Does it pay off to ‘buy’ well?
Empirical Evidence from an M&A
Perspective
By
J.J. VAN ESSEN
S2377942
ABSTRACT
Mergers and acquisitions (M&As) offer a framework to shed a new light on whether corporate social responsibility (CSR) performance enhances corporate financial performance (CFP). Using ASSET4 data as a measurement of CSR performance in a sample of worldwide deals for the period 2004-2017, I find evidence that the environmental performance of target firms enhances acquirers’ shareholder wealth. No influence is found for different value implications in different institutional contexts. Additionally, shareholders reward (disvalue) acquirers even stronger if the target is outperforming (underperforming) the acquirer in environmental performance. These findings suggest that shareholders reward the acquirer for making environmental investments and support the stakeholder view, which indicates that fulfilling stakeholder interests can be combined with shareholder wealth creation.
Keywords: Corporate social responsibility (CSR), M&As, stakeholder view, institutional
frameworks, abnormal announcement returns.
DD MSc International Financial Management (UoG/UU)
Faculty of Economics and Business
University of Groningen
Supervisor: Prof. dr. C.L.M. Hermes
Co-Assessor: Prof. dr. M. Ararat
JEL classification: G340
2 1. INTRODUCTION
An increasing number of firms worldwide are integrating corporate social responsibility (CSR) activities into various aspects of their businesses. Recent research demonstrates that firms increasingly invest a growing amount in CSR activities to satisfy market demands (Bhandari and Javakhadze, 2017). According to the Global Sustainable Investment Review (2016), assets in socially responsibility investing (SRI) have grown with 25.2% since 2014. In 2016, there were $22.89 trillion of assets in SRI portfolios, which is 26.3% of all assets under management. Given all these resources allocated to CSR activities, it is important to get a better understanding of the effects of CSR on corporate financial performance (CFP). Although there is recognition of the importance of CSR, a clear consensus in the current debate on the impact of CSR on CFP is missing (Margolis and Walsh, 2003). Some studies show a negative or
non-existentrelation (see, e.g., Griffin and Mahon, 1997; Waddock and Graves, 1997; Harrison and
Freeman, 1999), while others show a positive relation between CSR and CFP (see, e.g., Cochran and Wood, 1984, Roman, Hayibor, and Agle, 1999; Brammer and Millington, 2005). These mixed empirical results are mainly based on the theoretical foundations of the opposing classical shareholder expense and stakeholder view. In conclusion to the overall literature, meta-analyses and literature reviews indicate a slightly positive overall effect of CSR on CFP
(see, e.g., Orlitzky, Schmidt, and Rynes, 2003; Margolis, Elfenbein, and Walsh, 2009).1 These
studies usually try to answer whether firms do well by doing good.
In view of this contradictory evidence, the question whether CSR performance is beneficial or detrimental for CFP remains largely open. Therefore, this study takes a different approach
by trying to answer whether firms do well in terms of shareholder wealth2 by ‘buying’ well.
More specifically, this research conducts an analysis based on mergers and acquisitions (M&As) to shed light on the shareholder value implications of CSR. It analyses the role of target firms CSR performance and the difference in acquirer’s and target’s CSR performance (ATCSRD) on acquirers’ short-term announcement return. In doing so, it aims to find the answer to the question whether it pays off for firms to acquire other firms which perform well on CSR. A unique M&A market framework is used, with acquirer and target measures of CSR performance, for the following three reasons. First, M&As are important strategic investment decisions with a significant effect on CFP (Healy, Palepu, and Ruback, 1992), and specifically shareholders’ wealth (see, e.g., Doukas and Travlos, 1988; Agrawal, Jaffe, and Mandelker, 1992; Masulis,
1 Margolis et al. (2009) analysed 167 studies, of which only 22 use non-U.S. data. Of these 22 studies, only 3 studies use a multiple country sample with firm-level measures.
3
Wang, and Xie, 2007). Furthermore, M&A deals involve the support and challenges of various stakeholders in the approval and post-deal integration process between the acquirer and target firm (Deng, Kang, and Low, 2013). Finally, prior studies investigating the relation between CSR and CFP had some problems with reverse causality in both variables. Consequently, the question remains whether firms do good by doing well or do well by doing good (Waddock and Graves, 1997; McWilliams and Siegel, 2001). This omnipresent reverse causality issue can be alleviated by the M&A framework, as M&A deals are namely largely unanticipated events. By using short-term announcement returns the direct influence of CSR investments to shareholder wealth can be captured (Krüger, 2015). Additionally, the short-term market-based abnormal returns used in this study give better insights in the risk-adjusted discounted future cash flows in comparison with accounting-based measures, which measure historical performance and are highly sensitive for differences in accounting procedures and managerial manipulation (McGuire, Sundgren, and Schneeweis, 1988; Brammer and Millington, 2008).
Using a sample of global public firms and 309 completed deals, this study finds strong evidence of a positive effect of targets’ environmental performance on acquirers’ abnormal returns. The targets’ social and combined CSR performance has no impact on the acquirers’ abnormal returns. These findings suggest that acquirers are rewarded for environmental investments, but not for social investments. Moreover, shareholders reward (disvalue) acquirers even stronger if the target is outperforming (underperforming) the acquirer in environmental performance. Overall, the results provide further evidence that environmental specific CSR investments are value creating for shareholders.
This research contributes to the existing empirical work on the effect of CSR performance on shareholder wealth in multiple ways. First, prior studies focused on the empirical examination of the correlation between CSR performance and firm value (see, e.g., Jo and Harjoto, 2011; Servaes and Tamayo, 2013) or on the effect of CSR on CFP measures (see, e.g., Griffin and Mahon, 1997; Margolis and Walsh, 2003). This study, however, examines the causal link between both targets’ CSR performance and ATCSRD on acquirers’ short-term abnormal returns controlling for reverse causality. Hereby, a clear channel through which CSR performance can potentially influence shareholders wealth can be clearly identified. Moreover, to the best of my knowledge this is the first study that explicitly looks at the ATCSRD and
hence gives an interesting opening in the CSR-M&A field. Next, prior empirical evidence on
4
M&A sample, which includes 36 different acquirer nations to targets based in 33 nations is used in this paper. This makes this study, the first to examine cross-country variation in shareholder wealth as a result of CSR investments in the M&A field.
The remainder of this paper is structured as follows. The next section presents the theoretical and empirical foundations for this study. Section 3 discusses the data and the empirical methodology. Subsequently, section 4 presents and discusses the main empirical results of the univariate, multivariate regression, and portfolio analyses. The final section concludes, discusses implications and provides ideas for further research.
2. THEORETICAL BACKGROUND AND HYPOTHESES
2.1. CSR and CFP: A theoretical framework
Two opposing fundamental perspectives give an insight on the relation between CSR and CFP. The classical shareholder expense view sees CSR as costly and therefore value decreasing for firms. Accordingly, firms should focus on maximizing shareholders’ wealth and leave social
responsibility decisions to shareholders themselves (Friedman, 1970).3 In contrast, the
stakeholder view argues that the interest of shareholders should not be the only concern of firms. According to this view, firms should conduct CSR activities due their responsibility to any
entity or person that is affected by their activities (Freeman, 1984; Donaldson and Preston,
1995). This view therefore emphasizes a firm’s societal role. More specifically, the stakeholder view holds that firms benefit from developing stakeholder trust through reduced transaction costs (Williamson, 1989; Jones, 1995). This reasoning implies that CSR satisfies the interests of stakeholders and accordingly their willingness to support the firm (Donaldson and Preston, 1995). Hence, more CSR investments can be beneficial for all stakeholders, shareholders included. In the context of this study, deals including targets with high CSR performance (hereafter, high CSR targets) resulting in higher acquirers’ abnormal returns are in line with the stakeholder view due to the value creation for shareholders.
The stakeholder view is in alignment with the contract theory, which views a firm as a network of contracts between the owners of the firm (shareholders) and other stakeholders (Jensen and Meckling, 1976; Cornell and Shapiro, 1987). Cornell and Shapiro (1987) state that
stakeholders support the firm with critical resources in exchange for explicitand implicit claims.
