Factors of mergers and acquisitions’ timing

(1)

Factors of Mergers and Acquisitions’ Timing

Master Thesis

Student: Anastasia Gavrilova /Student № 10603247/

Supervisor: Bernardo Silveira Barbosa Correia Lima MSc Business Studies, Strategy Track University of Amsterdam, Amsterdam Business School

Final version Date: 31/07/2014

(2)

Abstract

The paper investigates the timing and motives for mergers and acquisitions’ in manufacturing industries. Based on the previous research, the model includes the factors on multiple levels: previous acquisition experience, firm performance and industry conditions. Stemming from behavioural theory of the firm, an assumption is made that the decision-making differs significantly in the domain of large deals, so they are analysed separately as well. The study aims at finding the factors that are most important for the decision to do M&A across different deal sizes and industries. The data represents a panel of US firms from 1985 to 2005 and is analysed using event history techniques. Results indicate that across the whole sample experience accumulation and high performance of the previous deals, abundance of financial slack and munificent environment significantly increase the chances of making new acquisitions. For the large deals, however, only slack and munificence make new acquisitions more likely. Also, considerable differences in M&A motives are found across industries.

(4)

Introduction

The market of corporate control is growing at a rate of almost 10% per year and will soon reach pre-crisis volume (the Economic Times, 2013). Mergers and acquisitions have been a popular research topic mainly in finance and management fields during past decades and were approached from different theoretical perspectives. Nevertheless, the results were often contradictory and the role acquisitions play in the companies’ strategies still remains underexplored, especially considering that on average acquirers lose money on M&A (Capron & Pistre, 2002). Undoubtedly, there is no “one size fits it all” explanation for M&A incentives and performance drivers, but certain robust patterns may be found for different types of deals.

As most of the deals destroy value, researches tried to explain why and when firms engage in such risky enterprises and why the efforts do not pay off, but the phenomenon still remains puzzling (Hayward, 2002). It goes without saying that strategic decisions such as M&A are very complex, so a lot of factors are taken into consideration by managers. However, the existing literature does not offer much research on how these factors interact, even though this would shed the light on the reasons for firms’ behavior.

There are several theories that are used to analyze M&A on different levels. One is organizational learning theory that states that previous experience has an influence on the subsequent activities. This happens through the development of the routines within the firm. Depending on how applicable previous experience is, they can either facilitate or hinder future activities, thus having positive or negative influence on firm’s performance. Therefore, M&A are seen from strategic perspective, elaborating on learning opportunities and knowledge accumulation from the deals and helping to analyze strategic decisions over time. The theory also emphasizes path-dependency of managers’ choices and creates a dynamic perspective of analysis. The theory of organizational learning is used to predict the timing and type of subsequent acquisitions too, as the developed routines influence not 4

(5)

only the success of these deals but the decisions to engage or not engage in them per se. Next important factor of M&A timing is firm-level conditions. On this level behavioral theory of the firm comes into play and states that M&A result from the company’s desire either to solve particular problem and increase performance or to seek new experience. It was shown by Iyer & Miller (2008) that these factors are taken into account by managers while making these decisions. Moreover, many authors studied the role of experience similarity, (Hayward, 2002; Schijven & Hitt, 2012)) and its influence on subsequent deals performance, but the relevance of such experience and its timing were left out. Last but not least, industry conditions seem to be of high importance for M&A. It is a well-known fact that M&A activity is highly dependent on the state the economy and acquirer’s industry are in, but this, too, is not yet well-researched. Overall, different levels of analysis were rarely studied together, even though these are the main factors of M&A timing and are all taken into consideration during decision-making. Current study incorporates deal-, firm-, and industry-level factors to build a comprehensive model predicting timing of acquisitions. A distinction between overall motives and those specific for the large deals is made, for the latter play a special role in firms’ strategy. Size is often used as a control variable because large deals are treated differently by the company than smaller ones (Ellis, Reus, Lamont, & Ranft, 2011).

The paper will answer the following questions. How do different factors interact when put together? How does decision-making process for large deals differ from the overall pattern? Do different industries vary in terms of strategic decision-making? The study aims not only to disentangle the effects of these factors, but also to understand the logic behind managers’ decisions.

The paper is organized as follows. First, Theory and Hypotheses part describes the state-of-the-art of the M&A research on each of the three levels, presents the research gap to be addressed and outlines the hypotheses. Second, the Methods section gives an overview of sampling criteria, explains the methods of data analysis and models used in the study. Third, the outcomes of the study are 5

(6)

presented under the Results. Finally, the Discussion section elaborates on contributions and limitations of the current study and suggests the directions for further research.

(7)

Theory and Hypotheses

The current study integrates different perspectives on M&A timing and develops a comprehensive multi-level framework, which includes deal-, firm- and industry-level factors. Organizational phenomena often require multi-level frameworks, as Morgeson & Hofmann (1999) note. Modern businesses are highly complex organizations and when managers make strategic decisions, many factors are taken into account and various mechanisms come into play, so variables from different levels are needed to fully understand how and why particular choices are made. The variables included in the model are taken from the organizational frameworks that are commonly used to analyze strategic decision processes and have received empirical support.

According to the organizational learning theory (J. J. Haleblian, Kim, & Rajagopalan, 2006), previous experiences have an influence on the future choices of the firm, because companies learn from what they have done. The main factors here are accumulated acquisition experience and the performance of previous M&A. They denote the deal level. Another set of works uses behavioral theory of the firm to see how conditions at the acquiring company determine the decision to engage in acquisition. This denotes firm level. Finally, several authors look at how profitability and dynamics of the industry influence the intentions to acquire. This approach is used as the industry-level of the framework.

The relationship between these factors and the intention to make acquisitions is very much dependent on the type of deal, since different mechanisms come into play. In particular, size of the deal is linked to the type of strategic goals the company is pursuing. Therefore, in the current study the distinction is made between overall significant factors triggering M&A activity and those attributable to large deals only. Somewhat similar approach was used by Ellis et al. (2011), who showed that experiences of large and small deals are qualitatively different. Large deals require

(8)

more sophisticated integration and managers’ effort in order to create value for the business (Ellis et al., 2011). Nevertheless, the uniqueness of large deals was previously studied only in terms of its influence on acquisition performance, not timing of and intentions behind those strategic actions. Current study aims at making one step further and incorporating different factors alongside deal-level to analyze the decision-making peculiarities in the context of different types of acquisitions. The factors and hypotheses are discussed in details further.

1. Deal level

According to the organizational learning theory (J. J. Haleblian et al., 2006), previous experiences influence the choices a company makes in the future. Haleblian, Devers, McNamara, Carpenter, & Davison (2009) note that acquisitions are a perfect context for studying learning mechanisms, because they are discrete repetitive and strategically significant events. There are two key factors moderating this relationship. First, experience accumulates within the company in the form of routines, so the more experience the company has the more likely it will repeat the familiar activity. Second, favorable experiences are more likely to be retained as routines and repeated. Therefore, deal level is represented by these two factors, which are explained in detail further.

