on current shareholders.Previous researchers have mostly analyzed relationships between CSR and M&A performance (and value creation in general) in US firms. Here, the focus is on European acquirers, adding to literature by extending the analysis towards the European region. Differences between regions can potentially be found because of cultural and/or geographical differences. Hence, a comparison with US data using the same databases is made, as well as a split sample for the Europe-based dataset, comparing UK acquirers and Eurozone acquirers. By analyzing European acquirers, implicitly a different CSR dataset is used. Previous literature has generally used the (US-based) KLD database, whereas the analysis here revolves around ASSET4 ESG data, which is further disaggregated into social, environmental and governmental ratings to enable more detailed analysis on the constituents of CSR performance.
Results from the empirical analysis are not conclusive. Univariate analysis points to significant negative relationships between CSR performance and abnormal returns. However, multivariate regression analysis cannot confirm the significance of relationships. Comparisons and sensitivity analyses implied that results are not robust for different regions, indexes, proxies for expected return and event windows. Thus, it is hard to make definitive conclusions on the empirical results of this paper. The negative relationships may imply that CSR is actually a value destructive process for shareholders. However, considering the insignificance and sensitivity of results, one may conclude that CSR performance actually does not influence M&A performance substantially and that the stock market is indifferent on CSR. Another reasoning may hold that substantial differences exist in the impact of CSR on shareholder value and returns in different regions and countries.
2.
Research framework
2.1 CSR and value creationTraditionally, firms were viewed as being solely responsible for generating profits for its shareholders. As Friedman (1970) stated, “the business of business is business.” However, in the past few decades firms have increasingly focused on corporate social responsibility. The stakeholder theory of the firm describes how managers generally have to balance the pressures of a multitude of stakeholders whilst making strategic and financial decisions. Furthermore, the theory asserts that stakeholders affect and are affected by the actions of the firm (Freeman, 1984). Examples of stakeholders, next to the shareholders of the firm, are customers, employees, suppliers, community groups, and governments. The increased attention on more socially responsible operations has come largely from pressures of these stakeholders. As such, one can conclude that firms that focus more on CSR implicitly devote more attention to the desires of stakeholders.
CSR can be defined as “actions that appear to further some social good, beyond the interests of the firm and that which is required by law” (McWilliams & Siegel, 2001). Thus this definition underscores that CSR goes beyond obeying the law, as well as going beyond solely focusing on bottom line performance. McWilliams & Siegel (2001) list the following examples of CSR: “going beyond legal requirements in adopting progressive human resource management programs, developing non-animal testing procedures, recycling, abating pollution, supporting local businesses, and embodying products with social attributes or characteristics.” Much research has been devoted to the implications of CSR policies on firm performance and value creation. Literature is not necessarily unanimous on whether CSR actually creates more benefits than costs. However, when analyzing existing CSR related research, benefits do seem to outweigh the costs (Malik, 2015). CSR can be utilized as an important strategic tool to maximize shareholder value by protecting other stakeholders’ interests. As mentioned in the introduction, benefits include matters such as enhanced operating efficiency, product market gains, improved employee productivity, risk management, and earnings quality. That implies that CSR can create value for firms and shareholders.
corporate social responsibility issues are voluntarily disclosed, whereas the results of El Ghoul et al (2011) suggest that a firm’s measured CSR score negatively affects cost of equity capital as well. Consequently, the discounted value of cash flows is higher, resulting in higher firm value, assuming future cash flows determine firm value.
It can be concluded that the stock market has been paying attention to CSR and ESG, considering the substantial growth of socially responsible investing and illustrated by the fact that the majority of asset management firms offer some form of SRI fund nowadays. Sparkes & Cowton (2004) describe how SRI has become a mainstream form of investment, and how the link with (the need for) corporate social responsibility is strong. Although Bauer et al (2005) reported similar risk-adjusted returns between SRI portfolios and regular funds, Derwall et al (2005) found that SRI actually improves portfolio performance when looking at overall economic value created (described as “eco-efficiency”). Several other researchers have found that socially responsible assets outperform its non-responsible counterparts on a stock market-based assessment (e.g. Flammer, 2013; Eccles, Ionnou & Serafeim, 2014). It is clear though that some ambiguity exists regarding the value that shareholders attach to socially responsible assets.
Likewise, literature is not unanimous on the result of a cost-benefit analysis on CSR. Some researchers have argued that the costs of acting and governing in a socially responsible manner put firms at an economic disadvantage (Aupperle, Carroll & Hatfield, 1985). Where CSR has been found to positively influence profitability and financial performance by a multitude of influential researchers (e.g. Porter & Van der Linde, 1995; Godfrey, 2005), some literature has described a negative association as well (Harrison & Freeman, 1999). The existing ambiguity signifies the need for more research regarding the value implications of CSR.
2.2 M&A and value creation
Although research on CSR and its value implications is plentiful, researchers have devoted little attention to the value creation potential of CSR through mergers & acquisitions. The literature that does exist generally finds a positive relationship between CSR and acquisition performance. Before this literature is elaborated on, M&A and its implications for shareholders are explained in more detail.
potentially create economies of scale and scope (Chakrabarti et al, 2009). Neoclassical theory suggests that acquisitions can be seen as an efficiency-improving response to various industry shocks, such as antitrust policy or deregulation (Mitchell & Mulherin, 1996; Jovanovic & Rousseau, 2002; Shleifer & Vishny, 2003). Furthermore, using the resource-based view on competitive advantages, acquisitions can create sustainable, firm-specific competitive advantages through access to new resources, routines, and repertoires, provided that those newly accessed resources, routines, and repertoires are valuable, rare, and inimitable (Wernerfelt, 1984; Barney, 1991; Morosini, Shane & Singh, 1998). Acquisitions that are done across borders provide a mechanism for firms to access valuable, rare, and inimitable routines and repertoires that are embedded in other national cultures, without having to develop them (Jemison & Sitkin, 1986; Morosini et al, 1998). From a strategy perspective, the advantages of M&A are often grouped under the term synergies. These synergies can be found in several fields. Economies of scale and scope often lead to cost synergies, partly due to increases in bargaining power versus suppliers and customers and through cutting overhead. Synergies may also be realized through an increase in market power, enabling the merged company to profit from the control of prices, quantities or the nature of products (Seth, 1990). Moreover, revenue synergies may be realized through cross- and up-selling of products and services (Calomiris, 1999).
