0
Downsizing and M&A
activity: closing the gaps
Name: Ruben van Dijken Student Number: 4231791 Address: Sophiaweg 216 k5 Phone: +316 81 03 29 05
E-mail: rubenvandijken3@hotmail.com First Supervisor: Dr. drs. H. L. Aalbers
1 Table of content
Table of content ... 1
1. Introduction ... 3
2. Theoretical background and hypotheses ... 7
2.1 Theoretical background ... 7
2.1.1 Downsizing ... 8
2.1.2 M&A Performance ... 12
2.2 Hypotheses ... 14
2.2.1 The effect of downsizing on M&A performance ... 14
2.2.2 Industry-relatedness ... 16
2.2.3 Liability of foreignness ... 17
2.3 Conceptual model ... 18
3. Methodology ... 19
3.1 Data collection and sample ... 19
3.2 Dependent variable ... 22
3.3 Independent variable ... 23
3.4 Moderator variables ... 25
3.5 Control variables ... 26
3.5 Data analysis strategy ... 28
3.6 Assumptions regression analysis ... 35
4. Results ... 36
4.1 Regression analysis ... 36
4.2.1 Effect of downsizing on performance in the year of the deal (t) ... 36
4.2.2 Effect of downsizing on performance in the year after the deal (t+1) ... 38
4.2.3 Effect of downsizing on performance two years after the deal (t+2) ... 41
4.2.4 Effect of downsizing on performance two years after the deal (t+3) ... 44
4.3 Summary regression analysis ... 46
5. Discussion ... 47
5.1 M&A performance ... 47
5.2 Effect of downsizing on M&A performance ... 48
5.3 The influence of industry relatedness and the foreignness of the acquirer and target ... 52
6. Conclusion, limitations, recommendations and managerial implications ... 55
6.1 Conclusion ... 55
2
6.3 Recommendations ... 56
6.4 Managerial implications ... 57
Reference list ... 59
Appendix ... 74
Appendix 1: Deals analysed & desciptives ... 74
Appendix 1A Deals with US target ... 74
Appendix 1B European targets ... 78
Appendix 2: Normality ... 82
Appendix 3 Regression analysis ... 85
Appendix 3a: ROA_t ... 85
Appendix 3b: ROA t+1 ... 94
Appendix 3c: ROA t+2 ... 106
3
1. Introduction
Recently, the U.S. pharmaceutical company Mylan announced that it will reduce its workforce in order to streamline its operations globally, after the acquisition of the Swedish drug maker Meda. Immediately, the shares of Mylan fell with more than 5 per cent (Dye, 2016). Why are the shareholders reacting in such a negative way towards this downsizing announcement? The obvious answer would be that, according to research, in almost all cases downsizing has negative financial (Cascio, 1993; Gandolfi, 2008; Guthrie & Datta, 2008; Love & Nohria, 2005) organizational (Cameron, Whetten, & Kim, 1987) and human consequences (Datta & Basuil, 2015; Cascio, 1993) and downsizing is therefore likely to reduce shareholder value. The next question that immediately arises is: why do firms then downsize?
Managers often have the deep-seated belief that downsizing increases organizational efficiency and performance (Datta & Basuil, 2015). However, in literature it is suggested that the influence of downsizing on market returns is contingent to the strategic intent underlying the employee downsizing. On the one hand it has been found that a company that has a more reactive downsizing intent, triggered by demand declines, increasing costs and poor performance, is subject to a more negative market response (Datta & Basuil, 2015; Marshall et al., 2012; Chen et al., 2001; Gunderson et al., 1997). On the other hand, research indicates that proactive downsizing, designed to enhance organizational efficiency or in responding to post-merger rationalization, is subject to a more positive market response (Datta & Basuil, 2015; Franz et al., 1998; Hallock, 1998; Cagle et al., 2009; Farber & Hallock, 2009; Hahn & Reyes, 2004). The discussion above is about the shareholder value, because shareholde value is most often reported in the media, like in the case of Mylan. However, the findings are consistent with findings of accounting-based measures of performance (Datta & Basuil, 2015; Farrell & Mavondo, 2005; Chalos & Chen, 2002; Iqbal & Akhigbe, 1997; Palmon et al., 1997).
Next to the strategic intent of downsizing, other factors like union presence (Abraham, 2006), the practices used to implement downsizing (Appelbaum, 2002; Appelbaum et al., 1999), nature of industry (Cagle et al., 2009; Gunderson et al., 1997; Marshall et al., 2012) and whether the firm is the first in its industry to downsize (Lee, 1997), are important determinants that influence market outcomes associated with downsizing (Datta & Basuil, 2015).
Thus, it can be concluded that managers believe that employee downsizing has positive implications for organizational efficiency and enhanced performance, but that is often not the
4 case and very dependent on the circumstances under which downsizing might be appropriate. This research will focus on one circumstance in particular: downsizing that is related to M&A activity.
There are several theories on what causes M&As: operating synergy, financial synergy, diversification, strategic realignment, hubris (managerial pride), buying undervalued assets (Q-Ratio), managerialism (agency problems), tax considerations, market power and misevaluation (DePamphilis, 2015). Anticipated synergy between the acquirer and target is most often cited as the primary motivation for M&A deals (DePamphilis, 2015). After an acquisition, firms often engage in cost cutting by consolidating the activities of both firms and this usually involves work force reduction within the combined firm (O’Shaughnessy & Flanagan, 1998; Labib & Appelbaum, 1994). In this paper, downsizing is defined as the planned set of organizational policies and practices aimed at workforce reduction with the goal of improving firm performance (Datta et al., 2010, Cameron, 1994; Cameron et al., 1993; Cascio, 1993; DeWitt, 1998; Freeman & Cameron, 1993). This workforce reduction is often necessary to eliminate redundancies so that operational synergies can be extracted in related (horizontal) acquisitions (Krishnan et al., 2007; Conyon et al., 2002; O’Shaughnessy & Flanagan, 1998). Therefore, there is theoretically sound reasoning for downsizing that is related to M&A activity, at least in related acquisitions.
There is a huge body of knowledge regarding the effect of downsizing on organization performance (Datta et al., 2010) and there is a huge body of knowledge regarding M&A performance (Zollo & Meier, 2008). However, there are not so many studies that assess the effect of downsizing on M&A performance. The first study that explored whether workforce reduction carried out after a major acquisition benefits the organization was conducted by Krishnan and Park (2002). Krishnan and Park (2002) found support for their hypotheses that workforce reduction after an acquisition will result in poor post acquisition performance. However, a major limitation of this study is that a sample of only 42 firms is used. In a later study, using a larger sample of 174 firms, Krishnan et al. (2007) found again support for the hypothesis that there is a negative relationship between workforce reduction and post-acquisition performance. In both studies, the sample consisted of only US based acquired and acquiring firms that are in related industries.
In general, the majority of downsizing research has been conducted in the United States (Gandolfi, 2009; Chadwick, Hunter & Walston, 2004), which is a Liberal Market Economy
5 (Harzing & Pinnington, 2015), although downsizing is not confined to firms in the United States, but occurs throughout the world (Ryan & Macky, 1998; Freeman, 1994; Thornhill & Saunders, 1998; Dolan et al., 2000; Lamsa & Takala, 2000), including Coordinated Market Economies (Harzing & Pinnington, 2015), and is even becoming more common in countries that have historically displayed very stable employment practices (Datta & Basuil, 2015; Mroczkowski & Hanaoka, 1997).
