The role of ESG in acquirers’ performance change after M&A deals

: This research aims to study the impact of the target’s ESG score on the acquirer’s ROA and stock price changes after M&A deals by regressing the percentage change of acquirer’s performance change against the target’s ESG score and a set of control variables. This research contributes to the current literature by exploring whether this impact is influenced by the acquirer’s pre-M&A ESG level through two methods—expressing the coefficient of the target’s ESG as s linear function of the acquirer’s ESG and dividing the deals into two groups according to the acquirer’s ESG level. The result of shows that the impact of the target’s ESG score on the acquirer’s ROA change is significant at 95% confidence level and varies for low-ESG and high-ESG acquirer groups. Although most acquirers suffer ROA declines one year after the deals, the ROA decline is aggravated for low-ESG acquirers but is relieved for high-ESG acquirers. This discrepancy can be attributed to the temporary integration costs that are higher for low-ESG acquirers than for high-ESG peers if the target’s ESG level increases. Besides, this research concludes that the impact of the target’s ESG score does not have a significant impact on the acquirer’s stock price change before and after an M&A deal.


Introduction
In recent years, sustainability is becoming an increasingly important concern all over the world. The United Nations (UN) listed 17 sustainable development goals (SDG) to promote sustainable economic growth. In 2004, the UN first used the ESG term in a report about the impact of financial markets on sustainable development (United Nations, 2004). Such a trend makes it essential to investigate the impact of sustainability concerns on global business. In the investigation of Franklin acquirer's performance change (both from the ROA and stock price change perspectives) and that the acquirer's ESG score does not have a significant influence on the coefficient of the target's ESG score. The second regression is devoted to studying further whether the overall coefficient of the target's ESG score is significantly different between high ESG (top 25%) acquirers and low ESG (lowest 25%) acquirers by running another interactive regression. The interaction term in the second regression model is constructed as the product of the target's ESG score and a new dummy variable that can classify acquirers into two groups-high ESG acquirers and low ESG acquirers. The result shows that, for the ROA change, the overall coefficient of the target's ESG score for low-ESG acquirers is lower (more negative) than that for high-ESG acquirers, i.e., low ESG acquirers suffer greater ROA decline than high ESG acquirers if the target's ESG score increases. However, the influence of the target's ESG score on the acquirer's stock price change is still not significant even though the acquirer's pre-M&A ESG level is considered.
The findings of this research have a set of practical implications for making M&A-related decisions. An acquirer should consider not only the target firm's ESG score but also the acquirer's own ESG status when making acquiring decisions. When selecting targets, an acquirer should pick firms whose ESG levels are in line with its strategic goals, rather than simply thinks that the higher the ESG score, the better. Moreover, an acquirer should evaluate the impact of the target's ESG level on its performance changes by considering various indexes such as ROA, stock price, etc. This research develops in the following outline. It first briefly introduces the M&A activities under the ESG background and lists the research objectives and contributions. Following that, the literature review section presents the current research achievements about the impact of ESG consideration on M&A and other financial activities. In light of these previous studies, the paper then moves into the methodology section, which designs a set of regressions to test the impact of targets' ESG level on acquirers' post-M&A performances. The key findings are summarized and analyzed in the "numerical analysis" section. This section also proposes some business suggestions from a practical perspective. Then the following "robustness analysis" section presents some additional analysis to validate the key findings of this research. For example, consider the environmental score, social score, and governance score separately in the regression models and consider US deals only to exclude the country effect. This research also conducts a 2SLS regression to exclude the endogeneity problem. Finally, the conclusion section reviews the whole study and discusses the implication of the findings to real business practices and future researches in the area of ESG/green M&A.

Literature review
How to improve economic sustainability is one main business challenge today. There has been an increasing social concern on sustainability issues not only among environmental advocacy groups and policymakers but also among enterprises that have begun to consider ESG factors parallel to their economic performance (Abbasi and Nilsson, 2016;Rao et al., 2015).

Benefits of good ESG practices
Good ESG practices usually can reflect a firm's corporate social responsibility (CSR) and improve a firm's business image among stakeholders (Godfrey, 2005;Martínez-Ferrero et al., 2016;Franklin, 2019). Thanks to the positive business image, both the firm and its stakeholders are satisfied (Freeman, 1984). For example, customers will become more loyal to the firm, and business partners will trust the firm more if a company has a positive image (Barney and Hansen, 1994;Fombrun et al., 2000). Besides, a good reputation can help the firm attract more capital investment (Cheng et al., 2014) and high-caliber talents (Fombrun et al., 2000;Branco and Rodrigues, 2006). Moreover, Ferrell et al. (2016) found that good ESG performances are helpful for increasing firm value and decrease the negative effect of managerial entrenchment. These impact of sustainability practices on stakeholders may, in turn, affect financial activities such as the M&A processes and the acquirers' post-M&A performance (Bettinazzi and Zollo, 2017). This relationship is just the underlying idea that guides this research, which is devoted to exploring whether and how good ESG practices determine post-M&A performances of bidders.

