Employment growth following takeovers

• Published date: 01 December 2019
• Date: 01 December 2019
• DOI: http://doi.org/10.1111/1756-2171.12300
• Author: Karen Geurts,Johannes Van Biesebroeck

RAND Journal of Economics

Vol.50, No. 4, Winter 2019

pp. 916–950

Employment growth following takeovers

Karen Geurts∗

and

Johannes Van Biesebroeck∗∗

We construct a comprehensive sample of takeovers in Belgium that shows they are remarkably

common. Takeovers involve both small and largefirms and, over a five-year period, 17% of private

sector employment. We estimate their impact on employment growth using a framework that

explicitly takes into account that takeovers involve pairs of firms and that post-merger outcomes

are heterogeneous. The average merger temporarily reduces employment of the combined entity

by −1.4%. Mergers likely to be motivated by market power show a stronger and permanent

employment reduction of −14%, whereasthose motivated by efficiency gains lead to employment

expansions of +10%.

1. Introduction

When employment effects of mergers or takeovers are mentioned in the media, the news is

almost invariably negative. Benefits for merging firms are expected to come from rationalizing

their operations, avoiding the duplication of fixed costs, and tackling inefficiencies. This will

boost their bottom line, but at the expense of their workers. When the authorities investigate a

proposed merger because they fear the exercise of market power, higher prices, and lower output,

this is implicitly expected to reduce employment further.

Existing research shows a range of estimates for the employment effects of mergers, but no

consensus has emerged. Studies that find a negative effect outnumber those that find a positive

effect, but all come with caveats.For example, in the context of foreign takeovers, Girma and G¨

org

(2004) find a reduction in employment growth for targetsin the United Kingdom (UK) electronics

sector, but not in the food sector; Lehto and B¨

ockerman (2008) find a similar reduction for Finnish

manufacturing plants, but not in the service sector. In contrast, Bandick and G¨

org (2010) find

evidence of employmentexpansion in Swedish targets, but only for exporters in vertical takeovers;

∗KU Leuven; karen.geurts@kuleuven.be.

∗∗KU Leuven and CEPR; jo.vanbiesebroeck@kuleuven.be.

We wouldlike to thank the Editor, Chad Syverson, and two anonymous referees for constructive comments that greatly

improved the article. We also thank the following people for helpful comments and suggestions: Tom ˚

Astebro, Eric

Bartelsman, Jan De Loecker, John Earle, Joep Konings, Otto Toivanen, and Frank Verboven. The authors are grateful to

the Belgian National Social Security Office, Statistics Belgium, and the National Bank of Belgium for providing the data.

This project has received funding from the European Union’s Horizon 2020 research and innovationprogram under g rant

agreement no. 649378 and the KU Leuven Methusalem grant on the Granular Economy.

916 C2019, The RAND Corporation.

GEURTS AND VAN BIESEBROECK / 917

McGuckin and Nguyen (2001) find more general positive effects in US manufacturing, except

for the largest plants.

At the same time, the literature has long noted the potential for efficiency gains from

mergers. Acquirers often argue that reduced variable costs can offset market power and lead to

lower prices, which in turn can raise optimal output and employment. This can work through

several channels, for example, sharing of assets (Maksimovic and Phillips, 2001), transferring

superior (intangible) assets of an acquirer, such as managerial expertise (Hortac¸su and Syverson,

2007; Bloom, Sadun, and Van Reenen, 2012), or combining complementary assets (Gans and

Stern, 2003). Working with detailed historical data for the Japanese cotton spinning industry,

Braguinsky et al. (2015) document large physical productivity effects of mergers and an even

larger boost to firm profitability. Blonigen and Pierce (2016) find robust support for higher

markups and a positive effect on revenue-based total factor productivity (TFP) for acquired

establishments across the entire US manufacturing sector, but no effect on output-based TFP, a

proxy for efficiency. However, none of these studies have looked at employment effects.

One possible reason why estimated employment effects are ambiguous is that most studies

only consider the effects on acquired firms, whereas effects on acquirers could go in the other

direction. Margolis (2006) finds, for example, that employees of acquired firms are more likelyto

leave than those of the acquiring firm in the years immediately following a merger. Stiebale and

Trax (2011) find positive effects on output and productivity in the acquirer’s home market after

cross-border mergers and acquisitions (M&A), but no significant effect on employment, that is,

no downsizing. The few studies that look at the evolution of total employment for the merged

entities are again inconclusive. Brown and Medoff (1988) find employment expansion following

a merger, at least when the firms integrate their workforces. Conyon et al. (2002) find sharply

reduced demand for labor, even after controlling for output changes. Also, Gugler and Yurtoglu

(2004) find negative employment effectsf orEuropean, but not for US firms.

To estimate a well-defined employment effect of mergers, we need to answer the following

three questions: (i) How can weidentify a merger? (ii) What is a proper counterfactual for a newly

created firm? (iii) Why did the firms merge? Answering these questions forces us to deal with

important measurement challenges that help explain the wide range of estimates in the literature.

We make a contribution on each dimension.