Next to that, relatively few studies investigated cross-border acquisitions as a specific group. It has been concluded in the literature that, because just relatively few studies investigated cross-border acquisitions as a group, we actually know very little about cross-cross-border acquisitions (Collins et al., 2008). Cross-border acquisitions are a manner in which firms can rapidly expand across national borders (Hitt et al., 1998; Nodolska & Barkema, 2007). However, cross-border acquisitions are more complex and characterized by higher levels of uncertainty and information asymmetry (Basuil & Datta, 2015; Collins et al. 2009; Johnson and Vahlne, 2009; Vermeulen & Barkema, 2001; Kogut & Singh, 1988; Shimizu et al., 2004; Kostova, 1999; Kostova & Zaheer, 1999; Zaheer 1995). Therefore, because of the cultural and institutional differences and the liability of foreignness, it is reasonable that this will moderate the effect of downsizing on M&A performance.
As previously argued, there is theoretically sound reasoning for workforce reduction that is related to M&A activity in related acquisitions, but there is a gap in knowledge on how industry-relatedness moderates the effect of downsizing on M&A performance (Datta & Basuil, 2015; O’Shaughnessy and Flanagan, 1998). According to Conyon et al. (2002), workforce reduction in an unrelated acquisition is problematic and is less likely to occur than in a related acquisition. Conyon et al. (2002) reasons that if an unrelated acquisition is made by managers who are primarily motivated by the desire for diversified firm earnings and a reluctance to unburden free cash flow, there will be no job losses. According to Conyon et al. (2002), downsizing may only realistically follow if the transaction is seen as a disciplinary one in which the market for corporate control operates to divert assets into the hand of the more talented managers. However, no study has been conducted on whether industry-relatedness between the acquirer and target moderates the effect of downsizing on M&A performance.
In sum, I identified two main gaps in the literature regarding the effect of downsizing on M&A performance. Firstly, there is no study that assessed foreignness as a moderator on the effect of downsizing on M&A performance, so I will include this variable in my research. Secondly, no
6 study has been conducted on whether industry-relatedness between the acquirer and target moderates the effect of downsizing on M&A performance and therefore I will also include industry-relatedness in my research. I will close these gaps by, in correspondence with the literature cited when I described the different gaps, adopting a connection with respectively the human resource literature, international business literature, transaction cost literature and the resource based view.
The goal of this paper is to address the gaps identified above. These gaps are part of a larger discours in literature, which I described above, namely the effect of downsizing on M&A performance. Therefore, the main research question is: What is the effect of downsizing on M&A performance? In order to close the gaps identified above, several sub questions arise:
1. What is M&A performance and how is it measured?
2. How does downsizing affect performance, especially post deal?
3. How is the effect of downsizing on M&A performance influenced by industry relatedness and foreignness between the acquirer and target?
Answering this research question and these sub questions will enhance our understanding of how workforce reduction in the context of M&As affects M&A performance and influences the ability of the acquiring firm to reimburse its investment in the deal (O’Shaughnessy and Flanagan 1998; Brauer, 2006).
The text below is structured as follows: first, I will give a brief review of relevant literature, necessary to subsequently draw hypotheses from (Paragraph 2). Next, I will describe my methodology, including sample selection, the operationalization of variables and the analytical methods used (Paragraph 3). In the following paragraph, I will describe the data analysis and discuss the results (Paragraph 4), followed by a discussion of those results in the next paragraph (Paragraph 5). Finally, I conclude by discussing the managerial implications, limitations and future research (Paragraph 6).
Table 1.
The gaps in M&A downsizing literature. Gap 1: No attention for
cross-border effects
Gap 2: No attention for industry-relatedness
7
2. Theoretical background and hypotheses 2.1 Theoretical background
In this paragraph, I will elaborate more on the processes underlying the effect of downsizing on M&A performance, in order to subsequently draw hypotheses. As stated in the introduction, after an acquisition, firms often engage downsizing activities (Datta et al., 2010). Downsizing is defined as the planned set of organizational policies and practices aimed at workforce reduction with the goal of improving firm performance (Datta et al., 2010, Cameron, 1994; Cameron et al., 1993; Cascio, 1993; DeWitt, 1998; Freeman & Cameron, 1993).
In a major review, Datta et al. (2010) developed a useful integrative framework, which I will use to structure my argumentation provided in this theoretical section (figure 1). I adapt the model to better connect the model with the objective of this paper.
In the first place, this means that I will not use every single point of the model, because some points do not fall within the scope of this research, as outlined in the introduction. To answer the research question, I will only take into account the organizational factor ‘strategy’, more specifically ‘M&A’ and ‘diversification level’, that have a direct effect on employee
Figure 1. Integrative
framework. Adopted from: Datta et al., 2010
8 downsizing. Furthermore, regarding the environmental factors, I will only consider the institutional environment and the type of industry as moderator effects of downsizing on performance. Regarding, the employee downsizing, I am only interested in the downsizing decision in this research. Regarding the outcomes, I am only measuring the M&A performance and I will elaborate on that in the methodology section (chapter 3).
In the second place, Datta et al. (2010) and the studies included in that article, give only an overview of the factors that influence downsizing. Below, I will focus on the reasoning and I will adopt different perspectives that explain why it is likely that certain effects described in the model occur.
2.1.1 Downsizing
As can be seen in the model (box 2), one of the organizational factors that, according to Datta et al. (2010) has a direct effect on employee downsizing, is strategy on the M&A level. The question that arises: what are the reasons for workforce reduction that is related to M&A activity? Another question that immediately follows is: To what extent is it likely that the workforce will be reduced after an M&A?
One reason for downsizing related to M&A activity is to eliminate redundancies, so that operational synergies can be extracted (Krishnan et al., 2007; Conyon et al., 2002; O’Shaughnessy & Flanagan, 1998; Labib & Appelbaum, 1994). Although this is most obviously the case for related M&As, because there are more redundancies in companies that are related, this may also be the case in unrelated M&As (O’Shaughnessy & Flanagan, 1998). I will elaborate more on this in paragraph 2.3. Another reason for downsizing related to M&A activity is that shareholders may require managers of firms in poorly performing industries to use downsizing as a means to increase shareholder wealth (Morck et al., 1989). When targets are poorly performing before the M&A, a third reason for downsizing is to correct, through downsizing, the inefficient application of the human assets (O’Shaughnessy & Flanagan, 1998; Chatterjee, 1992; Lichtenberg & Seigel, 1987). A fourth reason for downsizing is that, in the case that the deal is financed with debt, managers are required to cut costs through downsizing in the combined firm to meet de debt payments (O’Shaughnessy & Flanagan, 1998; Ofek, 1993).
Thus, although there are studies that did not find a significant relation between M&A activity and workforce reduction (Wagar, 1997), there is theoretically sound reasoning for workforce
9 reduction related to M&A activity. It can even be stated that downsizing after an M&A not a surprise, but rather expected (Shook & Roth, 2011).
Moreover, as can be seen in box 2 of figure 1, is has been found that merger relatedness has an effect on whether a firm downsizes or not. Buying a firm beyond a current line of business of a company is called diversification (DePamphilis, 2015). Diversification can lead to financial synergy (by reducing the cost of capital) or it can allow a firm to shift its core product lines/target markets to those with higher growth perspectives that are unrelated to the firm’s current products or markets. (DePamphilis, 2015).