Unlike explicit claims, implicit claims (such as job satisfaction and pollution reduction) are
5
imprecise and have no juridical standing. As such, the prices of implicit claims depend on
stakeholders’ expectations about a firm fulfilling these claims (Cornell and Shapiro, 1987).4
Acquiring firms which invest in CSR, by taking over high CSR targets, have a strong reputation for fulfilling implicit claims (Deng et al., 2013). This kind of acquisition can be seen as a signal to learn from the CSR performance of the target firm (Aktas, De Bodt, and Cousin, 2011). Stakeholders of these acquirers are therefore more willing to support the firm with critical resources (Russo and Fouts, 1997), which is beneficial for shareholders. Studies such as Shane and Spicer (1983), Fombrun and Shanley (1990), and Orlitzky et al. (2003) assert that CSR investments help to build a more positive reputation for firms. More specifically, the relation between valuable intangible resources of target firms and CFP is researched by Betton and Eckbo (2000). They report that one of the most important determinants of the acquirers’ abnormal announcement returns is the target’s reputation.
Adding the resource-based view (RBV) of Barney (1991) to this line of reasoning gives a comprehensive understanding of the potential value enhancement of acquiring high(er) CSR
targets. Barney (1991) argues that resources and capabilities can be a source of sustainable
competitive advantage if they are rare, valuable, inimitable, and non-substitutable. These criteria are often met by critical intangible resources such as human capital and firm reputation (Hall, 1992). These resources are respectively closely linked to the social and environmental dimension of CSR performance used in this study. Hart (1995) was among the first who linked the RBV framework to the CSR field by addressing the fact that CSR activities, in particular environmental performance, can constitute a critical resource that leads to a sustainable competitive advantage. M&As are a good opportunity for firms to take over or to develop these critical resources which can achieve and sustain competitive advantage (Cochran and Wood, 1984; Waddock and Graves, 1997). Among others, Wickert, Vaccaro, and Cornelissen (2017) researches this reasoning in practice and describes that Procter & Gamble’s CFP is creditable to their CSR behavior and reputation. They state that once gained, a pro-CSR reputation is a valuable inimitable resource. For firms, it is difficult to make or replicate these valuable resources in the short-term. Therefore, ‘buying’ such resources is a growing trend (Kearins and Collins, 2012) among firms to enhance their own CSR performance from targets (Mirvis, 2008).
6
The acquisitions of Unilever–Ben & Jerry’s and L‘Oréal-The Body Shop are good representations of acquirers buying CSR by taking over high CSR targets (Wickert et al., 2017). 2.2. CSR and CFP: Empirical evidence
Over the past decades, many scholars examined the relation between CSR and CFP, often based upon the aforementioned shareholder expense and stakeholder view. Although the lack of complete consensus, qualitative reviews (Margolis and Walsh, 2003) and meta-analyses
(Orlitzky et al., 2003; Allouche and Laroche, 2005; Margolis et al, 2009)5 conclude a
statistically strong but economically modest positive effect of CSR on CFP. Several other studies evaluate the CSR-CFP relation from an investment perspective, comparing SRI funds with conventional funds. Among others, Anderson and Frankle (1980), Statman and Glushkov (2009), and Derwall, Koedijk, and Ter Horst (2011) show that SRI funds outperform conventional ones. Conversely, other researchers find results consistent with shareholders paying a price for CSR (see, e.g., Renneboog, Ter Horst, and Zhang, 2008; Hong and Kacperczyk, 2009; Borgers et al., 2015). Finally, some other studies find no performance differences between SRI and conventional ones (see, e.g., Hamilton, Jo, and Statman, 1993; Bauer, Koedijk, and Otten, 2005; Schröder, 2007).
It is widely argued in the literature that firms with high CSR performance have certain benefits in the capital market, leading to better CFP. Taking an accounting approach, Watts and Zimmerman (1979) argue that CSR investments lead to a higher supply of information. This results in lower costs of obtaining information and consequently in lower cost of capital. For firms, this lower cost of capital can be used for more positive net present value (NPV) investments, which gives rise to higher shareholder wealth (Lamont, Polk, and Saaá-Requejo, 2001). For example, Cheng, Ioannou, and Serafeim (2014) use the environmental and social dimension of the ASSET4 database and discover that U.S. firms with better CSR performance have fewer capital constraints due to lower agency costs and less information asymmetry through stakeholder engagement. This relation is mainly driven by the environmental dimension of CSR performance. This is also the dimension where Chava (2014) focuses on. He finds that firms with high environmental concerns have a higher cost of debt and their shareholders require higher returns. Similarly, other studies show a significant positive influence of more CSR investments and the cost of equity capital (Richardson and Welker,
7
2001; Dhaliwal et al., 2011; El Ghoul et al., 2011). In addition, Goss and Roberts (2011) study the cost of debt and report a U-shaped relation between CSR involvement and the cost of capital. In accordance, Barnett and Salomon (2006) find a U-shaped relationship between CSR performance and CFP.
2.3. Linking CSR and shareholder wealth
The extant event-study literature claims a strong positive association between CSR involvement and short-term shareholder wealth as a measure of CFP. One of the first studies which employed the event study methodology in the field of CSR was Davidson and Worrell (1988). They use 131 announcements of corporate illegalities as a proxy for social irresponsibility and report a significant negative effect on stock returns. Following this study, Hall and Rieck (1998) investigate the effect of the announcement of voluntary positive CSR actions, measured by recycling, donation, social policy, and environmental-friendly activities on returns. They show no statistically significant returns for the whole sample, but a significant positive influence is found for announcing donations and environmental-friendly activities. A more direct relation of CSR events and shareholder wealth is researched by Krüger (2015). He finds value creation effects of CSR investments. More specifically, he argues that shareholders react negatively to negative related CSR news and concludes that positive CSR activities are in the shareholders’ interests. A focus on environmental investments is taken by Klassen and McLaughlin (1996) and Flammer (2013). They report that environmental responsible firms face a significant stock price increase, whereas environmental irresponsible firms have a significant decrease. A more social direction is investigated by Edmans (2011) who finds a positive relation between job satisfaction and stock returns. He adds to this evidence that engaging in CSR activities results in higher abnormal shareholder returns in the short-term. Overall, prior event studies in the CSR field indicate a significant influence of CSR performance on shareholder wealth. However, all studies use a U.S. sample. To draw generalized conclusions, it is essential to shift the focus to new non-U.S. evidence.
2.4. CSR in M&A context
8
106 deals includes financial and utility firms and is highly concentrated in the last three years (80.2%) and dominated by U.S. and U.K. acquirers (41.7%). The authors, controlling only for deal-specific characteristics, find evidence of a significant positive influence of target firms’ environmental and social performance on the acquiring firm’s returns. Moreover, they try to explain the source of the value creation by a learning view. Additionally, Deng et al. (2013) contribute to the CSR on shareholder wealth debate by focusing on the acquirers’ CSR
performance. Using KLD data6 in a sample of 1,556 U.S. mergers in the period 1992-2007,
they compare low with high CSR acquirers. In comparison with low acquirers, high acquirers have significant higher short-term stock returns, long-term stock returns, and long-term operating performance. In addition to this, deals including high acquirers have faster completion time and a higher probability to succeed. The findings of Deng et al. (2013) support the stakeholder view and are inconsistent with the shareholder expense view. However, their results are confined to U.S. mergers with similar market institutions, thereby neglecting cross-country differences. This study empirically elaborates upon the findings of Aktas et al. (2011) and Deng et al. (2013) by revisiting the small sample results of Aktas et al. (2011) through investigating the influence of targets’ CSR performance. Additionally, this paper goes beyond research on U.S. data and examines the impact of the ATCSRD on acquirers CFP.
