The impact of family ownership on both short-term
and long-term acquisition performance of NYSE and
NASDAQ listed companies
B.A.W. Nijland
Keywords: Acquisition Performance, Agency Theory, Event study, Family Firms and
Non-family Firms, U.S. Companies
JEL classification: G3, G32, G34
Abstract
Family firms represent a significant part of the large companies around the world and some clear characteristic differences between family firms and non-family firms are known. However, the existing literature is still inconclusive about the effect of family ownership on acquisition performance. This paper examines the relationship between family ownership and short-term as well as long-term acquisition performance of NYSE and NASDAQ listed companies. The event study methodology is applied to a sample of 449 (372) acquisitions made by family and non-family firms between 2008 and 2016 for the short-term (long-term) analysis. Univariate tests are performed and the (cumulative) abnormal returns are regressed against the family ownership variable in combination with some other determinants of acquisition performance. The results indicate that family ownership does not have a significant impact on short-term as well as long-term acquisition performance. Hence, this result suggests that the costs of family ownership are offset by the benefits.
MSc Finance Thesis June 2017
University of Groningen
Faculty Economics and Business Department Finance
1. Introduction
Family-controlled firms represent a significant part of all large listed corporations around the world. Villalonga and Amit (2006) report that 37% of Fortune 500 firms, a collection of the largest U.S. companies, are family-controlled. Approximately 29% of large listed companies in Europe’s richest countries are controlled by families, and in Canada 25% of large listed corporations are family-controlled (La Porta et al., 1999). Characteristic differences between family and non-family firms have been extensively researched. Family firms generally have large undiversified owners, family members that hold senior positions within the company, and usually maintain a long-term investment focus (Basu et al., 2009). Adhikari and Sutton (2016) state that the main characteristic difference between family and non-family firms stems from the sort of agency problems the two types of companies possess. They distinguish two types of agency problems: Within family firms an agency problem between large shareholders (families) and minority shareholders often arises since families may derive private benefits, whilst in non-family firms an agency problem more often arises between shareholders and managers due to conflicts of interest.
The effect of family ownership and control on firm performance has been ignored in the finance literature for several years. This subject has become of central concern since the early contribution of Holderness and Sheehan (1988; in Caprio et al., 2011). Because of the characteristic differences between family and non-family firms, the owners and managers of these two types of firms may have a different attitude towards and influence on their companies, which can possibly lead to dissimilarities in firm performance. This makes it an interesting subject to investigate. In this paper, the effect of family ownership on acquisition performance in large U.S. companies is examined. Mergers and acquisitions provide a suitable framework for investigating major investment decisions made by companies’ managers and/or owners. In this way, the effect of the agency problems and other characteristic differences between family and non-family firms mentioned above on acquisition decisions within large U.S. companies is implicitly tested. In other words, it is tested whether the costs of family ownership arising from those differences outweigh the benefits. Costs and benefits of family ownership are described in section 2.2.
What is the effect of family ownership on short-term as well as long-term acquisition performance in NYSE and NASDAQ listed companies?
Both a univariate analysis and a multivariate analysis are performed. From the results of earlier empirical work it can be inferred that NYSE and NASDAQ listed family firms outperform NYSE and NASDAQ listed non-family firms in both the short-run and long-run when acquiring (i.a., Ben-Amar and André, 2006; Basu et al., 2009; Bouzgarrou and Navatte, 2013).
Several studies in the finance literature investigate the differences between family firms and non-family firms regarding M&A performance, though the results are inconclusive. This paper differs from these earlier studies in a variety of ways. Firstly, the majority of the papers study acquisitions in periods up to 2008. This paper investigates the period between 2008 and 2016, which is a period of economic recovery after the global financial crisis. Hence, this research leads to new insights. Secondly, a major part of empirical results has been obtained from studies examining non-U.S. companies (Ben-Amar and André, 2006; Feito-Ruiz and Menédez-Requejo, 2008; Caprio et al., 2011; Shim and Okamuro, 2011; Bouzgarrou and Navatte, 2013). Thirdly, while this paper closely relates to Caprio et al. (2011) and Shim and Okamuro (2011), their main focus is on the likelihood of family firms merging while the focus of this research is on the performance resulting from M&As. Lastly, the majority of existing literature investigates the effect of family ownership and control on either long-term M&A performance (Shim and Okamuro, 2011; Adhikari and Sutton, 2016) or on short-term M&A performance (Ben-Amar and André, 2006; Bauguess and Stegemoller, 2008; Feito-Ruiz and Menédez-Requejo, 2008; Basu et al., 2009; Caprio et al., 2011). In this research Bouzgarrou and Navatte (2013) are followed in studying the effect of family ownership on both long-term performance and short-term performance.
The rest of this paper is structured as follows. Section 2 provides an overview of the existing literature on the topic. The methodology used in this paper is described in Section 3. Section 4 outlines the data and presents summary statistics. The results are presented in Section 5. Section 6 provides the conclusions, limitations and suggestions for future research.
2. Literature review
To provide a comprehensible overview of the literature concerning the effect of family ownership on acquisition performance, this chapter is subdivided into five sections. First, the definition of family firms is provided. Second, theories in the existing literature regarding family ownership and control are elaborated. Third, the effects of M&A announcements and M&A completions on firm performance are examined. Fourth, based on existing empirical evidence, the relationship between family ownership and control and M&A performance is investigated and hypotheses are formulated. At last, determinants of acquisition performance, which are used as control variables in this study, are described.
2.1. Definition of family firms
Existing literature shows that a considerable share of companies worldwide is categorized as family-controlled. However, these categorizations are based on a variety of definitions of family firms.
Even though differences between family and non-family firms have been extensively researched, this has never resulted in a universally accepted definition for family firms (Lau, 2010). Lau (2010) aims to resolve the diversity of the definitions in use by developing an operational definition for listed family firms that is consistent with agency theory. He notes that an ownership and control threshold and the presence of family members in management positions (so that there is control in the decision-making process) are often used to identify family firms, and puts forth that these criteria can serve as an operational definition.
must exceed a 10% threshold (Caprio et al., 2011). When this criterion is met, and the largest ultimate shareholder is an individual or family, the company is characterized as a family firm. Ben-Amar and André (2006) use the same definition in their research, although without any threshold.
To establish a definition for family firms, this paper combines the classifications from earlier studies mentioned above and the recommendations from Lau (2010). As a consequence, a family firm is defined as a firm with an individual or family as (largest) ultimate owner at a 50% threshold. A threshold of 50% is used since in this way the individual or family is the major shareholder, the individual or family is automatically the largest ultimate owner, and the individual or family has control over investment decisions. It must be acknowledged that, as no universally accepted definition for family firms exists, empirical results of studies into family ownership and control are often difficult to compare.
2.2. Costs and benefits of family ownership
The characteristic differences between family and non-family firms may lead to differences in acquisition performance. Some authors argue that the main characteristic difference between family and non-family firms stems from the sort of agency problems the two types of companies possess (i.a., Adhikari and Sutton, 2016), whereas others also emphasize other characteristic differences between family firms and non-family firms (i.a., Basu et al., 2009). In this paper, these theories are combined since some characteristic differences mentioned by one may lead to one of the agency problems stated by others. This section describes the costs and benefits of family ownership arising from those characteristic differences.
2.2.1 Costs of family ownership
Furthermore, family firms usually have large undiversified owners who tend to pursue risk reduction strategies (Andres, 2008). This is a good thing when family firms forgo value destroying acquisitions proposed by its managers, but this risk averse character of family firms due to low diversification may also lead to sub-optimal investment decisions (Caprio et al., 2011).
