EMPLOYMENT PROTECTION, TECHNOLOGY CHOICE, AND WORKER ALLOCATION

• Author: Eric J. Bartelsman,Joris Wind,Pieter A. Gautier
• DOI: http://doi.org/10.1111/iere.12176
• Date: 01 August 2016
• Published date: 01 August 2016

INTERNATIONAL ECONOMIC REVIEW

Vol. 57, No. 3, August 2016

EMPLOYMENT PROTECTION, TECHNOLOGY CHOICE, AND WORKER

ALLOCATION∗

BYERIC J. BARTELSMAN,PIETER A. GAUTIER,AND JORIS DEWIND1

Vrije Universiteit Amsterdam, the Netherlands and Tinbergen Institute,the Netherlands; Vrije

Universiteit Amsterdam, the Netherlands, and Tinbergen Institute, the Netherlands; CPB

Netherlands Bureau for Economic Policy Analysis,the Netherlands

We show empirically that high-risk sectors, which contribute strongly to aggregate productivity growth, are relatively

small and have relatively low productivity growth in countries with strict employment protection legislation (EPL).

To understand these findings, we develop a two-sector matching model where firms endogenously choose between

a safe technology and a risky technology. For firms that have chosen the risky technology, EPL raises the costs

of shedding workers in case they receive a low productivity draw. According to our calibrated model, high-EPL

countries benefit less from the arrival of new risky technologies than low-EPL countries. Parameters estimated through

reduced-form regressions of employment and productivity on exit costs, riskiness, and in particular their interaction are

qualitatively similar for actual cross-country data and simulated model data. Our model is consistent with the slowdown

in productivity in the European Union relative to the United States since the mid-1990s.

1. INTRODUCTION

In this article, we provide evidence that a change in the nature of technological opportunities

in the mid-1990s interacted with cross-region differences in employment protection to explain

part of the observed divergence in productivity between the United States and the European

Union (EU). Continuing improvements in computing power coupled with steepening adoption

rates of communication technology resulted in a large variance in realized productivity and

profits for firms choosing to use these technologies. The increase in variance is appealing to

individual firms because they can fully benefit from good draws while they can limit the loss

from bad draws through the option to shed workers. The increase can be good at the aggregate

level, if more firms choose the risky technology and resources flow to the firms that use the risky

technology successfully. When in the mid-1990s these technological opportunities arose, the

expected net benefits of adoptions were higher in countries with low employment protection

legislation (EPL) because the option to shut down was less costly.

Our article draws from and combines results from a variety of different literatures. The

model we use builds upon models available in the labor search literature. The relation between

Information and Communication Technology (ICT) and productivity is prominent in recent

literature on intangible investment and growth. Further, our use of model calibration and

comparison of model simulations with moments and parameter estimates from data draw on a

rich emerging macro literature. Finally, we expand on a sequence of empirical papers studying

the effects of EPL on labor markets, productivity, and macro outcomes using a new combination

of cross-country industry and firm-level data sources. We discuss these points in turn.

∗Manuscript received May 2012; revised January 2015.

1We would like to thank the referees, Bj¨

orn Br¨

ugemann, Wouter den Haan, Bart Hobijn, Marcel Jansen, Egbert

Jongen, Roland Luttens, Guido Menzio, and Julien Prat for useful comments and suggestions. We also would like

to thank participants of CAED, Tokyo, ECB/CEPR Labour Market Workshop, Frankfurt, the Search and Matching

Workshop in Konstantz as well as seminar participants at the Bank of England, University of Copenhagen, CPB,

Tinbergen Institute, VU University Amsterdam, University of Maastricht, and De Nederlandsche Bank.

Please address correspondence to: Pieter A. Gautier, Department of Economics, Vrije Universiteit Amster-

dam, De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands. Phone: +31-20-5986038/+31-6-41496310. E-mail:

p.a.gautier@vu.nl.