Impact of good ESG practices on M&A deals
It is reasonable to think that, after merging a target with good ESG performance, the acquirer's ESG level, reputation, and business image will be positively influenced. These positive influences will be finally reflected in the acquirer's market price and financial performances, which can be measured by various indexes such as ROA. Previous scholars have studied the interaction relationship between sustainable issues and M&A deals from various perspectives. Most researches focus on the influence of merging a good ESG firm on the acquirer's post-M&A performance. Tampakoudis and Anagnostopoulou (2020) studied 100 European M&A cases and concluded that acquires' post-M&A ESG level and market value increase after merging a target firm with high ESG performance. Deng et al. (2013) studied a sample of US M&A deals and found that deals taken by acquirers with high ESG scores are more likely to succeed, take less time to complete, and can improve the long-term post-M&A operating performance of these acquirers. Salvi et al. (2018) studied 171 firms in the US & EU 28 countries (including the UK) from 2001 to 2013. They conducted a simple OLS regression of acquirers' performance (ROA) on target firms' overall ESG score and a set of control variables. Their results rejected the null hypothesis that the target's ESG level did not significantly affect the acquirer's post-M&A performance and concluded that acquirers that choose "green" deals could gain better financial performances than acquirers that merge firms in other sectors. Most of these researches got a consistent result that merging a high-ESG target is good for the acquirer's post-M&A performance.
Some researchers studied the influence of ESG on acquires' risk rather than performance. For example, Arouri et al. (2019) mainly explored the relationship between the ESG and M&A uncertainty by focusing on 726 M&A deals worldwide from 2004 to 2016, and they found that the arbitrage spreads in M&A deals are decreased by 1.10 percentage points for each unit-increase of standard deviation in the acquirer's ESG score. Besides, they also concluded that 'Green' acquirers are more likely to avoid unsuccessful deals. Arouri et al. (2019) verified their results by conducting robustness tests.
Other relevant studies include , who investigated the positive impact of acquirers' engagement in employee issues on the M&A deals. Rydell and Leucht (2020) qualitatively studied the role of ESG in M&A Due Diligence. They found that Investigating ESG in the Due Diligence process can help the acquirer to assess risk, identify sources of value, and support the achievement of long-term goals. Researches conducted by Manocha and Srai (2020), Manocha et al. (2016), andFeng (2021) tried to study the role of ESG, especially the environment factor, in M&A deals in the supply chain industry. Outside of the M&A topic, Ng and Rezaee (2015) explored the interaction relationship between sustainability and cost of equity. They discovered that good sustainability performance is helpful for reducing a firm's cost of equity.
To test whether acquiring a high ESG target can improve the acquirer's post-M&A performance, this research first "repeated" the simple OLS regression method frequently used in previous studies such as Salvi et al. (2018). The OLS regression in this research is based on 124 deals worldwide in the last two decades. Referring to the conclusion of previous studies that acquiring a high-ESG target can benefit an acquirer, the expectation on the relationship between the target's ESG score and the acquirer's performance change is formulated in hypothesis: Hypothesis 1a: The increase of the target's ESG score will have a significant (confidence level no less than 95%) positive influence on the acquirer's ROA change (increase the acquirer's ROA).