Regarding diversification, DePamphilis (2015) identifies a firm’s primary diversification options as: 1) buying another firm with the same products that is active in the same market (no diversification), 2) buying another firm with new products that is active in the same market (related diversification), 3) buying another firm with the same products in a new market (related diversification) and 4) buying another firm in a new market with new products (unrelated diversification). The following question can be raised: How is the incidence of downsizing following M&As influenced by industry relatedness between the target and the acquirer? O’Shaughnessy and Flanagan (1998) and Conyon et al. (2002) found that related acquisitions, which are acquisitions in which the target and acquirer operate in the same industry are more likely to be followed by layoffs. Their reasoning is very sound: if related firms combine there is more opportunity for the combined firm to realize operational synergies by eliminating redundancies than when non-related firms combine (O’Shaughnessy and Flanagan, 1998; Conyon et al., 2002). Moreover, in related acquisitions, managers of the acquiring firm are better able to run the target and recognize inefficiencies, and therefore managers of the acquired firm are less needed (Walsh, 1989). Thus, in contrast to non-related acquisitions, related acquisitions are more likely to involve layoffs.
However, it is recognized by O’Shaughnessy and Flanagan (1998) that also in unrelated acquisition, it is possible that workforce reduction occurs in the form of the elimination of redundant activities. Would that be the only downsizing activity in an unrelated acquisition? A mechanism that could project layoffs in unrelated M&As is the following. The share prices of firms that are operating in a number of unrelated industries (conglomerates) are traded at a discount to their value if broken up, which is lower than for focused firms (Ammann et al., 2012; Khorana et al., 2011). These conglomerates also often exhibit poor governance practices (Hoeckle et al., 2012). Moreover, investors are reluctant that managers try to build empires
10 rather than to improve firm performance and are therefore less willing to invest in the conglomerate (DePamphilis, 2015). Next to that, the acquirer is less familiar with the business and has less ability to optimize the investment (DePamphilis, 2015). Because of all these negative effects, an unrelated M&A is more likely to show lower return (DePamphilis, 2015) and is therefore likely to reduce its workforce due to bad performance.
However, the majority of studies show that downsizing is more likely to occur in related M&As than in unrelated M&As (i. e. O’Shaughnessy and Flanagan, 1998; Conyon et al., 2002). Above, the reasoning regarding downsizing after an M&A has been explained. The question that subsequently raises is: how to determine whether a firm has reduced its workforce? This question has, to my best knowledge, not received any specific attention in literature yet (Datta et al., 2010). In the next subparagraph, I will, before further theoretical background about downsizing and M&A activity is discussed, pay specific attention to the determination of whether a firm has downsized or not.
2.1.1.1 Downsizing: measurement issues
In the introduction, downsizing was defined as the planned set of organizational policies and practices aimed at workforce reduction with the goal of improving firm performance (Datta et al., 2010, Cameron, 1994; Cameron et al., 1993; Cascio, 1993; DeWitt, 1998; Freeman & Cameron, 1993). In previous studies, downsizing is measured as the percentage reduction in the number of employees in the consolidated organization after the acquisition (e.g. Lehto & Böckerman, 2008; Krishnan et al., 2007; Conyon et al., 2002; Osterman, 2000). However, the data sources where this percentage is based upon, vary across different studies.
According to Datta et al. (2010), the three most common sources of data to determine whether a company engaged in downsizing activities are: 1) layoff announcements in newspapers (e.g. Krishnan & Park, 2002; Ofek, 1993; O’Shaughnessy & Falanagan, 1998, Hillier et al., 2006), 2) surveys that allow firms to self-identify themselves as downsizers (e.g. Osterman, 2000; Iverson & Pullman, 2000) and 3) archival data that is publicly available (e.g. Compustat or Orbis) to measure the employment level change (e.g. Said et al., 2007; Cascio & Young, 2003; Cascio, 1997). Furthermore, based on varying decision rules across different studies, the data is often dichotomized to label firms as either “downsizers” or “non-downsizers” and some studies use percentage change as a direct measure of downsizing (Datta et al., 2010). Because of this diversity of approaches, it is not clear that these operationalizations measure the same
11 underlying construct and this makes comparisons across studies difficult and, therefore, a careful consideration of the construct validity of “downsizing” is needed in my opinion. Construct validity is the extent to which a set of measured variables actually represent the theoretical latent construct, downsizing in this case, they are designed to measure (Hair et al., 2014). In my opinion, the previously mentioned methods have the following advantages and disadvantages regarding construct validity. An advantage of the first method, gathering layoff announcements, is that it is clear that the workforce reduction was a conscious decision of the board a not just a coincidence (e.g. a lot of people retiring). A disadvantage of layoff announcement is that it is not clear if the workforce reduction actually took place and when the workforce reduction took place and therefore this compromises the construct validity (Krishnan et al., 2007). Moreover, not every workforce reduction will be reported in newspapers. An advantage of the second method, surveys, is that it is more reliable that firms that actually proceed with workforce reduction are in the sample and not also firms that only announced it, but changed their minds. However, a disadvantage is that the method is rather subjective and dependent on the perceptions of the manager. An advantage of the last method, measure employment level change with public available data, is that it is an objective measure to assess whether the workforce has actually been reduced. However, a disadvantage is that it is not clear whether this reduction in the workforce was caused by an downsizing decision
As suggested by Lee & Hong (2016) and Griffin (2003), annual reports are subject to government regulations and are considered as one of the most credible documents about a company. Therefore, annual reports provide a rich source of reliable data.
A method to extract information from annual reports, which has been used before by, for example, Hajek & Henriques (2017) to find cases of financial statement fraud, is datamining. Datamining is a technique that allows to find patterns in these reports (Shirata et al., 2011) and is defined as ‘the process of analyzing large amounts of data in order to discover patterns and
other information. It is typically performed on databases, which store data in a structured format. By "mining" large amounts of data, hidden information can be discovered and used for other purposes’ (Techterms, 2017).
To my best knowledge, data mining has not yet been used to measure downsizing. However, this technique has been applied in, for instance, Corparate Social Responsibility literature (Liew et al., 2014; Saha & Nabareseh, 2015; Garechana et al., 2017), in the detection of financial statement fraud (Kirkos et al., 2007; Hajek & Henriques, 2017) and, more in general, in
12 different financial research domains (Kumar & Ravi, 2016). In, for instance, financial statement fraud studies, the technique of data mining has been found a very accurate method (even better than in-depth expert knowledge) to detect fraud (Hajek & Henriques, 2017; Dyck et al, 2010). If it is an accurate technique to detect fraud form linguistics, it is, reasonably, also an adequate technique to detect layoffs from linguistics.
In sum, it is not easy to determine whether a firm has engaged in downsizing activities and this particular topic has received not enough attention in literature. Above, I gave an overview of techniques that are currently being employed and I reasoned, based on other fields of research, that data mining of annual reports is also an accurate method to determine whether a firm has downsized or not. I will discuss the exact measurement of downsizing in the paragraph 3.3. Another area that requires specific attention, before any discussions about the effect of downsizing on M&A performance can be discussed, is: What is M&A performance? I discuss this question in the next paragraph.
2.1.2 M&A Performance
M&A performance is typically defined as the amount of value captured by the acquirer as a result of an acquisition (Cording et al., 2010; King et al., 2004). It is not overstated to say that the measurement of M&A performance is almost a separate field of study, considering the massive amount of research that focuses on how to measure M&A performance (e.g. Meglio & Risberg, 2011; Cording et al., 2010; Zollo & Meier, 2008; Thanos & Papadakis, 2011). M&A performance is an ambiguous construct that can be measured in a variety of ways and therefore it is important to clearly define what is meant with ‘M&A performance’ (Meglio & Risberg, 2011). In a literature review, Meglio and Risberg (2011) systematically analysed studies on M&A performance through the following questions: 1) where is M&A performance measured in terms of industry and geographic location? 2) how is M&A performance measured? 3) when is M&A performance measured? 4) what is the unit of analysis for M&A performance? 5) what is measured as M&A performance?