Lastly, it is common that family members hold senior positions within family firms, while non-family firms hire external seniors. Appointing a family member for a management position can have a negative impact if the individual does not have enough talent, knowledge, or expertise to fulfill this role (Ben-Amar and André, 2006).
2.2.2. Benefits of family ownership
The classic agency problem between shareholder and managers arises due to conflict of interest and will be generally less present in family firms for two reasons. Firstly, due to the fact that families often are large undiversified owners, family firms tend to pursue risk reduction strategies (Andres, 2008). Risk averse owners have the power and incentives to monitor managers, and thus decrease agency costs (Bouzgarrou and Navatte, 2013) and pass up value destroying acquisitions proposed by managers who want to maximize their personal utility and/or are overconfident (Caprio et al., 2011). Secondly, since it is common that family members hold senior positions within family firms, fewer conflicts of interest will arise between shareholders and the management.
7
2.3. The effect of M&A announcements on firm performance1
Before discussing the relationship between family ownership and M&A performance, the more general relationship between M&A announcements and firm performance is examined. Jensen and Ruback (1983) perform a literature review into the effect of M&A announcements on post-deal performance. They find that the announcement returns for acquiring firms are positive (see, e.g., Dodd, 1980; Eckbo, 1983). Vazirani (2012) performs a similar investigation and also finds that M&A announcements generally have a positive impact on stock price performance on the days around the announcement. Consequently, in this study a positive acquisition announcement effect is expected in the short-run for both family firms and non-family firms. Since this paper assumes that the stock prices fully incorporate and reflect all relevant information, the performance around the acquisition announcement is an estimate for long-term acquisition performance (Moeller et al., 2004).2 Hence, it is expected that family firms and non-family firms also gain from an acquisition in the long-term.
It must be acknowledged that there are also other findings in literature as regards the relationship between M&A completion and long-term firm performance. Doeswijk and Hemmes (2001) find that acquiring firms show positive abnormal returns in the period before and on the day of the announcement but that significant negative abnormal returns follow immediately thereafter. Moreover, Vazirani (2012) argues that it can be concluded from existing literature that the intended benefits of acquisitions are often not realized in the long-term.
2.4. Relationship between family ownership and control and M&A performance
This section presents an overview of the existing empirical work regarding the effect of family ownership and control on M&A performance. When empirical evidence shows that either family firms or non-family firms are performing better, it can be implicitly stated that the
1 As mentioned in section 3.1, the effect of family ownership on short-term acquisition performance is measured over some days around the acquisition announcement, whereas the effect of family ownership on long-term acquisition performance is measured over some months following acquisition completion. This paper assumes that the stock prices fully incorporate and reflect all relevant information, so that performance around the acquisition announcement is an estimate for term acquisition performance. Other factors that have an impact on long-term acquisition performance, such as systematic risk, are ignored in formulating expectations. Hence, this study’s expectations for long-term acquisition performance are based on empirical findings about the effect of M&A announcements on short-term performance.
benefits of family ownership outweigh the costs or vice versa. For the sake of clarity, Appendix 1 provides an overview of relevant papers examining the effect of family ownership and control on M&A performance, which are mentioned below.
Several studies in finance literature investigate M&A performance of U.S. family firms versus U.S. non-family firms, although the results are inconclusive. Bauguess and Stegemoller (2008) and Basu et al. (2009) examine the effect of family ownership and control on short-term acquisition performance. Bauguess and Stegemoller (2008) analyze 1,411 acquisitions of S&P 500 firms between 1994 and 2005. Their results show that U.S. family firms destroy value when acquiring.Basu et al. (2009) study short-term acquisition performance of 103 U.S. newly public acquiring firms between 1993 and 2004. They show that acquirers with low levels of family ownership have lower abnormal returns than acquirers with high levels of family ownership.
Adhikari and Sutton (2016) instead focus on the long-term acquisition performance of U.S. family and non-family firms. They expand the sample of Anderson and Reeb (2003) and Liu (2011), who examine the relationship between founding family ownership and firm performance and founding family ownership and firm cash holdings, respectively. This results in an investigation of acquisitions by 213 S&P 500 firms between 1993 and 2006. In their research they perform a cross sectional analysis of buy-and-hold abnormal returns over a one-year and three-one-year period, and in addition run a calendar time four factor and five factor model regression, also over a one-year and three-year period. Their sample consists of 223 acquisitions. They find that the long-term abnormal return in the post-acquisition period (one year after the acquisition) is approximately 17% higher for family firms than for non-family firms. Also the calendar time factor model regressions show that U.S. family firms perform better than U.S. non-family firms after acquisition completion.
Hence, findings of studies regarding M&A performance of U.S. firms are divergent. Results of Bauguess and Stegemoller (2008) suggest that the costs of family ownership outweigh the benefits in U.S. firms, whereas the results of Basu et al. (2009) and Adhikari and Sutton (2016) suggest that the benefits outweigh the costs.
Feito-Ruiz and Menédez-Requejo (2008) analyze family versus non-family firm returns under different legal environments when a merger or acquisition is announced. They investigate 143 M&A announcements of European firms in the period 2002-2004. Their results show that the cumulative average abnormal returns, in [-2; 2], are 2.82% and 0.92% for family firms and non-family firms, respectively. However, the difference between family firms and non-family firms is not significant. Similarly, Caprio et al. (2011) investigate 777 large Continental European companies in the period 1998-2008 and also find no evidence that family-controlled firms perform differently in the short-run when they acquire other companies.
Shim and Okamuro (2011) investigate 509 family firms and 693 non-family firms, all listed on Japanese stock markets. They study the long-term post-merger performance of these companies in the 1955-1973 period. In their research they conduct a difference-in-differences analysis to analyze relative merger performance over a three-year period. Their results show that non-family firms perform better when merging than family firms.
Lastly, Bouzgarrou and Navatte (2013) investigate the impact of family control on French acquirers’ performance. They study 239 French acquisitions between 1997 and 2006, and claim to be the first researchers who investigate both long-term and short-term acquisition performance. Their paper concludes that family firms outperform non-family firms around the announcement date. Univariate tests show that family firms realize cumulative abnormal returns of 2.81% and non-family firms realize cumulative abnormal returns of 0.08% over three days around the announcement date. This difference is statistically significant. Furthermore, they argue that family firms are more efficient over a three-year period, though the test results regarding this aspect are insignificant.
Thus, as for studies on large U.S. companies, results from researches on the relationship between family ownership and control and M&A performance for non-U.S. firms are inconclusive. Studies mentioned above showing that the benefits of family ownership outweigh the costs are those from Ben-Amar and André (2006), Feito-Ruiz and Menédez-Requejo (2008), and Bouzgarrou and Navatte (2013). In contrast, the research from Shim and Okamuro (2011) suggests that the costs predominate the benefits.
predominant than the costs. Consequently, for this study it can be expected that, in the short-run, NYSE and NASDAQ listed family firms outperform NYSE and NASDAQ listed non-family firms when acquiring. This can be expressed by the following hypothesis:
H1: Compared to NYSE and NASDAQ listed non-family firms, NYSE and NASDAQ listed family firms experience higher (cumulative) abnormal returns in the short-run when acquiring.
This paper assumes that the stock prices fully incorporate and reflect all relevant information, so that stock performance around the acquisition announcement is an estimate for long-term acquisition performance (Moeller et al., 2004).3 Consequently, it can be expected that, in the long-run, NYSE and NASDAQ listed family firms outperform NYSE and NASDAQ listed non-family firms when acquiring. This can be expressed by the following hypothesis:
H2: Compared to NYSE and NASDAQ listed non-family firms, NYSE and NASDAQ listed family firms experience higher (cumulative) abnormal returns in the long-run when acquiring.