Research gaps
Despite the various previous studies on the impact of sustainability on M&A deals, current literature has left unanswered two crucial questions about ESG activism. The first unanswered question is whether the impact of target firms' sustainability on acquirer's performance (the overall coefficient of target's ESG in regressions) differs between high-ESG and low-ESG bidders. This research, for the first time, studied the impact of acquirers before-M&A ESG level through two methods. The first method is to consider the overall coefficient of the target's ESG score as a linear function of the acquirer's pre-M&A ESG score by introducing an interaction term-the product of the acquirer's ESG score and the target's ESG score-into the regression. The findings of Franklin (2019) and  imply a negative relationship. They found that low ESG companies can significantly improve performance after merging high-ESG firms because the deals are helpful for them to build a positive business image, reduce cost, and mitigate legal risks. However, they argue that bidders that already have high-ESG performance may not improve their performance a lot. A high-ESG firm merges a firm with good ESG practices because it wants to continually keep a positive reputation, achieve economies of scale, or just select targets that are similar to themselves-reasons that do not directly lead to performance improvement. According to these findings, the expectation about the linear relationship is captured in the following hypothesis: Hypothesis 2a: The increase of the acquirer's pre-M&A ESG score will significantly (at 95% confidence level) decrease the impact of the target's ESG score on the acquirer's ROA (change), i.e., a negative relationship between the acquirer's ESG score and the overall coefficient of the target's ESG score.
Besides, the hypothesis 1a is also tested in this method. Although this method has some merit, the linear relationship between the acquirer's ESG score and the overall coefficient of the target's ESG score may not hold in real practices. To release this strict assumption of linear relationship, the second method studies the influence of the acquirer's own ESG score by measuring the impact of target's ESG score on low-ESG and high-ESG acquirers separately. This method sorted selected acquirers into high-ESG group and low-ESG group based on acquirers' ESG levels before deals by constructing an interaction term, which is the product of the target's ESG score and a dummy variable that represents the two group of acquirers (1 if the acquirer is a high ESG acquirer and 0 if the acquirer is a low ESG acquirer based on pre-M&A ESG scores). It is necessary to distinguish the low-ESG and high-ESG bidders because the coefficient that reflects the impact of the target's ESG may significantly differ (but not in a linear relationship) between these two kinds of acquirers. Also referring to the findings of Franklin (2019) and , the expectation about the different impact of the target's ESG score between high-ESG acquirers and low-ESG ones is captured in the following hypotheses: Hypothesis 3a: The increase of the target's ESG score will significantly increase the acquirer's ROA after the deal, but the ROA of high-ESG acquirers will increase significantly less (in percentage) than the ROA of low-ESG acquirers.
The second question is that what is a good proxy for the acquirer's performance change? And will the results converge if using different proxies? Salvi et al. (2018) only studied the impact of the target's ESG score on the acquirer's ROA change. And Deng et al. (2013) studied the impact of the targets' ESG levels on the acquirer's long-term operating performance. Both these two studies use the accounting performance as a proxy for performance without considering performance in the financial market, such as the stock price. Only Tampakoudis and Anagnostopoulou (2020) studied the influence on market value but only based on a European database. This research, for the first time, examines the impacts on both the accounting performance (ROA) and the stock price separately. This research studies the acquirer's stock price change by changing the dependent variable in regression models to the acquirer's stock price change. Therefore, the hypotheses about the impact of the target's ESG score on the acquirer's stock price change and the influence of the acquirer's own ESG score on this impact are defined as: Hypothesis 1b: The increase of the target's ESG score will have a significant positive influence on the acquirer's stock price change (increase the acquirer's stock price).
Hypothesis 2b: The increase of the acquirer's pre-M&A ESG score will significantly (at 95% confidence level) decrease the impact of the target's ESG score on the acquirer's stock price (change), i.e., a negative relationship between the acquirer's ESG score and the overall coefficient of the target's ESG score.
Hypothesis 3b: The increase of the target's ESG score will significantly increase the acquirer's stock price after the deal, but the stock price of high-ESG acquirers will increase significantly less (in percentage) than the stock price of low-ESG acquirers.
In addition, this research also advances the current literature by focusing on the detailed impact of sustainability on the acquirer's performance change. Moreover, following the suggestion of Fransen (2013), this research exploits the effects of various ESG dimensions-environment, society, and governance-separately in the robustness analysis section.
A brief comparison between previous studies and this research is summarized in the following Table 1.

Data and variable description
This section summarizes the overall observations and defines all the variables that are used in this research.

Sample selection
The data used in this research includes all important M&A cases that happened since 2000-from 1 st Jan 2000 to 31 st Dec 2020. We delete deals between 2007 and 2009 to exclude the influence of the abnormal merges trend caused by the financial crisis. Data about M&A transactions can be gained from the Zephyr database. Specifically, this research only selects M&A deals with the following prosperities: (1) the M&A deal is completed, (2) the deal value is at least 1 million USD, (3) the acquirer initially holds less than 50% of the target firm and acquires more than 50% finally, (4) the deal has standard terms and does not contain option-like contingent claims, and (5) both targets and acquirers' ESG scores can be gained from the Thomson Reuters ASSET4 ESG database. This research also excludes all financial (SIC code 6000-6999) and utilities (SIC code 4000-4949) companies from the study sample because M&A activities are highly regulated in these industries (Edmans et al., 2012). We delete all deals that have one or more unavailable values for variables defined in Table 3, Panel A. 124 deals satisfy all the above requirements and are used as inputs in this research. Table 2 briefly summaries these deals, and documents that included detailed information are available on https://github.com/XUAN-FENG9/ESG_Tar-and-performance_Acq.   This research uses two different proxies as the dependent variables to reflect the acquirers' performance change from different perspectives. The first one is an accounting proxy-the acquirers' Performance change from 1 year before the deal completion to 1 year later after the deal completion. The ROA is defined as EBIT divided by total asset in this research. Instead of the net income, the EBIT is used as the nominator of ROA because it is a more stable proxy for the return as the EBIT excludes the influence of deal characteristics on the acquirer's performance. For example, deals that are conducted by high-leverage LBO or deals in countries that charge high tax on income will lead to smaller net income for acquirers after the deal. But EBIT is not affected by these factors, and, therefore, ROA with EBIT as the nominator is a more suitable proxy to reflect acquirers' profitability. The second proxy-the stock price change from 3 months before the announcement to 1 month after the completion (∆StockPrice_-3to1)-is from a perspective of financial markets. The stock price change ∆StockPrice_-3to1 is highly positively related to the stock price change from the day before the announcement to the day after the completion (∆StockPrice). For the independent variables, a company's ESG score is chosen as the arithmetic mean of its available ESG scores from 2000 to the year before the announcement year. The ESG scores (DS Datatype: TRESGS) is directly available at the Thomson Reuters' Asset 4 database, in which a company's ESG score is defined as the overall company score based on the self-reported information in the environmental, social, and corporate governance pillars. We choose the average score before the announcement year, rather than the completion year, to exclude the abnormal ESG changes led by expected M&A activities. Another reason for using the historical average ESG score is that the ESG scores before the announcement year are not available for many companies, even though they have ESG scores years before the announcement year. Control variables that reflect characteristics of a deal include the deal size, the deal premium, the deal payment methods (cash or others), geographic information (cross-border deal or domestic deal), and industry information (cross-industry deal or inter-industry deal). For control variables reflecting firm characteristics (A for acquirers and T for targets), this research uses financial data one year before the deal announcement. These control variables reflect the firm from five different perspectives. Tobin's Q value can reflect whether the firm is overvalued or undervalued in the market. The Debt-to-Asset ratio is a proxy for the firm's financial leverage. The asset-turnover ratio is an indicator of a firm's efficiency to generate revenue from its asset. And the EBIT margin measures a firm's efficiency of operation and its capability to transfer sales revenue to profit. This research also includes the natural log value of the total asset to represent the firm size. There are no significant correlations among all independent and control variables, as shown in Figure 1. For further regression models, all continuous inputs are winsorized at 5% and 95% levels.