Although advantages of acquisitions seem to be apparent from a strategy-minded approach, the primary interest here is whether value is created for shareholders. Literature that focuses on value maximization through synergies and similar motives generally describes M&A as a process of value creation. However, there are also proponents of a view that describes M&A as a value-destructive process. This perspective revolves around the agency problems that are encountered with self-interested managers. Here, managers embark on acquisitions to maximize their own utility at the expense of shareholders (Seth, 1990). Referred to as empire-building, the agency problems of managers, described as systemically irrational bidders in acquisition processes, can lead to unexplained merger waves, significant overpayment and resulting value destruction (Rhoades, 1983; Black, 1989). Included in Jensen’s (1986) free cash flow hypothesis is the phenomenon that managers can realize large personal gains from empire building. Here, the likeliness of making value-destructive acquisitions is dependent on the level of cash flows the firm generates in combination with the investment opportunities available to the firm. Firms with abundant cash flows but few profitable investment opportunities are more likely to make value-destructive acquisitions than to return the excess cash flows to shareholders.
returns for the target’s shareholders, with average CARs up to +40% as concluded by a survey of some of the most influential papers on M&A returns by Bruner (2002). Mergers and acquisitions deliver a premium return to target shareholders. The ambiguity of the subject can be found within abnormal returns for acquirer shareholders, where generally average CARs between -5% and +5% can be found in studies (Bruner, 2002). Thus, research has not been able to ascertain that M&A actually creates value for the shareholders of acquiring firms, despite its widespread existence. As described above, an explanation for these results may be found in agency theory, suggesting that self-interested managers may select acquisitions that do not benefit shareholders. Later, this paper will give an explanation of how the managers of CSR-focused firms could be more profound in this matter, selecting more suitable acquisitions and creating more value for the acquiring firm’s shareholders. Generally, when forming a weighted average of returns between target and acquirers, a positive net gain has been found on overall abnormal returns around announcements, suggesting M&A pays off for investors in the combined target and acquirer firms (Bruner, 2002).
While CSR is considered as a potentially important mediating factor towards generating more positive or less negative abnormal returns for shareholders here, several other factors have already been found to have a significant effect on shareholder value during acquisitions. Generally, these factors can be characterized as either deal-specific or firm-specific, as described by Masulis, Wang & Xie (2007). Firm-specific characteristics may include firm size, Tobin’s Q, free cash flow to equity holders (FCFE), and leverage. Deal-specific characteristics that have been found to influence abnormal returns include relative deal size, method of payment, target and acquirer industry (relatedness), event year and whether the target is listed or not. Characteristics that have been proven to influence abnormal returns during acquisitions will be included as control variables in the empirical analysis applied here, as described in the methodology section.
2.3 Relationship between CSR and M&A
Although consensus on whether mergers & acquisitions actually create value for shareholders has not been reached, we can almost certainly conclude that M&A at least has the potential to create value. Considering the need for research on the value creation implications of CSR, there is sufficient basis to research the effect that CSR can have on M&A performance as a value-creating event. This paper will consider a stock market-based assessment on value creation, i.e. the reaction the stock market has to the M&A event is measured, considering the CSR scores of parties involved. Keeping relevant literature in mind, hypotheses are formulated accordingly.
operations (Deng et al, 2013). In firms with relatively much focus on CSR, the interests of managers, shareholders and stakeholders are in greater alignment than in those with less focus on CSR, which should lead to higher long-term profitability and thus shareholder value (Jensen, 2001). As described before, CSR can positively influence both operational matters, e.g. risk management, and capital market issues, e.g. the cost of equity capital. As such, CSR increases firm performance and shareholder value. However, it must be mentioned that ambiguity does exist regarding these conclusions.
Looking at the relationship between acquirers’ CSR and M&A performance, a positive relationship is expected. Mergers & acquisitions affect many stakeholders. Often some employees are fired and customers may be affected, which could lead to value destruction in a worst-case scenario. High CSR acquirers are expected to be more aligned with the interests of stakeholders than low CSR acquirers (Deng et al, 2013). The agency problems that exist between managers, shareholders and stakeholders alike could thus be significantly smaller for those firms that focus on CSR. Consequently, bids by high CSR acquirers are expected to be more supported by key stakeholders and deals that are undertaken are more likely to benefit stakeholders, increasing the probability of outperforming the market. Acquirers with a distinct focus on CSR are in a position to assess and control acquisition risks more comprehensively because of the integration of ecological and social aspects in their risk management approach (Meckl & Theuerkorn, 2015). This increases the likelihood of the M&A transaction to be successful. Moreover, a majority of researchers has found a positive relationship between CSR performance and stock performance (e.g. Flammer, 2013; Eccles et al, 2014). Important is that the assumption is made here that the capital market forms unbiased expectations for the potential of (future) value creation (Datta, Pinches & Narayanan, 1992; Lubatkin, 1987) It is expected here that shareholders will specifically anticipate outperformance by CSR-focused firms for M&A events as well, increasing abnormal returns.. Above expectations are supported by the findings of Deng et al (2013), whom concluded that high CSR acquirers indeed have significantly higher abnormal returns than low CSR acquirers. As such, the stock market positively prices acquirers’ CSR, leading to the following hypothesis:
H1: Acquirers’ CSR performance positively influences acquirers’ cumulative abnormal returns around the announcement date.
Higher abnormal returns after the announcement date compared to before the announcement date could hold a potential opportunity for investors to make substantial returns. Following a similar reasoning as outlined for hypothesis 1, one would expect that for high CSR acquirers outperformance in abnormal returns is significantly higher than for low CSR acquirers, leading to the following hypothesis:
H2: Acquirers’ CSR performance positively influences the outperformance of acquirers’ abnormal returns after the announcement versus before the announcement.
3.