I will adopt these five questions to define what M&A performance is in this study. First, I do not limit the M&A performance to a specific industry, but I do consider industry-relatedness between the target and acquirer as a moderator variable on the effect of downsizing on M&A performance. Meglio and Risberg (2011) found that in the majority of studies on M&A performance, it was impossible to identify the industries involved in the studied M&As. In this research I focus on M&A performance in all industries. Moreover, in terms of geographic
13 location, the US is the most commonly studied geographical studied area considering M&A performance (Meglio & Risberg, 2011). This study is limited to M&A deals where the target is either vested in the US or the UK. I do not only consider domestic M&As, but also cross-border M&As.
Second, M&A performance is predominantly measured through quantitative methods (Meglio & Risberg, 2011). This research will be no exception to this dominance of quantitative methods in M&A research.
Third, there are different time scales concerning M&A performance: short-term (< 1 year), medium (> 1 year < 3 years) and long-term (> 3 years) (Meglio & Risberg, 2011). Meglio and Risberg (2011) found that from a sample of 101 journal articles that studied M&A performance, 48 articles employed a medium or long-term time scale and 35 employ a short-term time scale, though authors often do discuss and argue why they use a certain time interval. There is no clarity in literature about the use of short-term versus long-term windows in event studies and the evidence is till inconclusive (for example Zollo & Meier, 2008; Meglio & Risberg, 2011; Dutta et al., 2013). Short-term event studies can lead the researcher to erroneous conclusions, especially when market-based measures are used, because only the dominant cognitive heuristic, related to the type of acquisition announced, is captured and not the real information on the economic value generation from the transaction (Zollo & Meier, 2008) However, the advantage of using a short-term time horizon is that other factors than the downsizing that may influence the performance can be largely eliminated (Cording et al., 2010). Central in this study is the effect of downsizing on M&A performance and because it is not really clear what is the best time measurement for the effect of downsizing on M&A performance, I differentiate between M&A performance at different points on time, from the year of the deal until three years after the deal. To measure performance over different time horizons also fits with the suggestions made by Meglio et al. (2015) and Cording et al. (2010), whom argue that in reality, the impact of human integration is likely to manifest at different times and in different ways. Fourth, the unit of analysis can be the acquiring firm, the acquired firm of both firms, although the acquiring firm is the most common unit of analysis (Meglio & Risberg, 2011). Below, I will also focus on the acquiring firm.
Fifth, M&A performance can be measured through financial, non-financial or mixed measures (Meglio & Risberg, 2011). It has been found that the majority of studies used financial measures, like accounting-based measures (Meglio & Risberg). To integrate this research into
14 the existing literature, this study will also measure performance through financial measures. The advantage of using accounting based measures is that these represent the actual returns of the firm and it is less sensitive to unexpected economic changes and information asymmetries in the market (Krishnan et al., 2007). Moreover, accounting based measures better reflect the internal effectiveness of managing resources in the merged firm (Krishnan et al., 2007). However, limitations are that managers may manipulate accounting returns and firms may employ different accounting policies, but then also the market based information would be influenced (Cording et al., 2010). Therefore, I will focus on accounting based measures. In sum, post-acquisition performance is thus a quantitative measure of the performance at different points of time after the M&A event, at the acquiring firm, in an (cross-border) acquisition with European and/or US targets, regardless of the industry.
2.2 Hypotheses
2.2.1 The effect of downsizing on M&A performance
If it can be assumed that firms that are involved in M&A activity will have a high probabability to resort to workforce reduction, the question that arises is: what are the outcomes of downsizing (box 4 and 5)? And thus: what is the effect of downsizing on M&A performance (box 5). As argued above, firms have different reasons to downsize after an acquisition. However, downsizing also has some costs. Cascio (2009) gives an overview of the direct and indirect costs that firms have to deal with when they downsize (table 2).
Direct costs Indirect costs
Severance pay Recruiting and employment costs of new hires Accrued vacation and sick pay Low morale, risk-averse survivors
Supplemental unemployment benefits Increase in unemployment tax rate
Outplacement Lack of staff when economy rebounds, training and retraining
Pension and benefit payouts Potential lawsuits from aggrieved employees Administrative procession costs Heightened insecurity, reduced productivity Costs of rehiring former employees Loss of institutional memory
15 The three most commonly cited streams of reasoning on the effect of downsizing on firm performance are: the resource-based view, the human resource literature and the knowledge-based view.
The majority of literature adopts a resource-based view to conclude that workforce reductions are unlikely to have positive effects on firm performance (Cascio et al., 1997; Collins & Montgomery, 1995). The essence of the resource-based view is that competitive advantage is derived from the resources and capabilities a firm controls that are valuable, rare, imperfectly imitable and not substitutable (Barney, 1991). The resource-based view has made important contributions to the field of Human resource management (Barney et al., 2001; Wright et al., 2001). Based on this view, is has been reasoned that workforce reduction leads to a negative effect on firm performance because workforce reduction erodes the human resource base. Subsequently, the most valuable, rare and difficult to imitate knowledge, namely the implicit knowledge held by a firm’s human capital, is affected and therefore the competitive advantage is likely to be harmed (Krishnan et al., 2007; Chadwick et al., 2004; Nixon et al., 2004; Cascio et al., 1997; Collins & Montgomery, 1995).
Another point of view, which stems from the human resource literature, is that downsizing conceals hidden costs (Datta & Basuil, 2015). These hidden costs are the costs of violating the psychological contract between employees and employers. The psychological contract between the employee and the employer is an employees’ belief in mutual obligations between that employee and the employer, which is based on the perception that a promise has been made (for example employment or promotion) and a consideration is offered (for example accepting the position), binding both parties to some set of reciprocal obligations (Rousseau & Tijoriwala, 1998). During restructuring, it is likely that the psychological contract is violated (Turnley & Feldman, 1998). Consequentially, the violation has a negative effect on the attitudes and behaviours of the employees remaining within the firm and that will result in a loss of motivation and commitment, greater work related stress and a lower morale within the firm (Datta & Basuil, 2015; Rousseau & Tijoriwala, 1998; Pugh et al., 2003; Travagliona & Cross, 2006; Kets de Vries & Balazs, 1997; Feldheim, 2007), which will trigger eventually an increase in absenteeism, increased turnover and lower productivity, and thus a lower performance (Datta & Basuit, 2015).
