Do Labor Regulations Affect the Formation of Corporate Groups?∗

Sharon Belenzon† Ulya Tsolmon‡

January 29, 2013

Abstract

This paper studies the effect of country employment regulations on the formation of corporate groupsin Europe. While employment regulations in some European countries impose substantial constraints onexternal labor markets, these regulations do not apply to internal labor markets of corporate groups. Usinga comprehensive firm-level dataset covering 15 West European countries, we study the strategic responseof firms to increased employment regulations. We document the presence of a specific mechanism throughwhich regulations affect corporate groups: internal labor markets. We find that the propensity of firms to beaffi liated with groups is larger where country regulations and industry turnover are jointly high. This resultis especially strong for smaller firms, which do not have their own internal labor market, and for affi liates oflarge and diversified groups. We complement our findings with direct evidence on managerial mobility, andshow that internal mobility becomes more likely than external mobility as country employment regulationsincrease.

Keywords: corporate groups, labor regulations, internal labor markets, managerial mobility

JEL Classification: J08, J40, L22, M51

1 Introduction

Classic works by Chandler (1962) and Penrose (1959) emphasize internal labor markets (ILM) as a central

mechanism through which firms can achieve sustained competitive advantage and growth. The present paper

investigates how this mechanism drives the strategic response of firms to increased country employment

regulations. Consistent with the logic of transaction-cost economics (Coase, 1937; Williamson, 1975), we

show that as transaction costs associated with using external labor markets rise, more of these transactions

would take place inside corporate groups– collections of legally independent firms (Leff, 1978; Khanna and

Yafeh, 2007). Supporting Chandler and Penrose’s view of the strategic importance of internal labor markets,

we underscore the conditions under which they affect firm boundaries. Our empirical approach is to use

the combined variation in country regulations and industry frequency of labor readjustment to test whether

∗Acknowledgement: We are grateful for valuable comments from Ashish Arora, Tarun Khanna, Will Mitchell, JoanneOxley, Jan Rivkin, Anand Swaminathan, Dennis Yao, Yishai Yafeh, and Bennet Zelner.†Duke University, Fuqua School of Business (sharon.belenzon@duke.edu)‡Duke University, Fuqua School of Business (ulya.tsolmon@duke.edu)

1

utilization of internal labor markets is an important mechanism through which country regulations affect the

organization decision of firms. We contend that corporate group affi liates in West Europe have an advantage

in redeploying labor because within-group coordination allows internal adjustment of labor in response to

external shocks without dismissing employees and incurring employment regulation penalties for dismissals.

This advantage is further enhanced by a legal provision which exempts intra-group labor adjustments from

country employment regulations.

Employment protection regulations comprise a set of legal rules, administrative procedures, and compen-

satory payouts that apply to employee dismissals. In contrast with unemployment benefits, which are funded

through payroll taxes, employment protection regulations impose direct costs on the employer responsible

for dismissals.1,2 We identify the effect of country regulations on corporate group affi liation by showing that

a specific mechanism through which employment regulations affect corporate group affi liation is present. To

identify this mechanism we perform the following test. We examine how the difference in the share of group

affi liates between industries with high and low frequency of labor adjustments varies by country employment

regulations. Consistent with the ILM mechanism, we expect this difference to increase with the level of

employment regulations.

Our difference-in-difference empirical strategy is thus to exploit exogenous country and industry condi-

tions, and test whether country regulations have the strongest impact on group affi liation in industries where

firms, for exogenous reasons, adjust their labor force more frequently. Using data from the U.S., we rank

industries according to their level of labor turnover in relatively regulation-free labor markets. We use the

annual establishment-level employment data from the U.S. Bureau of Labor Statistics to calculate average

turnover rate for each industry. Then we rank the 15 West European countries in our sample according to the

level of their employment regulations, using the OECD Employment Protection Index and as other measures

from complementary sources. Our formal test is whether the difference in group affi liates between high and

low industry labor turnover is higher in countries with high employment regulations than in countries with

low regulations.

There are three major challenges in identifying the effect of country regulations on corporate group

affi liation. We follow Belenzon et al. (2012) and explain how our methodology addresses each challenge.

First, a reverse causality argument implies that while groups may replace ineffi cient external labor markets,

they may also restrain the development of the institutions they mimic (Granovetter, 1995; Khanna and Yafeh,

2007). Our focus on a specific mechanism through which regulations affect affi liation substantially mitigates

this concern. For instance, to argue reverse causality in our context, countries where the difference in group

1For example, in Spain, individual dismissal procedures require 30-day written notice with a statement of reasons for dismissaland a written notification to the worker’s representatives at the workplace. Upon dismissal, Spanish employees are entitled toseverance pay equivalent to 33 days’salary for each year of service. Similarly, Austrian workers with more than 3 years of serviceare entitled to 8 weeks’notice and 6 months’salary as severance pay (OECD Employment Outlook, 2004).

2There is a wide variation in the employment protection in our sample of West European countries. Greece, Spain, and Franceare countries with the strictest employment protection regulations, while Great Britain, Ireland, and Switzerland have the fewestrestrictions on employee dismissals.

2

affi liates between high- and low-turnover industries is large should also have higher employment regulations.

To our knowledge, there is no evidence suggesting this should be the case, especially in the specific context of

European employment regulations.3 Second, omitted or latent macro variables can be correlated with both

employment regulations and group affi liation. By including industry and country fixed effects, selecting an

empirical setting where countries share a relatively similar level of economic development (and still have a

considerable variation in employment protection policies), and focusing on industry-country interaction rather

than “level”effects, we substantially mitigate country unobserved heterogeneity concerns. Moreover, our rich

firm-level data allow us to investigate how the regulation-turnover effect varies by different firm subsamples,

and perform additional tests based on the conditions under which the ILM mechanism should be especially

prominent. Third, group affi liates are often privately held corporations, and due to intricate ownership

arrangements (Faccio and Lang, 2001) they can be “relatively invisible” in some settings (Granovetter,

1995). To mitigate the invisibility problem, we capitalize on the strict reporting requirements of the EU,

where public and private firms have to file annual reports detailing ownership and financial information, to

construct comprehensive ownership and control hierarchies for groups.

The European Union is an ideal environment for testing the ILM theory for three main reasons. First,

EU countries exhibit a wide variation in employment regulations, but at the same time, they exist within

a narrow range of economic development, such that we can focus on developed economies and substantially

reduce cross-country heterogeneity. Second, EU countries share a clear and consistent definition of groups

based on historical, institutional, and economic traditions. Lastly, employment regulations apply only to labor

readjustments that use external markets, but not to ones that take place inside corporate groups. Under EU

law, mobility of workers within a corporate group is not considered a “market transaction.”This means that

intra-group mobility is not subject to country labor-market regulations and affi liates can transfer employees

to a different affi liate without incurring dismissal penalties set by employment protection regulations. The

European Union Directive 96/71/EC sets out to facilitate movement of human capital within and across the

Member States and allows group affi liates to “post workers to an establishment or to an undertaking owned

by the group in the territory of a Member State”(Offi cial Journal L 018, 01/21/1997 p. 1—6). This provision,

granted to group affi liates, allows unilateral transfer of employees among affi liates without having to dismiss

and rehire each transfer and thereby not being subject to employment protection regulations. This legal

provision provides a clear advantage to firms that are affi liated with corporate groups.

While the legal exception formalizes the flexibility of worker reallocation within groups, it is important to

emphasize that this provision is not a necessary condition for the ILM argument to hold. In the absence of

the provision, corporate groups are still likely to hold an advantage over unaffi liated firms in their ability to

reallocate labor. Voluntary deployment is another way for groups to get around dismissal regulations. Since

dismissal regulations do not apply for separations initiated by employees, voluntary basis of internal transfers

3Section 2 provides an overview of the relevant institutional context of employment regulations.

3

allows reallocations to be completed without costly dismissal penalties. Further, voluntary employee rede-

ployments are facilitated by coordination within the group, in contrast to external labor market exchanges,

where the burden of matching employees with jobs rests with the employee, which makes voluntary quits less

likely. Thus, the advantage of ILMs for corporate groups is particularly strong in countries where employ-

ment regulations are high, and in industries where labor readjustments are frequent. Under those conditions,

consistent with the Coasian perspective of firm organization, corporate group affi liation should also be higher.

This is the main point we make and test empirically in this paper.

Our findings strongly support the internal labor market mechanism. We find that the difference in the

share of group affi liates between industries with high and low labor turnover is significantly larger in countries

with high employment regulations than in countries with low regulations. The range of the effect is large.

Moving from the lowest decile to the highest decile of country employment regulations increases the difference

in group affi liation between high- and low-turnover industries by 4.2 percentage points, or about 20 percent

of sample average share of group affi liates. We confirm the robustness of our results by using alternative

measures of country regulations and industry labor mobility. The effect is strongest when exploiting only

variation in managerial mobility (rather than production workers), and for industries where tasks are less

routine (that is, more worker discretion is required).

To further test our mechanism, we explore how the country-industry effect varies by different firm sub-

samples. We find patterns that are highly consistent with the ILM mechanism. The country-industry effect

is very strong when restricting the sample to small firms, and is essentially zero when restricting the sample

only to large firms. This pattern is consistent with the theory, because large firms have internal labor mar-

kets of their own (which we do not observe) that are not subject to employment regulations, similar to the

internal labor markets of corporate groups. Thus, for large firms we should not expect their group affi liation

to be much affected by country employment regulation. This is not the case for small firms that do not

have their own internal labor markets, and thus their group affi liation decision should be strongly mediated

by employment regulations. Continuing this line of investigation, we also distinguish between affi liates of

diversified groups and affi liates of specialized groups. Theory predicts more active internal labor markets

for more diverse groups, because such groups are more likely to be able to smooth employment fluctuations

as industry-specific shocks arrive. Consistent with this theory, our results reveal stronger country-industry

effects when comparing standalone firms to affi liated firms of diverse groups than when comparing standalone

firms to affi liates of specialized groups. Lastly, we complement our findings with direct evidence on man-

agerial mobility. Using information on mobility patterns of more than 250,000 managers of group affi liates

between 2002 and 2007, we show that internal mobility becomes more likely than external mobility as country

employment regulations increase.

4

2 Institutional Background

2.1 Labor Markets in Europe

In this section we provide background on the sources of variation in employment protection regulations across

our sample countries. Our overview emphasizes sources of variations that stem from diverse traditions of

labor-industry relations in each country that are not clearly related to the prevalence of corporate groups,

thus mitigating reverse causality concerns.