As previously stated, investing in high CSR targets can be a direct manner to ‘buy’ critical and difficult to replicate resources from the target. Next, a deal involving high CSR targets can have an indirect impact by enhancing stakeholders’ expectations related to fulfilling implicit claims (Cornell and Shapiro, 1987), which in turn lead to more willingness to support the acquirer with critical resources (Hart, 1995; Russo and Fouts, 1997). Additionally, these obtained critical resources, such as the reputation of the target (Fombrun and Shanley, 1990), can act as a source of sustainable competitive advantage for the acquirer (Barney, 1991; Hart, 1995). Thus, it is a positive signal to stakeholders, including shareholders showing their willingness to invest in CSR (Aktas et al., 2011). More CSR investments can also enhance the acquirer’s access to capital (see, e.g., Cheng et al., 2014) making investments in positive NPV projects easier. The overall empirical evidence, using accounting-based and market-based measures, also demonstrates a slightly positive effect of CSR investments on CFP (Orlitzky et al., 2003). Thus, the interests of stakeholders and shareholders are in greater alignment if
acquirers invest in high CSR targets and as a result these investments enhance acquirers’
shareholder wealth. Therefore, deals including a high CSR target are rewarded by shareholders
9
resulting in higher abnormal returns around the M&A announcement. Building on the foundation of the empirical evidence of Aktas et al. (2011) and the aforementioned theories, the following hypothesis is developed:
Hypothesis 1: Higher CSR performance of the target has a positive effect on the acquirers’ CFP.
Furthermore, it can be expected that this positive influence on acquirers’ CFP is higher for larger ATCSRD. Investing in a higher CSR performing target can act as a positive signal to all stakeholders showing the willingness to invest in CSR (Aktas et al., 2011). This signal of willingness to invest can have more impact on all stakeholders if the ATCSRD is greater. Moreover, greater ATCSRD can lead to a higher probability of improving its relationships with target stakeholders. In addition, Aktas et al. (2011) find that the acquirer CSR rating increased significantly after the deals, without considering the differences between CSR performances. In this study, I adopt the view taken by Wang and Xie (2008), stating that greater differences mean higher learning potential for the acquirer. Accordingly, shareholders will notice this deal as a more wealth enhancing investment, leading to higher acquirer’ abnormal announcement returns. Wang and Xie (2008) indicate that shareholder wealth creation in M&As increases with a higher difference in corporate governance between the acquirer and the target. The expectation is that these synergistic gains for acquirers also results from larger ATCSRD. Moreover, the reputation effects for the acquirer can be more positive and therefore valuable in the case of greater relatively differences between the target and acquiring firm. Therefore, acquiring a relatively higher CSR performance target have a positive effect on the abnormal returns of the acquirer and these synergies becomes larger in the case of greater differences. This results in the following hypothesis:
Hypothesis 2: The larger the difference in CSR performance between acquirer and target, the higher the acquirers’ CFP if the acquirer has a lower CSR performance relative to the target. 2.5. The role of institutional frameworks
10
acquirer’ and target’ CSR performance and shareholder wealth for the constituted international sample. Institutional frameworks are considered weak if they fail to ensure effective markets and strong if they support the voluntary exchange which acts as a foundation for an effective market mechanism (Meyers et al., 2009). As a result of weak institutional frameworks, market failures occur. Firms have to search for strategic ways to overcome these market failures (Khanna and Palepu, 1997). El Ghoul et al. (2017) state that CSR involvement can be a solution to overcome these failures. They use ASSET4 data in a sample of 11,672 observations and 2,445 firms grounded in 53 countries over the 2003-2010 period and report a positive relation between CSR performance and firm value, measured by Tobin’s q. This study, by using a similar approach integrates the market-supporting institutions in the CSR-M&A framework.
11
be substituted by CSR investments. Therefore, similarly to El Ghoul et al. (2017), this study argues that the value of CSR investments differs for different country-level institutions. In specific, it is suggested that the influence of the targets’ CSR performance on acquirers’ shareholder wealth is strengthened by weaker institutional frameworks. Combining these arguments, it can be expected that:
Hypothesis 3: The CSR performance of target firms is valued more in countries with weaker institutional frameworks, which results in higher acquirers’ CFP.
3. DATA AND METHODOLOGY
3.1. CSR measurement
CSR is operationalized by taking an equal-weighted average of the environmental (ENV) and social (SOC) scores, which results in an overall CSR score- namely the corporate social performance score (CSP). Detailed definitions and specific computation methods of all the variables used in this study are described in Appendix A. The both measures are derived from the ASSET4 ESG Database provided by Thomson Reuters. This is in accordance with recent prior CSR studies (Ioannou and Serafeim, 2012; Cheng et al., 2014; El Ghoul et al., 2017), but in contrast to prior CSR empirical research in the context of M&As. Deng et al. (2013) obtained their aggregated absolute CSR rating from Kinder, Lydenberg and Domini (KLD) Research and Analytics Inc. STATS database. This data set contains negative (concerns) and positive (strengths) ES performance indicators and is one of the most comprehensive ES data time series available, but only contains U.S. firms. Subsequently, a major disadvantage of the KLD data, is the lack of adjustable weights for all the individual strengths and concerns (McGuire et al., 1988). Hence, the assumption of equal importance of the strength and concerns scores is inappropriate, because they are both conceptually and empirically different constructs (Mattingly and Berman, 2006). Next, Aktas et al. (2011) use the discrete IVA provided by Innovest as a measurement for CSR. This database links managerial ability of ES related risks and opportunities to long-term outperformance. IVA research combines 120 performance indicators under four pillars: environment, human capital, stakeholder capital and strategic governance. Companies are rated on a seven-point scale (‘AAA’-‘CCC’) relative to their
industry peers.7
ASSET4 gathers ES data on around 5000 global companies during the period 2002-2017. The ASSET4 framework compares and rates companies against over 750 publicly available
12
data points. These data are accumulated into 280 key performance indicators (KPIs), which serve as subcomponents of 18 categories (Thomson Reuters, 2013). The categories are grouped into four main pillars reflecting sustainability: economic, environmental, social and corporate governance. The pillar scores are calculated by equally weighting and z-scoring all data points. By using a z-score, a pillar score reflects the performance of one company compared with the average performance of all the companies included in the ASSET4 database. The resulting ES pillar scores are therefore a relative measure of CSR performance, which is in line with the IVA rating, but in contrast with the absolute KLD rating. The ES scores are presented as values between zero and 100, making them more precise than the KLD and IVA ratings. To the best of my knowledge, this is the first study using the environmental and social scores of ASSET4 as an explanatory variable in the context of M&A deals.
In addition to the aggregated CSP score, regressions are run on the disaggregated scores to analyse the differences in influence on abnormal returns. All these individual pillars reflect the generation of long-term shareholder value using best management practices and capturing environmental and social opportunities. More specifically, the environmental pillar (ENV) represents a firm’s influence on non-living and living natural systems, comprising water, soil, air and complete ecosystems. This measure includes for example resources and emission reduction, and beneficial product innovation for the environment. The social pillar (SOC) focuses on evaluating a firm’s capacity in the generation of trust and loyalty with its customers, society and employees. It displays the healthiness of a firm’s license to operate and its reputation. For example, investments in employee training and development, health and safety, diversity, human rights and customer/product responsibility are included in this measure (Thomson Reuters, 2013).
3.2. Institutional framework measurement
13
domestic product (GDP), total value of shares traded over GDP, and total value of shares traded over market capitalization. This combined indicator of stock market development is used by, among others, Pagano (1993). Credit market development is defined as the total volume of credits provided by the financial sector divided by GDP, following Fauver, Houston, and Naranjo (2003).