Our first contribution is to identify the universe of mergers between two domestic firms

in Belgium between 2005 and 2012, by following workers as they move between different firm

identifiers. One reason for conflicting results in the literature are differences in the samples of

mergers. Some studies only look at a relatively small number of mergers by large, listed firms

(e.g., Gugler and Yurtoglu,2004), whereas others use innovative methods or unique data to obtain

a record of all mergers in a target population (e.g., Brown and Medoff, 1988). Our objective is

to identify the universe of firm-level reorganizations that are an integral part of firms’ growth

process. Wecomplement filings at the Commercial Court with instances where we observe in the

Belgian social security records that workforces of two firms are combined or that a substantial

fraction of the workforce moves from one firm to another.1We do not include strategic mergers

that are often motivated by diversification or market power, where the original firms continue to

operate independently.Instead, we focus on the process of firm integration and the impact of such

restructurings on firm growth.

In the process of constructing a control sample of potential mergers, we document new

insights about which firms merge and we show that takeover activity is a more dynamic and more

widespread phenomenon than often portrayed, involving both small and large firms. Note that

the distinction between mergers and takeovers oracquisitions is essentially a legal one without a

clear-cut difference in an economic sense. Similar to previous work, we do not discern between

them and use the terms interchangeably.In our comprehensive sample of takeovers in the Belgian

domestic market, we count 2601 mergers over a seven-year period, involving 6000 firms as

1This analysis is based on the universe of employer-employee records for the private sector in Belgium.

C

The RAND Corporation 2019.

918 / THE RAND JOURNAL OF ECONOMICS

acquirer or target. In an average five-yearperiod, 17% of private sector employees workfor a fir m

involved in a takeover. If these events systematically increased or reduced firms’ labor demand,

it would imply a huge impact on the labor market.

A second difference with most previous studies is to focus on total year-on-yearemployment

growth for the two merged firms combined, rather than follow the employment evolution of

acquired firms only. This accounts for reallocations of jobs betweentargets and acquirers and the

fact that one of the original entities often ceases to exist when firms are combined. To deal with

endogeneity, we only look at changes in outcomes, combining a difference-in-differences setup

with matching on observables to construct an appropriate control group (Imbens and Wooldridge,

2009).2Given that mergers are formed by pairs of firms with different characteristics, we do

not use an individual firm as control, but construct a control group from simulated pairs of

nonmerging firms that match pre-merger characteristics of both the target and the acquirer. This

“dyadic” approach has been used in the strategic alliance literature, where alliance formation is

considered to be determined by both partners’ preferences and characteristics (Chung, Singh, and

Lee, 2000). It specifically allows for the characteristics of both the target and the acquirer and

even the combination of characteristics to affect both the decision to engage in a takeover as well

as the post-merger growth pattern.

The third and most important difference with the existing literature is to allowobservationally

distinct types of mergers, that is, involving firms with unique combinations of characteristics and

possibly initiated for unique reasons, to havedifferent employment effects. Wealready mentioned

that increased market power and the elimination of duplicated fixed costs are likely to reduce

employment. More efficient use of existing workers, for example, by renegotiating existing labor

contracts (Shleifer and Summers, 1988) or raising capacity utilization through scale economies

(Kalnins, Froeb, and Tschantz, 2017) are other motivations for mergers that will have similar

effects. Potential efficiency gains from mergers, as mentioned above, provide an opposite force:

they increase labor productivity, lower variable costs, and raise optimal output and employment.

Merely including control variables in a regression that estimates the employment effect of a

representative merger will not capture this richness.

We construct a control group of simulated firm pairs in two steps, in order not to be

constrained to estimate a single effect across all mergers. In a first step, we match merging firms

and control firm pairs exactly on a set of discrete characteristics that create cells with unique

combinations of values for all variables. In a second step, we use the propensity score estimated

within each of these cells to select nearest neighbors or construct weights for the performance

regression. As a result, we can estimate separate effects for each cell or, more practically, for a

group of cells that correspond to a specific merger “type”. Because the selection-on-observables

is performed independently by cell, it allows flexiblyfor different selection equations for different

types. We draw on the theoretical literature to select a few observable characteristics of acquirers

and targets that define two types of mergers. One type is expected to raise market power and

reduce employment, whereas a second type of efficiency-enhancing mergers is expected to raise

employment. This differs from Blonigen and Pierce (2016), who estimate the effect of mergers

on different dependent variables, such as productivity or the markup, but impose a uniform effect

for all mergers in their sample. In contrast, we focus on employmentas key performance variable,

but allow for different effects for different subsets of mergers.3

Beyondthese methodological innovations, our results show a number of interesting economic

findings. First, mergers that lead to workforce integration are a lot more common than one

might expect, affecting on average 0.74% of firms and 4.3% of workers each year. Second, the

average merger reduces employment by 1.4%, compared to the employment evolution of an

2Even though endogeneity of mergers is clearly important, it is sometimes ignored because it has provenvirtually

impossible to find a good instrument. Any variable correlated with a firm’sdecision to engage in a merger is likely to be

correlated with post-merger performance.

3In Table10, we also report takeover effects on value added and labor productivity.The effects for the two subsets

of mergers are in line with expectations, but they are estimated too imprecisely to carry a lot of weight.

C

The RAND Corporation 2019.