Methodology
Firstly, based on the study of Arouri et al. (2019), this research builds a benchmark model in section 4.1 by regressing the change of acquirers' performance against the target companies' ESG scores, the acquirers' ESG scores, and a set of control variables that reflect the characteristics of the deal, target firm, and acquirers. Different from the research of Arouri et al. (2019) that explores the impact on M&A uncertainty, this research mainly focuses on investigating the potential impacts of target companies' sustainability level on acquirers' post-M&A performances.
Furthermore, this research builds two interaction regression models to study the impact of acquirers' own ESG level on the causal effect from targets' ESG to acquirers' performance change. As presented in section 4.2, the first model considers the impact of targets' ESG score to acquirers' performance change as a linear function of acquirers' ESG score. As presented in section 4.3, the second model is devoted to exploring whether the causal effect's extent significantly differs between high-ESG acquirers and low-ESG acquirers. The acquirers can be sorted into two groups by introducing a dummy variable that equals 1 for high-ESG acquirers and 0 for low-ESG acquirers.

Model 1-The benchmark model
This research begins with a benchmark model without considering the impact of acquirers' ESG scores. The benchmark model can be written as: where the subscript represents the identity of deals. ∆ is the change of acquirers' performance defined in Table 3. This research runs the regressions separately based on two proxies of performance change-ROA change (one year before to one year after the announcement) and the stock price change (3 months before the announcement to one month after the completion).
-the intercept of this model-is meaningless because, in reality, all control variables cannot be zero at the same time. and are the initial observed ESG score of targets and acquirers, separately. As defined in section 4, is a set of control variables that reflect the acquirer's characteristics, is a set of control variables that represent characteristics relative to the deal, and represents control variables that represent the characteristics of the target company. , , and are fixed effects of time, country and industry, separately.

Model 2-Considering the coefficient as a linear function of acquirer' ESG
Then this research considers the impact of targets' ESG on acquirers' performance change as a simple linear function of the acquirers' ESG score. In the first step, this research builds an interaction term-the product of and . This interaction term in Equation (2) is meaningful by considering that merging a target company can have different impacts on post-M&A performance among acquirers with different ESG levels. This interaction regression is presented as: The Equation (2) can also be written as Equation (3): where the overall coefficient of target's ESG score ( ) is comprised of two parts-a constant ( 1 ) and another term that is determined by acquirer's ESG score ( 3 × ). 1 is the effect of targets' ESG on acquirers' performance change when the acquirers' ESG score is zero. 3 measures the impact of acquirer's ESG score on the overall coefficient of target's ESG score, i.e., the incremental impact of targets' ESG on acquirers' performance change when the acquirers' ESG score increases by 1. So, the overall impact of target's ESG score on the acquirer's performance change is presented as . As defined in section 3, is a set of control variables that reflect the acquirer's characteristics, is a set of control variables that represent characteristics relative to the deal, and represents control variables that represent the characteristics of the target company. , , and are fixed effects of time, country and industry, separately.