3.3. Control variables
To test the hypotheses, other factors than the firms’ CSR performances need to be controlled for. In particular, this study includes firm-specific and deal-specific characteristics following leading M&A research (Masulis et al, 2007; Deng et al., 2013). All acquirer and target characteristics variables are obtained from Datastream, while all deal-specific control variables are from the SDC database. Regarding firm-specific characteristics, five control variables are considered. Large firms often overestimate potential synergy gains and overpay for targets based on the hubris hypothesis (Rau and Vermaelen, 1998). Consistent with the hubris hypothesis Moeller, Schlingemann, and Stulz (2004) find that large firms pay higher premiums and enter deals with negative synergies, resulting in lower abnormal returns. Hence, I include the natural logarithm of the market value of equity to control for acquirer size (ASIZE). Second, the profitability of the acquirer, measured by the return on assets (AROA) is used as a control variable in this study, in line with leading prior studies (Easton and Harris, 1991). Moreover, Lang, Stulz, and Walkling (1991) find that acquirer returns are significantly negatively related to higher free cash flows (AFCF). This finding is built upon Jensen’s free cash flow hypothesis (1986), stating that managers of acquirers with large free cash flows are more likely to invest in less beneficial or value destroying M&As rather than paying it out to shareholders. In order to control for more profitable targets, I include the targets return on assets (TROA). This measure influences the abnormal returns of the acquirer by making targets more attractive for bidders and thus costlier (Shawver, 2002). Additionally, the target’s Tobin’s q (TTQ) is positively related with acquirer returns (Lang et al., 1991; Servaes, 1991) and therefore included as a control variable.
14
abnormal returns is the number of bidders. Competition among bidders (COMP) increases the bargaining power of targets, consequently drives up the premium and decreases the acquirer returns (Bradley, Desai, and Kim, 1988; Moeller et al., 2004). Furthermore, whether the deal is cross-border or domestic (DOM) has certain implications for the acquirer returns. Announcing a cross-border M&A can be seen as an exploitation of foreign market distortions and is therefore positively valued by shareholders (Eckbo, 1983; Doukas and Travlos, 1988). The last deal-specific control variable used in this study is the method of payment (METHOD). Stock-financed deals are known to have a negative influence on acquirer abnormal returns (Travlos, 1987; Servaes, 1991). These findings are generally attributed to the equity signaling hypothesis of Myers and Majluf (1984), which state that stock payment by the acquirer signals overvaluation of their equity by the market.
3.4. Sample selection and distribution
The initial M&A sample is extracted from Thomson ONE (SDC Mergers and Acquisitions database). The sample selection procedure and corresponding number of observations are
presented in Table 1. Initial bids announced between January 20028 and September 2017 are
selected according to the following criteria:
i. Completed merger or acquisition deals from public listed acquirers and targets to ensure the availability of financial data;
ii. Deal value is at least $1million and acquirer has a majority ownership after transaction to ensure the relevance of the data;
iii. The financial- (SIC codes 6000-6999) and utility (SIC codes 4900-4999) sectors are excluded, because the applied special regulations and the differences in debt levels make them hardly comparable.
These restrictions follow extant data criteria of M&A literature (Fuller, Netter, and Stegemoller, 2002; Deng et al., 2013), and result in an initial sample of 6,044 completed M&A transactions. Acquirers and targets which are not listed in the ASSET4 Database are excluded from the sample. Merging the M&A deals from SDC with the ASSET4 data results in a sample of 503 deals. From these 503 deals, both acquirer and target need to have ES data available prior the announcement date, which is the case for 361 deals. Out of the 361 deals, abnormal returns of 352 could be computed with stock prices obtained from Thompson Reuters
15
Datastream. In the end, the full sample encompasses 309 different deals, with an average deal value of $967.47 million, with all company-specific data available.
Table 1. Sample selection procedure.
Selection step Number of Obs. Number of missing Obs. Acquirer public status 441,419 Initial sample 352 Target public status 100,128 AROA 3
Deal status complete 49,962 AFCF 35
Deal value minimal 1US$ million 39,030 TROA 13 Majority ownership after transaction 19,637 TTQ 12 Date effective between 2002-2017 9,522 IF 43 Excluding financial and utility acquirers 6,044
Acquirer and target in ASSET4 503 ES data available for acquirer and target 361
Actual returns acquirer 352 Final sample 309
Panel A through D of Table 2 gives a comprehensive overview of the breakdown in countries, announcement years, and industries of both acquirers and targets in the full sample. Panel A presents the country distributions and shows that the sample contains deals from 36 different nations to targets based in 33 nations. Most of the acquirers are from the U.S. (36.6%),
Japan (9.4%) and U.K. (8.4%).9 This distribution is comparable to the primary unrestricted
sample of 6,044 deals obtained from the SDC database10, which contains deals from the U.S.
(29.5%), Japan (20%), and U.K. (6%). The most frequent target nation is the U.S. (42.4%), followed by U.K. (9.4%), Australia (9.1%), and Japan (6.2%). The initial sample has a distribution in these countries of respectively 30%, 5.6%, 6.9%, and 18.5%. Thus, the sample contains relatively a higher number of acquirers from the U.S. and U.K. in comparison with the initial sample. A reason for this is the higher inclusion of U.S. and U.K. companies in the ASSET4 database.
Panel B reports the distribution by year. The number of M&A deals increase gradually and peak in 2015. A concentration of deals in the later years of the observation period can be identified, around 68% of the deals are from the second halve of the sample period. In contrast, the initial sample shows a constant number of deals during the 2002-2007 period. The main reason is the availability of ASSET4 ES data. To be included in the database, the firms need to have at least three years of history available, and most firms are covered from 2005 onwards.
16
Furthermore, the global financial crisis in 2008 could explain the small downfall of deals in the years 2008 and 2009.
The industry distributions of the acquirer and target are presented in Panel C and Panel D respectively. The acquirers and targets are classified on the two-digit SIC codes and distributed into six and seven main industries. Panel C shows that the majority of the acquirers and targets
are active in the manufacturing industry (respectively 50.8%11 and 44.7%), while a relatively
small amount of acquirer and targets are from the construction (respectively 2.2% and 2.9%) and wholesale and retail trade (respectively 5.8% and 8%) industry. Note, the industry distribution of both acquirer and targets, except the 5% changes in manufacturing and wholesale and retail trade, are quite similar.