787

C

(2016) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

788 BARTELSMAN,GAUTIER,AND DE WIND

We develop a model where the decision to produce not only requires a fixed entry fee but

also requires some complementary factor input, namely, labor, with an associated flow of factor

payments. Firms can choose to enter in a risky or a safe sector that differ in their productivity

dynamics. Specifically, in the risky sector firms are modeled as in Mortensen and Pissarides

(1994), with a hazard of receiving a productivity shock. The safe sector firms are as in Pissarides

(2000, Chapter 1) and have a constant, known productivity. The sectors are connected with each

other through the pool of unemployed workers from which both sectors hire. This framework

is particularly useful to study labor market policies because it is simple and simultaneously

solves for the labor market stocks and flows.2Frictions are essential in our model to explain

the coexistence of vacancies and unemployed workers, but they are also needed to allow for

an equilibrium where both high- and low-productivity firms can simultaneously exist. As in

Mortensen and Lentz (2008), a key factor for aggregate productivity is the allocation of workers

to different firms.

Growth accounting exercises in the United States have shown most of the acceleration of

output growth to be due to ICT-capital deepening and to increases in total factor productivity

(TFP) associated with ICT use (for an overview of the findings, see Jorgenson et al., 2008).

Cross-region comparisons (van Ark et al., 2008) show that ICT production and use have been

much lower in the EU than in the United States and that this may explain much of the relative

slowdown. The growth accounting literature is not, however, capable of explaining why the ICT-

producing sector in the EU is smaller, why ICT investment and thus ICT-capital deepening is

lower, why the contribution from ICT-using industries is smaller, and thus why aggregate

productivity diverges. The link we make between technology choice and employment protection

and exit costs in general depends on the special nature of information and communication

technology, with sunk investments and risky market outcomes.3Consistent with the nature of

these technologies, Brynjolfsson et al. (2008) find that the cross-sectional variance of firm-level

profits in industries that use ICT intensively is higher, and has been increasing steadily since

1995, relative to the cross-sectional variance of profits in firms in low ICT-uptake industries.

Finally, Schaal (2012) gives evidence for an increase in the idiosyncratic variance of firm growth

in the United States. In this article, we document that the variance of productivity across firms

and the churn of jobs has become higher since 1995 in ICT-intensive industries. We therefore

interpret the ICT revolution as giving exogenous change in the variance of productivity shocks

that allows us to exploit and study the differential responses in high- and low-EPL countries.

The main conclusions of the empirical literature on employment protection are that the

effects of EPL on employment are negative but small. Labor force participation is typically

smaller in countries with strong EPL, and the effects on unemployment are essentially zero.

EPL reduces the flows in and out of employment and increases unemployment duration. Autor

et al. (2007) give some evidence that EPL reduces productivity at the plant level, but they

cannot rule out that their results are (partly) due to confounding economic shocks. Samaniego

(2006) gives evidence that EPL is negatively correlated with ICT diffusion, and he develops a

simple vintage capital model where a firm’s optimal size decreases over time when the firm’s

technology falls behind the frontier (of which the speed depends on the rate of technological

change). Bassanini et al. (2009) give evidence that productivity in high-turnover industries is

relatively low if EPL is strong, which is consistent with our findings (in our model turnover is

endogenous and depends on the choice of technology). Finally, Cu ˜

nat and Melitz (2012) show

that countries with flexible labor markets concentrate their exports mainly in sectors with higher

volatility. In our empirical exercises based on cross-country industry and firm-level data, we add

2The effects of EPL have been studied extensively in the search and matching literature using a single sector model.

See, for example, Br ¨

ugemann (2006), Ljungqvist (2002), and Mortensen and Pissarides (1999).

3A nice case study of such risky innovation is given by McAfee and Brynjolfsson (2008), where the benefits of

adopting an innovative ICT system arise in conjunction with a reorganization of the production process. This fits nicely

with the findings of Bloom et al. (2012) that U.S. multinational firms have high returns to investment in ICT in their

U.K. subsidiaries because they only transplant the ICT implementations that were first adopted successfully in the

United States.