Moreover, from a knowledge-based view of the firm, it has been reasoned that the human capital of a firm represents a social community for knowledge creation, diffusion and
16 utilization, which facilitates interactions among people in the firm (Nixon et al., 2004; Spender, 1996; Zander & Kogut, 1995). These interactions promote the creation and diffusion of (especially tacit) knowledge, which becomes embedded in the organization and enhance the firm’s knowledge inventory (Nixon et al., 2004; Levinthal & March, 1993). According to the knowledge-based view of the firm, knowledge resources have the distinctive properties of heterogeneity and immobility, and knowledge thus becomes a way of creating a sustainable competitive advantage (Mabey & Zhao, 2017; Kogut & Zander, 1996; King & Zeithaml, 2003; Horwitz et al., 2003). Accordingly, downsizing can lead to: 1) disruption of the operation of the social community for knowledge creation and/or sharing within the firm, 2) reduction managers’ options for the current and future use of this inventory, and 3) disruption of the social community through the elimination of the relationships in which the tacit knowledge was diffused, which will take a lot of time and focus to rebuild. Because of these effects, downsizing is likely to have a negative impact on firm performance (Nixon et al., 2004; Miller, 2002). In conclusion, based upon the resource-based perspective, human resource literature and the knowledge-based view of the firm, it is likely that downsizing has a negative effect on firm performance and therefore it will have a negative effect on M&A performance. Thus, I hypothesize that:
Hypothesis 1: Downsizing has a negative effect on M&A performance.
2.2.2 Industry-relatedness
As argued above, there is research that shows that in related M&A deals, the incidence of workforce reduction is higher than in unrelated M&As (O’Shaughnessy and Flanagan, 1998; Conyon et al., 2002). This does not imply that unrelated firms do not lay-off workers. As stated above, O’Shaughnessy and Flanagan (1998) recognize that also unrelated M&As can result in downsizing in, for example, overhead activities. Thus, both related and unrelated M&As can result in downsizing, but it is only more likely that related M&As will result in downsizing than unrelated M&As. The next question that arises is: how is the effect of downsizing on M&A performance moderated by industry-relatedness? This question has not been answered in literature yet (Datta & Basuil, 2015; O’Shaughnessy and Flanagan, 1998).
In unrelated M&As, it has been found that organizational autonomy is more allowed (Chakrabarti & Mitchell, 2013), whereas in related M&As, there is a greater level of integration
17 effort (Howell 1970, Pablo 1994). Therefore, the synergistic effects of downsizing in unrelated M&As will be in laying-off workers in (managerial) overhead functions or supporting activities (O’Shaughnessy and Flanagan, 1998). Neilson (1990) argued that the secret in downsizing practice is to layoff personnel in overhead functions. The majority of costs summarized in table 2 does not apply to overhead functions and supporting functions.
Moreover, in the reasoning behind hypothesis 1, I outlined three processes that explain why it is reasonable that downsizing has a negative effect on firm performance. In short, I argued that downsizing will have a negative impact on firm performance because 1) the implicit knowledge hold in a firm’s human capital gets lost, 2) violation of the psychological contract increase in absenteeism, increased turnover and lower productivity, and 3) there will be several negative effects on the firm as a social community to will negatively affect firm performance. However, in my opinion, these effects are grafted on workforce reduction within the, to speak with Porter, ‘primary activities’ within the value chain (Porter, 1991). As stated above, downsizing in unrelated M&As will occur within the overhead functions or supporting activities. Therefore, I reason that the processes that negatively influence the effect of downsizing on firm performance will be less present in unrelated M&As. Consequently, I hypothesize:
Hypothesis 2: The effect of downsizing on M&A performance will be less negative for M&As in which the industry of the acquired company is unrelated to the industry of the acquiring company.
2.2.3 Liability of foreignness
Recently, cross-border M&As have become increasingly important (DePamphilis, 2015; Erel et al., 2012). Motives for cross-border M&As are: geographic and industrial diversification (Seth et al., 2002), accelerating growth (Graham et al., 2008), industry consolidation (DePamphilis, 2015), utilization of lower raw material and labour costs (DePamphilis, 2015), leveraging intangible assets (Eun et al, 1996; Morck & Yeung, 1991), minimizing tax liabilities (Zodrow, 2010), avoiding entry barriers (DePamphilis, 2015), fluctuating exchange rates, Boateng et al., 2014; Erel et al., 2012) and following the customers (DePamphilis, 2015). Although O’Shaughnessy and Flanagan (1998) found that the probability that a layoff will be announced was not changed in a border deal in comparison to a domestic US-deal, cross-border acquisitions are more complex and characterized by higher levels of uncertainty and
18 information asymmetry (Basuil & Datta, 2015; Collins et al. 2009; Johnson and Vahlne, 2009; Vermeulen & Barkema, 2001; Kogut & Singh, 1988; Shimizu et al., 2004; Kostova, 1999; Kostova & Zaheer, 1999; Zaheer 1995). The international business literature suggests that national borders matter because borders induce a liability of foreignness (McCarthy & Aalbers, 2016).
The liability of foreignness is generally defined as all of the additional costs that a firm operating in a foreign market incurs compared to a domestic firm (Zaheer, 1995) and is a fundamental assumption in the international business literature regarding multinational enterprises (McCarthy & Aalbers, 2016; Miller & Richards, 2002). It has been found that host country and home country environments have more influence on a foreign firm’s performance in the European Union than regional economic group membership (Miller & Richards, 2002; Qian et al., 2013). Thus, liability of foreignness is also present within the European Union. According to Zaheer (1995), the additional costs of the liability of foreignness can arise from four different sources: 1) directly from the distance and geography (for example travel and transportation costs), 2) the unfamiliarity of a multinational enterprise with and lack of roots in the local environment, 3) costs intrinsic to the host country (for example the lack of legitimacy of foreign countries) and 4) costs connected to the home-country (for example restrictions on sales to particular countries). Moreover, Friebel and Heinz (2014) recently found evidence that a foreign downsizing firm receives twice as much, and more negatively toned, attention as a domestic downsizing firm (Flanagan & O’Shaughnessy, 2005). Thus, it seems that foreignness increases the costs of downsizing through the liability of foreignness and a more negative reputation than domestic downsizing firms. Consequently, the M&A performance will be negatively influenced when a foreign firm acquires a domestic firm and lay-off workers. Therefore, it is hypothesized that:
Hypothesis 3: The effect of downsizing on M&A performance will be more negative for M&As in which the acquired company is located in a different country than the acquiring company.
2.3 Conceptual model
19
3. Methodology
3.1 Data collection and sample
The M&A deal sample was built in two steps: 1) two list of M&A deals was compiled in ThomsonOne, 2) these lists are linked to the Orbis database.
In the first step, I used the ThomsonOne database (following for example Chakrabarti & Mitchell 2016; Hoechle et al., 2009; Boateng et al., 2014; Schildt & Laamanen 2006; Grote & Rücker 2007; Grote & Umber 2006), to compile two lists of M&A deals: one list with US targets and one list with European targets. This database contains an exhaustive list of M&A deals and also enables the user to enter limitations in the search criteria for deals in order to find deals with more specific characteristics. I used four criteria: 1) date effective, 2) publicly listed status, 3) percentage of shares 4) country of target.
I only searched for deals with an announcement date between 2008 and 2013. The reason to focus (following for example Chakrabarti & Mitchell 2016 and McCarthy & Aalbers, 2016) on the announcement date, and not on the date effective of the deal, is that the mere announcement already has effects on the performance (Nixon et al., 2004; Guthrie & Datta, 2008). I choose for deals between 2008 and 2013 to improve the likelihood that I could find adequate descriptive details covering the transaction and the performance. In previous M&A research, this has been a valid reason to scope the time range on which the research focuses (for instance O’Shaughnessy & Flanagan, 1998). Moreover, I did not search for deals in 2014-2017 because in those deals, the downsizing and the effect of downsizing on M&A performance is
20 not visible yet. Another criterion in the building of the sample is that I only focused on publicly listed acquirers and targets. The reason for this criterion is that information on public firms is more widely available than information on private firms (Capron & Shen, 2007). A third criterion in the building of the sample is that I only included deals in which the acquirer bought 100% of the shares of the target. The fourth criterion in the building of the sample is that I only included deals in which the target is either an European firm or an US firm. Acquisitions of firms in other regions of the world are not in the scope of this paper.
