The causal effect of religious piety on shareholder wealth: evidence from acquirer returns and historical religious identification

ABSTRACT

Prior research shows that religion promotes honesty. Honesty in turn motivates managers to view an expropriation from shareholders as self-serving, opportunistic and unethical, thereby alleviating the agency conflict. Religious piety is thus expected to discourage agency-driven acquisitions that reduce shareholder wealth. We exploit the variation in religious piety across US counties (and states) and show that firms located in a more religious environment are indeed less likely to make poor acquisitions, measured by the stock market reactions to the acquisition announcement. To draw a causal inference, we use historical religious piety as far back as 1952 as our instrument. The two-stage least squares (2SLS) analysis confirms that religious piety induces firms to make better acquisitions. Our analysis based on propensity score matching also corroborates the conclusion.

KEYWORDS:

• Mergers and acquisitions
• M&A
• religion
• religiosity
• religious piety
• corporate restructuring

JEL CLASSIFICATION:

• G34
• Z12

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The literature considers a firm more ‘religious’ if its headquarters are located in a more religious environment, measured as the percentage of religious adherents in the county where the headquarters reside (Hilary and Hui 2009; Grullon, Kanatas, and Weston 2009; Dyreng, Mayew, and Williams 2012). We follow the same method, although we measure religious piety both at the county as well as the state levels.

² Prior studies show that firms located in more religious areas exhibit stronger risk aversion and more conservative investment policies (Hilary and Hui 2009), less opportunistic timing of option grants (Grullon, Kanatas, and Weston 2009), less earnings management (Grullon, Kanatas, and Weston, 2009; McGuire, Omer, and Sharp, 2012), fewer class action lawsuits (Grullon, Kanatas, and Weston 2009), fewer financial restatements and financial reporting irregularities (Dyreng, Mayew, and Williams 2012; McGuire, Omer, and Sharp, 2012).

³ Additional recent studies that demonstrate the effect of religion on individuals’ behaviour include Kumar, Page, and Spalt (2011) and Benjamin, Choi, and Fisher (2010). Kumar, Page, and Spalt (2011) find that local religion influences the gambling attitude and portfolio choices of the individuals in the local area. Benjamin, Choi, and Fisher (2010), in a laboratory setting, show that religious identity salience affects individuals’ decision-making.

⁴ We obtain our sample of acquisitions over the 1998 to 2003 period that meets the following criteria: (1) the deal is announced between 1 January 1998 and 31 December 2003. The date restriction is imposed because complete data on directors are available beginning in 1998. (2) The deal is successful and completed in less than 1000 days after the announcement date. (3) The acquirer controls less than 50% of the shares of the target at the time of the announcement and obtains 100% of the target shares. (4) The deal value is equal to or greater than $1 million. The deal value threshold corresponds to 1% of the market value of the acquirer’s assets (defined as the book value of assets minus the book value of equity plus the market value of equity). We find similar results when we use alternative thresholds of 10% or 5%. (5) The target is a US public or private firm. (6) Stock return data and other required financial data on the acquirer are available from Center for Research in Security Prices and COMPUSTAT.

⁵ Our sample period is selected based on data availability. For instance, the governance variables such as board independence and CEO duality are available through 2003.

⁶ Because our sample period straddles the internet bubble, which burst after 2000, our results may be driven by the deflation of the internet boom after 2000. To minimize this concern, we execute two robustness tests. First, we run a regression excluding states with a high concentration of internet firms. This subsample is substantially less vulnerable to the deflation in the internet industry. We obtain a similar result. Second, we run a regression including only the years before 2000. This subsample is much less contaminated because the troubles in the internet industry began largely after 2000. Again, we obtain a consistent regression result.

⁷ An alternative approach to minimize the omitted variable bias is to control for county-specific factors, such as population, education, income etc. There are a few challenges to this approach, nevertheless. First, the county data are available in such a way that is particularly challenging to collect, both in terms of time and effort. Second, even if we control for these variables, it is still not possible to rule out the omitted variable bias because it is always possible that there are certain characteristics that may not be included in the model. Practical consideration suggests that our approach is more reasonable because it is substantially less demanding in terms of data collection but offers a test that should be just as stringent.

⁸ Data on religious piety is advantageous because it goes back as far as 1890. Nevertheless, the earliest religious data with relevant variables that allow us to merge the data into our sample are from 1952. Earlier data were gathered under different settings and do not report adequate data for matching with our current sample.

⁹ Further, to ensure that our instrument is truly exogenous, we totally exclude those firms that existed in 1952. The remaining firms came into existence only after 1952. Obviously, any corporate policies of these firms could not have been correlated with religious piety or any unobservable factors in 1952 because these firms simply were not in existence. Thus, for these firms, we can be highly confident that our instrument, that is, religiously piety in 1952, is genuinely exogenous. We rerun a two-stage least squares (2SLS) analysis using only these firms and obtain consistent results.

¹⁰ We also use religious piety from other years in between and obtain consistent results.

¹¹ To enhance the generalizability of our results, we execute a robustness check using an expanded sample. In particular, we extend the sample period as far as 2010. The results remain similar, although the degree of significance is smaller (results available upon request). The results based on the expanded should be viewed with caution; however, because a smaller set of control variables are included. Furthermore, the more recent period has been affected by the financial crisis, which may skew the results. Consequently, we employ the expanded sample as a robustness check, rather than as the primary sample.

¹² As further robustness check, we run a regression where we predict current religious piety using historical religious piety from 1952. We then split the sample into quartiles based on religious piety predicted from the regression. This approach helps control for endogeneity further as we do not use current religious pity but predicted religious piety based on religious piety form the distant past. We then classify firms in the top quartile as the treatment group. Again, the treatment effect remains significant. Thus, our analysis based on propensity score matching confirms that religious piety does have a favourable effect on the stock market reactions to the acquisition announcement.