Experience: developed routines lead to repetition

As the company becomes more proficient in a particular task, it tends to repeat this task more frequently. This happens because the proper routines have already been established, so the task requires fewer resources and is associated with less uncertainty. This effect has been present in multiple studies. Haleblian et al. (2006) analyze the decisions to make acquisitions in banking industry and find that as experience is accumulated, bank tends to make more deals. Shimizu (2007) comes to the same conclusion for the decisions to divest poorly performing units. The author shows that prior divestiture experience increases the chances of a new spin-off in the future. He argues that this happens due to the fact that the proper processes are already in place and the choices made

(9)

previously by the firm may be associated with less uncertainty and thus be preferred to the new options.

Next important finding is that the relationship between previous experience and current strategic choices is moderated by the extent to which the prior experience and current task are similar. The similarity can be measured in many different ways. For instance, Amburgey and Miner (1992) examine the type of the deal: horizontal, vertical, product extension or conglomerate deal. They demonstrate that the type of previous acquisition raises the probability of engaging in similar experience, because the firm develops necessary competencies. The authors call this phenomenon of repeating previous strategic choices a “repetitive momentum”.

If we look deeper at how the routines are established, the distinction between small and large deals becomes important. Routines can be easily developed for relatively small acquisitions only, because large acquisitions tend to be rare and unique events (Barkema & Schijven, 2008; Finkelstein & Haleblian, 2002). Large deals are more difficult to arrange and they are unlikely to be comparable to one another. Ellis et all (2011) find that experience from small deals does not help to make large ones more profitable and previous large deals do not increase the performance of subsequent large deals too. So the acquirers who have more experience with large deals have no reason to be more prone to seeking large deals again.

H1a: Previous experience increases the overall probability of acquisitions, but for the large deals

this effect is smaller.

Prior deal performance: favorable experiences are more likely to be retained

The second factor of experience accumulation is whether it was favorable or not. The importance of performance feedback in learning stems from the idea that firms adapt to the environment (Cyert & March, 1963) striving to perform the activities that are rewarded. M&A are not an exception: if 9

(10)

managers see them as a wise strategic move that usually leads to stronger performance, they are more likely to engage in the deals. Haleblian et al. (2006) find that likelihood of subsequent acquisition is positively related to the performance of previous deal. In part, having made successful acquisitions in the past, management has an argument to justify a new attempt for the stakeholders. However, for large deals this effect will not necessarily hold, for these are big strategic decisions that significantly change the way business operates and that are not very comparable to one another. Thus, large deals are expected to be less influenced by this factor.

H1b: Positive performance of previous deals increases the overall probability of acquisitions, but

for the large deals this effect is smaller.

If existent, this effect might have an important consequence: managers repeat rewarded behaviors, but are less likely to persist in punished behaviors (as in Haleblian et al. (2006)). However, as Hayward (2002) shows, this impedes learning, as more inferences from experience are made in case of small losses. He provides evidence that the performance of past deals influences the process of experience accumulation. In this regard, he finds that slightly negative performance in past deals is positively related to the performance of subsequent acquisitions. Possible explanation is that small losses stimulate managers to engage in search of available solutions, while positive performance can make them overlook problems. Positive outcomes also give birth to the heuristic logic (Dyer, Kale, & Singh, 2004): managers stick to ideas that brought success in the past even if the context has changed or if prior high performance was mainly the result of luck. At the same time, managers may try to attribute bigger losses to the circumstances and not analyze them deeply. Therefore, small losses stimulate fruitful inferences and lead to useful experience accumulation most, but these kinds of deals are unlikely to be repeated. This may be one of the reasons for M&A overall poor performance, which makes this topic so important to research.

(11)

2. Firm-level conditions: low risk for survival stimulates risk-seeking behaviors

Many authors use firm-level conditions as predictors of strategic decisions like acquisitions and divestitures. Firm-level factors are often viewed from the perspective of behavioral theory of the firm (Cyert & March, 1963). It states that managers seek changes in the company in two cases: they either have excess resources to engage in experiments (slack search) or need to solve a particular problem and increase performance (problemistic search). These ideas have already found support in M&A research. Iyer and Miller (2008) show that up to a certain point companies tend to accumulate slack and use it to support ongoing operations, while after this threshold excessive slack increases the probability of acquisition. Indeed, having abundant resources the company may consider significant changes such as large acquisitions (e.g. takeover of a major rival). Furthermore, the more slack a firm accumulates, the higher the probability of seeking qualitatively new experience rather than engaging in similar deals will be. Slack represents the funds that have relatively high liquidity and are readily available for reinvesting (especially as regards unabsorbed slack which is calculated as current assents divided by current liabilities), so it serves a good proxy for company’s spare resources at hand.

H2a: Abundant slack increases the probability of large deals to a greater extent than smaller ones.

In the case of problemistic search, if business is underperforming, managers are willing to try some risky activities that they would usually avoid. Acquisition is a perfect example of a problemistic search activity, so it is expected to be done more by underperforming businesses. However, Iyer and Miller (2008) find no evidence that companies engage in more acquisitions when their performance falls and the reasons for that are not very well-investigated. One possible explanation is that the problemistic search tends to be local for poorly performing businesses, as radical changes put firm’s survival at a greater risk. So in order “not to make things worse”, a poorly performing company is 11

(12)

expected to avoid large acquisitions, which might cause more distress. So while small deals can be an instrument of improving performance, large ones do not seem to be an option. Therefore,

H2b: As performance falls below aspirations, the probability of large deals decreases.

To sum up, strategic decisions such as M&A are highly dependent on the firm performance and available resources. Being more complex and demanding than smaller ones, large deals require more funds and are not expected to be used as potential solution to existing problems.

3. Industry conditions: M&A as a means of growth

Industry conditions represent another important, but not very well-researched factor in M&A activity. For example, Park (2003) finds that the type of acquisition depends on the acquirer’s performance and target’s industry profitability, thus integrating firm- and industry-level factors.

Going back to the behavioral theory of the firm, Iyer & Miller (2008) claim that it is important to distinguish between opportunity-driven and problem-driven strategic actions. In terms of the firm level these are slack search and problemistic search respectively. The same logic is applicable on the industry-level of analysis with munificence and industry growth variables.

Munificence (Boyd, 1995; Dess & Beard, 1984) denotes the abundance of resources available for the firms in the industry and seems extremely relevant for the purposes of M&A analysis: following the logic presented in the previous section, the availability of resources may encourage companies to take on greater risk of performing large acquisitions, because these are very demanding actions.

H3a: Industry munificence increases the probability of large deals to a greater extent than smaller ones.

(13)

Industry growth is the main factor determining the opportunities of a business within its industry. Acquisitions are often viewed as a means of fostering growth, so in the environments where opportunities to expand are limited this tool is expected to be used more frequently. It has been previously shown in the literature that M&A are the way of accelerating growth for businesses (Slatter, 1984).