Since the industrial revolution, the role of labor has evolved due to confluence of different organizing

paradigms and values rooted in each country’s political organization, and shaped the degree of government

involvement in functioning of labor markets (O’Sullivan, 2003; Botero et al., 2004; Pagano and Volpin,

2005). Different systems of capitalism emphasized to varying degrees the roles of economic effi ciency, scope

of government regulation, and separation of ownership and control. Economies with stronger emphasis

on competitive markets are less likely to impose restrictions on employee dismissals, while countries with

emphasis on non-market coordination have higher employment protection restrictions (Hall and Soskice,

2001).

Comparative political economics literature distinguishes between Anglo-Saxon capitalism (also known as

“liberal market economies”), centered over economic effi ciency, and Rhenish capitalism (also called “coordi-

nated market economies”), characterized by concerns over social welfare and collective action (Albert, 1993;

Hall and Soskice, 2001; Dore et al., 2003). Britain and Ireland are classified as belonging to the Anglo-Saxon

type of capitalism due to their primary reliance on competitive markets and organizational hierarchy, and

accordingly, employment protection regulations in these countries are very low. Northern European countries,

such as Germany, Switzerland, Belgium, Sweden, Norway, Finland, Austria, and Denmark, are thought to

function in the Rhenish type of capitalism, which is characterized by the prevalence of non-market, network-

based coordination. These countries have stricter employment protection regulations than countries with

liberal market economies.

In Britain, development of Anglo-Saxon managerial capitalism mirrored that of the U.S. in 1920s by

extending control to managers; however, managers in UK, keeping traditional social class distance, were

far removed from the production floor, thus yielding significant control to shop floor supervisors (Dore et

al., 2003). This led to increased organizing and strengthening of bargaining power by production workers,

and by the 1960s national productivity and competitiveness had severely declined due to massive increases in

strikes and labor conflict. This prompted drastic reforms to reduce the influence of organized labor, dismantle

employment protection regulations, and make it easier for firms to fire employees (O’Sullivan, 2003). As a

result, lifelong employment was replaced by careers marked by labor market mobility.

On the other hand, Germany, Netherlands, Denmark, and Sweden have supported the institution of code-

termination, which ensures worker representation on the supervisory boards of firms and encourages workers

and management to cooperate in planning and addressing changes in direction and nature of business. State

5

support for non-market coordination provides workers increased involvement in decision-making and allows

resolution of conflict within organizations. Rhenish capitalism often results in a high degree of employment

protection, which in turn leads to comparatively long periods of employment tenure. Development of labor re-

lations in France reflects another variation of institutional influence on organization of production. Although

state has always played a central role in ownership and control of business enterprise, as in Britain, workers

and unions have been removed from corporate resource allocation decisions, regardless of state-mandated

creation of work councils and pro-union incentives to encourage collaborative coordination, known as the

“Aurux Laws” of the 1980s (O’Sullivan, 2003). While the French were not able to achieve the same level

of labor-management collaboration as in Rhenish forms of capitalism; nonetheless, the system for social

democracy has ensured higher protection for employees than Anglo-Saxon forms of capitalism.

Proponents of Rhenish capitalism argue that strict employment regulations encourage firm-specific invest-

ment by employees, increased training by employers, and greater involvement of employees in the workplace

democracy. On the other hand, critics of Rhenish capitalism point out lower unemployment in Anglo-Saxon

economies and blame labor market rigidities created by labor regulations that prevent firms from quickly

adjusting to market fluctuations.

In this paper we provide systematic evidence that is consistent with the view that firms organize in

corporate groups partly to offset external labor market rigidities that are imposed by employment protection

regulations. We proceed to discuss how corporate groups are defined in the context of this paper, and how

our definition relates to alternative definitions that the literature has previously proposed.

2.2 Corporate Groups

Corporate groups are an integral part of the economic landscape in the European Union (EU). The concep-

tualization of a corporate group in academic literature has been varied. Since Leff (1978) defined groups as

“multicompany firm. . . under common entrepreneurial and financial control” (p. 663), scholars have identi-

fied many different organizational structures of “firms bound together in some formal and/or informal ways,

characterized by an ‘intermediate’ level of binding” (Granovetter, 1995, p. 95). Most of the work focused

on issues of concentrated ownership, familial control and reciprocal trading arrangements in the context of

emerging economies (Kester, 1992; Khanna and Rivkin, 2001, 2006). While there is no one overarching de-

finition in the literature, our definition of corporate groups is grounded in and consistent with the EU’s, as

most academic work on corporate groups also uses the EU’s control-based definition of corporate groups.

The legal definition of a corporate group is based on the concept of control between parent and subsidiary

companies as defined in Article 1 of the Seventh EU Directive (Offi cial Journal L 193, 07/18/1983 p.1), and

this definition is utilized in the EU Directive 96/71/EC granting exceptions to labor mobility within corporate

groups (Offi cial Journal L 018, 01/21/1997 p.1—6). The presence of at least one of the following establishes

control: holding the majority of the voting rights; a contract; or the ability to appoint and remove the

6

majority of the Board of Directors (Forum Europaeum Corporate Group Law 2000). Most of the academic

work also utilizes the EU’s control-based definition of corporate groups (Morck et al., 2005; Cestone and

Fumagalli, 2005), where control is determined based on the ownership stakes the controlling shareholder has

in each of the group affi liates (Windbichler, 2000). Therefore, following Faccio et al. (2009), we identify a

firm as a group affi liate if it is a subsidiary (that is, it has a controlling parent company), if it controls another

firm, or if it has the same controlling shareholder as at least one other firm. Our operational definition of

control is consistent with Faccio and Lang (2001) and La Porta et al. (1999), which we detail in the Data

section.

Several previous studies explain corporate group structure as resulting from ineffi cient institutional envi-

ronments and emphasize benefits that affi liate firms gain from being part of a corporate group, such as access

to capital, information, and labor. Mahmood and Mitchell (2004) and Chang, Chung and Mahmood (2006)

determine that groups facilitate innovation where institutional infrastructure is weak. Belenzon et al. (2010,

2012) find that in European corporate groups, flexibility with financial resources affects both the incentives

to form groups and their innovative performance. Khanna and Yafeh (2005) suggest that affi liates reduce

overall uncertainty through risk sharing within a corporate group. Especially in the context of labor markets,

Khanna and Palepu (1997) and Khanna and Rivkin (2001) suggest effi ciency of groups in recruiting and

internally allocating talent. Chang and Hong (2000) also observe that Korean business group affi liates draw

from internally developed pools of managers to offset the underdeveloped external markets for managerial

talent. In the context of developed economies, the corporate group literature has focused on the “pyramidal

groups,” their formal ownership structure, and the darker side of groups within developed economies, such

as expropriation of minority shareholders by dominant owners (Almeida and Wolfenzon, 2006). We take a

different stand by focusing on the institutional void explanation and propose that ineffi cient external labor

markets in Europe promote the formation of corporate groups.

We contend that corporate groups in Europe have an advantage in redeploying labor for two reasons:

first, the legal provision that allows penalty-free transfers of labor among affi liates, and second, internal

coordination of labor adjustment which avoids involuntary dismissals. Our discussions and consultations

with European labor law experts and HR executives have further validated the utility of this provision for

the purpose of unfettered mobility of personnel between affi liates of a corporate group. Second, internal

coordination of labor allows groups to manage adjustments of labor across its affi liates with less reliance on

external labor markets. For example, if an affi liated firm A and a standalone firm B are both hit with a

negative shock that requires cutting back on employment, firm A can utilize the group’s internal market to

place employees in an affi liate where a positive shock has increased demand for labor, thus avoiding involuntary

dismissal costs, while standalone firm B does not have that option and must let go of its employees and

incur various dismissal costs. The internal coordination capability bolstered by the legal provision provide

corporate groups an advantage in countries with strict employment protection regulations and in industries

7

where frequent adjustments of labor are needed.

While business scholarship has devoted considerable effort to understanding the economic consequences

of groups, their emergence, and their prevalence, our work provides more general implications for large

organizations in developed countries. This is because European corporate groups in many respects “obviously

resemble the American conglomerates” (Leff, 1978, p. 664). Using affi liates as a window into the internal

structure of complex and large organizations in Europe is likely to have generalizable implications for large

American firms. This is important because division-level data is not systematically available for American

conglomerates, which makes it impossible to compare standalone firms to divisions with similar characteristics.

Equivalent European “divisions,”on the other hand, are typically incorporated legal entities which are visible

to researchers and comparable to standalone firms. Specifically in the context of the present paper, identifying

the conditions under which large organizations are more likely to emerge due to ILM considerations (i.e., where

corporate affi liation is more likely) should be also relevant for better understanding the relative prevalence

of large American corporations.

3 Related Literature

Traditionally, the literature distinguishes between two types of labor market institutions: internal labor

markets, which operate inside firms/organizations, and external labor markets, which operate across organi-

zations. Workers search for jobs that can pay them higher wages, while employers search for workers who can

perform tasks more effi ciently. Under natural conditions, the only stable allocation of workers to jobs involves

positive assortative matching: the ablest workers are matched with the best jobs, and the least able workers

are matched with the worst jobs (Heckman and Sedlacek, 1985). Well-functioning labor market institutions

ensure that this effi cient allocation of workers to jobs is rapidly achieved (Borjas, 2002). However, employ-

ment regulations can substantially restrict the effectiveness of external labor markets in achieving an optimal

allocation of workers to jobs, thus giving rise to internal labor markets to organize employment relationships

and determine allocation of labor within a firm.

The literature identifies different advantages firms can gain from managing their labor internally. First,

firms can increase incentives for firm-specific investment by employees by insulating them from external

fluctuations and providing opportunities for internal career paths (Doeringer and Piore, 1985; Baker, Gibbs

and Holmstrom, 1994a, 1994b; Bertrand, 2004; Wang and Barney, 2006). Due to diffi culty of designing

complete employment contracts, once worker wages are negotiated and set upon her entry into a firm, firms

utilize informal agreements or relational contracting to govern the employment relationship (Macneil, 1985;

Baker et al., 1994a, 1994b; Levin, 2003; Bertrand, 2004). Informal wage-shielding agreements keep wages

stable relative to external market dynamics and employees are motivated by non—contractible promises for

promotions and other rewards, all of which deem credible only through relational contracting in the absence of

formal contracts. Thus, firms can strategically influence employee effort and firm—specific investment through

8

credible non-contractible promises within their internal labor markets. Increased firm—specific investment by

employees can enhance productivity and create opportunities for expansion (Penrose, 1959). Second, internal

mobility of labor facilitates transfer of tacit knowledge and expansion of social capital, activities critical

for firm performance and growth (Kogut and Zander, 1992; Burt, 1992). Third, firms with more internal

reallocation options can potentially attract better talent at lower cost relative to firms that cannot project

that level of stability (Doeringer and Piore, 1985; Prendergast, 1993).