2.5. Determinants of acquisition performance
Literature has shown that there are some other explanatory variables for acquisition
performance. This study uses these variables as control variables. These control variables can be grouped into two categories: acquirers’ characteristics and deal characteristics. Control variables in the category acquirers’ characteristics are firm age, firm size, leverage, relative deal size, return on assets, and Tobin’s Q. Control variables in the category deal characteristics are method of payment and relatedness of activities. The rest of this section describes these control variables and their expected influence on acquisition performance.
Firm age
It is expected that the market responds more positively to acquisition announcements made by young corporations than by old corporations, as the older companies are more likely to have
exhausted their growth opportunities (Moeller et al., 2004). Moreover, Anderson and Reeb (2003) use this firm age variable to explain firm performance, and they find that younger companies perform better than older companies. Hence, a negative sign is expected for the firm age variable.
Firm size
The same reasoning as for the firm age variable applies to the firm size variable in the sense that it is expected that the market responds more positively to acquisition announcements made by small companies than by large companies, as the larger companies generally have less potential to grow. Besides, Himmelberg et al. (1999) find that the managerial ownership stake is higher for small firms than for large firms, which suggests that agency problems between managers and shareholders is more pronounced in large companies. In line with these suggestions mentioned above, Moeller et al. (2004) find that the acquisition announcement return is higher for small acquirers than for large acquirers. So, in this study it is expected that the variable firm size has a negative impact on acquisition performance.
Leverage
Harrison et al. (2013), pp 571 argue that: ‘‘Leverage may have a positive effect on firm performance by limiting managers’ ability to allocate resources to unproductive uses, as well as increasing pressure on them to perform well’’. Moreover, Maloney et al. (2003) find that acquisition announcement returns are greater when the leverage of the acquirer is higher. Consequently, a positive sign is expected for this leverage variable.
Relative deal size
Various studies examine the impact of the relative size, the size difference between the acquirer and the target, on acquisition performance. Asquith et al. (1983) find that acquirers’ abnormal returns are positively related to the relative size of the deal. In other words, acquirer gains after an acquisition are larger when targets are larger. Furthermore, Kitching (1967) concludes that when the target’s sales volume is less than 2% of the acquirer’s sales volume, the number of acquisition failures equals 84%. He argues that a mismatch of size can be expected when the relative deal size is high. All in all, in this study a positive sign is expected for the relative deal size variable.
Strong operational performance leads to encouraging the establishment of an aggressive acquisition strategy (Halebalian and Kim, 2006) and to overconfidence of firm’s decision-makers. Moreover, Caprio et al. (2011) and Adhikari and Sutton (2016) also use the variable return on assets to explain M&A performance, and they find a negative, but insignificant, relationship between return on assets and M&A performance. Hence a negative sign is expected for the return on assets variable.
Tobin’s Q
Tobin’s Q, measured as the ratio of the sum of the market value of equity and the book value of total liabilities to the book value of total assets, is a commonly used measure for firms’ growth opportunities. It can be expected that the market responds more positively to acquisition announcements made by companies with high growth potential than by companies with low growth potential. Besides, Lang et al. (1989) find that acquirers with a high Tobin’s Q gain more than acquirers with a low Tobin’s Q. So, this study expects a positive relation between acquirers’ Tobin’s Q and acquisition performance.
Method of payment
Shleifer and Vishny (2003) argue that it is likely that if the acquirer is overvalued (undervalued), the acquirer prefers to pay with shares (cash). So, it is expected that when shares are used as the medium of payment that the abnormal returns around the acquisition announcement are negative, since the acquirer’s share price will decrease to its fair value. Furthermore, Huang and Walking (1987) and Myers and Majluf (1984) find a negative relation between the payment in shares and acquisition performance. As a result,a negative relation between the method of payment variable, which is a dummy variable equal to 1 if only shares are used for payment and zero otherwise, and acquisition performance is expected.
Relatedness of activities
activities and acquisition performance. As a result, this study expects a positive sign for the relatedness of activities variable.
The expectations for the signs of the control variables mentioned above apply for both short-term acquisition performance and long-term acquisition performance since performance around the acquisition announcement is an estimate for long-term acquisition performance. The definitions of all variables, their units of measurement, and expected signs are presented in Appendix 2.
3. Methodology
To test the formulated hypotheses the event study methodology suggested by MacKinlay (1997) is adopted. This event study methodology is further elaborated in the next section. Subsequently, a description of the univariate analysis and the multivariate analysis is provided followed by a discussion of the diagnostics tests that are performed.
3.1. Event study methodology
A key question in conducting event studies to measure long-term effects is whether to use CARs or buy-and-hold abnormal returns (BHARs) (Brooks, 2014). Fama (1998), pp 294-295 argues that: ‘‘BHARs can give false impressions of the speed of price adjustment to an event since BHARs can grow with the return horizon even when there is no abnormal return after the first period, and moreover CARs pose fewer statistical problems than long-term BHARs’’. He strongly suggests using CARs rather than BHARs. Furthermore, Bouzgarrou and Navatte (2013), whose study is similar to this paper, utilize CARs to determine long-term performance. For these reasons, this research uses CARs to measure long-term acquisition performance.
After defining the event and the measures for acquisition performance, the next step is to define the estimation window in which the expected returns are estimated, and the event window in which the abnormal returns are calculated. Typical lengths of the estimation period range from 100 to 300 days for daily studies and from 24 to 60 months for monthly studies (Peterson, 1989). Taking this into account, this study uses an estimation window between -261 and 11 days before the announcement date for the shortterm analysis, and between 26 and -3 months before the completion date for the long-term analysis. To be completely sure that anticipation (i.e. ‘leakage’) of the event does not affect estimation of the expected returns, often a gap is left between the estimation window and the event window (Brooks, 2014). As a result, the event window for short-term analysis is [-5; 5], and hence consists of 11 trading days. The event window for long-term analysis is [0; 11], and hence consists of 12 months. In this Adhikari and Sutton (2016) are followed. The returns in these windows are indexed in event time using t, denoting t = 0 as the event date (MacKinlay, 1997). For the short-term analysis the announcement date is considered to be the event date and for the long-term analysis the last trading day in the month following acquisition completion is considered to be the event date. The timeline for the short-term analysis and long-term analysis is illustrated in figure 1.
(Estimation window] (Event window]
Figure 1. Timeline for the event study. The estimation window is from t = T0 to t = T1 and is [-261; -11] for the short-term
analysis and [-26; -3] for the long-term analysis. The event window is from t = T1 to t = T2 and is [-5; 5] for the short-term
analysis and [0; 11] for the long-term analysis.