H3b: High industry growth decreases the probability of all acquisitions.

It is important to disentangle effects that are reflected in the hypotheses H3a and H3b. While munificence determines the resources available to the firm and that are necessary for making acquisitions, industry growth shows to what extend firm has an opportunity to spend these resources on organic growth of sales and other operations. If organic growth is restricted by industry conditions, the company is more likely to seek other ways to grow, e.g. through acquiring other businesses.

The logic explained above seems quite uniform for all kinds of businesses. However, different industries might vary considerably in terms of business models, operation practices, types of customers, the stage of the industry lifecycle, etc. Therefore, in each particular industry the set of the factors that play a role in decision-making is not the same. This idea has an important practical implication: even if universal laws and tendencies in business exist and are certainly worth studying, the context the particular firm operates in should be allowed for.

H3c: Factors of acquisitions' timing differ across industries.

Presented model gives an explanation why firms engage in different types of acquisitions, even though M&A deals are known to destroy value. It also lets to disentangle different effects in

(14)

managers’ decision-making process and potentially explain the contradictory results of previous studies.

(15)

Methods

The sample consists of panel data on US companies from 1985 till 20051. First, the “total population” of manufacturing firms with yearly financial data was obtained from the Compustat database. For the further analysis three industries were selected: Chemicals and Allied Products (SIC 2800), Industrial and Commercial Machinery and Computer Equipment (SIC 3500) and Electronic and other Electrical Equipment and Components, except Computer Equipment (SIC 3600). These are the industries that account for 14%, 16% and 16% of all deals in the initial sample respectively. Second, the information about acquisitions in this period was exported from Thomson One containing announcement date, completion date, acquirer’s and target’s name, their SIC-codes, deal value and status, acquired and final percentages of ownership. Third, the data on deal performance were uploaded from Eventus module of CRSP database based on the deal announcement date and acquirer’s CUSIP. They were then added to the list of deals from Thomson One. Fourth, the dates of fiscal year end from Compustat were added to this list of deals and the information on the deals was collapsed to the fiscal year level to match the financial data from Compustat. Thus, the number of deals of each type, average deal values, returns, etc. were computed for each year and these deal data were added to Compustat sample of firms. Finally, environmental measures from NBER-CES Manufacturing Industry Database were uploaded to this sample according to each company’s 3-digit SIC code. This main database contains 50,709 observations; each one is uniquely identified by CUSIP-fiscal year combination.

Only completed deals that resulted in at least 50% ownership of the target firm were analyzed (they are called “eligible” later in the text), because this enables decision-making for the acquirer and leads to the experience accumulation. This approach is common in the literature (e.g. Haleblian & Finkelstein (1999), Iyer & Miller (2008)). Also, each firm enters the sample when the first deal

1_{The timeframe was chosen due to the data availability: the Thomson One database starts in 1985 and the environmental} measures were available till 2005.

15

(16)

satisfying these two conditions was made, because previous deal performance is one of the explanatory variables in the model. This approach yielded 11,950 observations. Deals that satisfy these two criteria and have the value of at least 100mln were considered large (Ellis et al., 2011). To sum up, in this study two types of deals are compared: overall sample of deals satisfying two criteria (completed and leading of at least 50% ownership) and its subsample of acquisitions equal or larger than 100mln. Thus, the comparison is not purely between "small" and "large" deals, but rather between "all" and "large" ones. This is due to the fact that deal value is often unknown, so it is not possible to classify all deals while leaving the unknowns out would lead to a loss of a lot of observations. The correlation between the deal value and company size is only 0.14, so theoretically the differences in M&A motives cannot be attributed to the big companies' peculiarities. The unit of observation is CUSIP-year, which is typical for panel data on companies.

Dependent variable

There are two types of models in this study, which allows the comparison between the whole sample and the subsample of large deals. Thus, two different dependent variables are used: the number of

deals per year and the number of deals per year exceeding 100mln. Independent variables

Deal performance. There are two main ways of measuring companies’ performance after M&A

deals using secondary data: accounting studies and event studies. The former implies using data from financial statements. The most frequently used robust measure of combined firm’s performance is Return on Assets (ROA) (Ellis et al., 2011). However, in the context of multiple deals per year assessing their performance employing ROA is not possible. Moreover, in the current model deal performance should be disentangled from company’s performance and this measure does not allow doing that. The second way to measure deal performance is event studies, when acquirer’s stock returns are assessed. This widely used methodology grasps investors’ opinion about the deal success (J. Haleblian & Finkelstein, 1999; Hayward, 2002). According to the market efficiency 16

(17)

hypothesis (Malkiel & Fama, 1970) this is a valid measure of acquisition performance especially on the developed capital markets such as the USA. In this study market returns are obtained from Eventus module in CRSP2. Models with two types of returns were calculated. First, those in which the performance of the most recent prior deals was repeated year after year demonstrating that companies “remember” how well they did before irrespective of how much time has elapsed since this experience. Second, the models where deal performance was “retained” for only one year, thus influencing the decisions made only one year after the previous deal. In the subsequent years the performance was denoted as zero in case of absence of new deals. The former models with “unlimited memory” yielded insignificant coefficients for prior performance, demonstrating that this assumption was indeed unrealistic. Therefore, performance “memory” of only one year was used in the final model. Also, the models with different event windows were calculated (e.g. 1, 3 and 5 days before and after the announcement date). They all led to similar results, so the window of (-5;+5) was used (J. Haleblian & Finkelstein, 1999).

Prior experience. Number of deals announced in prior years is used as a proxy for accumulated

experience.

Financial slack. Most of the authors (e.g. Iyer and Miller (2008)) use three different categories of

slack: absorbed (SGA expenses divided by sales), unabsorbed (current assets divided by current liabilities) and potential (equity to debt ratio) slack. Financial measures from the year (t-1) were used for analysis, since managers are expected to base their decisions on the last available financial year results.

Firm performance. Following the approach of Iyer & Miller (2008), in order to assess firm’s

performance, its results relative to aspiration level were used. There are two types of aspiration commonly used in the literature: social and historical. The former shows how the company under 2_{The returns of the deals that were announced shortly one after another, thus present in each other’s estimation window,} were excluded, because the estimated returns would be otherwise biased. The problem is non-existed for the deals that were announced on the same day, since the market captures the performance of both of them.

17

(18)

consideration performs compared to the industry peers, and the latter – compared to its historical performance. In the first case the indicator (ROA – average ROA in the industry defined at the 3-digit SIC code level) was computed. Historical aspiration level was calculated as an average of the last 3 and 5 years. Both types of performance were measured for the fiscal year (t-1). The variables “above” and “below” aspiration were then calculated to distinguish between over- and underperforming firms. "Above" is zero if (ROA - aspiration level) is negative, and "below" is zero if (ROA - aspiration level) is positive.

Industry growth was measured as annual growth of shipments in the industry defined on the 3-digit

SIC level.