Methodology and data
3.1 Methodology
3.1.1. Research method
This study employs the event study methodology in order to obtain a proxy for shareholder value creation. An event study generally measures the impact of a specific event on the market value of a firm (MacKinlay, 1997). Event studies are rooted in efficient market theory and, as concluded by Fama et al (1969), stock markets are (semi-strong) efficient given the rapid adjustment of stock prices to a particular event. This implies that if certain information or news is made public, investors will immediately incorporate it into the stock price, i.e. the market is semi-strong efficient. As such, the rationale behind using the event study methodology is the assumption that the capital market forms unbiased expectations for the potential of value creation (Datta et al, 1992; Lubatkin, 1987). Abnormal returns may predict long-term performance well, because changes in stock price around the announcement reflect a change in the expectation of future earnings (Stahl & Voigt, 2005). Event studies cannot necessarily be used to make inferences about causality, but are effective in inferring whether something is actually going on (Brinkhuis & Scholtens, 2017).
returns do. Daily excess returns or abnormal returns have been found to be no different in this respect (Brown & Warner, 1985). Parametric test statistics in event studies generally assume normality of abnormal returns, creating a disadvantage of employing these methods. Non-parametric statistics generally do not require these stringent normality assumptions about return distributions and are consequently highly applicable to event studies (Cowan, 1992).
Further, univariate methods of analysis will be employed to assess the stand-alone relationship between abnormal returns (and the outperformance of those abnormal returns) and CSR performance. More specifically, differences in means and medians are to be analyzed for subsamples of deals based on CSR score quantiles. Consistent with Deng et al (2013), the sample will firstly be divided into two groups, based on median CSR scores. As a parametric test, differences in means between the subsamples are tested for significance using the Student T-test. In case Jarque-Bera tests indicate non-normal distributions of returns, the Mann-Whitney U-test is employed as a non-parametric test for differences between subsamples. Furthermore, subsamples based on quintiles will be analyzed, presenting means per quintile to illustrate differences between higher and lower CSR scores. Here, the Student T-test is used to test differences in means between the lowest and highest quintiles.
To isolate the relationship between abnormal returns and CSR performance, multivariate cross-sectional regression analyses are performed using cumulative abnormal returns or outperformance as a dependent variable and CSR scores from the ASSET4 database as an independent variable. Before the OLS regressions are performed, the White test is used in order to imply whether residuals display homo- or heteroscedasticity. Further, this analysis includes a check for serial correlation using the Breusch-Godfrey serial correlation test. Conclusions on heteroscedasticity and serial correlation will be followed by the usage of the appropriate standard errors. The regressions can be represented by the following equations:
𝑀𝐴𝑅𝑖 = 𝛼𝑖 + 𝛽𝑖𝐶𝑆𝑅𝑖,𝑡−1+ 𝛾𝑖𝐹𝑖,𝑡−1+ 𝛿𝑖𝐷𝑖,𝑡 + 𝜀𝑖,𝑡 (1) Here, 𝑀𝐴𝑅𝑖 represents a vector of abnormal return measures on the stock of acquirer i, being either cumulative abnormal returns around the announcement date (t0) of an acquisition or
3.1.2 Dependent variable
This paper primarily includes an analysis of the impact of merger- and acquisition events on the acquirer’s market value. To enable this, the abnormal stock returns of acquiring firms around the acquisition announcement (i.e. event) date are needed. Event studies commonly measure cumulative abnormal returns (CAR) around the day of the announcement, consistent with McWilliams & Siegel (1997) and MacKinlay (1997). The abnormal return is calculated as the difference between the actual stock return and the return that would be expected given the performance of the market (Stahl & Voigt, 2005):
𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡− 𝐸(𝑅𝑖,𝑡) (2)
Where 𝐴𝑅𝑖,𝑡, 𝑅𝑖,𝑡 and 𝐸(𝑅𝑖,𝑡) refer to the (daily) abnormal, actual and expected return for firm i between day t and the preceding day. Important to notice is that time t0 is the event date. The
expected return can be determined through a multitude of measures. Here, consistent with McWilliams & Siegel (1997) the market model is used to determine the expected return. The market model assumes a stable linear relationship between the return on the respective firm’s stock and the market return (MacKinlay, 1997). The linear relationship can be analyzed as follows:
𝐸(𝑅𝑖,𝑡) = 𝛼𝑖+ 𝛽𝑖𝑅𝑚,𝑡 + 𝜀𝑖,𝑡 (3)
Here 𝛼𝑖, 𝛽𝑖 and 𝜀𝑖,𝑡 represent the intercept, slope (measure of systematic risk) and residual error term for the linear relationship between the expected return for firm i and market return 𝑅𝑚,𝑡. Generally, the market return is proxied by the return on a broad market index, e.g. the S&P 500. Considering the fact that this study revolves primarily around European acquirers, the MSCI Europe Index is considered a sufficient proxy for the market return. The estimation window is set at (-209,-10), i.e. 200 days, consistent with Masulis et al (2007).
Both actual returns per firm and market returns are generated through Datastream. Here, the Total Return Index (RI) is used in order to account for dividends and capital appreciations. To compute continuous time returns, the natural logarithm from the division of RI and its lagged value, i.e. the RI value of the preceding day, is taken to derive daily returns:
Ri,t = lnRIRIi,t
i,t−1 (4)
After the calculation of the abnormal returns per day, cumulative abnormal returns are formed as the sum of abnormal returns for a certain range of days around the announcement date of the M&A event in order to draw overall inferences for an event (MacKinlay, 1997):
𝐶𝐴𝑅𝑖(𝑡−1,𝑡+1) = ∑ 𝐴𝑅𝑖,𝑡 𝑡+1
Here 𝐶𝐴𝑅𝑖(𝑡−1,𝑡+1) represents the cumulative abnormal return on the stock of firm i between days t-1 and t+1 around the announcement date, where t-1 is one period of days before the event and
t+1 one period of days after the event. In practice, a range of days (including the event date) is
analyzed instead of only the event date, because this captures the price effects of announcements that occur after the stock market closes on the announcement date. At the same time, the days prior to the announcement can also be of interest in case the market acquires information about the event prior to the actual announcement (MacKinlay, 1997). Consistent with literature, the event date range (-1,+1) is used as the primary dependent variable.