As mentioned above, abnormal returns are actual returns minus expected returns. The actual return for firm i on day (in month) t, Rit, is calculated as
Rit = LN (
𝑅𝐼𝑖𝑡+1
𝑅𝐼𝑖𝑡 ) ,
(1)
where 𝑅𝐼𝑖𝑡+1 is the return index of firm i on day (in month) t+1 and 𝑅𝐼𝑖𝑡 is the return index of firm i on day (in month) t in the case of the short-term (long-term) analysis. To estimate the expected returns, MacKinlay (1997) appoints two models that are of common use in event studies; the constant mean return model and the market model. Brown and Warner (1980) find that the constant mean return model, which does not explicitly adjust for market wide factors or for risk, generates results comparable to the market model. However, MacKinlay (1997) states that the market model reduces the variance of abnormal returns since this model removes the part of the return that is related to variation in the market’s return. Moreover, the market model is the most common approach in generating expected returns (Armitage, 1995). Consequently, in this study the market model is used to estimate the expected returns. The market model to compute the expected returns is
Rit = αi + βiRmt + εit , (2)
where Rit is the return of firm i on day (in month) t, αi is the intercept, βi is the slope, Rmt is the
return of the market index on day (in month) t and εit is the zero mean disturbance term in the
case of the short-term (long-term) analysis. Ordinary least squares (OLS) regression is used to estimate these parameters, and is under general conditions a consistent procedure (MacKinlay, 1997). To compute the market returns, the S&P 500 index is used as this index consists of NYSE and NASDAQ listed companies. Abnormal returns can then be calculated as
ARit = Rit - 𝛼̂i - 𝛽̂iRmt , (3)
where ARit is the abnormal return of security i on day (in month) t, Rit is the return of firm i on
the market index on day (in month) t in the case of the short-term (long-term) analysis. Brooks (2014) argues that it is likely that there will be some variation of the returns within the event window, with positive returns on some days (or in some months) and negative in others. This makes it difficult to draw overall inferences. Therefore, the cumulative abnormal returns are also calculated in this study. For long-term analysis only the CARs are discussed, since ARs in a particular month say nothing about long-term performance. The cumulative abnormal return from 𝑡1 to 𝑡2 is the sum of all included abnormal returns and is calculated as
𝐶𝐴𝑅𝑖(𝑡1,𝑡2)= ∑𝑡𝑡= 𝑡2 1𝐴𝑅𝑖𝑡 , (4)
where 𝐴𝑅𝑖𝑡 is the abnormal return of security i on day (in month) t in the case of the short-term (long-term) analysis.
One of the objectives of this event study is to test whether the companies across the samples, on average, generate abnormal returns around an acquisition announcement or following an acquisition completion, and whether these abnormal returns differ between family firms and non-family firms (see section 3.2 for this univariate analysis). To test this, the average abnormal returns (AARs) and cumulative average abnormal returns (CAARs) are calculated as
𝐴𝐴𝑅𝑡 = 1 𝑁 ∑ 𝐴𝑅𝑖𝑡 𝑁 𝑖=1 , (5) and 𝐶𝐴𝐴𝑅(𝑡1,𝑡2) = ∑𝑡𝑡= 𝑡2 1𝐴𝐴𝑅𝑡 , (6)
respectively, where 𝑁 is the number of events. The following section presents the test statistics used to test the significance of the AARs and CAARs.
3.2. Univariate analysis
event window. In literature, a wide range of different tests to assess this are mentioned. The two subsections below describe the tests which are used in this study.
3.2.1. Parametric tests
The first test which is used in this research is a parametric test suggested by MacKinlay (1997). The test statistic to test whether the AARs are significantly different from zero on a particular day or in a particular month is calculated as
t1 =
𝐴𝐴𝑅𝑖𝑡
√𝑉𝐴𝑅(𝐴𝐴𝑅𝑡)
~ N(1,0). (7)
To test whether the CAARs are significantly different from zero over a particular period, the test statistic is calculated as
t2 =
𝐶𝐴𝐴𝑅(𝑡1,𝑡2) √𝑉𝐴𝑅[𝐶𝐴𝐴𝑅(𝑡1,𝑡2)]
~ N(1,0). (8)
To test whether the AARs and CAARs are significantly different between family firms and non-family firms on a particular day, in a particular month or over a particular period, a two-sample t-test is performed. To ensure that the correct test statistic is used, it is first tested whether it can be assumed that the two distributions have equal variances. The most suitable test to assess this is the F-test. The F-statistic can be calculated as
t3 =
𝑠12
𝑠22
~ F (𝑛1− 1, 𝑛2− 1), (9)
where 𝑠12 and 𝑠
22 are the variances of the family firm and non-family firm sample, respectively and are calculated as
𝑠12 = 1
𝑛1− 1 ∑ (𝑥𝑖− 𝑥̅1)
𝑁
𝑖=1 2 , (10)
𝑠22 = 1
𝑛2− 1 ∑ (𝑥𝑖− 𝑥̅2)
𝑁
𝑖=1 2 , (11)
where 𝑛1 and 𝑛2 are the sample sizes of the family firm and non-family firm sample, respectively, and 𝑥̅1 and 𝑥̅2 are the (C)AARs of the family firm and non-family firm sample on a particular day, in a particular month or over a particular period within the event window, respectively. When the F-test shows that the two distributions have the same variance, the test statistic of the two-sample t-test can be calculated as
t4 = 𝑥̅1− 𝑥̅2 √𝑠𝑝2(1 𝑛1+ 1 𝑛2)
~ N(1,0), (12) where, 𝑠𝑝2 = (𝑛1− 1)𝑠12+ (𝑛2− 1)𝑠22 𝑛1+ 𝑛2− 2
.
(13)When it can be assumed that the two distributions do not have equal variances, the test statistic of the two-sample t-test can be calculated as
t5 = 𝑥̅1− 𝑥̅2 √𝑠12 𝑛1+ 𝑠22 𝑛2
~ N(1,0). (14) 3.2.2. Non-parametric tests
tests are preferred, since evidence shows that they perform better than parametric tests in case of non-normally distributed security returns (Armitage, 1995).
In this research the nonparametric rank test introduced by Corrado (1989) is performed. This is one of the most successful nonparametric tests in literature (Corrado, 2011). The rank test transforms each security’s time series of abnormal returns into its respective ranks. The test statistic can be calculated as
t6 = 1 𝑁 ∑ (𝑘𝑖𝑡− 𝑘) 𝑁 𝑖=1 𝑠𝑘
,
(15)where 𝑁 is the number of events, 𝑘𝑖𝑡 is the rank of security i’s abnormal return on day (in month) t in the case of the short-term (long-term) analysis, 𝑘 is the expected rank, and 𝑠𝑘 is the
time-series standard deviation of the sample mean abnormal return ranks (Campbell et al., 2010). The expected rank, 𝑘, can be calculated as
𝑘 = 𝑀
2
+
0.5, (16)where 𝑀 is the number of observed returns within the estimation plus event period. For the short-term analysis 𝑠𝑘 can be calculated as
𝑠𝑘 = √ 1 272 ∑ ( 1 𝑁 ∑ (𝑘𝑖𝑡− 𝑘) 𝑁 𝑖=1 ) +10 𝑡= −261 2 , (17)
and for the long-term analysis 𝑠𝑘 can be calculated as
𝑠𝑘 = √ 1 38 ∑ ( 1 𝑁 ∑ (𝑘𝑖𝑡− 𝑘) 𝑁 𝑖=1 ) +11 𝑡= −26 2 . (18)
t7 = d1/2 𝑘̅𝐷− 𝑘 [∑𝑀𝑡=1 (𝑘̅𝑡− 𝑘)2 / 𝑀] 1 2
,
(19)
where 𝑘̅𝐷 is the average rank across the number of securities and d days of the CAR-period, and 𝑘̅𝑡 is the average rank across number of securities on day t of the 𝑀 day combined estimation and event period (Cowan, 1992).