Industry munificence was measured using a standard index based on the annual expert survey. Control variables

As the analysis is on the firm level, the choice of the control variables is rather limited. Moreover, after including environmental measures in the model possible industry and time specifics are already taken into account, while the type of model (containing random effects) helps to allow for the differences between firms. The only important factor that has been left out is firm size, because large firms are expected to have more acquisitions on average. The size is measured by the logarithm of assets (J. J. Haleblian et al., 2006).

(19)

Models

In total 6 models were estimated: three for each type of performance aspiration (social, 3 years historical and 5 years historical) both for all sample and for large deals only. Plus additional 3 models were estimated for each of the 3 industries to see the differences across them.

Event history method was employed to analyze the data. The models used in this approach estimate the risk of an event happening given the factors that theoretically have an influence on this risk. Random effects model is used to take into account the differences across industries and firms.

Data will be analyzed using STATA statistical package.

Since the dependent variable is a count variable, the suitable models are Poisson and Negative Binomial model.

Table 1 Dependent variable (deals per year) distribution

Deals per year

Number of observations % of observations 0 7,367 61.65 1 3,130 26.19 2 846 7.08 3 306 2.56 4 158 1.32 5 53 0.44 6 29 0.24 7 20 0.17 8 18 0.15 9 5 0.04 10 5 0.04 11 3 0.03 12 4 0.03 15 1 0.01 16 2 0.02 18 1 0.01 20 1 0.01 28 1 0.01 11,950 100 19

(20)

Table 2 Dependent variable (large deals per year) distribution

Deals per year

Number of observations % of observations 0 2,585 73.69 1 783 22.32 2 94 2.68 3 24 0.68 4 12 0.34 5 3 0.09 6 3 0.09 7 1 0.03 8 1 0.03 12 1 0.03 21 1 0.03 3,508 100

The variance for the distribution of all deals exceeds the mean two times, which signals over-dispersion and requires the Negative Binomial model. However, for non-zero cases mean and variance are almost equal, so over-dispersion is the result of the excessive number of years when companies do not make deals. Therefore, Poisson Zero-Inflated model might be a good fit too. The same pattern is found in the distribution of large deals, so Poisson Zero-Inflated model can be a good fit there also.

Table 3 Mean and variance of the number of deals

Variable Mean Variance Observations

Eligible deals 0.68 1.32 11,950 Eligible deals, non-zero cases 1.61 1.83 4,583 Large deals 0.33 0.59 3,508 Large deals, non-zero cases 1.27 1.04 923

Choosing the model

The Negative Binomial as well as Poisson regression with random effects were estimated on the existing data (see Appendix). Random effects model specification is chosen because companies may have specific features that are not related to the factors present in the model. Both Negative

(21)

Binomial and Poisson models are significant on the 1% level. Therefore, the choice has to be based on the AIC and BIC criteria, which are widely used to choose the best fitting parametrical model.

As can be seen from Table 4, the values of both criteria are smaller for the Negative Binomial model. Therefore, it fits data better than Poisson.

Vuong’s statistics equals 1.56 (p-value 0.0594 ), which indicates that standard Negative Binomial is gives relatively better predictions than Zero-Inflated model. This is also proved by the Information Criteria, which are higher than those for the standard model (Table 4).

Table 4 Information criteria value for different models

Model AIC BIC Observations

Negative Binomial 15,250 15,339 7,050 Poisson 15,385 15,467 7,050 Zero-Inflated Negative Binomial 15,500 15,610 7,050

All in all, standard Negative Binomial model shows a better fit for the data and will be used further.

(22)

Results

The descriptive statistics and correlations are presented in Table 5, Table 6 and Table 7. Similar to what Iyer and Miller (2008) had in their data, historical and social aspiration levels are highly correlated, so the results of the three different models are expected to be consistent with each other. Also, the number of deals per year is positively related to the number of previous deals, which indicates the importance of including this factor as a predictor. Large firms seem to engage in more acquisitions and controlling for this relation was indeed needed.

Table 5 Descriptive statistics (all deals)

Variable Mean S.d. Min Max

Deals per year 0.62 1.15 0 28 Average deal performance 0.004 0.12 -0.75 1.63 Experience 5.48 8.91 1.00 151.00 Absorbed slack 0.68 6.20 0 387.44 Unabsorbed slack 3.38 3.79 0 137.67 Potential slack 2.08 3.62 -1 79.21 Above aspiration (social) 0.18 0.19 0 1.58 Below aspiration (social) -0.03 0.11 -0.97 0 Above aspiration (3 year historical) 0.04 0.09 0 1.32 Below aspiration (3 year historical) -0.04 0.08 -0.93 0 Above aspiration (5 year historical) 0.05 0.10 0 1.73 Below aspiration (5 year historical) -0.04 0.09 -0.95 0 Industry growth 1.04 0.09 0.65 1.29 Industry munificence 0.04 0.06 -0.19 0.15 Logarithm of assets 5.36 2.18 -6.91 12.27

(23)

Table 6 Descriptive statistics (large deals)

Variable Mean S.d. Min Max

Deals per year 0.33 0.77 0 21 Average deal performance -0.002 0.12 -0.63 0.87

Experience 11.63 14 1 151

Absorbed slack 3.13 4.91 1 74 Unabsorbed slack 0.30 0.29 0 7.28 Potential slack 1.15 1.67 -0.59 23.15 Above aspiration (social) 3.15 79.76 -396.00 4564.58 Below aspiration (social) 0.21 0.21 0 1.34 Above aspiration (3 year historical) -0.01 0.07 -0.80 0 Below aspiration (3 year historical) 0.03 0.08 0 1.32 Above aspiration (5 year historical) -0.02 0.06 -0.65 0 Below aspiration (5 year historical) 0.04 0.08 0 1.43 Industry growth -0.03 0.07 -0.77 0 Industry munificence 1.03 0.09 0.65 1.26 Logarithm of assets 0.03 0.06 -0.19 0.15

(24)

Table 7 Correlation matrix Deals per year Large deals per year Experi ence Experien ce (large deals) Perf. of previous deals Perf. of large deals Abs.

slack Unabs. slack Debt to equity Above asp. (social) Below asp. (social) Above asp. (3 yr historical) Below asp. (3 yr historical) Above asp. (5 yr historical) Below asp. (5 yr historical) Ind.

growth Ind. munif. Log (assets)