17
Table 2. Sample distribution. This table presents the sample distribution by country, year, and industry. The sample of the full
model consists of 309 observations from 36 acquirer countries to 33 target countries in 6 different industries over the 2004-2017 period. The following main two-digit SIC industry classification, obtained from SDC, is used: mining (10-14), construction (15-17), manufacturing (20-39), transportation (40-49), wholesale and retail trade (50-59), real estate (65) (only targets), and services (70-89). The sample is obtained from the Thomson ONE SDC Database. The selection criteria are described in Section 3.4. Panel A. Sample distribution by country Panel B. Sample distribution by year
Acquirer Target
Country N % Country N % Year N % Australia 21 6.80 Australia 28 9.06 2004 2 0.65 Austria 2 0.65 Austria 3 0.97 2005 6 1.94 Bahrain 1 0.32 Bahrain 1 0.32 2006 17 5.5 Belgium 2 0.65 Belgium 1 0.32 2007 26 8.41 Brazil 3 0.97 Brazil 4 1.29 2008 12 3.88 Canada 19 6.15 Canada 18 5.83 2009 12 3.88 Chile 1 0.32 China 2 0.65 2010 25 8.09 China 1 0.32 France 10 3.24 2011 23 7.44 Denmark 1 0.32 Germany 5 1.62 2012 26 8.41 Finland 4 1.29 Gibraltar 1 0.32 2013 12 3.88 France 11 3.56 Greece 2 0.65 2014 36 11.65 Germany 19 6.15 Hong Kong 1 0.32 2015 54 17.48 Gibraltar 1 0.32 India 4 1.29 2016 51 16.5 Greece 2 0.65 Ireland-Rep 1 0.32 2017 7 2.27 Hong Kong 1 0.32 Italy 4 1.29 Total 309 100 India 3 0.97 Japan 19 6.15
Ireland-Rep 2 0.65 Kuwait 1 0.32
Isle of Man 1 0.32 Luxembourg 2 0.65 Panel C. Sample distribution by industry acquirer Israel 1 0.32 Mexico 3 0.97 Industry N % Italy 3 0.97 Morocco 1 0.32 Mining 42 13.59 Japan 29 9.39 Netherlands 7 2.27 Construction 7 2.26 Mexico 3 0.97 New Zealand 2 0.65 Manufacturing 157 50.79 Netherlands 10 3.24 Norway 3 0.97 Transportation 46 14.88 Norway 2 0.65 Papua N Guinea 1 0.32 Wholesale & Retail trade 18 5.82 Poland 1 0.32 Singapore 2 0.65 Services 39 12.61 Saudi Arabia 1 0.32 South Africa 6 1.94 Total 309 100 Singapore 2 0.65 South Korea 2 0.65
South Africa 3 0.97 Spain 2 0.65
South Korea 3 0.97 Sweden 3 0.97 Panel D. Sample distribution by industry target Spain 4 1.29 Switzerland 7 2.27 Industry N % Sweden 1 0.32 Thailand 3 0.97 Mining 40 12.95 Switzerland 7 2.27 United Kingdom 29 9.39 Construction 9 2.91 Thailand 3 0.97 United States 131 42.39 Manufacturing 138 44.66 United Kingdom 26 8.41 Transportation 39 12.61
United States 113 36.57 Real Estate 2 0.64
Utd Arab Em 2 0.65 Wholesale & Retail trade 25 8.09
Services 56 18.10
18
3.5. Abnormal stock returns
To isolate the effects of M&A announcements on the acquirers’ abnormal returns, a standard event study methodology is applied (Fama et al., 1969; Brown and Warner, 1985). An event study measures the impact of the different M&A announcements on the value of firms. Assuming market efficiency, the effects of the announcements will be reflected in stock prices. In the first step, a statistical market model is constructed to calculate normal returns, thereby relating expected returns to the market portfolio when deal events are absent. The market model for any security 𝑖 is defined as follows:
𝑅𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑖𝑅𝑚,𝑡 + 𝜀𝑖,𝑡, (1)
where 𝑅𝑖,𝑡 is the expected daily return of stock 𝑖 on event day 𝑡, 𝑅𝑚,𝑡 is the return on the MSCI
World Index on event day 𝑡, 𝛼𝑖 and 𝛽𝑖are the OLS regression intercept and slope12, and 𝜀𝑖,𝑡 is
the zero-mean error term. In line with MacKinlay (1997), a broad stock index (MSCI World)
is used to proxy for the market portfolio. For each event the model parameters (𝛼𝑖 and 𝛽𝑖) are
estimated over the 250 trading days ending 10 days prior the announcement date, following Aktas et al. (2011) and MacKinlay (1997). A gap is left between the event window and the estimation period to prevent the anticipation of the announcement from having an effect on the normal return measure.
In the second step, abnormal returns are calculated to assess the impact of the
announcement. The abnormal returns, 𝐴𝑅𝑖,𝑡, of stock 𝑖 on event day 𝑡 are calculated by taking
the difference between actual returns and the normal returns and is expressed as follows:
𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡− (𝛼𝑖 + 𝛽𝑖𝑅𝑚,𝑡), (2)
To draw overall conclusions and capture the price effects of announcements, the daily abnormal return of each firm is accumulated over the period from the event window to obtain
the cumulative abnormal return (𝐶𝐴𝑅𝑡) from day 𝑡. An eleven-day event window (-5, 5) is used,
which is in line with prior CSR related event studies (Deng et al., 2013).13 In addition, an
eleven-day event window is better in a worldwide sample where holidays and different time
12 Nonsynchronous trading effects, which possibly occur by taking ‘closing’ prices with different time intervals induce biases in the moments of returns and thus into the intercepts and betas of the market model. This study, does not use the Scholes and Williams (1997) adjusted beta and intercept to account for this problem, because actively traded stocks are assumed in the sample. Therefore, the adjustments would be generally small and meaningless according to MacKinlay (1997).
19
zones influence the absorption of information by shareholders (Campbell, Cowan, and Salotti,
2010). Thus, the 𝐶𝐴𝑅𝑡 is the sum of the included abnormal returns over the eleven-day period
and is expressed by the following:
𝐶𝐴𝑅𝑡= ∑ 𝐴𝑅𝑡
𝑡+5 𝑡=𝑡−5
(3)
For a correct aggregation no clustering in the sample is assumed. In other words, the abnormal returns should be independent across securities, implying the absence of any overlap in the event windows of the different deal announcements (MacKinlay, 1997). Overlapping event windows can cause covariances different from zero between the abnormal returns, which influence the calculation of the variance of the CAR. As a result, the distributional results are no longer applicable (Bernard, 1987; MacKinlay, 1997). The sample in this study has some overlapping event windows, but since the deals are taken from a worldwide sample, no
clustering is assumed.14
3.6. Correlation matrix
Appendix C presents the Pearson correlation matrix of all variables used in the subsequent analyses. The environmental and social score are highly correlated with each other and the combined score (CSP). This is justified for the reason that both measures are used for the CSP score. The results of the other correlations indicate that no serious near multicollinearity exists between any two variables within the threshold level of 0.5, in compliance with Belsey et al. (2005). The correlations indicate that the CAR(-5,5) is negatively correlated with the size of the acquirer (SIZE) and the profitability of the target (TROA). The environmental score has a higher positive association with CAR(-5,5) than the social score (0.130>0.035). The CSR performance measures (ENV, SOC, CSP) are all positively correlated with the size of the acquiring firm (ASIZE). The profitability of the acquirer (AROA) and target (TROA) are respectively slightly negatively and positively correlated with the CSR measures. Furthermore, the strength of institutional frameworks (IF) is positively correlated with the CSR measures. Relative deal size (RELDS) and acquirer size (ASIZE) are highly correlated (-0.479), but within
the threshold level of 0.5.15
20
3.7. Empirical method
Ordinary least squares regression (OLS) is used to test the hypotheses. The individual dimensions environmental and social can have different effects on acquirer returns (Galema, Plantinga, and Scholtens, 2008). Therefore, they are included in separate regressions to capture their individual impact. All in all, this results in the following empirical models:
𝐶𝐴𝑅𝑖 = 𝛼0 + 𝛽1𝐶𝑆𝑅 + 𝛽2𝐴𝑆𝐼𝑍𝐸 + 𝛽3𝐴𝑅𝑂𝐴 + 𝛽4𝐴𝐹𝐶𝐹 + 𝛽5𝑇𝑅𝑂𝐴 + 𝛽6𝑇𝑇𝑄 + 𝛽7𝑅𝐸𝐿𝐷𝑆 + 𝛽8𝐼𝑁𝐷𝐷𝐼𝑉 + 𝛽9𝐶𝑂𝑀𝑃 + 𝛽10𝐷𝑂𝑀 + 𝛽11𝑀𝐸𝑇𝐻𝑂𝐷 + ∑ 𝐹𝐼𝑋𝐸𝐷 𝐸𝐹𝐹𝐸𝐶𝑇𝑆 + 𝜀𝑖𝑡, (4) 𝐶𝐴𝑅𝑖 = 𝛼0 + 𝛽1|∆𝐶𝑆𝑅| + 𝛽2𝐴𝑆𝐼𝑍𝐸 + 𝛽3𝐴𝑅𝑂𝐴 + 𝛽4𝐴𝐹𝐶𝐹 + 𝛽5𝑇𝑅𝑂𝐴 + 𝛽6𝑇𝑇𝑄 + 𝛽7𝑅𝐸𝐿𝐷𝑆 + 𝛽8𝐼𝑁𝐷𝐷𝐼𝑉 + 𝛽9𝐶𝑂𝑀𝑃 + 𝛽10𝐷𝑂𝑀 + 𝛽11𝑀𝐸𝑇𝐻𝑂𝐷 + ∑ 𝐹𝐼𝑋𝐸𝐷 𝐸𝐹𝐹𝐸𝐶𝑇𝑆 + 𝜀𝑖𝑡, (5) 𝐶𝐴𝑅𝑖 = 𝛼0 + 𝛽1𝐶𝑆𝑅 + 𝛽2𝐼𝐹 + 𝛽3(𝐶𝑆𝑅 × 𝐼𝐹) + 𝛽4𝐴𝑆𝐼𝑍𝐸 + 𝛽5𝐴𝑅𝑂𝐴 + 𝛽6𝐴𝐹𝐶𝐹 + 𝛽7𝑇𝑅𝑂𝐴 + 𝛽8𝑇𝑇𝑄 + 𝛽9𝑅𝐸𝐿𝐷𝑆 + 𝛽10𝐼𝑁𝐷𝐷𝐼𝑉 + 𝛽11𝐶𝑂𝑀𝑃 + 𝛽12𝐷𝑂𝑀 + 𝛽13𝑀𝐸𝑇𝐻𝑂𝐷 + ∑ 𝐹𝐼𝑋𝐸𝐷 𝐸𝐹𝐹𝐸𝐶𝑇𝑆 + 𝜀𝑖𝑡, (6)