The nonparametric tests described above are to test whether the AARs and CAARs across the samples are significantly different from zero within the event window. To test whether the AARs and CAARs across the samples differ between family firms and non-family firms within the event window, the nonparametric Mann-Whitney test is performed. In this Ben-Amar and André (2006), Feito-Ruiz and Menédez-Requejo (2008), and Bouzgarrou and Navatte (2013) are followed. The Mann-Whitney test transforms each security’s abnormal return at a particular day, in a particular month or over a particular period within the event window into its respective ranks. The test statistic can be calculated as
t8 = 𝑈− 𝑛1𝑛2 2 𝑠𝑅 , (20) where 𝑈 is calculated as 𝑈 = 𝑀𝑖𝑛 [𝑛1𝑛2+ 𝑛1(𝑛1+ 1) 2 − 𝑅1 ; 𝑛1𝑛2+ 𝑛2(𝑛2+ 1) 2 − 𝑅2] , (21)
where 𝑅1 and 𝑅2 are the sum of the ranks from the family firm sample and non-family firm sample at a particular day or in a particular month within the event window, respectively. The standard deviation of the ranks, 𝑠𝑅, can be calculated as
𝑠𝑅 =
𝑛1𝑛2(𝑛1+ 𝑛2+ 1)
12
.
(22)To examine whether family ownership, in combination with the control variables, has a significant impact on acquisition performance, this research performs a multivariate regression. The control variables used in this multivariate analysis are the method of payment (ALLSHARES), firm age (FAGE), firm size (FSIZE), leverage (LEVERAGE), relatedness of activities (RELATEDN), relative firm size (RELSIZE), return on assets (ROA), and Tobin’s Q (TOBINQ), which are presented and motivated in section 2.5. The model is estimated as
(C)ARi = α0+β1FAMFIRMi+β2ALLSHARESi+β3FAGEi+β4FSIZE+
β5LEVERAGEi +β6RELATEDNi+β7RELSIZEi+Β8ROAi+
β9TOBINQi+εi , (23)
where FAMFIRMi is a dummy variable equal to 1 if firm i is a family firm and equal to 0 if firm
i is a non-family firm, ALLSHARESi is a dummy variable equal to 1 if only shares are used for
payment and 0 otherwise, and RELATEDNi is a dummy variable equal to 1 if the acquirer and
the target have the same US SIC codes and zero otherwise. Ordinary least squares (OLS) regression is used to estimate the model. The definition of all variables used in this estimation, their units of measurement, and expected signs are presented in Appendix 2.
3.3.1. Diagnostic tests
The use of OLS regression involves some assumptions (e.g., Brooks, 2014). In this study, diagnostic tests are performed to detect potential violations of these assumptions. When violations are detected, the model and/or the data are modified so all assumptions underlying an OLS regression are met. To avoid skewed or misleading results, also the presence of multicollinearity is tested. In other words it is tested whether the regression model contains explanatory variables that are highly correlated to one another.
In this study, a White’s (1980) test is performed to test for heteroscedasticity. If the variance of the errors is not constant, the errors are said to be heteroscedastic. When it appears that the errors are heteroscedastic, test statistics are derived using heteroscedasticity-consistent standard errors using the approach of White (1980; in MacKinlay, 1997).
regression method instead of switching to another estimation method as the performance of an OLS regression within different contexts has been extensively researched. Moreover, when removing outliers, valuable pieces of information are lost. In addition to that, non-normally distributed error terms are nearly inconsequential when the sample size is large enough (N>~300) (Brooks, 2014). Consequently, this study employs OLS regression and only removes incidental outliers regardless of whether the error terms are normally distributed or not.
The assumption that the average value of the errors is zero is not violated since α0 is
included in the models. Furthermore, the explanatory variables in this study are non-stochastic, so this assumption is also met.
To detect multicollinearity, the correlation coefficients and the variance inflation factors (VIFs) are analyzed. When it appears that the model contains explanatory variables that are highly correlated to one another, one of these high-correlating variables is eliminated from the model.
4. Data
4.1. Sample selection
the acquisition is made by a family firm or not is measured on the moment the acquisition is registered in the Zephyr database.
To assess acquisition performance, stock data is obtained from Thomson Reuters Datastream. No stock data is available for 9 family firm acquisitions in the case of the short-term analysis. Thus, for the short-short-term analysis the final sample of family firm acquisitions consists of 103 events. In the case of the long-term analysis, no stock data is available for 11 family firm acquisitions. Moreover, to avoid overlapping results, the sample solely includes the first acquisition from a company within a two-year period in assessing long-term acquisition performance (Adhikari and Sutton, 2016). This screening results in a final sample of family firm acquisitions of 65 events for the long-term analysis.
As the initial sample of non-family firm acquisitions is quite large, it is desirable to take a random sample of these acquisitions as a control group. To ensure that there are no ‘family firms’ in the non-family firm acquisitions sample, first a subsample of widely held non-family firms is taken. A firm is categorized as a widely held firm if no shareholder owns more than 20% (Isakov and Weisskopf, 2014). After eliminating the acquisitions made by firms that are not widely held, 1,504 non-family firm acquisitions remain. To determine the sample size of a representative random sample of non-family firm acquisitions Moore and McCabe (2005) are followed. The required sample size for a small and finite population, 𝑛𝑆, can be calculated as
𝑛𝑆 =
𝑛𝐿
1+𝑛𝐿
𝑁
, (24)
where 𝑁 is the population size (1,504) and 𝑛𝐿 is the required sample size for a large and infinite population and can be calculated as
𝑛𝐿 =
(
𝑧∗ 𝜎
𝑚
)
2
, (25)
acquisitions consists of 346 events. In the case of the long-term analysis, no stock data and no completion data are available for 19 and 20 non-family firm acquisitions, respectively. In addition, as mentioned above, overlapping acquisitions are excluded from the sample in assessing long-term acquisition performance. This results in a final sample of 307 non-family firm acquisitions for the long-term analysis.
The final sample for the short-term analysis is described in Appendix 3. Appendix 3 provides the acquisition distribution of U.S. family and non-family firms by year and by industry. The table shows that the percentage difference between the number of acquisitions made by family firms and non-family firms are the largest in the years 2013 and 2014, and in the industries ‘‘Communications’’, ‘‘Computer, IT & Internet Services’’, ‘‘Industrial, Electric & Electronic Machinery’’, and ‘‘Personal Leisure & Business Services’’.
The data on the control variables, which are used in the multivariate analysis to test the impact of these variables in combination with family ownership on acquisition performance, is obtained from the Zephyr database. In the category acquirers’ characteristics, the variables firm size, leverage, relative deal size, return on assets, and Tobin’s Q are measured over the last available year before acquisition completion. The variable acquirers’ firm age is measured as the period between firm incorporation and the acquisition announcement. The variables in the category deal characteristics are measured at the time of the acquisition. The definitions of all control variables and their units of measurement are described in Appendix 2.
4.2. Summary statistics
in addition to the parametric t-tests since an assumption of normally distributed security returns underlies the use of t-tests.
Panel B shows that the mean long-term cumulative abnormal returns are negative for both family firms and non-family firms. When looking at one-year abnormal performance,
CAR [0; 11], the cumulative abnormal returns are -13.29% and -4.92% for family firms and non-family firms, respectively. Furthermore, the CARs are skewed to the left for the majority of the reported CAR-periods for the long-term analysis. The Jarque-Bera statistic in column 10 indicates that there is not enough evidence to conclude that the family firms’ cumulative abnormal returns are non-normally distributed, whilst for non-family firms’ cumulative abnormal returns there is.