With these search criteria, a list of 262 deals with an European target and a list of 822 deals with an US target in ThomsonOne were generated. For these deals, ThomsonOne generates columns of data with information per deal about, for instance, the advisor of the target and the Enterprise Value at Announcement. I will only use the following information, that is relevant for the analysis: Date of Announcement, the target name, the target macro industry, Acquirer Name, Acquirer macro industry. In ThomsonOne, it was not possible to gather information about the performance of the acquirer, the number of employees of the acquirer or combined firm and it also does not provide a country code for the acquirer or target.
Therefore, as a second step in the building of the sample, I linked the lists of acquirer that engaged in an M&A deal to a second database, Orbis, which provides information about the number of employees and performance of the companies. In Orbis, I used the ‘Batch search’ option to find information about the performance of the acquirer and the number of employees of the acquirer. For deals in which the target was vested in Europe, information of the acquirer about the number of employees and financial data (ROA, ROE) was available for 81 deals. For deals in which the target was vested in the US, information about finance (ROA, ROE) and number of employees was available for 177 deals. Because it is not possible in this research, in the case of multiple acquisitions, to attribute performance change to one deal or another, all repeat acquirers (acquirers that did more than 1 acquisition between 2008 and 2013) were excluded from the sample (McCarthy & Aalbers, 2016). This reduced the initial sample to 75 M&As in which the target was vested in Europe and 110 in the US. In sum, a total of 185 deals will be included in the analysis. A list of the deals included in this research is provided in appendix 1a and 1b.
However, not for every deal all the data was available for every year. Thus the sample size varies over the years. In the table 3, the sample sizes and frequency of downsizers in the different years to the M&A announcement year are given.
21 Table 3. Downsizing frequencies.
Year to M&A announcement Frequency of downsizers Frequency of no downsizers Total frequency T - 1 22 (15.8 %) 117 (84.2 %) 139 (100 %) T 13 (9.0 %) 132 (91.0 %) 145 (100 %) T + 1 24 (17.6 %) 112 (82.4 %) 136 (100 %) T + 2 15 (12.8 %) 102 (87.2 %) 117 (100 %)
Because I only selected M&A deals in which it was possible to obtain data on downsizing and financial data in at least one of the years t-1 to t+3, there is a risk of exclusion bias, which can lead to flawed conclusions (Certo et al., 2015). Below, I elaborate on how this problem is handled here.
In some studies in M&A research, it is however not visible how the problem of exclusion bias is handled (e.g. Krishnan & Park, 2002; O’Shaughnessy & Flanagan, 1998). In one study, for example, which is comparable to this study and also analysed the effect of downsizing on M&A performance, the mean revenues between firms excluded from the sample (due to unavailability of data on employees and lack of performance data) and firms included in the sample were compared using t-tests (Krishnan et al., 2007). However, Krishnan et al. (2007) remain unclear about how this has been done, because, according to their article, there was no financial data available.
In the sample used here, for the transaction excluded from the sample, no data was available on the number of employees and/or financials, so also no data on the mean revenues. However, the enterprise value at announcement was available. The idea of using the mean revenues is the same as the enterprise value (which is partly influenced by the mean revenues) and therefore, in order to make sure that there is no exclusion bias, I compared the enterprise value of firms included in the final sample and firms excluded in the final sample using T-tests. For European targets, there was no significant difference between the cases included and excluded from the final sample (t(228 = 1.944, p = .053), see table 4. For the US targets, there was a significant difference between the enterprise value of the cases included and the enterprise value of the
22 cases excluded (t(828) = 4.055, p < .001), see table 4. This bias is explained by the fact that companies with a greater enterprise value, are larger in size and have more accessible data available (DePamphilis, 2015). Because I excluded cases in which I could not obtain sufficient data on number of employees and/or financials, this explains the bias towards companies with a bigger enterprise value.
3.2 Dependent variable
The dependent variable is Post-acquisition performance. In paragraph 2.1.2, post-acquisition performance was described as a quantitative measure of the performance that can be measured at different points of time after the M&A event, at the acquiring firm, in an (cross-border) acquisition with European and/or US targets, regardless of the industry.
In their literature review regarding accounting-based measures in M&A performance, Thanos & Papadakis (2012) point out Return on Assets (ROA) is by far the most widely used accounting ratio in M&A literature (e.g. Zollo & Meier, 2008). This measure is less influenced by potential biases and is therefore considered to be the best measure (Thanos & Papadakis, 2012).
Table 4. Exclusion bias.
Sample t-value
European targets included 3.371**
European targets excluded 1.944
US targets included 7.339***
US targets excluded 4.055***
* p <.05 ** p <.01 *** p < .001
23 Consistent with the majority of studies on M&A performance, I will also use the ROA. Moreover, following the points and suggestions made by Meglio & Risberg (2010) about the lack of attention for the chosen timeframe in which the performance was measured, I measure the performance in different years (year t to t+3) and do not choose beforehand to measure for only a specific year after the deal. The measurement of M&A performance over a time horizon has been done before by, for instance, Zollo & Meier (2008) and Hitt et al. (1998), who measured the performance in year t-1 to t+3 respectively in year t-3 to t+3. As shown by Meglio & Risberg (2011), most articles do not specifically account for industry and the great majority of previous studies measure M&A performance at the acquiring firm. This study will follow this line in research.
Thus, in line with previous research, post-acquisition performance is quantitatively measured through ROA at different point of time from year t to t+3 after the M&A event. at the acquiring firm, in an (cross-border) acquisition with European and/or US targets, regardless of the industry.
3.3 Independent variable
The independent variable is downsizing. In paragraph 2.1.1.1, I argued that it is not easy to determine whether a firm has engaged in downsizing activities and that this particular topic has received not enough attention in literature. In the introduction, downsizing was defined as the planned set of organizational policies and practices aimed at workforce reduction with the goal of improving firm performance (Datta et al., 2010, Cameron, 1994; Cameron et al., 1993; Cascio, 1993; DeWitt, 1998; Freeman & Cameron, 1993) and in previous studies, downsizing is operationalized as the percentage reduction in the number of employees in the consolidated organization after the acquisition (e.g. Lehto & Böckerman, 2008; Krishnan et al., 2007; Conyon et al., 2002; Osterman, 2000).
Following for example Said et al. (2007), Cascio & Young (2003) and Cascio, (1997), publicly available data (from orbis) will be used to measure the employment level change. However, in paragraph 2.1.1.1, it has also been argued that this measure is not very accurate. Consistent with Lee & Hong (2016) and Griffin (2003), I argued that annual reports are a rich source of reliable data. To gather information from the annual reports, it has been found by for instance Hajek & Henriques (2017) and Dyck et al. (2010), that the technique of data mining is a very accurate method (even better than in-depth expert knowledge). This technique has not yet been applied in research considering downsizing, but it has been applied in other fields of research:
24 Henriques (2017) and Dyck et al. (2010) used this technique, for instance, to detect financial statement fraud, but it has also been applied in CSR-literature (Liew et al., 2014; Saha & Nabareseh, 2015; Garechana et al., 2017).