Building on existing literature, this paper also includes an analysis on the influence of CSR performance on an outperformance variable in order to assess attractive investment opportunities for outside investors in the case of an event. Outperformance is defined as the difference in CAR between a range of days before the announcement and a range of days starting the day of the announcement:
𝑂𝑃𝑖 = 𝐶𝐴𝑅𝑖(𝑡0,𝑡+1) − 𝐶𝐴𝑅𝑖(𝑡-2,𝑡-1) (6)
Where 𝑂𝑃𝑖 represents the outperformance of abnormal returns after the announcement versus before the announcement for acquirer i. Further, 𝐶𝐴𝑅𝑖(𝑡0,𝑡+1) is the cumulative abnormal return between event date t0 and one period of days after the event t+1, and 𝐶𝐴𝑅𝑖(𝑡-2,𝑡-1) is the cumulative abnormal return between two periods of days before the event t-2 and one period of
days before the event t-1. Specifically, 𝐶𝐴𝑅𝑖(-5,-3) is deducted from 𝐶𝐴𝑅𝑖(0,+2) to calculate the
outperformance created by the deal announcement. Consistent with literature and the primary CAR variable, 3-day windows are taken. The event window (-2,-1) is excluded to avoid the effects of any potential information leakage prior to the official announcement.
3.1.3 Independent variables
each major pillar (namely ENV, SOC & GOV). The table below illustrates the different categories per pillar (Thomson Reuters, 2013):
Table 1. ESG pillars and underlying categories
Environmental Social Governance
Emission reduction Community Board functions
Product innovation Diversity Board structure
Resource reduction Employment quality Compensation policy
Health & safety Shareholders policy
Human rights Vision & strategy
Product responsibility Training & development
3.1.4 Control variables
M&A performance, measured through abnormal returns around the announcement date, are influenced by a multitude of factors, which have to be controlled for to isolate the effect of CSR on abnormal returns. Here, it is important to control for firm-specific and deal-specific characteristics, consistent with Masulis et al (2007) and Deng et al (2013). Firm-specific control variables include acquirer’s firm size, Tobin’s Q, free cash flow and leverage, measured over the fiscal year prior to the event announcement in order to prevent analysis on the effect of the event itself.
Firm size is measured as the natural logarithm of the total book value of assets of the acquirer. Controlling for Tobin’s Q aims to ensure that results are not driven by management quality (Lang, Stulz, & Walkling, 1991). Tobin’s Q is defined here as the ratio of an acquirer’s market value of assets over its book value of its assets, measured as follows:
Tobin′s Q = BVA−BVE+MVE
BVA (7)
Where BV represents book value, MV market value, A assets and E common equity. Consistent with Jensen’s (1986) free cash flow hypothesis, it is also important to include free cash flow and leverage. Free cash flow (to equity holders), scaled by the book value of total assets, is defined as follows:
in formula 5). Leverage is generally found to have a significant effect on abnormal returns, related to the governance mechanism it implies (Masulis et al, 2007).
Lastly, it is important to control for prior outperformance of the analyzed stocks in order to attempt to remove endogeneity caused by prior stock performance (i.e. momentum as described by Jegadeesh & Titman, 1993) and to control once more for management quality. This variable is defined as the average abnormal returns over the estimation window used for the dependent variable, using the market return (i.e. return on the MSCI Europe Index) as a direct proxy for expected return, calculated as follows:
𝐴𝐴𝑅𝑖(𝑡−2,𝑡−1) = 1
𝑡−1−𝑡−2+1∑ 𝐴𝑅𝑖,𝑡 𝑡−1
𝑡−2 (9)
Where 𝐴𝐴𝑅𝑖(𝑡−2,𝑡−1) represents the average abnormal returns for firm i within the time range t-2 to t-1, i.e. two periods of days before the event t-2 and one period of days before the event t-1. For
consistency, the estimation window of average abnormal returns is set at (-209,-10), i.e. 200 days, which corresponds to the momentum range of 3 to 12 months described by Jegadeesh & Titman (1993).
Next to firm-specific variables, there is a need to control for deal-specific characteristics. Consistent with Deng et al (2013) and Masulis et al (2007), relative deal size, a public target dummy, method of payment, and target and acquirer industry (relatedness) are included here.
Relative deal size is defined as deal size scaled by the book value of total assets of the acquirer. To take stock market effects into account, a dummy variable is included to control for listed target firms. Method of payment is of importance considering the effect of paying in cash or stock in an M&A deal. Here, a dummy variable for all-stock payments is created to take the method of payment into account. Target and acquirer industry are controlled for in several ways. First, this study uses industry-fixed effects in order to control for acquirer industry, using two-digit SIC codes. Second, target and acquirer industry relatedness are controlled for by creating a dummy variable for the instance where target and acquirer have the same two-digit SIC codes. Third, a dummy variable is created to specifically control for the effects of targets active in high-tech industries, as defined by Loughran & Ritter (2004). Finally, time-fixed effects are applied to (partially) control for the effect (the economic situation of) different years have on the variables analyzed.
3.1.5. Sensitivity analysis
into account, a US sample is analyzed and compared to the primary European data on the results of univariate and multivariate analyses. To calculate the expected return using the market model, the S&P 500 Index is used as a proxy for the market return, consistent with Deng et al (2013). Further, the European sample is split into UK-based and Eurozone-based samples. The respective market indexes applied here are the FTSE 100 Index and the MSCI EMU Index. The two samples are compared through the results on multivariate analyses.
To test for robustness, the empirical analysis is extended through using a different proxy for the expected return. Here, the expected return is proxied directly by the return on the relevant market index, so in formula (2), 𝐸(𝑅𝑖,𝑡) is proxied by the return on the MSCI Europe Index, whereas other formulas are unchanged. Furthermore, a sensitivity analysis is done on the event window in the CAR calculation. Whereas the primary dependent variable is in the event window (-1,+1), results for event windows (-2,+2) and (-5,+5) are also presented, in order to ensure that the announcement is factored in completely. Lastly, robustness is tested through winsorizing and mean-centering. Winsorizing is employed at the 95% level, where observations above the 97.5 percentile and below the 2.5 percentile are replaced by the observations at the 97.5 and 2.5 percentiles for the CAR, outperformance and ESG variables. This ensures that the bias caused by possibly spurious outliers is accounted for (Tukey, 1962). Mean-centering is employed for the CAR and outperformance variables for its scaling advantages.
3.2 Data collection
For the research presented here, multiple databases have been used to conduct a proper empirical analysis. M&A data has been collected from the Zephyr database, operated by business intelligence provider Bureau van Dijk (recently acquired by Moody’s). Completed deals during the period 2010-2016 with an enterprise value above €10 million are analyzed. Pre-2010 deals are excluded to avoid the ambiguity caused by the financial crisis of 2007-2008 and the economic boom preceding it. The deals selected are limited to a maximum pre-deal acquirer stake of 10% and a minimum post-deal stake of 100% to be able to answer the research question in a valid and reliable manner.