Table 1
Summary statistics of (cumulative) abnormal performance of family firms and non-family firms within the event window
Variables Group N Mean (%) Max (%) Min (%) SD Skewness Kurtosis JB Panel A: Short-term abnormal performance
AR (-5) F 103 0.19 9.01 -6.09 0.021 0.596 7.004 74.891 NF 346 103 0.06 18.45 -19.84 0.028 0.316 19.615 3985.539 AR (-4) F 103 -0.35 4.86 -9.04 0.020 -0.816 6.254 56.870 NF 346 0.02 9.66 -9.51 0.020 0.265 7.399 283.058 AR (-3) F 103 -0.03 14.02 -8.30 0.023 2.071 15.829 779.973 NF 346 103 -0.03 11.36 -11.46 0.021 0.499 9.967 714.079 AR (-2) F 103 -0.04 9.76 -7.28 0.025 0.385 6.257 48.084 NF 346 -0.12 12.38 -12.97 0.025 -0.243 9.615 634.225 AR (-1) F 103 1.23 28.85 -11.31 0.046 2.430 15.259 746.315 NF 346 103 0.61 33.09 -13.58 0.036 2.573 25.370 7595.823 AR (0) F 103 0.55 43.30 -13.20 0.052 5.382 47.058 8827.990 NF 346 0.52 23.98 -30.19 0.041 0.422 18.600 3518.893 AR (1) F 103 0.15 18.52 -14.71 0.036 0.771 12.046 361.356 NF 346 103 -0.19 20.29 -12.44 0.026 1.264 15.958 2512.597 AR (2) F 103 0.07 10.14 -6.13 0.023 0.640 6.582 62.106 NF 346 0.19 12.18 -12.78 0.024 0.860 10.782 915.788 AR (3) F 103 -0.07 7.57 -7.37 0.021 0.598 6.463 57.598 NF 346 103 -0.09 8.84 -9.00 0.021 -0.146 6.914 222.089 AR (4) F 103 0.09 12.23 -8.41 0.025 1.231 9.600 212.999 NF 346 -0.21 17.04 -11.60 0.022 0.856 17.204 2950.665 AR (5) F 103 -0.46 3.58 -11.84 0.022 -2.364 11.887 434.842 NF 346 103 0.16 11.60 -6.50 0.019 0.912 8.526 488.127 CAR [-5; 5] F 103 1.32 46.14 -19.54 0.081 1.720 11.317 347.620 NF 346 0.91 55.40 -47.76 0.092 0.063 11.259 983.655 CAR [-3; 3] F 103 1.85 41.36 -14.57 0.078 1.520 8.576 173.095 NF 346 103 0.88 39.62 -37.29 0.073 -0.339 9.237 567.375 CAR [-2; 2] F 103 1.96 39.46 -13.84 0.073 1.375 8.549 164.603 NF 346 1.01 30.50 -30.92 0.064 -0.120 8.093 374.763 CAR [-1; 1] F 103 1.93 46.33 -12.25 0.073 2.725 16.216 877.055 NF 346 103 0.93 24.20 -21.48 0.057 0.178 7.442 286.333 CAR [-1; 2] F 103 2.00 41.28 -14.44 0.071 1.811 11.405 359.487 NF 346 1.13 29.81 -21.61 0.058 0.423 6.987 239.539 CAR [-1; 3] F 103 1.92 41.70 -14.92 0.076 1.816 10.692 310.534 NF 346 103 1.04 26.96 -28.66 0.063 0.078 6.843 213.309 CAR [-1; 5] F 103 1.55 45.70 -15.81 0.081 1.825 10.929 326.958 NF 346 0.99 42.88 -28.08 0.069 0.493 8.691 480.849
Panel B: Long-term abnormal performance
CAR [0; 11] F 65 -13.29 155.26 -275.22 0.791 0.813 2.684 0.572 NF 307 -4.92 265.22 -360.97 0.612 -0.326 9.628 567.426 CAR [0; 8] F 65 -13.76 134.92 -170.81 0.653 -0.914 2.737 0.711 NF 307 -6.18 261.11 -329.97 0.533 -0.510 11.690 979.240 CAR [0; 5] F 65 -5.00 146.89 -204.58 0.501 -0.011 1.991 0.212 NF 307 -5.15 221.53 -323.46 0.447 -0.680 14.705 1776.177
Table 2
Summary statistics of independent (control) variables used in the multivariate analysis of the effect of family ownership on short- and long-term acquisition performance
Variables Group N Mean Max Min SD
Panel A: Short-term analysis
FSIZE F 103 14.008 17.638 9.881 1.555 NF 346 14.455 19.277 9.287 1.915 FAGE F 103 30.330 113.000 3.000 23.820 NF 346 103 33.087 127.750 3.000 28.435 LEVERAGE F 103 0.474 1.298 0.080 0.227 NF 346 0.512 1.261 0.088 0.228 RELSIZE F 103 0.181 0.963 0.000 0.184 NF 346 103 0.131 1.188 0.000 0.207 ROA F 103 0.018 0.237 -0.898 0.162 NF 346 0.038 0.279 -0.635 0.113 TOBINQ F 103 2.087 5.877 0.562 1.405 NF 346 103 1.900 6.673 0.723 1.062 ALLSHARES F 103 0.078 1.000 0.000 0.269 NF 346 0.069 1.000 0.000 0.254 RELATEDN F 103 0.369 1.000 0.000 0.485 NF 346 103 0.318 1.000 0.000 0.466
Panel B: Long-term analysis
FSIZE F 65 13.940 17.920 9.511 1.765 NF 307 14.457 19.285 9.710 1.894 FAGE F 65 27.908 113.000 3.000 21.278 NF 307 1 32.908 116.940 3.000 27.422 LEVERAGE F 65 0.471 0.726 0.227 0.171 NF 307 0.513 1.310 0.102 0.227 RELSIZE F 65 0.243 3.497 0.000 0.447 NF 307 1 0.132 1.258 0.000 0.215 ROA F 65 0.014 0.241 -1.041 0.200 NF 307 0.039 0.269 -0.530 0.104 TOBINQ F 65 1.847 5.726 0.514 1.085 NF 307 1 1.890 6.588 0.720 1.038 ALLSHARES F 65 0.108 1.000 0.000 0.312 NF 307 0.068 1.000 0.000 0.253 RELATEDN F 65 0.323 1.000 0.000 0.471 NF 307 0.316 1.000 0.000 0.466
Panel A reports the summary statistics of independent (control) variables used in the short-term multivariate analysis to test the impact of these variables in combination with family ownership (FAMFIRM) on short-term acquisition performance of U.S. companies within the 2008-2016 period. Panel B presents the summary statistics of independent (control) variables used in the long-term multivariate analysis to test the impact of these variables in combination with family ownership (FAMFIRM) on
long-term acquisition performance of U.S. companies within the 2008-2016 period. Family firms are defined as a firm with an
individual or family as ultimate owner at a 50.01% threshold. The definitions of the independent (control) variables, their units
of measurement, and expected signs are described in Appendix 2. To reduce the impact of incidental outliers, all variables are
winsorized at 0.01 and 0.99. The F in column 2 represents the family firm sample and NF the non-family firm sample.
This means that the growth opportunities for family firms are higher than for non-family firms. Finally, statistics show that family firms more often make acquisitions in related industries than non-family firms do, 36.9% (32.3%) and 31.8% (31.6%) of the family firm and non-family firms acquisitions, respectively, are in related industries in case of the short-term (long-term) analysis.
5.1. Acquisition performance for family firms versus non-family firms
In this section the results of the univariate analysis are presented. Table 3 shows the short-term and long-short-term acquisition performance of both family firms and non-family firms and the difference between family firms and non-family firms regarding short- and long-term acquisition performance.