Model 3-Sorting acquirers into high-ESG and low-ESG groups
The ESG level of acquirers can have an influence on the impact of targets' ESG level on acquirers' performance change after deals. The coefficient 3 in Equations (2)-(3) is devoted to reflecting this impact, but the Equations (2)-(3) in the last step are based on all deals and do not directly distinguish the different effect of targets' ESG level on acquirers' performance change between high and low ESG acquirers. The overall coefficient ( 1 + 3 ) in Equations (2)-(3) implies a linear relationship between the acquirer's ESG score and the overall coefficient/impact of the target's ESG level. However, this linear increase relationship may not hold true in practice. To study the influence of the acquirer's own ESG score but release the strict assumption of linear relationship, this model explores the different impact of the target's ESG on two opposite acquirer groups (high-ESG acquirers and low-ESG ones). These two groups should have an obvious difference in their ESG scores. Therefore, this model only selects deals in which the acquires' total ESG scores are ranked in the top 25% and the lowest 25% among all acquirers. Acquirers that ranked in the top 25% are defined as high ESG acquirers, and those ranked in the lowest 25% are called low ESG acquirers. A dummy variable-_ is created to distinguish the acquirers based on their overall ESG levels. _ equals 1 if the acquirer is a high ESG acquirer and equals 0 if the acquirer is a low ESG acquirer. The product of this dummy variable and the target's ESG score is used as an interaction term to distinguish the impact of the target's ESG score on performance change between high-ESG and low-ESG acquirers. This model can be presented as: × _ + 4 + 5 + 6 + + + + The Equation (4) can also be written as Equation (5): where in Equations (4)-(5), the _ replaces the in Equations (2)-(3) to reflect the acquirers' difference in terms of their ESG level. For low-ESG acquirers, _ equals 0, and the coefficient 1 reflects the the impact of targets' ESG level on the acquirers' performance change. For high-ESG acquirers, the impact of targets' ESG on acquirers' performance change is reflected by the sum 1 + 3 , because the _ equals 1. As defined in section 3, , , represents control variables that represent the characteristics of the target company, acquirer, and the deal separately. , , and are fixed effects of time, country and industry, separately.

Numerical analysis
This section discusses the main results gained from Equations (1)-(5). The output of the fundamental OLS regression is first presented in Table 4. Then Table 5 records the result of the regression that considers the coefficient of the target's ESG score as a linear function of the acquirer's ESG score. Table 6 presents the result of the regression that introduces an interaction term to sort the acquirers into two groups (high-ESG and low-ESG acquirers).
All the regression models in this research are run on the R Studio environment. An Intel Core i5-8250U CPU (1.60 GHz) laptop with 8 GB RAM is applied for carrying out all the computations. Table 4 presents the result of Equation (1)-the fundamental OLS regression model. The result shows that targets' ESG scores do not have a significant influence on acquirers' post-M&A ROA or stock price change at 95% confidence level, a result that denies the hypotheses 1a and 1b. This result is also contradictory to the conclusion of most previous scholars such as Salvi et al. (2018), who found that acquirers could increase their ROA after deals by merging "green" targets. This discrepancy can be attributed to three possible reasons. The first reason is that merging a high-ESG target has opposite influences on low and high ESG-level acquirers. The opposite influences are offset in the OLS regression based on all deals, so the coefficient of the target's ESG score is not significant. Whether the opposite influences exist are explored in the third model that runs the regression on low-ESG and high-ESG acquirer groups separately. The second reason may be the data limitation. This research only collects 124 deals, and the study of Salvi et al. (2018) is based on 171 deals. Such small databases can not support a stable and reliable conclusion. So, the discrepancy between this research and previous studies may be attributed to random errors. The third reason is that different ESG proxies may lead to conflicting ESG scores for a company. In this research, firms' ESG score is calculated as the average ESG scores before deals, and the ESG scores are gained from the Thomson Reuters ASSET4 ESG database. Salvi et al. (2018) define a firm's ESG level according to its industry attributes (green industry or sin industry). There are also other ESG databases such as the MSCI (Formerly KLD). As Gibson et al. (2021) studied, seven different data providers have considerable disagreements on ESG ratings. Therefore, the discrepancy of the result of this research and previous studies may be attributed to the variance of methods used to measure firms' ESG level.  Table 5 presents the result of the regression that consider the overall impact of the target's ESG score on the acquirer's performance change as a linear function of the acquirer's ESG score. The coefficient of the interaction term ( 3 ) is not significant at the 95% confidence level. So, the hypotheses 2a and 2b cannot be held, i.e., there is not a significant linear relationship between the overall impact of the target's ESG score and the acquirer's ESG score. From this result, we cannot clearly recognize the influence of the acquirer's own ESG level. Besides, the coefficient of the target's ESG score in Table 5 is also insignificant. This result is consistent with the conclusion gained from Table 4-the targets' ESG scores do not have a significant influence on acquirers' post-M&A ROA or stock price change. Similar to the coefficient of the target's ESG score in the first OLS regression model, these insignificant coefficients in this regression can also be classified into the three potential reasons: random errors due to data limitation, ESG-proxy selection bias, and the non-linear relationship.   i.e., the heteroscedasticity does not significantly exist.