Despite the large theoretical literature on internal labor markets, empirical research on the topic is limited

(Baker and Holmstrom, 1995; Osterman and Burton, 2004; see Gibbons and Waldman, 1999, for a survey).

Notable exceptions include the following studies. Khanna and Palepu (1999) use survey data to show that

advantages in the labor and product markets were the main factors behind business group expansions in

Chile and India, as the groups repositioned themselves to take advantage of new opportunities in light of

new deregulation. The importance of internal labor reallocation is also highlighted by two recent studies on

productivity differences across U.S. firms. Schoar (2002) uses plant-level data to demonstrate that productiv-

ity increases following acquisition by a diversified conglomerate are due to more capable management being

reallocated to the new plants. Hortaçsu and Syverson (2009) find that the plants of vertically integrated U.S.

firms have higher productivity levels than the plants of non-vertically integrated firms due to a more effi cient

transfer of intangible inputs, such as managerial resources, in integrated firms rather than effi cient transfers

of goods along the production chain. In related work, Capron et al. (1998) stress the importance of resource

redeployment (including managerial redeployment) as a mechanism to acquire organizational capabilities.

Focusing more specifically on top managers, an important literature has examined the determinants of CEO

and top management team turnover (Fredrickson et al., 1988; Wiersema and Bantel, 1993). Together this

evidence suggests that diversified organizations (corporate groups and conglomerates) gain advantage from

internally redeploying their labor.

Understanding conditions under which ILM emerge provides insight into when an ILM becomes a strate-

gically important function of a firm. While there are many ways ILM may provide significant advantages over

external labor markets even when labor regulations are relatively low, we argue that the potential advantages

of internal labor markets should be more pronounced in environments where regulations are more severe.

We do not directly examine the different mechanisms by which ILM impart advantage to firm outcomes in

this paper. Instead, we focus on the wider question of whether increased labor adjustment costs due to

labor regulations contribute to the creation of internal labor markets, especially in conditions where frequent

adjustment of labor is necessary. However, we provide additional tests exploiting variation in industry task

routines and firm-specific training in the hope of highlighting potential strategic functioning of internal labor

markets that would be helpful to guide future research in this field.

9

4 Data

We construct our sample from the Bureau van Dijk’s (BvDEP) Amadeus financial database, which provides

a wide and representative coverage of both private and public companies in Europe. BvDEP standardizes

financial items across the various countries’filing regulations and captures a wide range of firm sizes. In

this paper we exploit cross-sectional variation, such as employment and sales, across firms, industries, and

countries. Our data include three main sections, which we describe in detail below: ownership, country

measures of labor regulations, and industry measures of labor turnover.

4.1 Ownership

Our dataset relies on detailed ownership links between European firms from the 2007 version of Amadeus.

In this section, we explain our methodology for constructing the data and describe our sample. Amadeus

provides information on equity links between firms in Europe. For these inter-firm dyadic ownership links,

we determine which represent a controlling interest. Following Belenzon et al. (2012), we make the following

assumptions for minimum equity holding representing control and ensure our results are not sensitive to

alternative specifications of control thresholds. For private firms, control is established if the shareholder has

at least 50% of the voting rights. Consistent with previous research, the threshold for public firms is set at

20%, because public firms typically have a more dispersed ownership (e.g., La Porta et al., 1999; Faccio and

Lang, 2001; Belenzon et al., 2012). Also, ownership links between firms can be indirect: if firm A owns 50%

of firm B, and firm B owns 50% of firm C, then firm A has a 25% ownership link to C.

We define a corporate group as a collection of at least two legally distinct firms in which one of them is

a controlling ultimate shareholder, where control is identified according to the equity links described above.

A firm is classified as a group affi liate if: i) the firm has a controlling parent company (it is a subsidiary), or

ii) it is a parent company of another firm (it has a subsidiary), or iii) it has the same controlling shareholder

as at least one other firm.4

4.2 Country Employment Regulations

Our main measure of country labor regulation is the OECD employment dismissal protection index for the

1998—2008 period. We refer to this index as employment protection. This index measures how diffi cult it is to

dismiss workers across countries. It includes different procedural inconveniences, severance pay, and overall

diffi culty of dismissal for economic and performance reasons. The index ranges from 0 to 6, with higher values

indicating stricter regulations. We use the average annual dismissal protection index for each country in our

sample. There is wide variation in dismissal protection across countries, from highly protected countries like

Greece (3.11) and Spain (3.01), to least-protected countries like the Great Britain (1.07) and Switzerland

4Details on the methodology used to construct ownership links are available upon request, and are also summarized in Belenzonand Berkovitz (2010).

10

(1.60).5

Besides dismissal regulations, we utilize the OECD’s data on country’s labor expenditures (labor expen-

ditures over GDP) to measure employment protection. Labor expenditures are different ways countries can

protect employees from dismissals. Instead of letting firms internalize the costs of turnover themselves, coun-

tries tax employers and provide benefits, such as unemployment insurance, from a common pool. The main

difference between labor expenditures and dismissal regulations is that firms bear costs of the former regard-

less of whether they dismiss employees, while the latter apply to each dismissal occurrence. Examining labor

expenditures is important because it provides a useful test of the ILM mechanism. We expect no effect for

employment expenditures because this protection is not specific to labor mobility across external or inter-

nal markets. There is wide variation in country spending on labor expenditures. Denmark (4.15%) spent

the largest share on labor market expenditures, while the UK (0.58%) spent the smallest share of its GDP.

Interestingly, there is a very low correlation between dismissal protection and labor expenditures (0.07).

We utilize several alternative country measures which we use mostly as robustness to the OECD em-

ployment protection index. The first is the World’s Bank’s rigidity of employment index from the Doing

Business rankings publication for the period 2004—2007. The rankings are based on comparison of countries’

labor regulations. The index is an average of three sub-indices for diffi culty of hiring on fixed-term contracts,

rigidity of work-hour restrictions, and diffi culty of terminating redundant employees. This index ranges from

0 to 1, where higher values indicate more rigid employment regulations. This measure is comparable to the

OECD’s employment protection index, as it incorporates individual and collective dismissal rigidities, but

differs by measuring restrictions on working hours, such as irregular work-hour limitations and number of

paid vacation days. While these on-the-job limitations may not directly increase dismissal costs, they can

increase costs of employing workers and therefore affect hiring decisions.

The second measure is flexibility in hiring and firing workers. This measure is from the Executive Opin-

ion survey (2008—2009), which is administrated by the World Economic Forum to collect country-specific

information on local business executives’perceptions on labor regulations. The index ranges from hiring and

firing decisions being flexibly determined by employers (1) to being strictly constrained by regulations (7).

Our final regulation measure is firing costs from the World Bank’s Doing Business report (2009). This

is a direct measure of the costs associated with dismissal in the form of number of weeks of wages paid as

severance to dismissed employees.

Table 1 presents the values of the above measures by country.6

5With the exception of Switzerland, there is little variation in employment protection index within countries over time. WhileGreece has relaxed its employment protection since 2003 from 3.46 to 2.81, Norway decreased its index from 2.72 to 2.61 in 2000,only to go back up to 2.72 in 2005. More systematically, a variance decomposition exercise shows that more than 80 percent ofthe country-year variation is attributed to between-country variation.

6The correlation between the employment protection index and the rigidity of employment index is 0.808, and the correlationbetween employment protection index and the measure of flexibility in hiring and firing workers is 0.659. However, the correlationbetween firing costs and employment protection index is much lower at 0.254. This suggests that firing costs comprise only onedimension in overall employment protection index, and different mixes of policies exist in our sample countries. We try to capturethe variation through these different measures.

11

4.3 Industry Employment Measures

We follow the methodology first used by Rajan and Zingales (1998) and rank industries according to their

level of labor turnover. We use U.S. data for ranking industries because the U.S. market is probably the least

regulated market in the developed world; thus, firing and hiring decisions are not likely to be significantly

affected by U.S.-specific employment regulations. Additionally, groups are less common in the U.S., so U.S.

labor turnover makes a good substitute for turnover without options for group ILM. Lastly, U.S. industry

data is separate from European firms, but major industries are structurally similar, so a U.S. industry’s labor

turnover is likely to be a good measure of that industry’s turnover in Europe. We make two main assumptions

on the nature of the industry measures: the first is that structural reasons (as opposed, for example, to local

demand and supply conditions) explain why some industries have higher labor turnover than others, and the

second is that these differences persist across countries.

Our main industry variable is labor turnover. We construct labor turnover for each industry using an-

nual establishment-level employment data from the U.S. Bureau of Labor Statistics’Current Employment

Statistics Survey (1977—2003). Following Autor et al. (2007) and Bozkaya and Kerr (2009), we calculate

firm-level labor turnover rate as the average of absolute change in annual employment at the firm divided by

the average firm employment across two years. The industry labor volatility measure is the average of firm

turnover rate in each two-digit SIC industry.7

We create a measure for industry involuntary labor turnover by utilizing the data on employer-initiated

dismissals from the Bureau of Labor Statistics’ Job Openings and Labor Turnover Survey (2003—2007).

Using monthly industry-level data on layoffs and involuntary dismissals due to reorganizations, elimination

of positions, and firings, we calculate industry-level involuntary labor turnover rate as average rate of layoffs.

Since employment protection regulations do not apply to voluntary separations, such as quits and retirements,

this measure provides additional ranking of industries by separating voluntary and involuntary turnover.

Managers and production workers. To further explore the nature of the ILM mechanism, we distin-

guish between managerial and production employee turnover by using the U.S. Bureau of Labor Statistics’

Occupational Employment Statistics Survey (1999—2007). Managerial turnover data is based on the Offi ce

of Management and Budget (OMB) Standard Occupational Classification (SOC) system classification of

management occupations, which include top management and other supervisory occupations. The produc-

tion and non-supervisory turnover data is based on the Current Employment Statistics Survey (1977—2003),

which reports establishment-level employment data for production employees in goods-producing industries

and non-supervisory employees in service industries. The production and managerial turnover measures for

each industry are calculated in the same way as the total labor turnover rate.

The next two measures rank industries according to two dimensions that are important to the ILM

7 Industries with highest labor turnover include Apparel (SIC 23) with 0.087 and Transportation Services (SIC 42) with 0.079;industries ranked with lowest turnover rates include Paper Products (SIC 26) with 0.019 and Printing and Publishing (SIC 27)with 0.020. Table 4 presents more examples of high- and low-turnover industries.

12

mechanism. The first is how routine tasks are, and the second is the importance of firm-specific training.

We use these measures later in the econometric analysis to split the sample between industries that rank

high and low on these measures, and explore the sensitivity of the regulation-turnover effect for the different

industry subsamples being used.