5.1.1. Short-term acquisition performance
Panel A of table 3 presents the results of the (cumulative) abnormal returns on the days around the acquisition announcement. Both the parametric and the non-parametric test results show that there are positive and significant average abnormal returns of 1.228% and 0.605% for family firms and non-family firms, respectively, on day -1. This positive AAR on day -1 may be a result of an information leakage prior to the acquisition announcement (Brunnermeier, 2005). Non-family firms also realize a significantly positive average abnormal return on day 0, namely 0.520%, while family firms realize a positive, though insignificant, average abnormal return of 0.550% on day 0. These positive AARs on days -1 and 0 are in line with what is expected from literature (i.a., Vazirani, 2012). Significant results are also obtained on day 4. Non-family firms realize a negative average abnormal return of -0.211% on day 4, significant at the 1% level.4 This negative AAR on day 4 is not as expected, though not in conflict with findings from previous research (see, e.g., Doeswijk and Hemmes, 2001). The results of the tests on cumulative average abnormal returns imply that family firms as well as non-family firms realize significantly positive CAARs around an acquisition announcement.5 For example, over the three days around the acquisition announcement family firms have a CAAR of 1.928%, while non-family firms have a CAAR of 0.934%.
The positive CAARs around the acquisition announcement are in line with what is expected from literature (i.a., Vazirani, 2012).
4According to the Corrado test results only
Table 3
Univariate analysis of short- and long-term acquisition performance of family firms versus non-family firms
Family firms Non-family firms Difference
Day/CAR-period (C)AAR(%) Parametric test Corrado test (C)AAR(%) Parametric test Corrado test (C)AAR(%) Parametric test Mann Whitney test Panel A: Short-term abnormal performance
-1 1.228 2.686*** 1.970** 0.605 3.149*** 2.879*** 0.623 1.256 -1.171 (0.007) (0.049) (0.002) (0.004) (0.209) (0.242) 0 0.550 1.080 0.888 0.520 2.370** 3.400*** 0.031 0.055 -0.904 (0.280) (0.374) (0.018) (0.001) (0.956) (0.366) 1 0.150 0.426 -0.486 -0.191 -1.361 -1.471 0.340 0.900 -1.184 (0.670) (0.627) (0.174) (0.141) (0.368) (0.236) 2 0.068 0.304 -0.109 0.192 1.520 0.652 -0.124 -0.473 -0.047 (0.761) (0.913) (0.129) (0.515) (0.636) (0.963) 3 -0.074 -0.364 -0.280 -0.087 -0.783 -0.268 0.013 0.057 -0.853 (0.716) (0.779) (0.434) (0.789) (0.955) (0.394) 4 0.091 0.366 -1.218 -0.211 -1.819* -2.771*** 0.302 1.103 -0.375 (0.714) (0.223) (0.069) (0.006) (0.270) (0.707) [-1; 1] 1.928 2.682*** 1.370 0.934 3.048*** 2.776*** 0.994 1.272 -0.310 (0.007) (0.171) (0.002) (0.006) (0.203) (0.757) [-1; 2] 1.996 2.848** 1.132 1.126 3.604*** 2.730*** 0.870 1.133 -0.823 (0.004) (0.258) (0.000) (0.006) (0.257) (0.411) [-1; 3] 1.922 2.552** 0.887 1.039 3.092** 2.322** 0.883 1.071 -0.701 (0.011) (0.375) (0.002) (0.020) (0.284) (0.484) [-1; 4] 2.013 2.473** 0.312 0.828 -1.209 0.988 1.185 1.331 -0.912 (0.013) (0.755) (0.227) (0.323) (0.183) (0.362)
Panel B: Long-term abnormal performance
[0; 11] -13.291 -1.354 -0.032 -4.920 -1.409 -0.436 -8.371 -0.948 -0.125 (0.176) (0.975) (0.159) (0.663) (0.343) (0.900) [0; 8] -13.764 -1.699* -0.633 -6.184 -2.033** -1.351 -7.580 -0.999 -0.130 (0.089) (0.527) (0.042) (0.177) (0.318) (0.896) [0; 5] -5.002 -0.806 0.409 -5.151 -2.018** -1.368 0.148 0.022 -0.076 (0.420) (0.683) (0.044) (0.086) (0.982) (0.940)
The tests examining differences between family and non-family firms regarding its (C)AARs show that family firms generally realize higher (cumulative) abnormal returns around the acquisition announcement than non-family firms. However, these results are insignificant and hence are inconsistent with expectations as formulated in hypothesis 1.
5.1.2. Long-term acquisition performance
Panel B of table 3 reports the results of the cumulative abnormal returns over the months following acquisition completion. The table shows that for family firms as well as non-family firms the long-term cumulative abnormal returns following acquisition completion are all negative. This is not in line with expectations, but it is also not contradictory to results from earlier studies (i.a., Jensen and Ruback, 1983). The results imply that family firms realize a marginally significantly negative CAAR of -13.764% over a 9 month period following acquisition completion. Non-family firms realize a negative CAAR of -6.184% and -5.151% over 9 months and 6 months following acquisition completion, respectively, both significant at the 5% level.6 Fama (1998) argues that long-term abnormal return anomalies often occur in the literature on M&A performance. He argues that these long-term anomalies arise due to bad-model problems: There is no perfect asset pricing bad-model that completely describes the expected returns, and even if there were a prefect model, any
sample period has a systematic risk causing the returns to deviate from the model’s predictions.7 The results of the tests examining differences between family firms and non-family firms regarding long-term acquisition performance show that there is no significant difference between family firms and non-family firms. Accordingly, no supporting evidence is found for the hypothesis that family firms outperform non-family firms over the months following acquisition completion.
6 Since these results are obtained from parametric tests and ARs and CARs are not normally distributed, these results should be treated with caution.
32
5.2. Multivariate analysis of the effect of family ownership on acquisition performance
This section presents the results of the multivariate analysis. Table 4 shows the results of the OLS regression as specified in equation (23).8 Appendix 4 shows the results of the tests for heteroscedasticity and non-normality in the error terms. Appendix 5 shows the correlation coefficients for the independent (control) variables used in the multivariate analysis to test the impact of these variables in combination with family ownership on acquisition performance, and Appendix 6 shows the variance inflation factors of these variables. In short, heteroscedasticity is only detected for the models of the short-term analysis, and the Jarque-Bera statistic indicates that the error terms for both the short-term and long-term analysis are not normally distributed. Furthermore, for the short-term (long-term) analysis no correlation coefficient is higher than 0.306 (0.296) or lower than -0.215 (-0.341), and the VIFs are lower than 5 across all the estimations presented, which implies that multicollinearity is not a problem in this study (De Vaus, 2013).
5.2.1. The effect of family ownership in the short-term
The dependent variables to test the effect of family ownership on short-term acquisition performance that are shown in this paper are AR(-1), AR(0), 1; 1], 1; 2], CAR[-1; 3], and CAR[-CAR[-1; 4] as these variables provide the strongest and most interesting results for the univariate analysis. Table 4 shows that the adjusted R2 ranges from 0.014 to 0.147, which implies that 1.4% to 14.7% of the variation in the dependent variables can be explained by the estimations presented. These results are consistent with other similar studies (see, i.a., Bauguess and Stegemoller, 2008; Caprio et al., 2011). The F-statistics indicate that the joint null hypothesis that all slope parameters are equal to zero has to be rejected for the majority of the reported estimations. This means that at least one of the slope parameters within the estimations is significantly different from zero and that the estimations presented are not ‘junk regressions’.
To examine the effect of family ownership on short-term acquisition performance, the coefficients of the variable FAMFIRM are analyzed. The majority of the FAMFIRM coefficients reported in table 4 are positive, indicating that family firms realize higher returns around the acquisition announcement than non-family firms. However, these results are insignificant. Hence, no supporting evidence is found for the hypothesis that family firms realize higher (cumulative) abnormal returns in the short-run.