Cameron (1994a) provide the following list of words that managers have used to describe downsizing activities in organizations: building-down, compressing, consolidating,
contracting, declining, de-hiring, de-massing, de-recruiting, dismantling, downshifting, functionalizing, leaning-up, ratcheting-down, rationalizing, reallocating, reassigning, reduction-in-force, re-engineering, renewing, reorganizing, reshaping, resizing, restructuring, retrenching, revitalizing, slimming, slivering, and streamlining. The stems of these words will
be used in a datamining program to analyse the annual reports. Gandolfi and Hansson (2011), more recently, describe, in their literature review on the causes and consequences of downsizing, a similar, but less exhaustive list of words to describe downsizing.
In the case that the workforce has been reduced with >5% and there is also a significant increase in the usage of (synonyms of) the word ‘downsizing’ in the annual report, it can be assumed that the company has reduced its workforce. The 5% hallmark regarding workforce reduction is consistent with previous research (Guthrie & Datta, 2008; Ahmadjian & Robinson, 2001; Cascio et al., 1997) and it has been assumed in previous research that a 5 % reduction represents a significant event and likely indicates an intentional reduction in employees (Guthrie & Datta, 2008; Freeman & Cameron, 1993).
The threshold for ‘significant increase in words that indicate downsizing’ is also 5%. By my best knowledge, this is the first time this variable has been used and therefore no other threshold is available. I tried different threshold (1%, 5% and 10%), but the cases in which I could conclude that there has been downsized did not differ much between the different thresholds. Therefore, in my opinion, it is most convenient to embrace the 5% threshold that is also used for the workforce reduction. Brauer & Laamanen (2014) tried different thresholds for the workforce reduction (3%, 5% and 10%), and concluded that these threshold rules resulted in no substantive change in study results. Therefore, they decided to use the 5% threshold. The same reasoning is applied here for the increase in words. I tried different thresholds, but these resulted in no substantive change in study results, so I decided to use the 5%.
I will thus use the information gathered by datamining of annual reports in combination with the archival data on workforce reduction. In that way, a combination of an 5% workforce reduction with an 5% increase in words that indicate downsizing, will be used as a measure for
25 downsizing. This is a new measure that is on one hand, regarding the 5% workforce reduction from archival data, in line with previous downsizing research (Guthrie & Datta, 2008; Ahmadjian & Robinson, 2001; Cascio et al., 1997) and on the other hand, regarding the 5% increase in words indicating downsizing in the annual reports, in line with techniques that have been applied in other fields of research such as financial statement fraud (Kirkos et al., 2007; Hajek & Henriques, 2017) and CSR-literature (Liew et al., 2014; Saha & Nabareseh, 2015; Garechana et al., 2017).
Since the organisational impact of mergers on the workforce may not be felt immediately or may be apparent before the deal announcement, in line with Conyon et al. (2002), continuous substitution of the basic model will be used to study the lagged effects of downsizing for years t-1 to t+2.
For the downsizing variable, consistent with previous research (e.g. Guthrie & Datta, 2008), a dichotomous measure of downsizing will be used, because this is easier to interpret than a continuous measure, which reflects decreases and increases in the workforce (Guthrie & Datta, 2008; Ahmadjian & Robinson, 2001). Therefore I label firms that meet the criteria above with ‘1’ and those that did not with ‘0’.
3.4 Moderator variables
In this study, two moderator variables are considered: industry relatedness and foreignness. These variables are less ambiguous than the dependent and independent variables and thus require less explanation.
Industry relatedness is assessed through the Standard Industrial Classification code of the acquirer and the target, which is available in the ThomsonOne database. If the acquirer and the target have the same SIC-code, the industry relatedness variable is given the value ‘1’, and if not then the industry relatedness variable is given the value ‘0’. This measurement of
industry-relatedness has previously been employed by, for example, Basuil & Datta (2015). The moderator variable of foreignness is based on the country codes that were extracted from the Orbis database. If the acquirer and the target have a different country code, which means that it is a cross-border M&A, the value ‘1’ is given to this variable and if the acquirer and target have the same country code, the value ‘0’ is given to this variable. This is consistent with, for instance, McCarthy & Aalbers (2016).
26
3.5 Control variables
Three control variables are taken into account in the analysis: ‘Target country of origin’, ‘Geographical Distance’ and ‘Size of Target’. One of the environmental factors in the model that affects downsizing is the institutional environment (figure 1, box 1). It has been argued by Aguilera & Dencker (2004) that, based on contingency theory, firm strategies and structures and the incidence of layoffs will vary among environmental contexts. (Cascio, 2009). On a supranational level, the European Union made legislation especially directed at labour protection in the case of an M&A. According to directive 2001/23/EG, which works for companies in all countries of the European Union, a company is not allowed to lay-off workers after an M&A transaction because of the transaction (Veldmaat & Assendelft de Coningh, 2012; Roukema, 2016). Considering that the sample consists of European targets and US targets, I control for the target’s geographic origin. The location of the acquired firm (US or Europe) is visible in the different lists generated from the databases. In order control for the effect of the location of the acquired firm, I made a dummy variable and I gave US-based targets the value ‘1’ and Europe-based targets the value ‘0’. This is in line with previous research on the effects of mergers on company employment in the US and Europe, conducted by Gugler & Yurtoglu (2004).
In previous research, it has been found that geographic distance between the acquirer and target has an impact on M&A performance due to transaction costs (McCarthy & Aalbers, 2016, Green & Cromley, 1984; Schildt & Laamanen, 2006). Transaction costs are the costs of finding and negotiating with an appropriate partner, and the costs of monitoring the performance of the partner firm and one of the underlying mechanisms is opportunism (Brouthers, 2012; Hill, 1990). According to transaction cost literature, geographic distance impacts firm performance because 1) it increases the monitoring and agency costs and 2) it reduces the benefits of soft information (McCarthy & Aalbers, 2016). I will therefore control for geographic distance. In line with, for example, Chakrabarti & Mitchell (2016), the geographical distance is measured in terms of the distance between the city of the headquarter of the the acquirer and the city of the headquarter of the target. To measure this, the cities of the headquarters are extracted from the Orbis database. Subsequently, the distance between the two cities is calculated via
https://www.mapdevelopers.com/distance_from_to.php and it is measured in kilometres. According to DePamphilis (2015), M&A performance is situational and also depends partly on the size of the target. Shrivastava (1986) states that the larger the size of the target, the more diverse and intensive integration problems tend to be. Therefore, I will control for the size of
27 the target, which is measured as the Enterprise value at the announcement date. The Enterprise value is the sum of the market value of a firm’s equity, preffered shares, debt, and nonconrolling interest less total cash and cash equivalents and is considered as an accurate representation of firm value (DePamphilis, 2015).
The variables that are explained above are summarized in table 5. Table 5.
Summary of the variables.
Variable name Variable description
Independent variable Downsizing 5% reduction in workforce in combination with a 5% increase in frequency of words indicating downsizing in the annual report of the acquirer, measured at t-1, t, t+1 and t+2. If this is the case the variable receives value ‘1’, otherwise value ‘0’.
Dependent variable M&A performance ROA at the acquirer in year t, t+1, t+2 and t+3.
Moderator variables Industry-relatedness SIC codes, extracted from
ThomsonOne, are either the same for the acquirer and target (value = 1) or not the same (value = 0).
Foreignness Country codes, extracted from Orbis, are either the same for the acquirer and target (value = 0) or not the same (value = 1).
Control variables Location of the target The target is either vested in the US (value = 1) or in Europe (value = 0), extracted from ThomsonOne. Distance Geographical distance, measured in
kilometer, between the HQ of the acquirer and target via
Mapdevelopers.com.