Task non-routineness. We create a measure for how routine tasks are in the industry, based on the

importance of making decisions and solving problems in various occupations within an industry. We follow

Costinot et al. (2011) to measure the importance of problem-solving skills in each occupation by using the

U.S. Department of Labor’s Occupational Information Network (O*NET). The O*NET has information on

more than 200 worker and occupational characteristics in about 800 six-digit occupations. The measure

of the level of decision-making and problem-solving importance ranges from 0 to 100, with 100 being most

important. The most recent available O*NET data (version 16.0, July 2011) was used to calculate routineness

of each occupation. We then matched the occupation-level measure of routineness to the share of employment

of six-digit occupations in an industry using the Bureau of Labor Statistics (BLS) Occupational Employment

Statistics 2007 data to compute the average non-routineness of work in each industry. The final measure of

task non-routineness in an industry ranges from 0 to 1, with 0 being most routine and 1 being least routine.8

Firm-specific training. Following Coff (1999), we create a measure for the level of firm-specific knowledge

by using the 2008 version of the Bureau of Labor Statistics’National Longitudinal Survey of Youth (NLSY).

The annual survey follows a panel of 12,000 individuals from 1979 onward. The survey asks the respondents

to specify the number of hours a week spent in training provided by their employer to maintain and upgrade

employee skills. Apart from regular schooling hours, which captures general employee knowledge, this measure

ranks industries by the level of employer-specific knowledge required. The hours of training are aggregated

to respondents’two-digit SIC industries and range from 0 to 50 hours.

4.4 Descriptive Statistics

Our estimation sample includes firms for which we have non-missing sales for 2006 or the most recent year

available and non-missing ownership information based on 2007 ownership structure. Later in the paper

we report robustness checks for alternative restrictions of the estimation sample. Countries in our sample

have different reporting requirements for small firms, which may result in underreporting of small firms in

certain countries, but reporting should not vary systematically within countries. Our results are robust to

cross-country variation in reporting requirements, because our empirical approach relies on the interaction of

industry and country measures and controls for country- and industry-level effects.

Table 2 provides summary statistics for firms and affi liates in our sample. On average, our firms (including

affi liates and standalones) have 92 employees (8 median) and generate $25.7 million in annual sales ($1.1

8Average non-routineness of a task in an industry is (µs)=1-∑bs (t) µ(t); where µ(t)= 1-

∑τ α (τ ,t)

P (τ)100

; bs(t) is the shareof employment in six-digit occupation in an industry s ; α (τ ,t) is the employment share of occupations τ (6-digit SOC 2010) intask t; P (τ)

100is the importance of problem-solving for occupation τ .

13

million median). Our sample firms belong to 68,137 unique groups. The average group has a total of 4

affi liates (with a 90th percentile of 6). The average group holds around $1 billion in assets; however, this

seems to be driven by groups at the highest end of the distribution, since the median is $10 million, and the

90th percentile is $271 million.

Table 3 presents summary statistics separately for affi liates and standalone firms. Affi liates tend to be

larger in terms of sales, total assets, and the number of employees, but quite similar to standalone firms in

terms of age. Interestingly, we find that affi liates have much higher turnover than standalone firms, consistent

with the basic premise of this paper of higher labor adjustment costs for standalone firms.

5 Econometric Specification

We investigate the effect of country labor regulations on group affi liation by testing whether corporate groups

substitute for more constrained external labor markets. We focus on a specific mechanism through which

group membership can mitigate market frictions– internal labor markets.9 If groups form as a substitute

for highly regulated labor markets, we should observe a higher probability of group affi liation for firms that

operate in industries where firing and hiring are frequent, that is, industries where labor turnover is high. In

these industries, government labor regulations should matter the most.

Thus, consistent with the ILM theory, we expect that the difference in the probability of group affi liation

between industries with high relative to low labor turnover would be higher in countries with high employment

regulations than in countries with low employment regulations. We estimate a Linear Probability Model for

affi liation probability. The dependent variable is a dummy that receives the value of unity for firms that are

affi liated with a corporate group, and zero for standalone firms.

The econometric specification is given as

Pr(Affiliate = 1)i = β1Salesi + β2EmpRegc × Turnoverj + β3Industry sales sharejc + ϕj + ηc + εi (1)

i denotes firms — the unit of observation, Salesi is annual sales of firm, EmpRegc is the employment

regulation for country c, Turnoverj is a measure of labor turnover for industry j, ϕj and ηc are complete

sets of industry and country dummies, and εi is an iid error term. Similar to Rajan and Zingales (1998), we

control for the share of industry sales in each country: Industry sales sharejc —the share of total sales of

industry j (in which the focal firm operates) in country c. This measure is computed using all firms in the

complete sample where we make no restrictions on sales.9 In a related paper and using a similar setting, Belenzon et al. (2012) examine the effects of country financial development

on propensity of firms to organize in corporate groups by exploiting the industry-level differences in dependency on externalfinancing. To determine the relationship between financing and labor mechanisms, we examined the correlations between thekey measures and controlled for financing country-industry effects in our main specification. We find no significant correlationbetween measures of country labor regulation and levels of financial development (-0.35), between industry-level measures of laborturnover and dependence on external financing (-0.03), and between labor and financing sets of country-industry interactions(-0.05). As well, the results for labor and financing effects remain robust when both are included, thus suggesting independenceof each mechanism in affecting the propensity of firms to organize in groups. Taken together, these two papers emphasize theimportance of internal markets in shaping firm boundaries.

14

Consistent with the hypothesis that the difference in share of affi liated firms between high— and low—

labor turnover industries would be larger in countries with higher employment regulation, we expect β̂2 >

0. The interpretation of β̂2 can be easily explained as follows. Taking the first difference in probability

of affi liation with respect to labor turnover, holding fixed country employment regulation, yields ∆Pc =

β̂2EmpRegc×∆Turnover. Next, taking the difference in ∆Pc between high and low country labor regulation

yields ∆P = β̂2∆EmpReg×∆Turnover. β̂2 thus measures how much higher the likelihood of affi liation is at

a high level of industry labor turnover with respect to an industry with low labor turnover, when it is located

in a country with a high level of country regulation rather than in one with the low level of development.

In all regressions we report ∆P , where industry and country differences are computed by moving from the

lowest to the highest quartile.

Table 4 illustrates our empirical approach. This table presents the highest- and lowest-ranked industries

with respect to labor turnover for countries with high and low labor regulations (split by median). Columns

1 and 2 present the share of group affi liates in each industry by two levels of country employment protection:

low and high, respectively. Column 3 presents the difference in group affi liation in respective industries

between countries with high and low regulations, after removing country and industry effects. There is a

clear difference in group affi liation between highest and lowest turnover industries: a larger share of firms

in industries with highest turnover are affi liates, especially in countries with stricter (above the median)

employment protection regulations. The differences in affi liation ranges from 22.4% to 40.5% in highest

turnover industries; in lowest turnover industries, the difference between high regulation and low regulation

countries is much smaller, between 6.2% and 16.4%.

6 Estimation Results

6.1 The Effect of ILM on Group Affi liation

Table 5 presents the estimation results for the interaction between country labor regulation and industry labor

turnover. The results are consistent with the ILM hypothesis (β2 > 0). Column 1 includes an interaction

between OECD index of country employment protection and industry labor turnover. As expected, the

coeffi cient estimate is positive and is highly significant. Based on this estimate, the differential effect of

ILM, ∆P, is 4.2 percentage points. This means that the difference between the highest and lowest deciles of

industry labor turnover rises by 4.2 percentage points, or 20 percent of the sample average share of group

affi liates, when moving from the lowest to the highest country employment regulation.

Column 2 adds the OECD index of labor expenditures. The measure of country labor expenditures is not

related to firm-specific firing or hiring decisions, and thus should not affect group affi liation. As expected,

the coeffi cient estimate of labor expenditures is zero.

Columns 3 to 5 present the estimation results for the alternative country regulation variables: rigidity of

employment, flexibility in hiring and firing, and firing costs. In all cases we find a positive and significant

15

interaction term with industry labor turnover. For the first two measures, the estimated differential effect is

almost twice as large as the estimated effect for labor turnover.

Lastly, in our main specification, the unit of observation is the individual firm. Column 6 checks the

sensitivity of our results for unity of observation selection by estimating a collapsed model where the unit of

observation is country-industry. For each country and industry, we calculated the average share of affi liates

and collapsed the data to the country-industry level, resulting in 3,457 observations for 288 three-digit SIC

industries in 15 countries (some countries have fewer industries represented than others). This estimation

yields similar coeffi cient estimates on the country-industry interaction term.

6.2 Alternative Industry Measures

We proceed by examining how our results change when using different industry measures that could potentially

inform on the specific functions of ILM mechanism. Table 6 presents the estimation results. Column 1 includes

industry involuntary turnover. We expect our results to be especially strong for involuntary separations,

because these are most likely to be affected by dismissal regulations. Our results are consistent with this

expectation. This is an important result because it rules out an alternative explanation where industry

turnover is simply driven by short-term jobs which are likely to have a high turnover rate because they do

not offer real career opportunities to employees.

Columns 2 and 3 distinguish between managerial turnover and production-worker turnover. We find that

our results are mostly driven by managerial turnover. This finding suggests that firm-specific knowledge

and skills may be more important for ILM considerations than generic interchangeable labor. To further

explore this interpretation, we utilize industry-level measures of task routine and the amount of on-the-job

training provided by employers. Columns 5 and 6 examine how the turnover-regulation coeffi cient varies

by our measure of task routineness. By splitting the sample by high- and low-routine industries, we find a

much stronger turnover-regulation interaction effect for the non-routine industries. Columns 6 and 7 split the

sample by high and low training. The coeffi cient estimates in the two subsamples are essentially the same;

however, the estimate is more precise in the high-training industries. Taken together, our findings show that

the ILM mechanism is stronger for managerial mobility, and in industries where worker discretion is large.

These findings highlight decision-making responsibilities, rather than low-skill temporary jobs, in the way in

which ILM influences firm group affi liation decisions.

6.3 Firm and Group Characteristics

To further test our prediction, we exploit variations in firm and group characteristics. Table 7 presents the

estimation results.

Starting with firm characteristics, we examine how the turnover-regulation interaction varies by firm

employment. Large standalone firms are likely to have their own internal labor markets and therefore should

be less sensitive to country regulations as compared to small standalone firms that have limited employment

16

pools to draw from when readjusting labor. Thus, we expect the effect of ILM to be more pronounced

when comparing smaller standalone firms with affi liated firms of similar size. Columns 1 and 2 confirm this

prediction when splitting the sample by firms with above and below 100 employees. Results are robust to

alternative employment cutoffs.