Sorting acquirers into high-ESG and low-ESG groups
However, the result is different if we release the linear relationship assumption and consider the influence of acquirers' ESG scores by running regressions for low-ESG acquirers and high-ESG acquirers separately. As defined in Equations (4)-(5), acquirers are separated into high-ESG and low-ESG groups before M&A deals by introducing an interaction term-the product of targets' ESG scores and the acquirers' property according to their pre-M&A ESG level (a dummy variable separating acquirers into high-ESG and low-ESG groups). As presented in Table 6, the result shows that the targets' ESG score has a significant impact (at 90% confidence level) on acquirers' post-M&A ROA change, but not on their post-M&A stock price change. The direction of this impact varies between high-ESG acquirers and low-ESG acquirers. For low-ESG acquirers, the coefficient that measures the extent of this impact is 1 , which equals 0.01 (1%). This value shows that, for low-ESG acquirers, if the target's ESG score increases 1 unit, the acquirer's ROA change will decrease 1%. The decrease of acquirers' ROA change also means the decrease of acquirers' ROA. In this research, most acquirers' ROA changes are negative (as shown in Table 3, the mean value of the ROA change is −37%), so the decrease of ROA change implies that the ROA change becomes more 'negative', i.e., the acquirer's ROA decreases. But the increase of targets' ESG score has a positive impact on high-ESG acquirers. This impact is measured by 1 + 3 , which equals 1.3% (−0.01+0.023), as presented in Table 6. This value shows that if the target's ESG score increases 1 unit, the ROA change of a high-ESG acquirer will increase 1.3% (absolute value) after the M&A deal. The increase of acquirers' ROA change also means the increase of acquirers' ROA.
The result-the target's ESG score has a positive impact on high-ESG acquirers' ROA but a negative impact on low-ESG acquirers' ROA-is just opposite to hypothesis 3a and previous studies such as Salvi et al. (2018). Previous researchers found that a high-ESG target can increase the ROA of both high and low ESG acquirers, but this research concludes that the increase of the target's ESG score has the opposite impacts on these two kinds of acquirers. This discrepancy can be potentially explained by the short-term cost of integration that is not considered in previous studies. In this research, the acquirer's ROA change is measured from a short-term period perspective-the difference of ROA one year before the announcement and one year after the completion. In the short term, high-ESG acquirers can integrate with high-ESG targets more efficiently than low-ESG peers because they have similar ESG-oriented corporate cultures, strategic goals, and governance organizations. As implied by the negative ROA change (ROA decrease) for all deals in this research database, the cost of integration cost cannot be avoided after M&A deals, and this integration cost can lead to the shortterm decrease of acquirers' ROA. However, in the short term, high-ESG acquirers will cost less and decrease less in their ROA after acquiring a similar high-ESG target than low-ESG acquirers do. This difference is especially significant in the short term because there will be no more integration costs after completing the integration. For low-ESG acquirers, the benefit of acquiring a high-ESG target that may lead to ROA increase can only be observed in the long term when the integration is completed. The long-term ROA increase is found by previous scholars, such as Salvi et al. (2018), who studied the acquirer's ROA change several years after the completion of M&A deals.
Even though this method sorts acquirers into high-ESG and low-ESG groups, the impact of the target's ESG score on the acquirer's stock-price change is not significant, and the hypothesis 3b cannot be verified. One main possible reason is that the considerable stock-price volatility brings lots of noise into the regression models. The stock price has greater volatility (daily change) than accounting indexes such as ROA, which is measured annually. So, in order to study the impact on acquirers' stockprice change, future researches should monitor each day's stock price during a broader period, rather than just select the stock prices at two points-the day 3 months before the deal and the day 1 month after the deal, as this research does.
The result in Table 6 can also explain why the fundamental model (Equation 1, result shown in Table 4) and linear relationship model (Equations 2-3, result shown in Table 5) do not show a significant impact of targets' ESG score on acquirers' post-M&A ROA change. The targets' ESG score has opposite impacts on high-ESG and low-ESG acquirers, but both these two models do not distinguish acquirers according to their pre-M&A ESG level. If pooling all acquirers together in the regression, the overall coefficient/impact of targets' ESG score is not significant because the negative impact on low-ESG acquirers offsets the positive impact on high-ESG acquirers.