where 𝐶𝐴𝑅𝑖 is the cumulative abnormal return of acquiring firm i, 𝐶𝑆𝑅 is the proxy of interest
of the target firm (for 𝐸𝑁𝑉, 𝑆𝑂𝐶, 𝐶𝑆𝑃), |∆𝐶𝑆𝑅| indicates one of the CSR difference measures
of interest (representing |∆𝐶𝑆𝑃| , |∆𝐸𝑁𝑉| , |∆𝑆𝑂𝐶| ), and 𝐼𝐹 refers to the strength of the
institutional framework of the acquirer nation. All the equations contain the firm and
deal-specific controls 𝐴𝑆𝐼𝑍𝐸 , 𝐴𝑅𝑂𝐴 , 𝐴𝐹𝐶𝐹 , 𝑇𝑅𝑂𝐴 , 𝑇𝑇𝑄 , 𝑅𝐸𝐿𝐷𝑆 , 𝐼𝑁𝐷𝐷𝐼𝑉 , 𝐶𝑂𝑀𝑃 , 𝐷𝑂𝑀 ,
21 4. RESULTS
4.1. Descriptive statistics
Table 3 reports the summary statistics of the firm and deal characteristics. Several outcomes are worth mentioning. The targets in the sample are on average slightly
underperforming with respect to CSR relative to other firms included in ASSET4.16
Furthermore, the central tendency of the social score is higher than the environmental score. In addition, both scores have a great standard deviation and the environmental score has a higher positive skew, shown by a lower median. Comparing the profitability measures shows that the acquirers have larger return on asset ratios than the targets. As for the deal-specific characteristics, the majority of deals involved only one bidder (92%), around 42% of the deals are fully paid by cash, 41% of the deals are cross-border, 34% diversifying, and the mean of the relative deal size is 0.69.
16 The z-scores are normalized to a scale of 100, which implies that the mean score of all the included firms is 50.
Table 3. Summary statistics. This table shows summary statistics for the main variables used in the analyses.
The full sample of M&A deals covers 309 observations in 36 acquirer countries for the period 2004-2017 and is obtained from the Thompson ONE SDC Database. The selection criteria are described in Section 3.4. The event-study methodology used to calculate the CAR (-5,5) is described in Section 3.5. All variables are described in Appendix A.
Variable Obs Mean Median Std. Dev. Min. Max.
22
The descriptive statistics of both subsamples A (A<T) and B (A>T), taken from the CSP differences (|∆𝐶𝑆𝑃|) are reported in Appendix D and E for the sake of brevity. Some important differences in variables can be noticed. For example, the results of subsample A and B show a positive mean CAR(-5,5) of 0.01 and a negative mean CAR(-5,5) of 0.01 respectively. Furthermore, the absolute mean of the CSP difference is higher in subsample B (33.22%) than in subsample A (17.76%), same as the spread around the mean (25.26%>16.39%). Next, around 47% of the deals in subsample B are paid in cash, while this percentage is relatively smaller in subsample A, namely 26%. Another important point to distinguish is the higher relative deal size in subsample A (1.29) in comparison with subsample B (0.49).
4.2. Univariate analyses
The parametric results in conjunction with the nonparametric results show whether the announcement of deals have statistical impact on the distribution of abnormal returns. Panel A in Table 4 reports the mean and median CARs for the full sample and subsample A and B of the CSP differences (|∆𝐶𝑆𝑃|) among acquirers and targets. The mean CAR(-1,1), CAR(-2,2), and CAR(-5,5) for the full sample are negative, where the mean CAR(-1,1) of -0.5% is statistically significant. This is consistent with prior studies, where the CARs are on average slightly negative or at best zero, although often insignificant (Fuller et al., 2002; Andrade, Mitchell, and Stafford, 2001). More in line with this research, Deng et al. (2013) find a negative mean CAR(-5,5) of -0.445, which is significantly different from zero at the 5% level. Aktas et al. (2011) report a lower statistically negative mean of -1.16% with a three-day abnormal return. The results of subsample A and B show that the negative returns are mainly driven by subsample B, which includes deals whereby acquirers have higher CSP scores than targets. More specifically, the mean CAR(-1,1), CAR(-2,2), and CAR(-5,5) in subsample B are significant and negative. On the contrary, the mean CARs of subsample A are higher and even positive for the five- and eleven-day window, although not significant. The results of the median
CARs for the full- and subsamples are akin.The univariate method of analysis shows that the
23
Panel B and C present four additional subsamples next to subsample A and B, based on the sample medians of target CSP and acquirer CSP respectively, following Deng et al. (2013). Subsequently, low and high CSP performing targets and low and high CSP performing acquirers are distinguished. The results of CSP scores of targets in Panel B show that the differences between the means are significant for CAR(-1,1) and CAR(-2,2). Moreover, all the acquirer CARs are statistically negative related to low CSP targets. In contrast, the high CSP targets generate higher mean CARs than the low CSP targets. This is partly in alignment with the results of Aktas et al. (2011). They find an insignificant positive relation between high CSP targets and acquirer CAR(-1,1), and a statistically significant negative relation between low CSP targets and CAR (-1,1). Next to this, they find a statistically significant difference between the means of low and high CSP targets at the 5% level. Panel C displays the differences between
Table 4. Acquirers’ Cumulative Abnormal Returns (CARs) and CSP ratings. This table reports the mean
and median CARs (in percentages) of acquirers during the three-day (CAR(-1,1)), five-day (CAR(-2,2)), and eleven-day (CAR(-5,5)) windows. The event-study methodology used to calculate the CARs is described in Section 3.5. The table the full sample, subsample A and B, and four additional subsamples. Panel A shows the mean and median of the full sample and both subsample A(A<T) and B(A>T), based on the |∆CSP|. Panel B presents the mean and median of the full sample and two additional subsamples based on the sample median of target CSP. Panel C reports the mean and median for the full sample and two additional subsamples based on the sample median of acquirer CSP. The full sample consists of 309 completed deals over the 2004-2017 period, and is extracted from Thompson ONE SDC. The selection criteria are described in Section 3.4. Tests of differences in means are based on a two-Sample t-Test. N denotes the number of observations. *, ** and *** denote statistical significance at the 0.10, 0.05 and 0.01 levels (2 tailed), respectively.
Panel A: Subsample based on the difference between CSP acquirer and target (|∆CSP|)
Full Sample (N=309) Subsample of targets having higher CSP-score: A (N=76) Subsample of acquirers having higher CSP-score: B (N=233) Test of Difference (A-B) Mean Median Mean Median Mean Median Mean CAR (-1,1) -0.005* -0.004 -0.002 -0.002 -0.006* -0.005 -0.005 CAR (-2,2) -0.005 -0.004 0.004 -0.001 -0.008** -0.006 -0.011 CAR (-5,5) -0.004 -0.007 0.007 -0.005 -0.007* -0.010 -0.014* Panel B: Subsample based on target CSP
Full Sample (N=309) High CSP targets (N=155) Low CSP targets (N=154) Test of Difference (High-Low) Mean Median Mean Median Mean Median Mean CAR (-1,1) -0.005* -0.004 0.001 -0.004 -0.012** -0.004 -0.013* CAR (-2,2) -0.005 -0.004* 0.001 -0.004 -0.011** -0.004 -0.012* CAR (-5,5) -0.004 -0.007 0.002 -0.007* -0.009* -0.007 -0.011 Panel C: Subsample based on acquirer CSP
24
the low and high CSP acquirers in relation with CAR. The results indicate that also the differences between the mean of the low and high CSP acquirers are significant for CAR(-1,1) and CAR(-2,2), which is comparable with the results of Deng et al. (2013). Their results indicate a significant difference between the mean CAR(-1,1) of low CSR performing acquirers and high CSR performing acquirers. Nevertheless, their other event windows give no significant differences. Comparing the findings in Panel B and C shows that the difference between mean CARs of low and high CSP targets are greater than the differences between mean CARs of low and high CSP acquirers. This suggests that shareholders value the CSP rating of targets more than the CSP ratings of acquirers.