Deals per year 1 Large deals per year 0.58 1 Experience 0.55 0.35 1 Experience (large deals) 0.42 0.46 0.8 1 Perf. of previous deals 0.03 0 0.01 0 1 Perf. of large deals 0.03 0.01 0.05 0 0.35 1 Absorbed slack -0.01 -0.01 -0.01 -0.01 0 -0.16 1 Unabsorbe d slack 0 0 -0.01 0 0 -0.14 0.02 1 Debt to equity 0 0 0 0 0 -0.03 0 0 1 Above asp. (social) 0.03 0.05 0.06 0.07 -0.01 -0.01 -0.01 0.02 -0.01 1 Below asp. (social) 0.07 0.04 0.08 0.05 0.01 0 -0.04 0 0 0.33 1 Above asp. (3 yr historical) -0.06 -0.03 -0.07 -0.04 -0.01 -0.05 0.01 0 -0.01 0.22 0.07 1 Below asp. (3 yr historical) 0.06 0.04 0.06 0.04 0.01 0.03 -0.01 0 0.01 0.24 0.6 0.21 1 Above asp. (5 yr historical) -0.06 -0.03 -0.08 -0.04 -0.01 -0.01 0.03 0 -0.01 0.3 0.1 0.82 0.19 1 Below asp. (5 yr historical) 0.06 0.03 0.07 0.04 0.01 0.03 -0.04 0.01 0 0.27 0.65 0.2 0.87 0.23 1 Ind. growth -0.02 -0.01 -0.07 -0.05 0 0.07 0.01 0.02 0.01 -0.07 -0.01 0.03 0.08 0.04 0.1 1 Ind. munif. -0.04 -0.02 -0.13 -0.08 -0.01 0.05 0.01 0.03 0 0.03 -0.01 -0.03 -0.03 -0.01 0 0.53 1 Log (assets) 0.22 0.16 0.33 0.26 0 -0.01 -0.06 -0.01 0 0.17 0.31 -0.18 0.19 -0.2 0.18 -0.06 -0.12 1

Correlations with absolute values higher than 0.07 are significant at the p<0.01

(25)

Estimation results are presented in the Table 8. They are reported in the form of Incidence Rate Ratios to make the interpretation easier. This means that the reported values show by what percent the probability of acquisition rises if the factor under consideration grows by 1 unit. So if the coefficient x is above 1, this denotes an increase by (x-1) percent, and if it is lower than 1, the probability is expected to drop by (1-x) percent if the factor augments by 1 unit.

For all deals sample having an additional deal in the past increases the average number of deals in the current year by 1%. This result is consistent across all the three models. As predicted by hypothesis H1a, this does not hold for the large deals: coefficients in the models with 3 types of aspiration levels are insignificant. Therefore, H1a is supported.

Similarly, prior deal positive performance increases the number of acquisitions per year more than two times on average for all deals. For the subsample of large ones this result is not found, thus fully supporting Hypothesis 1b.

Next, different types of slack give somewhat contradictory results regarding Hypothesis 2a. For the whole sample absorbed slack seems to decrease the probability of acquisition, while the unabsorbed and potential lead to its increase. Same effect was found by Iyer and Miller (2008) too. For the whole sample unabsorbed slack raises the probability of acquisition by 2% on average and for the large subsample – by approximately 4%. Given the different results for absorbed slack, Hypothesis 2a is partially supported.

In all models the odds of making a deal rise as performance falls below aspirations. There is a difference between the samples, since for the large deals the coefficients have higher variance. The most striking result is present in the model for 3-year historical aspirations for the large deals subsample: as underperforming firms increase their ROA the probability of making a large 25

(26)

deal plummets 20 times. For the 5-year aspirations the relation is also positive. Thus the Hypothesis 2b is partially supported.

Table 8 Results for 3 industries, SIC codes 2800, 3500, 3600

Factors All deals Large deals

Social Historical - 3yr Historical - 5yr Social Historical - 3yr Historical - 5yr

Experience 1.01*** 1.01*** 1.01*** 1.01 1.01 1.01

Prior deal performance 1.07*** 2.45*** 2.92*** 1.01 1.82 2.72 Absorbed slack (t-1) 0.99 0.8*** 0.71*** 0.53*** 0.5** 0.54** Unabsorbed slack (t-1) 1.02** 1.02** 1.02 1.03* 1.03* 1.03* Potential slack (t-1) 1.01* 1.02** 1.03** 1.04*** 1.05*** 1.05***

Social aspiration (t-1)>0 0.78** 1.48

Social aspiration (t-1) <0 2.34*** 0.91

Historical aspiration (3yrs)>0 0.89 1.18

Historical aspiration (3yrs)<0 4.09*** 20.19***

Historical aspiration (5yrs)>0 0.95 1.53

Historical aspiration (5yrs)<0 2.84*** 3.74*

Industry munificence 13.17*** 13.18*** 12.77*** 49.15*** 97.9*** 93.86*** Industry growth 0.94 0.72 0.66* 1.44 0.63 0.63 Ln (assets) 1.16*** 1.19*** 1.23*** 1.06 1.11** 1.16*** Number of obs. 7640 7050 6120 2045 1966 1809 Number of firms 1134 1038 929 305 301 277 Probability chi2 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Log likelihood -8492.16 -7609.668 -6383.47 -1692.58 -1582.3 -1421.02 *p<0.1 **p<0.05 ***p<0.01

Industry munificence seems to be the most important factor of all, since it increases the probability of concluding a deal more than 12 times in the whole sample. In line with the predictions, it plays an even more important role for the subsample of large firms, augmenting the probability more than 50 times on average, which is a striking and robust result. It gives full support to the Hypothesis 3a.

In the whole sample industry growth decreases the propensity of acquisitions in all models, but this result is significant only in the 5 year historical aspiration model. In the subsample of large

(27)

deals the coefficients for this factor are neither consistent nor significant. All in all, Hypothesis 3b finds no support in the data.

In order to test the last hypothesis, similar models with different types of aspiration levels need to be estimated for each of the three industries represented in the sample. The results are presented in Table 9, Table 10 and Table 11. The industries demonstrate strikingly different patterns.

To start with, producers of Chemicals do not make decisions based on prior deal performance or experience. Munificence seems to play a huge role, because it increases the probability of acquisition more than 70 times in the chemicals industry. The size of the firm is positively related to the propensity of acquisition: each additional 2,7mln of assets raises the chance of acquisition by around 15%. Aspirations do not show a single tendency, as most of the coefficients are insignificant.

(28)

Table 9 Chemicals and allied products industry subsample, SIC 2800 (All deals)

Social Historical - 3yr Historical - 5yr

Experience 1.67 1 1

Prior deal performance 1 1.84 2.12* Absorbed slack (t-1) 0.81*** 0.79*** 0.78*** Unabsorbed slack (t-1) 1.01 1.02 1.01 Potential slack (t-1) 1.02 1.02 1.03** Social aspiration (t-1)>0 0.46***

Social aspiration (t-1) <0 1.6

Historical aspiration (3yrs)>0 0.82 Historical aspiration (3yrs)<0 4.82**

Historical aspiration (5yrs)>0 1.2 Historical aspiration (5yrs)<0 1.39 Industry munificence 80.03*** 67.11*** 107.83*** Industry growth 0.67 0.5 0.4 Ln (assets) 1.13*** 1.14*** 1.18*** Number of obs. 2264 2084 1822 Number of firms 348 317 285 Probability chi2 0.0000 0.0000 0.0000 Log likelihood -2542.74 -2285.95 -1994.22

Next, totally different pattern is present for Machinery and Computer equipment. Companies are more prone to acquire if they have done more deals before. Although significant in all models, this coefficient is very small – the increase is by 1% only. Potential slack is significant in all models and increases the propensity to acquire around 13%. As regards aspirations and propensity to acquire, the findings here are in line with Iyer & Miller’s (2008) findings: underperforming firms tend to acquire more as their performance improves and for highly performing ones there is no universal tendency.