Separate ESG scores
To explore whether the target's individual ESG scores have different impacts on the acquirer's performance change, this research also separately uses the environment score, the social score, and the governance score to replace the total ESG score in the fundamental OLS regression model. The Thomson Reuters' Asset 4 database records three indexes relative to environment (Resource Use Score, Emissions Score, and Environmental Innovation Score), four indexes relative to governance (Management Score, Shareholders Score, CSR Strategy Score, and Workforce Score), and three indexes relative to society (Human Rights Score, Community Score, and Product Responsibility Score). In this research, the environment score in each year is defined as the arithmetic mean of the three indexes relative to the environment topic in that year. The society score and the governance score in each year are defined in the same way. Like the total score, a company's environmental, social, or governance score is calculated as the average value of each score before the announcement year. The whole Thomson Reuters' Asset 4 ESG database is available at https://github.com/XUAN-FENG9/ESG_Tar-and-performance_Acq/blob/main/Original%20ESG%20database.xlsx.
As presented in Table 7, the result shows that each of the three separate scores of a target does not have a significant impact, as the target's total ESG score does. This result is in line with the coefficient matrix in Figure 1, which shows that the three separate scores are highly positive related to the total score. There is no need to explore the impacts of these individual scores in other models.     Note: All continuous variables are winsorized at the 5% and 95% level. T-statistics are in parentheses. ***, ** and * denote significance at the 1%, 5% and 10% level respectively. The NCV (non-constant-variance) test is used to test whether heteroscedasticity exists. If the p-value of NCV test is greater than 0.05, we cannot reject the null hypothesis that the variance of the residual constantly equals to zero,

Continued on next page
i.e., the heteroscedasticity does not significantly exist.

Only US deals
Despite considering the country effect, the regression models used in this research can be biased because the main analysis used only 124 deals as inputs-some countries' acquirers only joined one deal. Among the 124 deals, there are 95 (over two-thirds) all-US deals in which both the acquirer and the target are US firms. Therefore, we only consider these all-US deals in this step to further control the country effect. As presented in Table 8, the results are in line with the result of the main analysis based on all deal data. For US firms, the coefficient of the target's ESG score is not significant at the 95% confidence level under the OLS model, and this coefficient does not have a significant linear relationship with the acquirer's ESG level under model 2. However, the result of model 3 is more meaningful. Although both the low-ESG and high-ESG acquirers will suffer a short-term ROA decline after the M&A deal, a low-ESG acquirer will decrease 1.3% more in its ROA while a high-ESG US acquirer will decrease 1.5% (−0.013+0.028) less in its ROA if the target's ESG score increase by 1 unit. These results verify the findings in the main analysis of this research.     i.e., the heteroscedasticity does not significantly exist.

Endogeneity problem
In previous researches, the causal effect from targets' ESG score to acquirers' performance change may suffer endogeneity problems. The study of Salvi et al. (2018) is under skeptical because they only simply regressed the acquirer's performance change on the target's ESG, without considering the endogeneity problems that may be caused by (1) the reversal effect from the acquirer's expected performance change to the selected target's ESG score; (2) the omitted variables that may influence both the dependent and independent variable; (3) measurement errors (Edmans et al., 2012). The first endogeneity problem is the most remarkable because an acquirer may acquire a certain ESG-level target because it knows that this action will improve its performance (ROA or stock price) after the deal. Therefore, the commonly expected performance change of acquirers may determine the acquirers' desirable ESG scores of targets, and, in practice, acquirers choose to merge targets with ESG levels similar to what they want. This reversal influence weakens the explanation power of the causal relationship between targets' ESG score and acquirer's post-M&A performance change.
To enhance the result that it is the targets' ESG level that leads to the change of acquirers' performance change, this research constructs a two-stage least squares (2SLS) model by introducing two instrumental variables-the average ESG scores of the targets' industry ( ) and the average ESG scores of the targets' country ( ), respectively. The use of these two instrumental variables is referred from Arouri et al. (2019). The two instrumental variables only determine the target company's ESG level but do not directly influence the performance change of acquirers (Arouri and Pijourlet, 2017;Gomes and Marsat, 2018).
For the first step, the target companies' ESG scores are regressed against the target's country average ESG scores, its industry average ESG score, and -a set of control variables that reflect target firms' characteristics, as defined in Section 3. The regression model is presented in Equation (6). And the regression result is recorded in column (1) in Table 9.
where the 0 represents the target-firms' observed ESG score. The residual term can have non-zero correlation with the dependent variable or the residual term in Equation (6) If the endogeneity problems exist. Therefore, in the following second step, we use the fitted value ̂ in Equation (7) as the independent variable to represent the ESG level of the target company.
Then in the second step, instead of the real ESG score of target companies, the fitted targets' ESG score -̂ -is used as the independent variable in the model 3 (Equations (4)-(5)) to mitigate the biases caused by endogeneity problems. The second-stage regression results are presented in column (2) in Table 9.
The result in Table 9 is consistent with the main conclusion from the model 3. Although both the low-ESG and high-ESG acquirers will suffer a short-term ROA decline after the M&A deal, a low-ESG acquirer will decrease 1.5% more in its ROA while a high-ESG US acquirer will decrease 0.4% (−0.015+0.019) less in its ROA if the target's ESG score increase by 1 unit. The impact on ROA change is significant at 90% confidence level, but the impact on stock-price change is not significant. These results verify the findings in the main analysis of this research.