The analysis is primarily focused on listed European acquirers, while listed US acquirers are included to be able to make a comparison. The initial sample of European deals consists of 1,381 events, whereas the US sample consists of 1,386 deals.
less than 1% of abnormal return data is missing completely at random (MCAR), posing no issues. Control variables that cannot be found in the Zephyr data are gathered through Datastream. These primarily include accounting data from the fiscal year prior to the announcement and are used to generate relevant ratios, consistent with control variable descriptions. Again, less than 1% of data is missing completely at random.
4.
Empirical results
In this section, the results of the empirical analysis are presented. First, sample characteristics and descriptive statistics will be discussed. Next, the results of univariate analyses on the relationship between abnormal returns and CSR performance are presented. The primary cross-sectional regression analyses are discussed as well. Finally, a variety of comparisons and sensitivity analyses are made to improve the validity and completeness of the empirical analysis. Important to notice is that results are primarily focused on the Europe-based dataset, whereas the US-based dataset and UK- and Eurozone-based subsamples are used solely for comparison purposes.
4.1 Sample characteristics
Table 2 presents relevant summary statistics for the dependent, independent and control variables. A common sample includes 707 observations, indicating that 7 observations, i.e. <1% of observations, have missing data (MCAR). Further, it can be seen that ESG scores are fairly high for Europe-based firms, with average scores ranging from 66.6 to 72.5 out of 100, and median scores ranging from 74.8 to 84.6 out of 100. Standard deviations are generally quite acceptable when compared to means and medians. Standard deviations for the dependent variables are about 4 times higher than mean returns, whilst minimum and maximum returns are extreme, indicating that the stock market shows quite some variance in reactions to M&A announcements. This is not surprising considering the significance of M&A events in the lifetime of a firm. Absolute value measures such as total assets and deal value encompass a high standard deviation by nature, considering the only restriction here is a minimum deal value of €10 million. The average firm in the sample has €37.81 billion in assets, although the median firm has assets at a book value of €4.62 billion. This indicates that some extremely large firms skew averages of absolute financial indicators, further justifying the usage of relative and logarithmic control variables in the empirical analysis. The average deal in the sample has an enterprise value of €420 million.
(42.6%;34.2%) and services industry (22.7%;29.7%). The deals included in the sample are fairly evenly distributed over the period of analysis.
Average and median values for Tobin’s Q and FCFE are similar to those of Masulis et al (2007) and Deng et al (2013). Almost all variables show potentially non-normal values for either skewness or kurtosis. Jarque-Bera tests are all highly significant, indicating that the variables analyzed here indeed all seem to be normally distributed. This justifies the usage of non-parametric statistics in the empirical analysis.
In the common sample, average cumulative abnormal returns are 0.82%. In the abundance of literature written on abnormal returns realized during M&A events, acquirer CARs tend to range from -5% to +5%, which indicates that average CARs here are not out of the ordinary. The CARs and outperformance of the full sample are tested for significance, parametrically and non-parametrically. Using both methods, it can be concluded that means and medians are significantly positive, as can be seen in table 4. As such, the results indicate that M&A events do create some value for shareholders (although not economically substantial), pointing to a positive reaction of the stock market to acquisitions. This potential value creation is further illustrated by an average outperformance of abnormal returns of 1.13% in the sample. This implies that outside investors
Table 2. Summary statistics
This table illustrates the summary statistics of the Europe-based sample used in this paper. Deals are included as described in the section regarding data collection. JB test refers to the Jarque-Bera test for normality.
N Mean Median Std. Dev. Max. Min. Skew. Kurt. p-value
could generate (a limited level of) excess returns by investing in and possibly speculating on M&A bound acquirer stock.
Table 4. Test statistics regarding abnormal returns
This table represents the test statistics for significance of the parametric and non-parametric tests of the abnormal returns being equal to zero. Deals are included as described in the section regarding data collection. The test statistics correspond to the Student T-test and the Wilcoxon signed rank-test.
CAR(-1,+1) Outperformance
Student T-test Wilcoxon Student T-test Wilcoxon
Mean 0.82% 1.13%
Median 0.48% 0.76%
t-statistic 5.211 5.574 5.851 5.699
p-value 0.000 0.000 0.000 0.000
Table 3. Sample characteristics Europe
This table illustrates the composition of the Europe-based dataset. The table includes the home country of the acquirer, as well as the primary industry of both the acquirer and target. Moreover, the event year is included. Both the absolute and relative amount of firms in a respective category are displayed.
Acquirer Acquirer Target
Country No. % Industry No. % No % Year No. %
Austria 3 0.4% Construction 14 2.0% 14 2.0% 2010 106 14.9%
Belgium 10 1.4% Finance, insurance & real estate 86 12.0% 89 12.5% 2011 118 16.5%
Denmark 16 2.2% Manufacturing 304 42.6% 244 34.2% 2012 94 13.2%
Finland 17 2.4% Mining 41 5.7% 25 3.5% 2013 92 12.9%
France 75 10.5% Public administration 1 0.1% 1 0.1% 2014 112 15.7%
Germany 40 5.6% Services 162 22.7% 212 29.7% 2015 90 12.6%
Greece 4 0.6% Transport, comm. & utilities 60 8.4% 66 9.2% 2016 102 14.3%
Hungary 2 0.3% Wholesale & retail 46 6.4% 63 8.8%
Pearson correlation matrixes for the full sample can be found in table 10 and 11 in Appendix A. As expected, different ESG measures are highly correlated. This indicates that the disaggregates of the ESG score cannot be included simultaneously in regression models. Size and Tobin’s Q are correlated with leverage at about 0.60, which could indicate multicollinearity. However, when included in models simultaneously, Variance Inflation Factors are far below the threshold of 10, indicating that there should be no issues regarding multicollinearity. Interestingly, CAR and outperformance are both negatively correlated with all CSR measures. This may indicate a negative relationship between CSR performance and abnormal returns, although direct inferences should not be made from correlations.
4.2 Univariate analysis
Acquirers are divided into subsamples to analyze differences in CARs on the basis of CSR quantiles. Univariate analysis on low and high CSR performance (divided on median) can be found in table 5 for CAR and table 6 for outperformance. For ESG and social ratings, higher rated acquirers have a relatively lower positive CAR than lower rated acquirers. This difference is significant for both parametric and non-parametric tests. This would indicate an opposite effect compared to the expectations of hypothesis 1.