We proceed to examine how our results vary by group characteristics. Columns 3—11 present the estimation

results for distinguishing between groups that are small or large, diversified or specialized, family held or

widely held, domestic versus multinationals. There are several interesting findings worth pointing out.

First, we expect our results to be especially strong when comparing standalone firms to affi liates of large

and diversified groups. Larger groups have more internal resources to maintain steady growth and offer less

risky employment prospects to their workforce. Diversified groups can provide an additional insurance from

external business fluctuations and redistribute labor as needed. The results support this view. We split groups

into three size categories based on total assets; we classify groups as specialized if all group affi liates operate

in a single three-digit SIC code industry, and diversified if group affi liates operate in at least two distinct

three-digit SIC code industries. We estimate the turnover-regulation effect separately for each subsample of

group affi liates. In each estimation, we compare all standalone firms that the respective set of group-affi liated

firms. The estimated coeffi cient on the country-industry interaction is much larger for the samples of large

and diversified group affi liates than for specialized group affi liates (columns 3—7).

Second, family-owned groups, which account for 5 percent of the groups in our sample, are likely to

have different considerations of internal promotions than widely-held groups. Family-controlled groups may

appoint managers to key positions not by merit but by family relation, which can reduce opportunities and

incentives for high-quality personnel (Bertrand and Schoar, 2006). We do not find large differences in the

magnitude of the effect for the different samples of family-held and widely-held firms (columns 8—9).

Lastly, we separately compare domestic and multinational group affi liates with standalone firms. All

affi liates in a domestic group operate within a single country, while multinational groups own affi liates in

more than one country in our sample. The estimated coeffi cients on the country-industry interaction term are

large and positive in both subsamples, but the coeffi cient for the domestic group affi liates is more significant

(columns 10—11). We find similar results when using alternative measures of group internationalization, such

as group affi liate-sales or assets concentration by country.

6.4 Robustness Checks

We perform several robustness checks for the effect of ILM on group affi liation. Table 8 presents the estimation

results for alternative ways of constructing industry labor turnover and for alternative restrictions on the final

estimation sample.

Non-linearity in industry ranking. We begin by exploring the sensitivity of our results to non-linear effects

of industry turnover (column 1). We break up industry turnover into quartiles and interact each quartile

17

with country labor regulations in order to check for non-linear industry effects, as well as to have a more

stable classification of industries by turnover level. The results show that the country-industry effect is quite

monotonic, with the strongest for the highest quartile of industry turnover.

Recently acquired firms. Evidence suggests that acquired units are less likely to take part in intra-group

mobility of talent (Belenzon and Patacconi, 2012), and managers are less likely to be deployed to acquired

units, because their firm-specific expertise is more valuable in internally developed units (Karim and Williams,

2012). Using Bureau van Dijk’s Zephyr database, we identify acquired affi liates and exclude them from the

sample. The results continue to hold (column 2).

Alternative industry data sources. Our main industry data are from the U.S. Bureau of Labor Statistics’

Current Employment Statistics Survey. Columns 3 and 4 examine the robustness of our results to using

alternative data sources on employment turnover. We construct two alternative industry labor turnover

measures using all firms in our sample countries in Amadeus (1997—2007) and all firms in Compustat (1977—

2007). We follow the same procedure to determine the industry rate of turnover using these different datasets.

The correlations between our primary industry labor turnover measure and these alternative measures are

0.36 and 0.35, respectively. The same pattern of results continues to hold when using these alternative

measures of industry turnover.

Removing outliers and firms with missing employment information. The concern is that our results are

sensitive to including very small or very large firms in our estimation sample. We limit the estimation sample

to firms with a number of employees between the 1st and 99th percentile of the employment distribution to

drop the smallest firms that are usually not subject to employment protection regulations and largest firms

which are likely to have internal labor markets and thus less likely for their affi liation with other firms to be

affected by the ILM considerations. The results hold for this sample as well (column 5). The effect is also

robust to including firms with missing employment information (not reported).

Missing ownership information. In our analysis thus far we exclude firms that do not have ownership

information. We check the sensitivity of our results to dropping these firms by including all firms with missing

ownership information as standalones and assume that firms are not affi liated with groups, unless we have

ownership information that indicates group membership. This procedure more than doubles the number

of observations in our estimation sample. The results are robust to inclusion of firms with no ownership

information (column 6).

Labor unions. Prevalence of collective labor agreement contracts may influence the U.S.-based turnover

rates, as union contracts may act as a substitute for country employment regulations. We use the Current

Population Survey data from the Bureau of Labor Statistics between 2003 and 2007 to determine industries

with higher share of employees covered by labor union contracts. The results are robust to exclusion of firms

in industries at the 95th percentile and above of union membership distribution (column 7), where 18.41 to

70.75 percent of all employees are covered by collective labor agreement contracts. The results continue to

18

hold also for lower thresholds of industry unionization.

Removing individual countries. Our last robustness check is to remove individual countries from the

sample and re-estimate our model separately for each excluded country (not reported). This procedure

allows us to examine whether a single country is driving our results. We confirm that our results continue to

hold. The estimate of the coeffi cient on the interaction between industry labor turnover and country labor

regulation is always positive and significant. The differential effect varies from a low of 2.9 percent when

France is excluded, to a high of 6.2 percent when Great Britain is excluded (this constitutes about 14.2

percent of the sample average share of group affi liates when France is excluded and a much stronger effect of

31.2 percent without Great Britain in the sample).

6.5 The Dynamics of Group Affi liation

Thus far we have shown that the propensity of firms to be affi liated with groups is larger where country

regulations and industry turnover are jointly high. In this section we provide additional evidence of the

dynamics of group affi liation by documenting patterns of changes in ownership structure.

We investigate how many of the firms which we classify as standalones in 2007 change their affi liation and

become part of a group in a later period. Our empirical exercise in this section is to match all standalone

firms used in our baseline estimation as of 2007 to the 2011 cross-section of firms. This allows us to identify

firms that changed their ownership structure from being classified as standalones in 2007 to being potentially

classified as group affi liates in 2011.

Our detailed analysis is as follows. We start with 948,968 standalone firms in our original 2007 sample and

match all of them to the 2011 version of the data. We find considerable changes in ownership: 8.5 percent of

standalone firms become group affi liates in this five-year period. A key potential driver of ownership change

is growth. Indeed, we find a strong correlation between growth and becoming group affi liates. Average

employment growth rate between 2007 and 2011 is -1.1 percent. This negative growth rate is driven mostly

by standalone firms with an average growth rate of -2.4 percent, as compared to positive growth of 5.9 percent

for firms that became part of groups as of 2011.

We proceed by examining different mechanisms through which firms change affi liation status from stand-

alone to group affi liate. We distinguish between firms that became group affi liates by forming new groups,

and firms that became group affi liates by joining existing groups. We observe 31,823 standalone firms in

2011 that are classified as parent company with at least one subsidiary (this comprises 39.5% of all firms that

changed affi liation). Next, we classify firms that join existing groups by using the 2011 ownership structure

and by identifying those that were acquired between 2007 and 2011 using Bureau van Dijk’s Zephyr data-

base, or listed as subsidiaries in the 2011 ownership database. We determine that 48,683 firms joined existing

groups between 2007 and 2011 (this represent 60.5% of firms that changed ownership).

We further examine the types of standalone firms that become affi liated with a corporate group between

19

2007 and 2011 by isolating small standalone firms and distinguishing them by their growth prior to affi lia-

tion (2005—2007). In countries where employment protection regulations restrict labor adjustments, smaller

standalone firms on growth trajectory may benefit more from becoming a part of a group and tapping into its

internal labor market, especially in industries with higher labor turnover. Table 9 shows that out of stand-

alone firms with fewer than 50 employees in 2007, a larger share of high-growth firms are likely to become

affi liated with a corporate group than low-growth firms (8.5 and 6.5 percent, respectively), and firms are

more likely to join an existing group than form a new one.

We next examine whether the change in affi liation is skewed towards industries and countries where both

turnover and regulation are jointly significant. Table 10 presents the estimation results. The within-firm

results are consistent with the ILM prediction: changes toward affi liation are higher in countries with stricter

employment regulation and higher industry turnover (column 1). The results also hold when estimating

separately for firms that formed new groups (column 2) and firms that joined existing groups (column 3).

Then we limit the estimation to small standalones. While the results for all small standalone firms with fewer

than 50 employees are consistent with results for all standalone firms (column 4), group affi liation is driven

mostly by high-growth firms (columns 5 and 6). This suggests that smaller firms on growth trajectory become

affi liated with groups to take advantage of internal labor markets when external labor market frictions are

high.

6.6 Direct Evidence on Intra-firm Managerial Mobility

In his seminal article, Leff (1978) argues that the redeployment of high-level management across business lines

is an important advantage of group affi liation. It has long been recognized that inputs such as “honesty and

trustworthy competence on the part of high-level managers”are inherently diffi cult to market effi ciently (Leff,

1978, p. 666) and that “it is self-evident that such management cannot be hired in the market-place”(Penrose,

1959, p. 41). Therefore, groups can mitigate the problems associated with managerial selection by creating

internal markets for managerial talent. Since firm-specific knowledge and know-how are diffi cult to acquire on

the market and costly to develop internally (Kogut and Zander, 1992; Wang and Barney, 2006), managerial

mobility facilitates effi cient reallocation of tacit organizational knowledge, skills, and social capital (Burt,

1992; Gelatkanycz et al., 2001). Further, research on managerial cognition suggests that cognitive models

of top managers have significant influence on firm response to environmental discontinuities and change

(Tripsas and Gavetti, 2000; Kaplan et al., 2003); thus, managerial rotation can also introduce fresh ideas

and perspectives in sense-making and formulating firm strategy. To our knowledge, we are first to provide

direct systematic information on patterns on intra-firm managerial mobility while distinguishing between

movements which are internal and external to the group.

We document the prevalence of internal labor markets for top managers as follows. Our data provides

information on firms’upper management, so we select managers for whom we have complete employment

20

information between 2002 and 2007 to determine managerial mobility patterns (Belenzon and Patacconi,

2012). Within each corporate group, we track whether a particular manager stayed with the same affi liate,

moved to another affi liate within a group, or moved out of the group between 2002 and 2007. If a manager

stayed with the same affi liate for the entire five-year period, this manager is classified as a non-mover. If a

manager in 2007 was employed in a different affi liate of the same group in the five-year period, this manager

is classified as a within-group mover. If a manager did not have continuous employment with a group for five

years, and was employed outside the group when not employed in the group, then this manager is classified

as an out-of-group mover.