Conclusions
The topic of this research is to explore the impact of the target's ESG score on the acquirer's post-M&A ROA change and stock price change. This impact is studied through a set of regression models (regress the acquirer's performance change against the target's ESG score and a set of control variables). As most previous researches did, this study firstly runs a simple OLS regression model, and the result shows that there is no significant impact of the target's ESG score on the acquirer's performance (ROA and stock price) change. Then this research contributes to the current literature by, for the first time, studying how the acquirer's pre-M&A ESG level can influence the impact of the target's ESG score on the acquirer's performance change. This influence is studied through two methods. The first method is to consider the coefficient/impact of the target's ESG score on the acquirer's performance changes (ROA change and stock price change) as a linear function of the acquirer's pre-M&A ESG score. The result of this method shows that there is not a significant linear relationship between these two variables. The second method is to measure the impact of the target's ESG score on the acquirer's performance change for high-ESG acquirers and low-ESG acquirers separately. The result of this method shows that the increase of the target's ESG score will deteriorate the ROA of low-ESG acquirers but benefit the ROA of high-ESG acquirers at 95% confidence level and that the influence on acquirers' stock price change is not significant. This result is opposite to the conclusion of previous scholars such as Franklin (2019) and , who found that the performance of high-ESG acquirers increased less than the performance of low-ESG acquirers. In this research, the input database for all regression models is comprised of 124 international M&A deals that happened during the last two decades. All acquirers and targets in these deals are list companies whose financial information is collected from the Zephyr database and ESG scores from the Thomson Reuters ASSET4 ESG database.
The findings of this study provide some valuable implications to real business practices. Bidders should consider their own ESG levels before making M&A decisions. Although it is reasonable to think that acquiring a target with good ESG performance is conducive to improve the acquirer's public image and mitigate its compliance risk, the stereotype 'the higher target's ESG, the better' is inaccessible and even harmful to acquirers. When acquiring a high-ESG target, a bidder should evaluate and balance both the benefit of the target's good ESG performance and the potential cost of integration due to the differences in corporate system and strategic goals. Low-ESG firms and high-ESG peers usually have different corporate policies, strategic goals, cultures, and organizational structures. A bidder should carefully consider these differences caused by ESG variation when evaluating a potential target firm.
This research is subject to a set of limitations in need of further investigation. Firstly, this research studies the impact of the target's ESG score on the acquirer's short-term performance change without considering the long-term influences. Although this study finds that acquiring a high-ESG target can deteriorate a low-ESG acquirer's short-term ROA (in one year), it is reasonable to think that such a situation may reverse in the long-term when the acquirer makes enough adjustments to finish the strategic and cultural integration. It is interesting to verify this long-term impact in future researches. Secondly, this research only uses acquirers' short-term ROA and stock price changes as proxies for their performance change. But only considering these two indexes is far from sufficient for an acquirer to make M&A decisions. Therefore, it is attractive for future scholars to study the impact of targets' ESG level on acquirers from other perspectives such as risk, bid premium, and credit rating. It is also interesting to study whether a potential target company can have a higher valuation by improving its ESG score before M&A. Thirdly, in this research, all firms' ESG scores are gained from the Thomson Reuters ASSET4 ESG database. But the proxy used to represent the ESG level/score varies across different rating agencies. This variation may lead to estimation biases, and, therefore, further scholars are encouraged to find a method that can better reflect the real ESG level of a firm. Last but most important, the database available for this research is small. This research only gathered 124 deals in which both the acquirer and the target are listed companies and have available ESG score before M&A.
Results based on such a small database may be biased and do not have sound persuasion. In order to verify the results of this study, further researches are recommended to repeat the method in this research but based on a larger deals database if available. Besides, because the available data volume is limited, this research does not explore the different impacts on acquirers across different regions/economic entities. Most of the deals used in this research come from the US market. In further researches, it is worthful to study M&A deals in various economic entities separately. The separate impacts are worth exploring because the impact of ESG practices on M&A deals may vary because the economic properties -such as legal systems, social preferences, and institutional arrangements -differ among US, European, and emerging markets . Moreover, it is also interesting to investigate the impact on acquirers that are not public firms. The result may vary because private companies usually face less regulation and stakeholder engagement than public firms. But it is very challenging to obtain data of non-public firms.

Ethical considerations
Research ethics is highly essential in research. Although primary researches are supposed to follow more ethics, secondary researches are subject to rigorous ethics as well. The research process of the project obeys all the ethical rules and principles that secondary researches must follow. This research is based on secondary data mainly collected from the Zephyr and Thomson Reuters ASSET4 ESG databases. These sources are all public and freely accessible through the online Erasmus University library. The data in this research only hold information about the data, are correctly cited with references, and do not expose any personal details. This research does not contain any confidential data, and the "Data Protection Act of 1998" is maintained in the data collection process. This research conducts the data analysis mainly through programming tools such as R and Python. Advanced models are established according to statistical theories described in this research. The output of this research will be in a dissertation style, and the APA citation style will be maintained in the whole paper. Overall, this research has followed all the business research rules and gives a clear view to readers.