Table 5. Univariate analysis CAR: high CSR vs. low CSR acquirers
This table reports the average CAR(-1,+1) for subsamples based on the ASSET4 ratings. Here, the Europe-based sample is divided into low and high rating subsamples based on the sample median of the respective rating. Deals are included as described in the section regarding data collection. The
difference between subsamples is analyzed through parametric and non-parametric tests, respectively the Student T-test and Mann-Whitney U-test.
Student T-test Mann-Whitney U-test
Mean N t-statistic p-value t-statistic p-value
Table 6. Univariate analysis outperformance: high CSR vs. low CSR acquirers
This table reports the average outperformance for subsamples based on the ASSET4 ratings. Here, the Europe-based sample is divided into low and high rating subsamples based on the sample median of the respective rating. Deals are included as described in the section regarding data collection. The difference between subsamples is analyzed through parametric and non-parametric tests, respectively the Student T-test and Mann-Whitney U-test.
Student T-test Mann-Whitney
U-test
Mean N t-statistic p-value t-statistic p-value ESG Low rating High rating 1.16% 0.68% 358 356 -2.366 0.018 1.665 0.096 ENV Low rating High rating 1.27% 0.96% 358 356 -0.839 0.402 0.673 0.501 GOV Low rating High rating 1.45% 0.78% 358 356 -1.809 0.071 1.688 0.092 SOC Low rating High rating 1.50% 0.74% 358 356 -2.040 0.042 1.775 0.076
Acquirers with a high governmental ranking have a fairly lower CAR than their lower ranked counterparts. This difference is significant in a Student T-test, but not in a Mann Whitney U-test. Results are thus ambiguous. A negative difference exists for environmental ratings as well, but it is insignificant both parametrically and non-parametrically.
Regarding outperformance, similar negative differences between high and low ratings can be found for all CSR measures. Parametrically, differences are significant for ESG and social ratings and marginally significant for governance. Non-parametrically, one can find marginally significant differences for ESG, governmental and social ratings. Similarly, differences for environmental ratings are insignificant.
This may indicate that a degree of randomness exists in the results, rather than a systematic difference in returns due to CSR performance.
Table 7. Quintile analysis: differences in effect for ESG quintiles
This table reports the average CAR(-1,+1) for subsamples based on the ASSET4 ratings. Here, the Europe-based sample is divided into quintiles based on the sample percentiles of the respective rating. Deals are included as described in the section regarding data collection. The differences in average CARs between the lowest and highest quintiles are analyzed through the Student T-test.
Quintile Student T-test
Variable 1 2 3 4 5 Diff. (5-1) t-statistic p-value
ESG 1.44% 0.87% 1.21% 0.41% 0.20% -1.24% -1.633 0.005
ENV 1.43% 0.33% 0.79% 0.45% 0.68% -0.75% -2.845 0.104
GOV 1.65% 0.52% 0.87% 0.64% 0.43% -1.22% -2.633 0.009
SOC 1.75% 0.96% 0.73% 0.28% 0.38% -1.37% -3.327 0.001
4.3 Cross-sectional regression analysis
Before actual analyses are made, the assumptions of OLS are checked. Jarque-Bera tests for normality are significant for all tests, as can be expected from the summary statistics, indicating that the null hypothesis of normally distributed residuals can be rejected. Further data adaptations such as winsorizing and mean-centering are applied in the sensitivity analyses to check the robustness of the regressions. From White tests for heteroscedasticity, it can be concluded that for the majority of regression models, homoscedasticity can be rejected. Breusch-Godfrey tests for serial correlation do not indicate further issues. Hence, Huber-White heteroscedasticity-consistent standard errors are applied in the OLS models.
4.3.1 Cumulative abnormal returns
CSR performance may become insignificant. Across all models, relative deal size has a significantly positive effect on CAR. Size has a (marginally) significant negative coefficient in models 2 and 3. There are no further significant relationships between control variables and the dependent variable here. The F-statistics are highly significant, indicating that the models explain a significant portion of the variance of CAR. Specifically, R2 values indicate that models explain around 12.5% of the variance of CAR.
4.3.2 Outperformance
Table 8. Multiple regression: explaining CAR(-1,+1) MM with ESG ratings (Europe)
This table represents the results of the cross-sectional multivariate OLS regression for the Europe-based sample using the market model-based CAR within the (-1,+1) event date range as the dependent
variable, whilst including all control variables and one ASSET4 rating per regression. Deals are included as described in the section regarding data collection. The top number shows the regression coefficient, whilst the bottom number represents the respective t-statistic under White-Hinkley heteroscedasticity-consistent standard errors. The intercept is not included for brevity. *, **, *** indicate statistical significance at the 10%, 5% and 1% levels.
Variable (1) (2) (3) (4) Independent variables ESG -0.053 -0.795 ENV 0.014 0.229 GOV -0.087 -1.355 SOC -0.067 -0.846 Control variables Size -0.003 -1.522 -0.003 -1.965 ** -0.003 -1.945 * -0.002 -1.280 Leverage 0.016 1.134 0.016 1.108 0.015 1.120 0.016 1.119 FCFE 0.125 1.585 0.123 1.561 0.124 1.586 0.125 1.585 Tobin’s Q -0.003 -0.874 -0.003 -0.892 -0.003 -0.905 -0.003 -0.852 AAR 1.831 0.724 1.785 0.705 1.809 0.724 1.848 0.729
Relative deal size 0.026
2.050 ** 0.027 2.086 ** 0.026 2.016 ** 0.026 2.060 **
Listed target dummy -0.005
-0.767 -0.005 -0.762 -0.006 -1.158 -0.005 -0.771
Stock deal dummy -0.012
-1.105 -0.012 -1.050 -0.013 -1.158 -0.012 -1.081
Industry relatedness dummy 0.005
1.456 0.005 1.432 0.005 1.446 0.005 1.436 High-tech dummy -0.003 -0.814 -0.003 -0.773 -0.002 -0.675 -0.003 -0.847
Year dummies Yes Yes Yes Yes
Industry dummies Yes Yes Yes Yes
F-statistic 3.853 *** 3.824 *** 3.914 *** 3.866 ***
R2 0.124 0.123 0.126 0.124
Table 9. Multiple regression: explaining Outperformance with ESG ratings (Europe)
This table represents the results of the cross-sectional multivariate OLS regression for the Europe-based sample using the outperformance between event date ranges (0,+2) and (-5,-3) as the dependent
variable, whilst including all control variables and one winsorized ASSET4 rating per regression. Deals are included as described in the section regarding data collection. The top number shows the regression coefficient, whilst the bottom number represents the respective t-statistic under White-Hinkley
heteroscedasticity-consistent standard errors. The intercept is not included for brevity. *, **, *** indicate statistical significance at the 10%, 5% and 1% levels.