We end up with 259,748 unique managers in 53,501 corporate groups in 2002—2007 period, out of which

12.8% moved internally. Using a share of internal moves in each corporate group, we show in Table 11

a pattern of higher within-group managerial mobility in larger groups and in countries with higher labor

protection. We use the share of within-group moves to test the difference between managerial moves in

countries of high and low employment protection (by median). Consistent with the ILM theory, we expect

the share of intra-group mobility to be higher when a group operates in countries with high employment

protection, and this is what we find.

The comparison-of-means test shows a consistently higher share of internal moves for affi liate managers

in countries with high employment protection. The difference grows with the size of the group: for small

groups the difference in means is 0.8%, while managers in largest groups move internally on average 12.9%

more in countries with high employment protection than in countries with low employment protection, and

the difference is highly significant.

Further analysis shows that diversified groups (whose affi liates generate sales from more than one three-

digit SIC industry) and multinational groups (in which at least one of the affi liates operates in a country

different from the rest) have higher internal mobility of managers than specialized and domestic groups (not

reported). The pattern holds when we use the share of internal moves out of all managers, including those

that did not move and stayed employed with the same firm for five years. These patterns provide a direct

evidence of active internal labor markets in corporate groups, especially when external labor market frictions

are higher. Groups take advantage of internal labor markets in countries with higher external labor mobility

restrictions and actively reallocate managerial resources.

7 Conclusion

This paper provides a detailed empirical examination of the effect of country labor regulations on corporate

group affi liation in West European countries. We argue that corporate groups have an advantage in readjust-

ing their labor through internal labor market coordination, aided by a provision granted to corporate groups

that allows their intra-group labor mobility to be exempt from country labor laws. This provision means

that while in some countries standalone firms are “penalized”for modifying their labor force using external

21

labor markets, group affi liates can tap into the group internal labor markets to freely adjust their labor.

To the extent that labor regulations pose a significant constraint on labor allocation and reallocation, we

expect a greater share of economic production to be organized in corporate groups, especially when country

constraints are significant, and where the need to readjust labor force is high.

Our results strongly confirm the presence of an ILM mechanism through which country employment

regulations positively affect the formation of corporate groups. The ILM is especially strong for managerial,

rather than production-worker, mobility, and in industries where the use of worker discretion is high. We

complement our indirect approach with direct evidence on managerial mobility. Lastly, we complement our

findings with direct evidence on managerial mobility, and show that internal mobility becomes more likely

than external mobility as country employment regulations increase.

More generally, our work contributes to understanding the strategic role of internal labor markets. How

internal labor markets affect firms’ competitive advantage in product and technology markets remains a

challenge for future research.

References

[1] Albert, M. 1993. Capitalism against capitalism. London: Whurr.

[2] Almeida, H., and D. Wolfenzon. 2006. A theory of pyramidal ownership and family business groups.Journal of Finance, 61: 2637—2680.

[3] Autor, D. H., W. R. Kerr, and A. D. Kugler. 2007. Does employment protection reduce productivity?Evidence from U.S. states. The Economic Journal, 117(521): F189—F217.

[4] Baker, G., Gibbs, M., and B. Holmstrom. 1994a. The wage policy of a firm. The Quarterly Journal ofEconomics, 109(4): 921-955.

[5] Baker, G., Gibbs, M., and B. Holmstrom. 1994b. The internal economics of the firm: Evidence frompersonnel data. The Quarterly Journal of Economics,109(4): 881-919.

[6] Baker, G., and B. Holmstrom. 1995. Internal labor markets: Too many theories, too few facts. TheAmerican Economic Review, 85(2): 255—259.

[7] Belenzon, S., and T. Berkovitz. 2010. Innovation in business groups. Management Science, 56(3): 519—535.

[8] Belenzon, S., T. Berkovitz, and L. Rios. 2012. Capital markets and firm organization: How financialdevelopment shapes European corporate groups. Management Science (forthcoming).

[9] Belenzon, S., and A. Patacconi. 2012. Managerial redeployment, ownership structure, and family ties.Working Paper.

[10] Bertrand, M. 2004. From the invisible handshake to the invisible hand? How import competition changesthe employment relationship. Journal of Labor Economics, 22(4): 723—765.

[11] Bertrand, M., and A. Schoar. 2006. The role of family in family firms. The Journal of Economic Per-spectives, 20(2): 73—96.

[12] Borjas, G. J. 2002. Labor economics. Singapore: McGraw-Hill.

[13] Botero, J. C., S. Djankov, R. La Porta, F. Lopez-de-Silanes, and A. Shleifer. 2004. The regulation oflabor. The Quarterly Journal of Economics, 119(4): 1339—1382.

22

[14] Bozkaya, A., and W. R. Kerr. 2009. Labor regulations and European venture capital. NBER WorkingPaper 15627.

[15] Burt, R. S. 1992. Structural holes: The social structure of competition. Cambridge, MA: Harvard Uni-versity Press.

[16] Capron, L., P. Dussauge, and W. Mitchell. 1998. Resource redeployment following horizontal acquisitionsin Europe and North America, 1988—1992. Strategic Management Journal, 19: 631—661.

[17] Cestone, G., and C. Fumagalli. 2005. The strategic impact of resource flexibility in business groups.RAND Journal of Economics, 36(1): 193—214.

[18] Chandler, A. D. 1962. Strategy and structure. Cambridge, MA: MIT Press.

[19] Chang, S. J., and J. Hong. 2000. Economic performance of group-affi liated companies in Korea: In-tragroup resource sharing and internal business transactions. Academy of Management Journal, 43(3):429—448.

[20] Chang, S. J., C. N. Chung, and I. P. Mahmood. 2006. When and how does business group affi liationpromote firm innovation? A tale of two emerging economies. Organization Science, 17(5): 637-656.

[21] Coase, R. H. 1937. The nature of the firm. Economica 4: 386—405.

[22] Coff, R. W. 1999. How buyers cope with uncertainty when acquiring firms in knowledge-intensive indus-tries: Caveat emptor. Organization Science, 10(2): 144—161.

[23] Costinot, A., L. Oldenski, and J. Rauch. 2011. Adaptation and the boundary of multinational firms. TheReview of Economics and Statistics, 93(1): 298—308.

[24] Council Directive 96/71/EC of the European Parliament and of the Council of 16 December 1996 con-cerning the posting of workers in the framework of the provision of services. Offi cial Journal of theEuropean Union. L 018, 01/21/1997 p. 1—6.

[25] Doeringer, P., and M. Piore. 1985. Internal labor markets and manpower analysis. Armonk, NY, London:M. E. Sharpe, Inc.

[26] Dore, R., W. Lazonick, and M. O’Sullivan. 2003. Varieties of capitalism in the twentieth century. OxfordReview of Economic Policy, 15(4): 102—120.

[27] Faccio, M., and L. H. Lang. 2001. The ultimate ownership of Western European corporations. Journalof Financial Economics, 65(3): 365—395.

[28] Faccio, M., L. H. Lang, and L. Young. 2009. Pyramiding vs. leverage in corporate groups: Internationalevidence. Journal of International Business Studies, 41(1): 88—104.

[29] Fredrickson, J., D. Hambrick, and S. Baumrin. 1988. A model of CEO dismissal. Academy of ManagementReview, 13: 255—270.

[30] Gelatkanycz, M. A., B. K. Boyd, and S. Finkelstein. 2001. The strategic value of CEO external directoratenetworks: Implications for CEO compensation. Strategic Management Journal, 22(9): 889—898.

[31] Gibbons, R., and M. Waldman. 1999. Careers in organizations: Theory and evidence. In Orley C.Ashenfelter and David Card (Eds.), The handbook of labor economics (Ch. 36). Amsterdam: ElsevierB.V.

[32] Granovetter, M. 1995. Coase revisited: Business groups in the modern economy. Industrial and CorporateChange, 4(1): 93—130.

[33] Hall, P. A., and D. Soskice. 2001. Varieties of capitalism: The institutional foundations of comparativeadvantage. New York: Oxford University Press.

23

[34] Heckman, J. J., and G. L. Sedlacek. 1985. Heterogeneity, aggregation, and market wage functions: Anempirical model of self-selection in the labor market. Journal of Political Economy, 93(6): 1077—125.

[35] Hortaçsu, A. and C. Syverson. 2009. Why do firms own production chains? U.S. Census Bureau Centerfor Economic Studies Paper No. CES 09-31.

[36] Kaplan, S., F. Murray, and R. Henderson. 2003. Discontinuities and top management: Assessing the roleof recognition in pharmaceutical firm response to biotechnology. Industrial and Corporate Change, 12:203—234.

[37] Karim, S., and C. Williams. 2012. Structural knowledge: How executive experience with structuralcomposition affects intrafirm mobility and unit reconfiguration. Strategic Management Journal, 33(6):681—709.

[38] Kester, W. C. 1992. Industrial groups as systems of contractual governance. Oxford Review of EconomicPolicy, 8(3): 24—44.

[39] Khanna, T., and J. W. Rivkin. 2001. Estimating the performance effects of business groups in emergingmarkets. Strategic Management Journal, 22(1): 45—74.

[40] Khanna, T., and J. W. Rivkin. 2006. Interorganizational ties and business group boundaries: Evidencefrom an emerging economy. Organization Science, 17(6): 774—775.

[41] Khanna, T., and K. Palepu. 1997. Why focused strategies may be wrong for emerging markets. HarvardBusiness Review, 75(4): 41-48.

[42] Khanna, T., and K. Palepu. 1999. Policy shocks, market intermediaries, and corporate strategy: Theevolution of business groups in Chile and India. Journal of Economics and Management Strategy, 8(2):271—310.

[43] Khanna, T., and Y. Yafeh. 2005. Business groups and risk sharing around the world. Journal of Business,78: 301—340.

[44] Khanna, T., and Y. Yafeh. 2007. Business groups in emerging markets: Paragons or parasites? Journalof Economic Literature, 45(2): 331—372.

[45] Kogut, B., and U. Zander. 1992. Knowledge of the firm, combinative capabilities, and the replication oftechnology. Organization Science, 3(3): 383—397.

[46] La Porta, R., F. Lopez-de-Silanes, and A. Shleifer. 1999. Corporate ownership around the world. Journalof Finance, 54(2): 471—517.

[47] Leff, N. H. 1978. Industrial organization and entrepreneurship in the developing countries: The economicgroups. Economic Development and Cultural Change, 26: 661—675.

[48] Levin, J. 2003. Relational incentive contracts. American Economic Review, 93(3): 835-857.

[49] Macneil, I. R. 1985. Relational contract: What we do and do not know.Wisconsin Law Review, 483-525.

[50] Mahmood, I. P., and W. Mitchell. 2004. Two faces: Effects of business groups on innovation in emergingeconomies. Management Science, 50(10): 1348-1365.