As for the control variables in the estimations explaining the impact of these variables on short-term acquisition performance, interesting and significant results are found for the variables ALLSHARES and ROA. The variable ALLSHARES, which is a dummy variable equal to 1 if only shares are used for payment and zero otherwise, has a significant coefficient ranging between 0.023 and 0.032. This implies that when only shares are used for payment, the acquirers’ (cumulative) abnormal return is between 2.3 and 3.2 percentage points higher. This result is notable since it is inconsistent with this study’s expectations and results of similar papers. A possible explanation for this positive effect of a share payment on short-term acquisition performance is that a share payment is a sign of confidence of the target company in the value of its acquirer (Rappaport and Sirower, 1999). The variable ROA, acquirers’ return on assets, has a significant coefficient varying between -0.121 and -0.080. This implies a decrease in the (cumulative) abnormal return between 12.1 and 8.0 percentage points when the return on assets of the acquiring firm increases with 1 unit. This is in line with the expectation that strong operational performance leads to the establishment of an aggressive acquisition strategy (Halebalian and Kim, 2006) and to overconfidence of firm’s decision-makers.
Table 4
Multivariate analysis of the effect of family ownership on short- and long-term acquisition performance
Short-term abnormal performance Long-term abnormal performance
Variable AR(-1) AR(0) CAR[-1; 1] CAR[-1; 2] CAR[-1; 3] CAR[-1; 4] CAR[0; 11] CAR[0; 8] CAR[0; 5]
FAMFIRM 0.002 -0.001 0.005 0.003 0.003 0.007 -0.092 -0.082 -0.015 (0.502) (-0.122) (0.749) (0.492) (0.433) 0.812 (-1.051) (-1.077) (-0.236) ALLSHARES 0.023** 0.010 0.032** 0.026* 0.025* 0.026 -0.214 -0.216* -0.091 (2.147) (0.865) (2.207) (1.875) (1.674) 1.460 (-1.632) (-1.901) (-0.966) FAGE 0.000* 0.000 0.000 0.000 0.000 0.000 -0.001 -0.001 0.000 (1.731) (0.250) (0.864) (0.694) (0.841) 1.101 (-0.981) (-0.645) (-0.443) FSIZE -0.002 -0.001 -0.003 -0.004* -0.005** -0.006** 0.056*** 0.041** 0.015 (-1.619) (-0.509) (-1.230) (-1.842) (-2.146) -2.448 (2.816) (2.405) (1.061) LEVERAGE 0.002 0.012 0.003 -0.007 0.000 0.004 -0.228 -0.222* -0.061 (0.220) (1.103) (0.221) (-0.510) (-0.020) 0.197 (-1.491) (-1.683) (-0.552) RELATEDN 0.003 -0.004 0.000 -0.006 -0.005 -0.005 -0.108 -0.068 -0.060 (0.660) (-0.777) (-0.070) (-1.009) (-0.705) -0.647 (-1.494) (-1.089) (-1.147) RELSIZE 0.022 -0.049 0.016 0.015 0.012 0.006 0.055 0.043 0.111 (1.360) (-0.919) (0.965) (0.852) (0.576) 0.300 (0.408) (0.370) (1.145) ROA -0.080** 0.002 -0.121** -0.104** -0.114** -0.095 -1.107*** -0.875*** -0.514** (-2.589) (0.186) (-2.304) (-2.135) (-2.105) -1.363 (-3.829) (-3.498) (-2.466) TOBINQ 0.003 0.000 0.003 0.004 0.004 0.003 -0.022 -0.012 0.008 (1.029) (0.137) (1.036) (1.211) (1.255) 0.710 (-0.672) (-0.439) (0.328) C 0.019 0.012 0.039 0.063** 0.072** 0.091*** -0.569** -0.430* -0.209 (1.232) (0.516) (1.235) (2.084) (2.254) 2.710 (-1.973) (-1.724) (-1.005) N 449 449 449 449 449 449 372 372 372 F. Stat 9.604*** 1.179* 7.152*** 6.295*** 6.314*** 5.158*** 2.985*** 2.476*** 1.215 Adjusted R2 0.147 0.014 0.110 0.096 0.096 0.077 0.046 0.035 0.005
The table presents the results of the multivariate analysis to test the impact of the independent (control) variables in combination with family ownership (FAMFIRM) on short-term as well as long-term acquisition performance of U.S. companies within the 2008-2016 period. The dependent variables, presented in row 1, are the ARs and CARs around the acquisition announcement
(over the months following acquisition completion) in the case of the short-term (long-term) analysis. The (cumulative) abnormal returns are calculated using the market model, which is
In conclusion, no evidence is found for the hypothesis that family firms outperform non-family firm in the short-run when acquiring. However, a significantly positive relationship is found between a share payment (ALLSHARES) and short-term acquisition performance, and a significantly negative relationship is found between acquirers’ return on assets (ROA) and short-term acquisition performance.
5.2.2. The effect of family ownership in the long-term
The dependent variables used in the regressions to test the effect of family ownership on long-term acquisition performance are CAR[0; 11], CAR[0; 8], and CAR[0; 5]. The adjusted R2 ranges from 0.005 to 0.046, which implies that 0.5% to 4.6% of the variation in de dependent variables can be explained by the estimations presented. The F-statistics for the estimations with CAR[0; 11] and CAR[0; 8] as the explanatory variables imply that these estimations are not ‘junk regressions’ since at least one of the slope parameters within these models is significantly different from zero.
The coefficients of the variable FAMFIRM are analyzed to examine the effect of family ownership on long-term acquisition performance. The coefficients of the FAMFIRM variable are all negative in explaining long-term acquisition performance, indicating that family firms realize lower returns over the months following acquisition completion. Although, these results are insignificant. This is not in line with hypothesis 2.
The other control variables do not appear to affect long-term acquisition performance. For two of the other control variables the coefficient is significantly different from zero, but only marginally significant and in one of the three reported estimations. This indicates that these significant results are coincidental. Thus, there is no strong evidence that the control variables ALLSHARES, FAGE, LEVERAGE, RELATEDN, RELSIZE, and TOBINQ are individually significantly different from zero in explaining long-term acquisition performance. This is, as with the short-term analysis, not in line with expectations based on theories and certain empirical results. However, the majority of the relevant papers examining the effect of family ownership on M&A performance described in section 2.4 and in Appendix 1 also have no significant results for these variables (i.a. Basu et al., 2009; Bouzgarrou and Navatte, 2013; Adhikari and Sutton, 2016).
In conclusion, no supporting evidence is found for hypothesis 2. Furthermore, the variables firm size (FSIZE) and acquirers’ return on assets (ROA) have a significantly positive and negative impact on long-term acquisition performance, respectively.
6. Conclusion
This chapter is subdivided into two sections. First, the paper is summarized. Second, some limitations subject to this study are provided and suggestions for future research are mentioned.
6.1. Summary
The effect of family ownership on acquisition performance has been extensively researched in literature since early contribution of Holderness and Sheehan (1988; in Caprio et al., 2011), although the results are inconclusive. Hence, this paper adds a valuable empirical finding to the literature on this subject. This paper examines the effect of family ownership in NYSE and NASDAQ listed companies on short-term as well as long-term acquisition performance. The event study methodology suggested by MacKinlay (1997) is applied to a sample of 449 (372) acquisitions made by family and non-family firms between 2008 and 2016 for the short-term (long-term) analysis. With this data both univariate and multivariate tests are performed.