All in all, the reported results in Panel A of Table 4 show that the acquirers’ abnormal returns are significantly higher in subsample A in comparison with subsample B, which indicates that acquirers announcing a deal with a target with relatively higher CSP scores are valued more by the market than acquirers announcing deals with a target with relatively lower CSP scores. These univariate findings support hypothesis 2. In addition, the difference between mean CARs of low and high CSP targets are greater than the differences between mean CARs of low and high CSP acquirers, suggesting that shareholders value the CSP rating of targets more than the CSP rating of acquirers. Finally, the findings in Panel B of Table 4 support hypothesis 1, stating that higher CSP of the target has a positive effect on the acquirers’ shareholder wealth.
4.3. Regression analyses
The univariate findings do not control for important firm and deal factors that possibly affect the abnormal returns of the acquiring firms. Therefore, several OLS regression analyses are carried to investigate whether the influence of target CSR performance and the difference
between CSR performance between acquirer and target remains after including controls.17 In
each model, the acquirers’ abnormal return with the eleven-day window is the dependent variable. All the main models include year, industry, and country fixed effects to filter away
macroeconomic shocks and differences.18 Statistical significance is based on robust standard
errors. Only the interaction effect with institutional framework strength is tested without country-fixed effects to capture the cross-country interaction impact. Table 5 represents the
17 Additionally, I run tests to check for non-linearity in the full and subsamples (Barnett and Salomon, 2006). However, no statistical evidence is found for a non-linear relation.
25
regression of the full sample with the independent variables ENV, SOC and CSP represented in models 1, 2, and 3 respectively. The interaction effects of the CSR proxies with IF are presented in models 4, 5, and 6.
Table 5. Full sample regressions of Acquirers’ Cumulative Abnormal Returns (CARs) and Targets’ CSR ratings. This table reports the OLS regression results of the full sample with Acquirers’ CAR(-5,5) as dependent
variable and the Targets’ CSR proxies (ENV, SOC, CSP) as the main independent variables. The event-study methodology used to calculate CAR(-5,5) is described in Section 3.5. The full sample contains 309 observations from 36 to 34 unique countries over the 2004-2017 period. The sample selection is described in Section 3.4. The models (1), (2), and (3) include the full sample and regress ENV, SOC, and CSP respectively. The models (4), (5), and (6) include the interaction with IF and consist of 266 observations. Models (1)-(3) include year, country, and industry fixed effects. Models (4)-(6) exclude the country fixed effects to capture the cross-country interaction impact. The t-statistics based on robust standard errors are in parentheses. Appendix A presents definitions and data sources of all used variables. *, ** and *** denote statistical significance at the 0.10, 0.05 and 0.01 levels (2 tailed), respectively.
Independent variables (1) (2) (3) (4) (5) (6) CSR proxy ENV 0.000*** (2.69) 0.000** (2.31) SOC 0.000 (.23) -0.000 (0.03) CSP 0.000 (1.58) 0.000 (1.35) Firm characteristics: ASIZE -0.008** (-2.04) -0.005 (-1.35) -0.006 (-1.66) -0.007* (-1.83) -0.006 (-1.46) -0.007 (-1.63) AROA 0.085 (1.42) 0.067 (1.12) 0.076 (1.27) 0.119* (1.69) 0.112 (1.57) 0.115 (1.62) AFCF 0.017 (0.31) 0.031 (0.54) 0.026 (0.45) -0.028 (-0.51) -0.024 (-0.42) -0.024 (-0.42) TROA -0.110* (-1.95) -0.095 (-1.63) -0.102 (-1.77) -0.135** (-2.14) -0.128* (-1.89) -0.132** (-2.02) TTQ 0.007 (1.50) 0.004 (0.94) 0.006 (1.23) 0.007 (1.45) 0.005 (0.99) 0.006 (1.22) Deal characteristics: RELDS -0.004 (-0.76) -0.002 (-0.36) -0.003 (-0.54) -0.004 (-0.67) -0.003 (-0.61) -0.003 (-.62) INDDIV -0.007 (-0.80) -0.006 (-0.67) -0.007 (-0.80) -0.014 (-1.39) -0.015 (-1.45) -0.015 (-1.51) COMP 0.009 (0.45) 0.009 (0.48) 0.010 (0.48) 0.006 (0.27) 0.004 (0.19) 0.005 (0.21) DOM -0.017* (-1.88) -0.015 (-1.58) -0.015 (-1.60) -0.008 (-0.84) -0.010 (-0.99) -0.008 (-0.79) METHOD -0.010 (-1.03) -0.012 (-1.22) -0.011 (-1.07) -0.011 (-1.03) -0.012 (-1.17) -0.011 (-1.06) Strength of institutions IF -0.011 (-0.85) 0.004 (0.29) -0.005 (-0.37) IF*CSR score 0.000 (0.82) -0.000 (-0.58) -0.000 (0.23) Constant 0.188*** (3.06) 0.201*** (3.29) 0.199*** (3.23) 0.177*** (2.67) 0.200*** (3.01) 0.192*** (2.87) Year fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes No No No Adjusted R2 0.109 0.080 0.090 0.107 0.080 0.088
26
In model 1, I find a positive and statistically significant impact of the environmental score (ENV). The positive coefficient is significant at the 1% level. This finding suggests a positive, although small, relation between environmental performance of the target and acquirer abnormal returns. The coefficient remains significant after including the strength of institutional frameworks (𝐼𝐹) in model 4. This finding suggests that the environmental performance of targets is valued positively by shareholders in the acquiring firms. The economic significance of environmental corporate investments can be quite substantial. Note, an increase of 1 point (scaled from zero to 100) in the targets environmental rating results in an increase in acquirer abnormal returns by 0.049%. In contrast with model 1, the findings in models 2 and 3 indicate no statistical evidence for a positive influence of the social (SOC) and total CSR score (CSP)
on the acquirer abnormal returns.This is in contrast with the findings of Aktas et al. (2011),
who find a significant positive influence of the environmental, social, and combined IVA score. However, also their results indicate a stronger influence of the environmental score in comparison with the social. Thus, the social performance and CSP of targets is not valued by shareholders in this sample. Therefore, it can be concluded that hypothesis 1 is supported for the environmental performance of the target, but not supported for the CSP and social performance of the target. Indicating that shareholders value environmental investments. This finding is in alignment with prior empirical evidence (see, e.g., Hall and Rieck, 1998; Cheng et al., 2014).
27
Models 4, 5, and 6 represent the interaction effect of 𝐼𝐹 with ENV, SOC, and CSP
respectively. The findings indicate that the institutional framework variable and the interaction
effects with the CSR proxies are insignificant in all the models.19 The signs of the interaction
effect are consistent with the expectation in the SOC and CSP model, although not in line with the expectation in the ENV model. Thus, no reasonable statistical evidence is found to state that the relation between CSR and abnormal returns is weaker (stronger) in countries with stronger (weaker) institutional frameworks. Therefore, hypothesis 3 is not confirmed for the environmental, social, and CSP performance.