Variable (5) (6) (7) (8) Independent variables ESG -0.053 -0.613 ENV 0.038 0.477 GOV -0.121 -1.484 SOC -0.030 -0.324 Control variables Size -0.005 -2.806 *** -0.006 -3.328 *** -0.006 -3.233 *** -0.006 -2.687 *** Leverage 0.045 2.727 *** 0.045 2.704 *** 0.045 2.726 *** 0.045 2.707 *** FCFE 0.011 0.130 0.008 0.094 0.009 0.114 0.009 0.114 Tobin’s Q 0.003 0.886 0.003 0.858 0.003 0.830 0.003 0.892 AAR -1.901 -0.563 -1.965 -0.584 -1.917 -0.567 -1.911 -0.566
Relative deal size 0.024
1.992 ** 0.024 2.046 ** 0.023 1.934 * 0.024 2.013 **
Listed target dummy -0.014
-1.493 -0.014 -1.492 -0.014 -1.536 -0.014 -1.493
Stock deal dummy -0.021
-1.419 -0.021 -1.381 -0.022 -1.536 -0.021 -1.392 Industry relatedness dummy 0.005
1.190 0.004 1.151 0.005 1.190 0.004 1.172 High-tech dummy -0.006 -1.254 -0.006 -1.203 -0.005 -1.121 -0.006 -1.245
Year dummies Yes Yes Yes Yes
Industry dummies Yes Yes Yes Yes
F-statistic 2.602 *** 2.593 *** 2.869 *** 2.588 ***
R2 0.087 0.087 0.090 0.087
4.4 Data comparisons
4.4.1 US-based sample
To be able to make more valid conclusions on the certainty of the regression results and to be able to see whether differences may exist between regions, a US-based sample is analyzed on the basis of univariate statistics and cross-sectional regressions. Results for these analyses can be found in table 12 through 15 in Appendix B. Univariate analysis on the relationship between CSR performance and CAR, as well as on the relationship between CSR performance and outperformance, delivers no significant results parametrically or non-parametrically. Interestingly though, subsamples of high ESG, environmental and governmental rated firms have higher abnormal returns and outperformance than their low-rated counterparts, contrary to the results seen in the Europe-based sample. The differences are thus consistent with hypothesis 1 and 2, although results are insignificant. For the social rating, both CAR and outperformance are lower for higher-rated companies, consistent with the results of the Europe-based sample, but inconsistent with hypothesis 1 and 2. Once more, the differences are insignificant for parametric and non-parametric tests.
As expected from univariate analysis, multivariate regression analyses do not deliver significant results for CSR scores. The signs of coefficients are largely consistent with univariate analysis. However, the ESG rating is found to be negatively related to both measures of M&A performance here, similar to results of the Europe-based sample. Considering the insignificance of results, no convincing proof is found for hypothesis 1 and 2 in the US-based sample. Control variables generally display insignificant relationships as well. An exception is the marginally significant negative effect of FCFE on CAR. A major difference in regression analyses between the Europe- and US-based samples can be found in the variance of the dependent variables explained by the models. R2 values for CAR are around 5.2% and those of outperformance around 3.1%, with F-statistics indicating that the variance explained is insignificant for all models. This implies that the empirical model is not sufficiently explanatory for the US-based sample.
4.4.2. Split sample UK vs. Eurozone
consistent with hypothesis 1 and 2. Again, these relationships are insignificant. Regarding control variables, size is significantly negatively related to CAR and outperformance, except for model 31 (where GOV is related to outperformance). The industry relatedness dummy consistently has a (marginally) significantly positive effect on both CAR and outperformance. Relative deal size is significantly positively related to CAR. F-statistics are highly significant for all models and R2 values indicate that regression models explain around 13.3% of CAR variance and 10.3% to 10.8% of outperformance variance. This implies that the empirical model is somewhat more explanatory for the Eurozone-based sample than the Europe-based sample, although evidence for hypothesis 1 and 2 is still not convincing.
The UK-based sample shows quite different results. Indeed, F-statistics for all models indicate significant variance explained. For CAR, R2 values are around 27.4% to 29.1% and those of outperformance between 19.3% and 20.7%, substantially higher than any prior analysis. Regarding outperformance, all models indicate negative relationships for CSR measures, contradicting hypothesis 2, although all relationships are insignificant. Here, relative deal size and leverage are significantly positively related to outperformance. Results for CAR are striking. From models 17 through 20, it can be concluded that ESG and social ratings have a significant positive influence on CAR, and governmental ratings have a significant negative effect for UK-based acquirers. Thus, signs contrast with those of the Europe-based sample and relationships are now actually significant. For the environmental rating, the relationship is found to be insignificantly negative. Again, relative deal size has a significant positive relationship with CAR for all models, while FCFE has marginally significant positive coefficients. For CSR and social ratings, results indicate that hypothesis 1 and 2 cannot be rejected in the UK-based sample, while no proof is found for environmental ratings. The negative relationships of governmental ratings contradict hypothesis 1 and 2. It can be concluded that substantial differences exist between relationships in different regions, and potentially also through the usage of different indexes, as split samples have appropriate indexes allocated to them. Further elaboration can be found in the discussion.
4.5 Sensitivity analysis
winsorized variables and around 12.3% to 12.6% for mean-centered CAR models. Overall, relevant results do not seem to be substantially sensitive to mean-centering and winsorizing.
Further, a sensitivity analysis is done on the definition of expected return for CAR calculations, illustrated by table 22 in Appendix C. When applying the return of the relevant market index, i.e. the MSCI Europe Index for the Europe-based sample, directly as expected return, all CSR measures are found to be negatively related to CAR. ESG, environmental and social ratings do not display significant relationships, governmental ratings are marginally significantly related to CAR. The latter conclusion provides weak evidence against hypothesis 1, although not substantially. R2 values indicate somewhat improved models, with explained variance of around 14.3% to 14.7%. Considering models have improved, more control variables are (marginally) significant, the sign of environmental ratings has changed, and governmental ratings have become marginally significant, results do seem to be sensitive to the proxy for expected return.