Also, compared to the chemicals industry, munificence plays a much less important role here.

(29)

Table 10. Industrial and Commercial Machinery and Computer Equipment industry subsample, SIC 3500 (All deals)

Experience 1.01*** 1.01*** 1.01*

Prior deal performance 1.17 1.8 1.65 Absorbed slack (t-1) 0.68 0.58* 0.40** Unabsorbed slack (t-1) 0.96 0.96** 0.95 Potential slack (t-1) 1.11*** 1.12*** 1.14*** Social aspiration (t-1)>0 1.21

Social aspiration (t-1) <0 3.11**

Historical aspiration (5yrs)>0 1.22 Historical aspiration (5yrs)<0 5.41*** Industry munificence 21.5*** 26.53*** 14.91*** Industry growth 0.99 0.74 0.0 Ln (assets) 1.2*** 1.26*** 1.32*** Number of obs. 2497 2310 2000 Number of firms 373 339 312 Probability chi2 0.0000 0.0000 0.0000 Log likelihood -2752.44 -2490.93 -2063.17

Electronic and other Electrical Equipment and Components industry presents yet another pattern of M&A incentives. Prior deal performance is extremely important here, as it increases the propensity of acquisition by 2 to 4.5 times, depending on the model. Unlike Machinery and Computer Equipment producers, companies in this industry do not seem to take into account the amount of experience that has been accumulated. Unabsorbed slack raises the probability of M&A by approximately 3%. Aspiration levels coefficients are mostly insignificant and do not let to draw a robust conclusion regarding their role in strategic decision-making. Another peculiarity is that Electronic and other Electrical Equipment and Components’ producers pay considerably less attention to the industry munificence than those in other two industries under consideration.

(30)

Table 11. Electronic and other Electrical Equipment and Components, except Computer Equipment industry subsample, SIC 3600 (all deals)

Experience 1 1.01 1.01

Prior deal performance 2** 3.37*** 4.53*** Absorbed slack (t-1) 1.01 0.86 0.58*** Unabsorbed slack (t-1) 1.03** 1.03** 1.03**

Potential slack (t-1) 0.99 1 1

Social aspiration (t-1)>0 1.82*** Social aspiration (t-1) <0 1.31

Historical aspiration (5yrs)>0 0.68 Historical aspiration (5yrs)<0 1.72 Industry munificence 5.56*** 6.61*** 7.22*** Industry growth 1.35 0.83 0.67 Ln (assets) 1.19*** 1.24*** 1.27*** Number of obs. 2879 2656 2298 Number of firms 468 427 370 Probability chi2 0.0000 0.0000 -2309.63 Log likelihood -3107.05 -2767.03 -2309.63

In every industry firms’ size appears to be positively related to the number of acquisitions, so the tendency to engage in more M&A as the firm grows is a uniform pattern.

The differences attributable to each industry do not bias the estimates of the sample of large deals, though. As shown on the Figure 1, the number of large deals is almost the same in each industry in the sample. It offers a possible explanation why most of the coefficients turned out to be insignificant for the large deals subsample: each industry has certain main factors that downplay each other being combined in one sample. Nevertheless, it also highlights that factors such as unabsorbed slack, industry munificence and company size are important for any large deal, since they were found significant in all three industries.

(31)

Figure 1. Distribution of the large deals by industry

The results obtained from three industries separately show apparent differences, so Hypothesis 3c is supported.

Presented research contains several limitations. First, separate models for the whole sample and for the large deals were not estimated for each industry. This is due to the fact that the Negative Binomial model does not converge in most cases, because of the large number of independent variables and the reduced number of observations. The models that were estimated did not show any deviations from the results presented above, though. Using a different estimation method may let researches study the topic in more detail and probably reveal industry-specific patterns of strategic behavior. This goes far beyond the objectives of the current paper and hence presents a promising direction for future research. Moreover, the underlying reasons for cross-industry differences were not studied here, so this is another topic to be addressed separately.

Second, the definition of “large” deals as having the value of at least 100mln can be considered controversial. For a relatively small company a deal of 10mln will appear large, so decision-making process will be the same as for a bigger company decision-making a 100mln deal. Therefore, such

(32)

indicators as deal value divided by acquirers’ assets can be used to distinguish between “small” and “large” ones. Although logical, this approach needs certain prior study to determine the threshold between these two categories. Here the absolute value was chosen following the approach used by Ellis et al. (2011) that has already shown the peculiarities of large M&A.

Third, the design of the study is such that it requires the deal information to be aggregated to the year level to use companies’ annual financial data. For this reason the indicators such as accumulated experience or average deal performance per year are used as proxies for the acquisition propensity and the performance of the given deal. This approximation and aggregation might distort certain relationships. Therefore, a study on the deal level should be conducted to analyze the effects found here in a greater detail.

Lastly, current study is only the first step to analyzing the interrelatedness and industry specifics of M&A incentives. There are many factors that might be combined in such models: e.g. the level of integration in previous acquisition (Singh & Zollo, 1998), distance to bankruptcy (Iyer & Miller, 2008), similarity of previous acquisition (Hayward, 2002), etc. These factors are not studied here due to the complexity of data collection methods and time limitations.

(33)

Discussion

Presented research provides an insight on decision-making in M&A. Contrary to the most previous studies it disentangles different levels of factors: deal, firm and industry to determine how significant each one is and explain the logic behind managers’ decisions. All in all, the study shows that factors triggering M&A vary across industries and different types of deals. This demonstrates the importance of the context in which acquisitions are analyzed and urges the researches to take those factors into account to arrive at comparable meaningful results. Such factors as available funds, environmental munificence and company size are the most important determinants of firms’ decisions and they play a special role for large deals. The analysis also yielded several interesting results that are discussed below.

As in previous studies (J. J. Haleblian et al., 2006), here positive prior deal performance significantly increases the propensity of the firm to acquire. Nevertheless, companies learn more from the deals that were associated with small losses, since it makes managers thoroughly analyze their mistakes (Hayward, 2002). As mentioned in the theoretical part of the paper, this is one of the reasons why on average acquirers lose money engaging in M&A. The result in the current paper presents another important implication for the decision-makers: acquisitions need to be thoroughly analyzed in order not to lead to heuristic logic and result in losses.