[51] Morck, R., D. Wolfenzon, and B. Yeung. 2005. Corporate governance, economic entrenchment, andgrowth. Journal of Economic Literature, 43(3): 655—720.

[52] O’Sullivan, M. 2003. The political economy of comparative corporate governance. Review of InternationalPolitical Economy, 10(1): 23—72.

[53] Osterman, P., and M. D. Burton. 2005. Ports and ladders: The nature and relevance of internal labormarkets in a changing world. In S. Ackroyd , R. Batt, P. Thompson, and P. S. Tolbert (Eds.) The OxfordHandbook of Work and Organization. Oxford: Oxford University Press.

24

[54] Pagano, M., and P. F. Volpin. 2005. The political economy of corporate governance. American EconomicReview, 95(4): 1005—1030.

[55] Prendergast, C. 1993. The role of promotion in inducing specific human capital acquisition. The QuarterlyJournal of Economics, 108(2): 523-534.

[56] Penrose, E. 1959. Theory of the growth of the firm. New York: J. Wiley and Sons.

[57] Rajan, R., and L. Zingales. 1998. Financial dependence and growth. American Economic Review, 88:559—586.

[58] Schoar, A. 2002. Effects of corporate diversification on productivity. The Journal of Finance, 57(6):2379—2403.

[59] Seventh Council Directive 83/349/EEC of 13 June 1983 based on the Article 54 (3) (g) of the Treaty onconsolidated accounts. Offi cial Journal of the European Union. L 193, 07/18/1983, p. 1.

[60] Tripsas, M., and G. Gavetti. 2000. Capabilities, cognition, and inertia: Evidence from digital imaging.Strategic Management Journal, 21(10—11): 1147—1162.

[61] Wang, H. C., and J. B. Barney. 2006. Employee incentives to make firm-specific investments: Implicationsfor resource-based theories of corporate diversification. Academy of Management Review, 31(2): 466—478.

[62] Wiersema, M., and K. Bantel. 1993. Top management team turnover as an adaptation mechanism: Therole of the environment. Strategic Management Journal, 14: 485—504.

[63] Williamson, O. E. 1975. Markets and hierarchies. New York: Free Press.

[64] Windbichler, C. 2000. Corporate group law for Europe. European Business Organization Law Review, 1:265—286.

25

Table 1. Country Measures of Employment Regulations (1) (2) (3) (4) (5) (6)

Country Number of

Sample Firms Employment

Protection Labor

Expenditures

Rigidity of Employment

Index

Flexibility in Hiring and Firing

Workers Firing Costs

Austria 9,651 2.25 1.93 0.36 4.70 1.10 Belgium 13,544 2.50 3.43 0.26 5.00 2.83 Denmark 9,539 1.86 4.15 0.10 2.10 0.00 Finland 2,201 2.13 3.10 0.56 4.30 3.30 France 252,351 2.88 2.60 0.67 5.20 3.50 Germany 529,215 2.45 3.13 0.70 5.50 4.25 Great Britain 115,944 1.07 0.58 0.12 3.90 3.14 Greece 5,394 3.11 0.51 0.65 5.00 3.22 Ireland 2,363 1.25 1.85 0.20 4.20 2.94 Italy 28,258 2.55 1.25 0.47 5.50 2.48 Netherland 4,602 2.30 3.38 0.54 5.00 2.89 Norway 107,028 2.66 1.34 0.47 5.10 2.64 Spain 114,916 3.01 2.21 0.66 5.30 4.04 Sweden 21,901 2.46 2.90 0.50 4.80 3.30 Switzerland 2,665 1.60 1.53 0.20 2.40 2.64

Average 81,305 2.27 2.26 0.43 4.53 2.84 Notes: This table presents measures of employment protection regulations for the countries in our sample. Columns 2 and 3 contain measures from the OECD, averaged for 1997–2007. Employment protection index measures how strict employee dismissal regulations are in a country and ranges from 0 to 6, with 6 being most difficult to dismiss an employee. Labor expenditures measure is the percent of GDP spent by countries on labor market expenditures, which primarily comprise of unemployment benefits. Column 4 contains World Bank's annual measure of rigidity of employment, averaged over 2004–2007 (range 0–1, 0 being most flexible and 1 being most rigid). Column 5 contains measure of flexibility firms have in hiring and firing employees (1=flexibly determined by employers; 7=impeded by regulation) from 2008–2009 Executive Opinion Survey. Column 6 contains a measure of firing costs in terms of weeks of wages paid (natural log), from World Bank's Doing Business rankings (2009).

Table 2. Summary Statistics for Main Firm and Group Variables Distribution

Variable # firms/ groups Mean Std. Dev. 10th 50th 90th

Panel A: Firm level Sales ($,`000) 1,213,681 25,744 763,338 177 1,134 16,282 Employees 897,808 92 1,947 1 8 75 Assets ($,`000) 720,811 44,834 1,339,466 108 934 22,673 Firm Age 1,200,956 17 19 2 11 36 Employment Turnover 523,921 0.139 0.141 0 0.100 0.325 Sales Growth 2,986,704 0.051 0.361 -0.225 0.072 0.353

Panel B: Corporate group level # of Affiliates 68,137 4 10 2 2 6 Sales ($, mm) 68,137 592.372 8,413,815 0.730 17.891 338.312 Assets ($, mm) 68,137 1,082.014 36,300,000 1.235 10.248 270.766 Employment 68,137 973 10,840 2 55 789 Industry Concentration Index (HHI) 68,137 0.74 0.23 0.43 0.77 1 Notes: This table provides summary statistics on main firm and group variables in the estimation sample. In the upper panel the unit of observation is a firm, and in the lower panel the unit of observation is a corporate group.

Table 3. Firm Characteristics: Affiliates vs. Standalone Firms

Affiliates Standalones

Variable Affiliates – Standalones # Firms Mean Median Std. Dev. # firms Mean Median Std. Dev.

Sales (`000) 86,797** 269,202 93,289 5,037 1,582,324 944,479 6,492 868 182,921

Employees 288** 207,450 314 27 3,965 690,358 26 6 433

Assets (`000) 115,275** 237,008 122,206 4,567 2,321,344 483,803 6,931 523 170,093

Firm Age 1.8** 265,748 18.2 13 19.2 935,208 16.4 11 19.3

Employment Turnover 0.034** 137,726 0.164 0.120 0.148 386,195 0.130 0.095 0.138

Sales Growth 0.018** 1,288,739 0.061 0.072 0.391 1,697,965 0.043 0.072 0.337 Notes: This table reports mean comparison tests for affiliates and standalones. The unit of observation is a firm. ** implies that the difference in means between affiliates and standalones is significant at the 1% level.

Table 4. Percentage of Group Affiliates in High- and Low-Turnover Industries by Country Employment Protection

(1) (2) (3)

Countries with Low

Employment Protection Countries with High

Employment Protection High Minus Low

Employment Protection

Panel A: Examples of high-turnover industries (% of group affiliates)

Health and personal care services 18.5 59.0 40.5 Transportation services 20.7 50.5 29.8 Management services 24.2 49.6 25.4 Insurance carriers 23.6 46.8 23.2 Apparel stores 11.4 33.8 22.4

Panel B: Examples of low-turnover industries (% of group affiliates)

Chemicals 51.1 57.3 6.2 Heavy construction 21.5 30.0 8.5 Wholesale trade 21.9 35.0 13.1 Printing and publishing 32.6 46.7 14.1 Paper products 36.4 52.8 16.4 Notes: This table presents patterns of affiliation in selected industries with high and low labor turnover, in countries with high and low employment regulations. Columns 1 and 2 present the share of affiliates by industry and country. Column 3 reports the difference between share of affiliates in countries with low and high employment protection after separating industry and country effects.

Table 5. The Effect of Country Employment Regulations on Group Affiliation Dependent Variable: Dummy for Group Affiliation

(1) (2) (3) (4) (5) (6)

Variables All All All All All

Country-industry

estimation

Country employment protection × Industry total labor turnover

1.153** (0.240)

1.098** (0.373)

1.505** (0.434)

Country labor expenditures × Industry total labor turnover

0.065 (0.261)

Rigidity of employment index × Industry total labor turnover

3.934** (0.932)

Flexibility in hiring and firing workers × Industry total labor turnover

1.326** (0.279)

Firing costs × Industry total labor turnover

0.755** (0.230)

ln(Sales) 0.074** 0.074** 0.074** 0.074** 0.074** (0.015) (0.015) (0.015) (0.015) (0.015)

Industry Sales Share 0.206 0.205 0.222 0.257 0.228 0.788** (0.406) (0.407) (0.414) (0.414) (0.418) (0.184)

Country Dummies (15) Yes Yes Yes Yes Yes Yes

Two-digit SIC Dummies (74) Yes Yes Yes Yes Yes Yes

Differential in Affiliation Probability (%): 4.2 4.0 8.1 7.5 4.3 5.4

% Affiliated 22.2 22.2 22.2 22.2 22.2 45.8

R2 0.272 0.272 0.272 0.272 0.272 0.520

Observations 1,213,681 1,213,681 1,213,681 1,213,681 1,213,681 3,457 Notes: This table reports the results of Linear Probability Model regressions that examine the effect of employment regulation protection levels on corporate group affiliation. The estimation is cross-sectional (at the firm level) and is based on the 2007 ownership structure. Sales data is for 2006 or the most recent year that data is available. Industry Sales Share is three-digit industry sales as a share of total country sales, computed over all firms in the estimation sample. Differential in affiliation probability measures how much higher the likelihood of affiliation is at the 90th percentile level of industry labor turnover with respect to an industry at the 10th percentile level when it is located in a country at the 10th percentile level of employment protection rather than in one at the 90th percentile level of employment protection. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by country. * significant at 5%; ** significant at 1%.