Lastly, a sensitivity analysis is done on the event window, calculating CAR using the (-2,+2) and (-5,+5) windows instead of (-1,+1). Regression analyses on subsequent CAR variables can be found in table 23 and 24 in Appendix C. R2 decreases as the event window widens, with about 11.2% variance explained for CAR(-2,+2) and around 9.9% for CAR(-5,+5). For CAR(-2,+2), ESG, environmental and social ratings have a positive effect, whereas governmental ratings have a negative coefficient. For the (-5,+5) event window, all CSR measures are positively related to CAR. For both event windows, all CSR measures have insignificant coefficients. Thus, no substantial proof for hypothesis 1 and 2 is provided using either event window. Considering that, although results remain insignificant, the signs of CSR measures change through different event windows of CAR, the relationships found for the (-1,+1) event window are not robust, expect for the positive effect of environmental ratings on CAR.
5.
Discussion
acting and governing in a socially responsible manner put firms at an economic disadvantage (Aupperle et al, 1985). Furthermore, although some advantages of CSR are widely recognized, there is no unanimity on a cost-benefit analysis. Literature has indeed displayed some ambiguity on the relationship between CSR and financial performance. As asserted by Friedman (1970), engaging in CSR may be symptomatic of an agency problem or a conflict between the interests of managers and shareholders. Research by Wright & Ferris (1997) interpreted the negative stock reactions to divestments as being consistent with agency theory.
Important to notice however, is that results are insignificant and not robust to the change of event windows and proxies for expected return. Moreover, the sample is split into UK-based and Eurozone-based subsamples. It is found that ESG and social ratings have a significant positive influence on CAR and that governmental ratings have a significant negative effect for UK-based acquirers. For the Eurozone-based sample, ESG, governmental and social ratings display positive effects, though insignificant, on CAR and outperformance. Important to notice here that the split samples had the appropriate index allocated to them. Thus, results are not robust across different regions and perhaps simultaneously sensitive to the use of different indexes. Two conclusions can be made from the somewhat ambiguous results presented in this paper. First, the insignificance of the large majority of results and the sensitivity of the signs of relationships to various changes in methodology may point towards a non-existent relationship. Quintile analysis indicated that a degree of randomness exists in the distribution of abnormal returns, rather than systematic differences in returns due to CSR performance. Although univariate analysis resulted in significant relationships, the inclusion of various control variables in the multivariate regressions removed significant effects of CSR measures. Endogeneity in univariate analysis may exist, where variables such as firm size, relative deal size and free cash flow actually have a relationship with CSR performance, and also directly with M&A performance, whereas CSR performance does not. As such, results are consistent with Revelli & Vivianni (2015), who found substantial heterogeneity in results between CSR and stock-market performance in a meta-analysis of 85 studies and 190 experiments.
Results may also differ from those of Deng et al (2013) through the usage of a different database in order to measure CSR. Their research implied significant positive relationships between CSR performance, M&A performance and concurrent value creation by using the KLD database to research US firms. Here, the empirical model applied to a US-based sample found no significant results. Different databases may implicitly value different aspects of CSR in different fashion and as such influence results. More research should be devoted to the consistency of results using different CSR databases.
It has to be recognized that this paper has some limitations. First, the usage of the ASSET4 database for CSR research is relatively novel. Most research has focused on the KLD database. This paper cannot ascertain that the usage of the ASSET4 database is as credible as the KLD database, since not much research has been done on it. Furthermore, there may be potential endogeneity in causality in relationships between CSR and M&A performance. Here, unmeasurable variables explain relationships, e.g. firms with high quality management team are proficient in CSR and select the right targets. However, the empirical analysis did aim to control for this specific effect by including Tobin’s Q and a historical average of abnormal returns to proxy management quality. Also, most relationships were not significant in multivariate analysis so this should not have large implications for the conclusions made. Still, it may be valuable to use a 2SLS approach in order to deal with endogeneity, as Deng et al (2013) successfully did. Finally, the research method may be limited in its rather broad primary focus on Europe-based acquirers. Considering the differences found in comparisons between different countries and regions, it may be more suitable to research the value implications of CSR on a country-to-country basis.
6.
Conclusion
The relationship between CSR performance and M&A performance was primarily examined using a sample of 714 mergers & acquisitions by Europe-based acquirers. CSR performance was measured through ESG ratings from the ASSET4 database, using disaggregated components of ESG as well, respectively environmental, social and governmental ratings. To be able to make comparisons across regions, a US-based sample of 585 deals was created and the Europe-based sample was split into UK-based and Eurozone-based subsamples. The samples were taken from the period 2010-2016, with a minimum enterprise value of €10 million and strict restrictions on maximum pre-deal acquirer stake (10%) and minimum post-deal acquirer stake (100%).
Results were ambiguous. Univariate analysis pointed to significant negative relationships between CSR performance and abnormal returns, converse to the positive expectations of the hypotheses. However, multivariate regression analysis could not confirm the significance of relationships. Comparisons and sensitivity analysis implied that results are not robust for different regions, indexes, event windows and proxies for expected return. Thus, it is hard to make solid conclusions on the empirical results of this paper. The negative relationships may imply that CSR is actually a value destructive process for shareholders. However, insignificance and sensitivity of results could lead to the conclusion that CSR performance actually does not substantially influence M&A performance and that the stock market is indifferent on CSR.
Another reasoning may hold that substantial differences exist in the impact of CSR on firm value between different regions and countries. For some countries, CSR may be more important, or the focus on aspects of CSR may differ and consequently its value implications. Therefore, it is advisable for future research to analyze on a country-to-country basis and to apply the relevant index. In addition, considering the existing ambiguity in literature of the value implications of M&A in general, it may be useful to find other events as a research basis to analyze the value effects of CSR. Moreover, research should be devoted to finding out whether using the ASSET4 database is consistent with usage of the KLD database, or other CSR databases for that matter, or whether substantial differences exist. Finally, research into long-term measures of value creation would be helpful and more consistent with the concept of sustainability, considering the short-term perspective of event study methodology.