EMPLOYMENT PROTECTION,TECHNOLOGY CHOICE 789

a new set of findings to the literature on the effects of EPL on labor market performance and

productivity: Risky, aggregate productivity enhancing activities are harmed relatively heavily

by EPL.4

The economic intuition in our article is related to a number of other papers that study the

effect of exit costs on risky technologies. In Saint-Paul (2002), countries with high EPL specialize

in secure goods at the end of their product cycle with stable demand whereas countries with low

EPL specialize in more innovative goods. Bertola (1990, 1994) shows how labor mobility costs

reduce the social returns to irreversible investments, and in Poschke (2009) growth is driven by

entrants who imitate the firms at the frontier (EPL reduces growth by reducing entry). Berdugo

and Hadad (2012) have a model where EPL makes innovators choose medium-tech projects,

which are more flexible in their human capital requirements than high-tech projects. Those

papers neither use the rich data sources that we use nor do they calibrate their model and

quantify the effects of EPL on productivity and the allocation of workers.5Our article is, to the

best of our knowledge, unique in attributing the divergence in productivity of the United States

and EU to the interaction between exit costs and the increasing riskiness of technology.

We calibrate our model for the United States using a variety of sources including the EU-

KLEMS data set (O’Mahony and Timmer, 2009) and a novel data set built up from firm-level

sources (Bartelsman et al., 2009, 2013; labeled BHS). By exploring new data sources, we are

able to get more information on primitives that previously had to be fixed at arbitrary values in

model calibrations. For example, we use our model to derive a relation between the underlying

ex ante mean and variance of the productivity distribution in the risky sector and the observed

(truncated) mean and variance. We can simulate the model across the empirically observed

variation across countries in EPL and a range of productivity shock variances consistent with

evidence from firm-level data, generating simulated panel data on employment and productivity

of the risky and safe sectors. Reduced-form regressions of the relative size and productivity of

the risky sector on EPL, riskiness, and their interaction using simulated data gives qualitatively

similar results as regressions using actual cross-country panel data.

We next present robust empirical evidence that high-risk innovative sectors, which are often

associated with intensive use of information and communication technologies, are relatively

small in countries with high EPL. Moreover, we confirm that productivity in the risky sectors

increased relatively slowly in high-EPL countries from the mid-1990s onward. The negative

relationship also holds for other exit frictions (e.g., low cost recovery of capital for exiting

firms).

Overall, both the simulated data from the calibrated model and our empirical country/industry

panel data exercises show that it is more advantageous to exploit new risky opportunities in

low-EPL countries. Relative to the literature, our main contribution lies in the interaction

between theory and evidence, which allows us to quantify the effects of EPL. We find that if

we increase EPL in the United States from one month of output to EU levels (five months of

output), aggregate productivity will be about 8% lower.

The article is organized as follows: Section 2 discusses our theoretical model. Section 3 gives

an overview of the data sources used for calibration and empirical exercises. The calibration

of the model and the moments we match are discussed in Section 4. The model predictions

are presented in Section 5. In this section, we also compare reduced-from regressions from

simulated and actual cross-country data. Section 6 shows that our main finding that risky

sectors are relatively smaller and have lower productivity growth in high-EPL countries can be

identified with a difference-in-differences estimation. We conclude with some reflections on the

importance of the link between EPL and productivity and with ideas for future research.

4We want to emphasize that our article looks at a firm’s decision to invest in risky or safe technologies. In a paper

considering the effort of workers in creating new knowledge, Acharya et al. (2013) argue that EPL may serve as a

commitment device to induce workers to take more risk. They present evidence that countries with high EPL have

more patenting. By contrast, and in line with our findings, Alesina and Zeira (2006) give evidence that countries with

less regulations have relatively more patents in high-tech sectors.

5Poschke (2009) is the only other paper that we are aware of that also calibrates his model, but he uses different data

sources and has a different mechanism.