Table 6. Alternative Industry Measures Dependent Variable: Dummy for Group Affiliation

(1) (2) (3) (4) (5) (6) (7) Types of industry turnover Industry characteristics

Variables

Layoffs and

dismissals Managerial

turnover Production turnover

Highly non-

routine Highly routine

High Training

Low training

Country employment protection × Industry involuntary labor turnover

0.035** (0.009)

Country employment protection × Industry managerial labor turnover

0.616** (0.138)

Country employment protection × Industry production labor turnover

0.290 (0.515)

Country employment protection × Industry total labor turnover

2.104** (0.240)

0.068 (0.194)

0.666** (0.218)

0.679 (0.374)

ln(Sales) 0.096** 0.074** 0.080** 0.074** 0.095** 0.121** 0.119** (0.013) (0.015) (0.016) (0.010) (0.013) (0.016) (0.012)

Industry Sales Share 0.008 0.209 -0.247 0.277 -1.004* -0.695 -0.816 (0.209) (0.411) (0.473) (0.282) (0.378) (0.429) (0.390)

Country Dummies (15) Yes Yes Yes Yes Yes Yes Yes

Two-digit SIC Dummies (74) Yes Yes Yes Yes Yes Yes Yes

Differential in Affiliation Probability (%): 4.4 3.0 0.8 7.6 0.2 2.4 2.5

% Affiliated 32.7 22.2 20.7 40.9 22.43 44.0 40.7

R2 0.239 0.272 0.275 0.227 0.218 0.225 0.239

Observations 576,009 1,207,638 982,849 164,999 189,708 108,003 130,769 Notes: This table reports the results of Linear Probability Model regressions that examine the effect of employment regulation protection levels on corporate group affiliation by alternative industry measures. Column 1 reports estimates from using involuntary dismissal turnover measure. Columns 2 and 3 use managerial and production worker turnover measures. Columns 4 and 5 report estimates from 4th (highly non-routine) and 1st quartile (highly routine) distribution of industries. Columns 6 and 7 split the sample by 1st and 4th quartile of industry-level training, respectively. The estimation is cross-sectional (at the firm level) and is based on the 2007 ownership structure. Sales data is for 2006 or the most recent year that data is available. Industry Sales Share is three-digit industry sales as a share of total country sales, computed over all firms in the estimation sample. Differential in affiliation probability measures how much higher the likelihood of affiliation is at the 90th percentile level of industry labor turnover with respect to an industry at the 10th percentile level when it is located in a country at the 10th percentile level of employment protection rather than in one at the 90th percentile level of employment protection. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by country. * significant at 5%; ** significant at 1%.

Table 7. Variation by Firm Size and Group Characteristics Dependent Variable: Dummy for Group Affiliation

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Firm size Group characteristics

Variables

Under 100 emp.

Over 100 emp.

Assets Industry Diversification Family Ownership Multinationals

Small Medium Large Specialized Diversified Family-

held Widely-

held Multinationals Domestic Country employment protection × Industry total labor turnover

1.434** (0.288)

-0.041 (0.236)

0.387 (0.294)

0.884** (0.146)

0.751** (0.249)

0.136* (0.068)

1.185** (0.277)

0.853 (0.512)

1.031** (0.176)

1.280* (0.568)

1.195** (0.202)

ln(Sales) 0.067** 0.072** 0.023** 0.046* 0.057** 0.018** 0.071** -0.009 0.074** 0.057** 0.001 (0.020) (0.016) (0.004) (0.017) (0.018) (0.001) (0.015) (0.009) (0.015) (0.012) (0.015)

Industry Sales Share 0.744 -0.098 0.772 0.273 0.045 -0.109** 0.242** 1.354 0.013 0.037 0.229 (0.413) (0.181) (0.373) (0.522) (0.419) (0.025) (0.425) (0.634) (0.351) (0.277) (0.312)

Country Dummies (15) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Two-digit SIC Dummies (74) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Differential in Affiliation Probability (%): 5.2 -0.1 1.4 3.2 2.7 0.5 4.3 3.1 3.7 4.6 4.3

% Affiliated 19.1 70.6 8.7 8.7 8.9 3.5 19.8 2.5 21.4 17.4 28.9

R2 0.253 0.209 0.102 0.191 0.284 0.061 0.268 0.130 0.270 0.257 0.197

Observations 828,232 69,576 1,026,181 1,034,048 1,034,217 978,296 1,179,022 444,854 1,202,346 525,392 610,848 Notes: This table reports the results of Linear Probability Model regressions examining how the effect of countries’ employment protection regulations on corporate group affiliation varies by firm and group characteristics. The estimation is cross-sectional (at the firm level) and is based on the 2007 ownership structure. Sales data is for 2006 or the most recent year that data is available. Industry Sales Share is three-digit industry sales as a share of total country sales, computed over all firms in the estimation sample. Differential in affiliation probability measures how much higher the likelihood of affiliation is at the 90th percentile level of industry labor turnover with respect to an industry at the 10th percentile level when it is located in a country at the 10th percentile level of employment protection rather than in one at the 90th percentile level of employment protection. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by country. * significant at 5%; ** significant at 1%.

Table 8. Robustness Checks Dependent Variable: Dummy for Group Affiliation

(1) (2) (3) (4) (5) (6) (7)

Industry Labor

Turnover Estimation Sample

Variables All

Excl. acquired affiliates Amadeus Compustat

Excl. employment

outliers

Inc. missing

ownership

Excl. high union

industries Country employment protection × Industry total labor turnover

1.145** (0.235)

0.630** (0.208)

0.179** (0.058)

1.170** (0.244)

1.003** (0.134)

1.249** (0.422)

Industry total labor turnover, Interacted:

2nd Quartile -0.006 (0.009)

3rd Quartile 0.025 (0.019)

4th Quartile 0.053** (0.017)

Industry Sales Share 0.212 0.282 0.264 0.188 0.310 0.497 0.136 (0.401) (0.433) (0.341) (0.419) (0.314) (0.329) (0.414)

ln(Sales) 0.074** 0.070** 0.081** 0.074** 0.084** 0.065** 0.073** (0.015) (0.015) (0.017) (0.015) (0.018) (0.009) (0.015)

Country Dummies (15) Yes Yes Yes Yes Yes Yes Yes

Two-digit SIC Dummies (74) Yes Yes Yes Yes Yes Yes Yes

Differential in Affiliation Probability (%): 5.4 4.1 4.0 3.6 4.2 1.9 2.3

% Affiliated 22.2 21.1 23.7 22.3 23.9 11.8 22.7

R2 0.272 0.251 0.326 0.271 0.314 0.224 0.262

Observations 1,213,681 1,196,405 661,846 1,210,995 791,131 2,779,622 955,509 Notes: This table examines the robustness of the effect of employment protection regulations on group affiliation. The estimation is cross-sectional (at the firm level) and is based on the 2007 ownership structure. Sales data is for 2006 or the most recent year available. Industry Sales Share is three-digit industry sales as a share of total country sales, computed over all firms in the estimation sample. Differential in affiliation probability measures how much higher the likelihood of affiliation is at the 90th percentile level of industry labor turnover with respect to an industry at the 10th percentile level when it is located in a country at the 10th percentile level of employment protection rather than in one at the 90th percentile level of employment protection. Columns 3 and 4 use alternative industry turnover measures calculated using data from all firms in Amadeus (1997–2007) and Compustat (1977–2007). Column 5 drops firms below 1st and above 99th percentile of employment distribution. Column 6 includes firms with missing ownership information. Column 7 excludes firms in industries with higher share of employees with labor union contracts (at 95th percentile and above). Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by country. * significant at 5%; ** significant at 1%.

Table 9. Change in Standalone Firm Ownership Between 2007 and 2011 All Low Growth High Growth

Affiliated in 2011 7.2% 6.5% 8.5%

Formed New Group 2.7% 2.4% 3.3%

Joined a Group 4.7% 4.3% 5.5% Notes: Share of 2007 standalone firms with fewer than 50 employees that became group affiliates by 2011. Firms formed new groups by either creating or acquiring subsidiaries. Firms joined existing groups by getting acquired. Low-growth companies are firms that experience growth in the 1st quartile of sales growth distribution in 2005–2007. High-growth companies are firms that experience growth in the 4th quartile of sales growth distribution in the same period.

Table 10. The Effect of Country Employment Regulations on Ownership Changes

Dependent Variable: Dummy for Change in Ownership (1) (2) (3) (4) (5) (6)

All standalone firms Standalone firms with under 50 employees

Variables All Form new

groups Join existing

groups All High growth Low growth Country employment protection × Industry total labor turnover

0.738** (0.234)

0.740** (0.226)

0.711** (0.007)

0.578** (0.194)

0.885** (0.282)

0.474 (0.281)

ln(Sales) 0.045** 0.044** 0.031** 0.035** 0.045** 0.026* (0.010) (0.010) (0.211) (0.009) (0.013) (0.010)

Industry Sales Share -0.008 -0.003 -0.231 -0.030 -0.191 0.066 (0.304) (0.302) (0.259) (0.278) (0.233) (0.265)

Country Dummies (15) Yes Yes Yes Yes Yes Yes Two-digit SIC Dummies (74) Yes Yes Yes Yes Yes Yes

Differential in Affiliation Probability (%): 2.7 2.7 2.6 1.9 3.0 1.6

% Affiliated 8.8 8.6 5.7 7.3 8.6 6.6

R2 0.184 0.180 0.122 0.159 0.165 0.143

Observations 690,425 689,065 580,054 558,333 104,548 123,709 Notes: This table reports the results of Linear Probability Model regressions that examine the effect of employment regulation protection levels on the change of corporate group affiliation. The sample includes all firms that are classified as standalones in 2007 and merged to the 2011 version of the ownership data. An ownership change is coded as a firm being classified as a standalone in 2007 and as a group affiliate in 2011. The estimations in columns 1–3 include all standalone firms in this subsample. The estimations in columns 4–6 include standalone firms with under 50 employees in 2007. Low-growth companies are firms that experience growth in the 1st quartile of sales growth distribution in 2005–2007. High-growth companies are firms that experience growth in the 4th quartile of sales growth distribution in the same period. Sales data is for 2006 or the most recent year that data is available; includes firms with non-missing employment information. Industry Sales Share is three-digit industry sales as a share of total country sales, computed over all firms in the estimation sample. Differential in affiliation probability measures how much higher the likelihood of affiliation is at the 90th percentile level of industry labor turnover with respect to an industry at the 10th percentile level when it is located in a country at the 10th percentile level of employment protection rather than in one at the 90th percentile level of employment protection. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by country. * significant at 5%; ** significant at 1%.

Table 11. Internal Managerial Mobility by Group Characteristics and Country Labor Regulation (1) (2) (3)

Countries with Below-Median Employment

Protection

Countries with Above-Median Employment

Protection Difference in Means

All Groups 12.6% 20.2% 7.6%**

Group Size by Assets

Small groups 10.5% 11.2% 0.80%

Medium groups 9.6% 20.7% 11.1%**

Large groups 15.1% 28.0% 12.9%**

Group Size by Sales

Small groups 9.5% 12.5% 3.0%**

Medium groups 9.4% 19.4% 10.0%**

Large groups 15.5% 27.8% 12.3%** Notes: This table provides direct evidence on internal labor market mobility. Columns 1 and 2 report the percentage of internal managerial mobility out of all managerial moves (the sum of within-group and out-of-group mobility) between 2002 and 2007 in our sample groups. Group size by assets and sales is determined by distribution tertiles. The unit of observation is a manager. ** implies that the difference in means between comparison samples is significant at the 1% level.