Size Target Target Enterprise value at the announcement date
28
3.5 Data analysis strategy
The dependent variable is a ratio variable (ROA). The independent and moderator variables are all nonmetric, more specific dummy-variables, except for distance (which is a metric variable). Considering these measurement scales, considering the tests of assumptions below and considering the research question, the most appropriate data analysis techniques is regression analysis. In table 6, the descriptive statistics for the variables and the correlations are given. There are no significant correlations visible between the downsizing variables and the ROA variables. Since downsizing is a dichotomous variable and ROA is a continuous variable, there is no lineair relationship visible in the correlation matrix and the regression analysis will be used to analyse the effects. The data is standardized using z-scores.
34
Table 6. Descriptive statistics.
S.N. Variables Mean s.dev 1 2 3 4 5 6 7 8 9 10 11 12 13
1 Downsizing t-1 .16 .366 1 2 Downsizing t .09 .287 -.010 1 3 Downsizing t+1 .14 .343 -.119 -.044 1 4 Downsizing t+2 .18 .383 -.082 .119 .220* 1 5 ROA t 4.2757 12.389 -.089 .141 -.015 .015 1 6 ROA t+1 2.9816 12.579 -.138 .034 -.120 .086 .518** 1 7 ROA t+2 4.3316 9.355 -.103 -.067 -.146 -.032 .456** .592** 1 8 ROA t+3 5.1072 8.793 -.096 .040 -.097 -.101 .383** .454** .620** 1 9 Domestic .6324 .483 .196* -.003 -.018 .067 -.109 -.034 -.019 -.161* 1 10 Industry relatedness .2865 .453 -.014 .113 .008 -.010 .059 .023 -.045 -.109 .062 1 11 Country target 1.4054 .492 -.126 .012 -.094 -.199* -.125 -.092 -.183* -.201 -.010 -.158* 1 12 Distance in Km (Log) 3276.810 3.116 -.089 .028 .096 .201** .114 .144 .180 .266** -.544** -.096 -.462** 1 13 Size 1876.497 .818 -.126 .041 -.066 -.109 .066 .211** .076 .018 .008 .146 -.088 -.007 1 * Correlation is significant at the .05 level
35
3.6 Assumptions regression analysis
In order to conduct a regression analysis, several assumptions have to be met (Field, 2013): 1) normal distribution of the variables 2) metrically scaled data 3) linear relationship between the dependent and independent variable 4) no multicollinearity and 5) homoscedastisity. The outcome of this assessment is summarized in table 7.
The assumption of normality of the dependent variable, is met for the ROA at the different time points (ROA_t, ROA_t1, ROA_t2 and ROA_t3), because for every ROA the skewness/standard error skewness < 2 and Kurtosis/Standard error kurtosis < 2, and therefore it can be concluded that the ROA is normally distributed. An output of the frequencies and histograms is provided for in appendix 2. Moreover, regarding the second assumption, the dependent variable is metrically scaled while the the independent variables are either also metrically scaled or dummy variables. Thus, the first two assumptions are met.
The third assumption, linearity between the dependent and the independent variable, should be assessed by using a scattorplot and assessing whether any patterns can be discovered (Field, 2013). However, since the independent variable that is used, is a dummy variable with only two values, the assumption of linearity cannot be assessed. The assumption of linearity can only be assessed if there are at least three different values in the variable (Field, 2013).
The fourth assumption, multicollinearity, is assessed through an analysis of the tolerance levels in tables 3 of Appendix 3a, 3b, 3c and 3d. According to Field (2013), multicollinearity becomes problematic when the value is below .10. As can be seen in tables 3 of each appendix 3a-3d, no tolerance level is lower than the .10 threshold. Thus, this assumption is also met.
The fifth assumption, the assumption of homogeinity of regression slopes, is assessed by an analysis of the scatterplots in figures 1 of appendix 3a-3d. As can be seen, there is no pattern visible in the residuals. Therefore, the variance can be assumed to be constant and the assumption of homoscedasticity is met (Field, 2013; Hair et al., 2014).
36 Table 7.
Summary of assumptions regression analysis.
Technique Assumption Is the assumption violated?
Regression-analysis Normal distribution Not violated Metrically scaled variables Not violated Linearity between the
dependent and independent variable
N/A
Multicollinearity Not violated Homoscedasticity Not violated
4. Results
In this chapter, I report the results of the analysis. I have split this chapter into paragraphs, according to the different points in time.
4.1 Regression analysis
I conducted analyses for each of the different dependent variables: ROA t, ROA t+1, ROA t+2 and ROA t+3. The results are described below. Table 12, in paragraph 4.3 gives an overview of the hypotheses supported/not supported by the analysis. The hypotheses that are supported are also visualized below in figures 3-9.
4.2.1 Effect of downsizing on performance in the year of the deal (t)
As can be seen in tables 1a-1g of appendix 3a, the models in the first regression analysis show a low, but given the complexity (Field, 2012), acceptable level of explained variance (Adj. R2 = .045 (model 1); Adj. R2 = .054 (model 2); Adj. R2 = .042 (model 3); Adj. R2 = .042 (model 4); Adj. R2 = .069 (model 5); Adj. R2 = .044 (model 6); Adj. R2 = .066 (model 7)). Moreover, the models are, except for model 6, also all significant, as can be seen in tables 2a-2g of appendix 3a (F(3, 171) = 3.740, p < .05 (model 1), F(4, 128) = 2.871. p < .05 (model 2), (F(4, 170) = 2.930, p < .05 (model 3), F(4, 170) = 2.916, p < .05 (model 4), F(6, 126) = 2.621, p < .05 (model 5), (F(6, 126) = 2.020, p = .068 (model 6), F(8, 124) = 2.175, p < .05 (model 7)).
37 In table 8, that is shown below, none of the coefficients is significant in model 1-4 at p < .05. Only the control variable Target Size is significant. In model 5, when the interaction effect Downsizing t-1 * Domestic deal is added, the variable downsizing t-1 becomes significant (b = .-480, p < .05. In this model, the interaction effect Downsizing t-1 * Domestic deal is also significant (b = 1.300, p < .05). The effect of downsizing in year t-1 on performance in year t thus only becomes significant, when the moderator Downsizing t-1 * Domestic deal, which is also significant, is added. The coefficient of the interaction term is positive. Therefore, it can be concluded that hypothesis 3 is supported. The interaction effect is visualized in figure 3 below.
Table 8.
Results regression analysis with ROA in year t as dependent variable.
Model Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
Independent variables
Downsizing t-1 -.082 -.480* -.037 -.466*
Domestic deal -.070 -.093 -.102
Industry-relatedness .052 .060 .070
Moderators
Downsizing t-1 * Domestic deal 1.300* 1.486 *
Downsizing t-1 * Industry-relatedness -.424 -.691 Control variables Country Target -.082 -/173 -.106 -.074 -.186† -.172 -.190 Geographical Distance .073 -.005 .024 .071 -.027 .006 -.029* Target size .223** .221 ** .214** .228** .213* .227* .220 Intercept .007 -.016 .007 .006 -.193 .001 -.190 Observations 174 132 174 174 149 161 132 F-Statistics 3.740 2.871 2.930 1.916 2.848 2.756 2.175 R-squared .062 .082 .064 .064 .073 .066 .123 R-squared change .02 -.18 0 .009 -.007 .057 † p < .10 * p < .05 **p < .01 *** p < .001 Figure 3.