Further, similar to the findings of Iyer & Miller (2008), the analysis of different types of slack produces contradictory results. Most likely the explanation is rooted in the definitions of different types of slack. Absorbed slack represents the ratio of SGA expenses to sales, which is expected to be a proxy for the funds that can be potentially released and used as an acquisition funding. However, this might not always be the case, because SGA expenses may be difficult to

(34)

change in the short-term given the fixed nature of many of them (e.g. salaries, advertising contracts, distribution centers’ maintenance, etc.). High SGA may even be a sign of firms’ inefficiency that negatively influences the opportunities to make M&A for the business. Unabsorbed slack, on the other hand, represents the ratio of current assets to current liabilities, so these are the funds that are actually available to be used to pay for acquisitions and that is why their amount is positively related to the propensity to acquire. Potential slack, which is calculated as equity to debt ratio, estimates firms’ capacity to borrow capital to finance new acquisitions. For that reason unabsorbed and potential slack give similar results.

Aspirations demonstrate another puzzling pattern in the data. Underperforming firms tend to increase the amount of acquisitions as their performance grows. This might indicate the validity of the argument presented in the theoretical section of current study: firms engage in problemistic search if that does not hinder their survival. So when performance improves, businesses can afford to engage in M&A deals that are risky enterprises. Iyer & Miller (2008) found only partial support to the hypothesis that proximity to bankruptcy decreases the intentions to acquire, which is a very similar idea. Perhaps, their results were not robust, since the size of the deal was not taken into account and, following the arguments presented in the current paper, this is a significant factor. Indeed, in the large deals subsample in the model with 3 year historical aspirations the underperforming firms are 20 times more prone to acquire as their performance increases.

For the whole sample the effect for underperforming firms is the strongest for 3 year historical aspiration as well and there are multiple reasons why this particular model yields the most significant results. To start with, for managers 3 year comparison looks more valid than that of 5 years, because both internal and external factors change over the years making the results from 5 34

(35)

years ago non-comparable to what the firm strives for at the moment. The most important internal factor is that management team itself goes through changes: new people come, some veterans leave the company. This way the business loses its ability to look too far in the past, which explains its “limited memory” that was also found in comparison of different performance indicators: those that still reflect the performance of previous deals several years after those deals were made are an insignificant predictor of managers’ decisions. As for the external factors, industrial and overall economic conditions transform: the industry or the company enters the maturity phase of its lifecycle, the main product is no longer in demand or economic crisis has hindered all company’s operations. In those cases the expectations that used to be valid 5 years ago may not be realistic anymore and do not play any role in decision-making. Firms’ “memory” is a perspective branch of research that can be conducted in relation to M&A as well. One of the examples is experience discounting: the model where the performance of previous deals is discounted by a certain factor over time, so that the latest deals play the highest role in decision-making should be tested. The same may be applicable to the experience accumulation: e.g. the company takes into account only the number of deals in the previous 5 years, and no longer. Next, 3 year historical aspiration appears to be more significant than the social one. A possible explanation is high heterogeneity of firms in the industry (Greve, 2003). This factor can also be studied in the future.

This study does not have a goal to explain the differences between industries, but possible explanations to some interesting findings are presented below. For Electronic and Electrical Equipment firms’ prior deal performance is an important factor, while the amount of experience is absolutely insignificant. One possible explanation is the fast changing nature of the industry, where previous experience becomes obsolete very fast, so companies are unable to establish and

(36)

retain sophisticated routines (Zollo & Winter, 2002). However, prior deal performance acts more as a psychological factor, making managers repeat successful choices, not because they necessarily led to learning, but because they as associated with high performance per se and seem “right”. Apart from that, munificence is of less importance here than in any other industry under consideration. In his work about environmental dimensions Boyd (1995) calls the computer production environment relatively munificent, since the resources are relatively abundant there, so munificence is “taken for granted” and does not play major role here. The same might be applicable to the Electronic Components producers too, though this topic requires a detailed research.

Apparently, cross-industrial differences should be studied in a more thorough way. For now, this topic is not extensively covered in the literature. There are only separate studies dedicated to certain industries (e.g. Haleblian et. al (2006)), but they do not explain why and how the incentives for strategic decisions differ in dissimilar contexts. Therefore, this is another promising field for future research.

Conducted analysis has important implications for researches and managers. It shows the main factors that are taken into consideration by managers who decide to engage in M&A. Being aware of the common drives of acquisition activity, practitioners can escape the trap of making unprofitable deals only because they have spare funds, the environment seems favorable or they had been successful in making similar deals in the past, for instance. Also, cross-industrial differences suggest that managers’ experience and best practices are not easily transferrable across different contexts and should be used with caution. For the researches the results give an explanation for some discrepancies in findings of previous studies (e.g. (Iyer & Miller, 2008)),

(37)

indicate that the context of industry and deal value should allowed for to produce robust results and spot the factors that require thorough research in the future.

(38)

Conclusion

This study is aimed at unveiling the factors of M&A timing on three levels: experience of prior deals, firm-level conditions and industry conditions. The analysis was run both on the whole sample of firms and separately on those having deals of at least 100mln. Large deals are often pivotal for company’s strategy, so the factors that managers take into consideration in this case were expected to differ. The analysis presented in this study confirms this supposition. While whole sample indicates the importance of M&A experience accumulation, prior deal performance, available financial resources and environmental munificence, in the large deals subsample only the latter two factors seem to trigger acquisitions. Their effect, however, is considerably stronger, since large deals are more sensitive to the state the firm and its industry are in. The fact that previous M&A activity does not have much influence on the probability of making new large deals indicates that managers do not compare them with each other, treating them as rare and unique events. Another finding proves that decisions to acquire are made differently across industries. Although favorable environmental conditions, for example, increase firms’ propensity to acquire regardless of the industry, other factors play a role only in particular settings.

Alongside the main focus of this study, several interesting findings in the related domains were made that also suggest the direction for future research. First, managers tend to repeat rewarded behaviors even though the learning process is more efficient for the choices that led to small losses. This offers a possible explanation why M&A frequently do not create value for the acquirers. Second, what managers consider to be funds available to be used in M&A financing is controversial. In line with previous studies, in the current model unabsorbed and potential slack turned out to be good proxies for the spare funds the company has, while the increase of the 38

(39)

absorbed one, on the contrary, is seen as a drop of available funds. Third, the concept of problemistic search needs to be studied further in the context of strategic decisions and firm-level conditions, as its implications appear to be highly dependent on degree of risk to survival the business has. Forth, the research produced the evidence of company’s limited “memory”: past experiences on which managers build their aspirations retain their validity for only couple of years. Hence, future studies can be aimed at assessing the length of this memory horizon in a greater detail. Lastly, industrial context seems to determine the factors relevant to acquisition decisions. The existing literature does not offer much research on cross-industrial differences, so this is another promising topic for M&A analysis.

(40)

References

Amburgey, T. L., & Miner, A. S. (1992). Strategic momentum: The effects of repetitive, positional, and contextual momentum on merger activity. Strategic Management Journal, 13(5), 335-348.

Barkema, H. G., & Schijven, M. (2008). How do firms learn to make acquisitions? A review of past research and an agenda for the future. Journal of Management, 34(3), 594-634.

Boyd, B. K. (1995). CEO duality and firm performance: A contingency model. Strategic Management

Journal, 16(4), 301-312.