Table 6 shows the regression outcomes of the subsamples A(A<T) and B(A>T) with the explanatory variables |∆𝐸𝑁𝑉|, |∆𝑆𝑂𝐶|, |∆𝐶𝑆𝑃|. For clarity, the subsamples A and B indicate the absolute value of the difference in CSR proxies between acquirer and target (A-T). Three main findings emerge. First, the coefficient of |∆𝐸𝑁𝑉| is positive and significant at the 5% level in subsample A. In contrast, the coefficient of |∆𝐸𝑁𝑉| is negative and significant at the 5% level in subsample B. Furthermore, both subsample A and B |∆𝐸𝑁𝑉| coefficients are higher than the full sample ENV outcomes, namely 0.002 and -0.004 respectively. Hence, a greater difference between acquirer and target environmental performance seem to matter for the effects on acquirer abnormal returns. All in all, shareholders appear to value (disvalue) acquirers taking over targets with a relatively higher (lower) environmental performance, which is in line with hypothesis 2. To conclude, the regression results presented in Tables 5 and 6 partly confirm the univariate results reported in Table 4. The significant interaction effect of the ENV and SOC difference with 𝐼𝐹 is presented in Appendix F. The interaction coefficient is slightly positive (negative) and significant at the 10% level in the ENV (SOC) model. The significant positive influence of ENV holds after including the interaction effect. This indicates that larger environmental differences between acquirer and target are valuated more positively in strong institutional frameworks if the acquirer scores lower compared to the target in environmental performance. This is contradicting hypothesis 3. All the interaction effects are insignificant in subsample B.
28
Table 6. Subsample regressions of Acquirers’ Cumulative Abnormal Returns (CARs) and Targets’ CSR ratings. This table reports the OLS regression results of the subsamples A(A<T) and B(A>T) with Acquirers’
CAR(-5,5) as dependent variable and the ATCSRD proxies (|∆ENV|, |∆SOC|, |∆CSP|) as the main independent variables. The event-study methodology used to calculate CAR(-5,5) is described in Section 3.5. Subsample A contains the effects of |∆ENV|, |∆SOC|, and |∆CSP| over respectively 86, 90, and 76 observations. Subsample B contains the effects of |∆ENV|, |∆SOC|, and |∆CSP| over respectively 220, 219, and 233 observations. The subsample selection is described in Section 3.6. All models (7)-(12) include year, country, and industry fixed effects. The t-statistics based on robust standard errors are in parentheses. Appendix A presents definitions and data sources of all used variables. *, ** and *** denote statistical significance at the 0.10, 0.05 and 0.01 levels (2 tailed), respectively. Subsample A Subsample B Independent variables (7) (8) (9) (10) (11) (12) ATCSRD proxy |∆ENV| 0.002** (2.21) -0.004** (-1.98) |∆SOC| -0.000 (-0.31) -0.000 (-0.04) |∆CSP| -0.000 (-0.09) -0.000 (-0.81) Firm characteristics ASIZE -0.011 (-1.36) -0.004 (-0.42) -0.009 (-0.88) 0.001 (0.10) -0.006 (-1.11) -0.001 (-0.22) AROA 0.102 (0.67) 0.285 (1.81) 0.175 (0.88) 0.067 (0.97) -0.029 (-0.38) -0.013 (-0.19) AFCF 0.148 (1.22) 0.012 (0.08) 0.228 (1.49) 0.012 (0.19) 0.047 (0.78) 0.033 (0.56) TROA -0.185 (-1.45) 0.191 (1.07) 0.002 (0.01) -0.109* (-1.68) -0.120** (-2.07) -0.112* (-1.84) TTQ 0.014 (1.35) -0.011 (-0.89) 0.003 (0.18) 0.004 (0.81) 0.005 (1.12) 0.004 (0.93) Deal characteristics RELDS -0.011 (-1.27) 0.003 (0.39) -0.001 (-0.14) -0.001 (-0.07) -0.004 (-0.33) 0.004 (0.36) INDDIV -0.018 (-0.68) -0.035 (-1.12) -0.020 (-0.65) 0.000 (0.03) 0.009 (0.83) 0.005 (0.45) COMP -0.088** (-2.30) -0.004 (-0.08) -0.030 (-0.59) 0.030 (1.47) 0.014 (0.65) 0.023 (1.12) DOM -0.026 (-1.12) -0.026 (-1.07) -0.042* (-1.74) -0.017* (-1.73) -0.019* (-1.78) -0.016 (-1.62) METHOD -0.022 (-1.08) -0.016 (-0.62) -0.021 (-0.72) -0.014 (-1.26) -0.019 (-1.63) -0.013 (-1.16) Constant 0.150 (1.17) 0.034 (0.27) 0.083 (0.63) 0.104 (1.08) 0.230 (2.81)*** 0.162 (2.10)** Year fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes Yes Yes Yes Adjusted R2 0.210 0.096 0.042 0.075 0.061 0.084
Observations 86 90 76 220 219 233
4.4. Portfolio results
29
that the target CSP considerably increases for portfolios Q1-Q3. Moreover, the score only marginally decreases in the highest CARs (Q4), exhibiting an inverse U-shaped relation. The portfolios of Panel A provide evidence for the increasing relation between target CSP and CARs, suggesting that shareholders significantly price higher target CSP values, from an economic point of view.
Table 7. Portfolio results of Acquirers’ Cumulative Abnormal Returns (CARs) and Acquirers’ and Targets’ CSR ratings. This table reports the portfolio results of the full sample and both subsamples.
Observations are assigned into equally weighted quartile portfolios according to Acquirers’ CAR(-5,5). The events with the highest CARs are exhibited in Q4 and the lowest CARs in Q1. The event-study methodology used to calculate CAR(-5,5) is described in Section 3.5. The subsample selection is described in Section 3.6. Panel A shows the portfolio results of the full sample. Panel B reports the results of subsample A(A<T) and Panel C of subsample B(A>T). The subsample selection is described in Section 3.6. Appendix A presents definitions and data sources of all used variables. The t-statistics based are in parentheses. N denotes the number of observations. *, ** and *** denote statistical significance at the 0.10, 0.05 and 0.01 levels (2 tailed), respectively.
Panel A: Full sample Quartiles 1 2 3 4 ENV_A 61.43 (3.44) 75.10 (3.23) 71.60 (3.37) 63.45 (3.59) SOC_A 62.37 (3.58) 74.51 (3.16) 72.70 (2.80) 63.09 (3.36) CSP_A 61.90 (3.39) 74.81 (3.09) 72.15 (3.00) 63.27 (3.38) ENV_T 39.06 (3.19) 42.57 (3.63) 51.18 (3.76) 50.49 (3.63) SOC_T 45.87 (3.49) 48.86 (3.24) 50.59 (3.54) 50.17 (3.42) CSP_T 42.46 (3.11) 45.72 (3.19) 50.89 (3.38) 50.33 (3.30) N 78 77 77 77
Panel B: Subsample A (A<T) Quartiles 1 2 3 4 |∆ENV| 10.72 (3.86) 10.71 (4.69) 12.90 (3.64) 28.44 (5.21) |∆SOC| 19.05 (4.32) 19.03 (3.59) 15.74 (5.59) 20.24 (3.76) |∆CSP| 14.88 (3.62) 14.87 (2.59) 14.32 (3.96) 24.34 (3.84) N 22 16 14 24
30
With reference to Panel B, which includes the absolute differences between acquirer and target CSR proxies, it can be observed that the difference in ENV rises considerably, with values highest for highest CARs (Q4). However, this positive correlation is not found for SOC, consistent with the regression results. Moreover, the difference in CSP is roughly the same for Q1-3, though it shows a sizeable increase for the highest CARs (Q4). Hence, this result indicates that CAR is economically large for deals including firms with high CSP values.
Portfolio analysis of panel C exhibits a hump-shaped and decreasing relation between differences in CSP values and CARs. The largest differences in CSP values are found in Q2 and Q3 respectively. This suggests that the lowest (Q1) and highest (Q4) CARs have the smallest differences between acquirer’s CSP and targets’ CSP value. Thus, when the CSP of the acquirer is larger than the CSP value of the target (panel C), the M&As with the smallest differences earn the lowest and highest CARs. Consequently, the portfolio analysis provides economically significant evidence for a positive relation between CSP values and CAR. 4.4. Robustness tests