Capron, L., & Pistre, N. (2002). When do acquirers earn abnormal returns? Strategic Management

Journal, 23(9), 781-794.

Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, NJ, 2

Dess, G. G., & Beard, D. W. (1984). Dimensions of organizational task environments. Administrative

Science Quarterly, , 52-73.

Dyer, J. H., Kale, P., & Singh, H. (2004). When to ally and when to acquire. Harvard Business Review,

82(7-8), 109-115.

Ellis, K. M., Reus, T. H., Lamont, B. T., & Ranft, A. L. (2011). Transfer effects in large acquisitions: How size-specific experience matters. Academy of Management Journal, 54(6), 1261-1276.

Finkelstein, S., & Haleblian, J. (2002). Understanding acquisition performance: The role of transfer effects. Organization Science, 13(1), 36-47.

(41)

Greve, H. R. (2003). Organizational learning from performance feedback: A behavioral perspective on

innovation and change Cambridge University Press.

Haleblian, J. J., Kim, J. J., & Rajagopalan, N. (2006). The influence of acquisition experience and performance on acquisition behavior: Evidence from the US commercial banking industry.

Academy of Management Journal, 49(2), 357-370.

Haleblian, J., Devers, C. E., McNamara, G., Carpenter, M. A., & Davison, R. B. (2009). Taking stock of what we know about mergers and acquisitions: A review and research agenda. Journal of Management,

35(3), 469-502.

Haleblian, J., & Finkelstein, S. (1999). The influence of organizational acquisition experience on acquisition performance: A behavioral learning perspective. Administrative Science Quarterly,

44(1), 29-56.

Hayward, M. L. (2002). When do firms learn from their acquisition experience? evidence from 1990 to 1995. Strategic Management Journal, 23(1), 21-39.

Iyer, D. N., & Miller, K. D. (2008). Performance feedback, slack, and the timing of acquisitions. Academy

of Management Journal, 51(4), 808-822.

Malkiel, B. G., & Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work*.

The Journal of Finance, 25(2), 383-417.

Morgeson, F. P., & Hofmann, D. A. (1999). The structure and function of collective constructs:

Implications for multilevel research and theory development. Academy of Management Review,

24(2), 249-265.

(42)

Park, C. (2003). Prior performance characteristics of related and unrelated acquirers. Strategic

Management Journal, 24(5), 471-480.

Schijven, M., & Hitt, M. A. (2012). The vicarious wisdom of crowds: Toward a behavioral perspective on investor reactions to acquisition announcements. Strategic Management Journal, 33(11), 1247-1268.

Shimizu, K. (2007). Prospect theory, behavioral theory, and the threat-rigidity thesis: Combinative effects on organizational decisions to divest formerly acquired units. Academy of Management

Journal, 50(6), 1495-1514.

Singh, H., & Zollo, M. (1998). The impact of knowledge codification, experience trajectories and integration strategies on the performance of corporate acquisitions.

Slatter, S. S. P. (1984). Corporate recovery: Successful turnaround strategies and their implementation Penguin Books.

Zollo, M., & Winter, S. G. (2002). Deliberate learning and the evolution of dynamic capabilities.

Organization Science, 13(3), 339-351.

The Economic Times, 2013, Global merger and acquisition deal volume hits $ 2.3 trillion in 2013, viewed on 27 January 2013 <

http://articles.economictimes.indiatimes.com/2013-10-24/news/43365989_1_dealogic-global-merger-and-acquisition-ma-deal-volume>

(43)

Appendix

Table 12 Model choice: Negative Binomial model output

Likelihood-ratio test vs. pooled: chibar2(01) = 384.99 Prob>=chibar2 = 0.000 s 4.339545 .4217982 3.586811 5.250249 r 23.74825 2.874289 18.73308 30.10605 /ln_s 1.46777 .0971987 1.277264 1.658276 /ln_r 3.167509 .1210316 2.930291 3.404726 _cons 1.640882 .4126724 1.97 0.049 1.002317 2.68627 ln_at 1.190931 .0174086 11.95 0.000 1.157294 1.225544 muniff_3dig 12.96405 4.265382 7.79 0.000 6.802708 24.70585 indgrowth_3dig .7241661 .1466459 -1.59 0.111 .4869307 1.076984 below_asp_3yr_diff 4.035078 1.333595 4.22 0.000 2.1112 7.712131 above_asp_3yr_diff .8907078 .2223506 -0.46 0.643 .5460681 1.45286 debt_to_equity .9999923 .0001881 -0.04 0.967 .9996237 1.000361 slack_unabs 1.035752 .0060959 5.97 0.000 1.023872 1.047768 slack_abs .8049568 .0490054 -3.56 0.000 .714417 .9069709 cumul_el_deals_l 1.007115 .002253 3.17 0.002 1.002708 1.01154 avCAR6_el_zero 2.44342 .5400879 4.04 0.000 1.584348 3.768302 el_deals IRR Std. Err. z P>|z| [95% Conf. Interval] Log likelihood = -7612.3808 Prob > chi2 = 0.0000 Wald chi2(10) = 360.77 max = 20 avg = 6.8 Random effects u_i ~ Beta Obs per group: min = 1 Group vari able: a_CUSIP_6dig Number of groups = 1038 Random-effects negative binomial regression Number of obs = 7050

(44)

Table 13 Model choice: Poisson model output

Likelihood-ratio test of alpha=0: chibar2(01) = 667.95 Prob>=chibar2 = 0.000 alpha .3040968 .0249167 .2589805 .3570726 /lnalpha -1.190409 .0819368 -1.351002 -1.029816 _cons .2910374 .0610178 -5.89 0.000 .1929698 .4389433 ln_at 1.220436 .0172836 14.07 0.000 1.187027 1.254786 muniff_3dig 14.0467 4.189374 8.86 0.000 7.829011 25.2024 indgrowth_3dig .6902115 .1257595 -2.03 0.042 .4829361 .9864492 below_asp_3yr_diff 3.68655 1.10007 4.37 0.000 2.054097 6.616365 above_asp_3yr_diff .9285007 .2188732 -0.31 0.753 .5849659 1.473784 debt_to_equity 1.000003 .0001825 0.02 0.987 .9996453 1.000361 slack_unabs 1.032358 .0061407 5.35 0.000 1.020392 1.044464 slack_abs .7965277 .0487641 -3.72 0.000 .7064632 .8980741 cumul_el_deals_l 1.006966 .0019063 3.67 0.000 1.003237 1.01071 avCAR6_el_zero 2.20346 .4592187 3.79 0.000 1.464563 3.315144 el_deals IRR Std. Err. z P>|z| [95% Conf. Interval] Log likelihood = -7680.4734 Prob > chi2 = 0.0000 Wald chi2(10) = 464.96 max = 20 avg = 6.8 Random effects u_i ~ Gamma Obs per group: min = 1 Group variable: a_CUSIP_6dig Number of groups = 1038 Random-effects Poisson regression Number of obs = 7050

Factors of mergers and acquisitions’ timing