ArticlePDF Available

Tenuous Link: Labour Market Institutions and Unemployment in Advanced and New Market Economies

Authors:

Abstract and Figures

International organizations and mainstream economists have consistently promoted the view that labour market rigidities are responsible for high unemployment, and that wide-ranging institutional deregulation is an appropriate policy response. Yet, as demonstrated by recent literature, the empirical support for the deregulatory view is ambiguous. This paper re-assesses this debate by bringing in new evidence from a larger group of countries, which includes advanced and new market economies. Using new data and paying special attention to the robust- ness of estimation results, we find rather thin support for the deregulatory view. The sensitivity analysis demonstrates that in most cases the adverse effects of institutions disappear with small changes in the sample or the use of alternative estimators and specifications. The impact of institutions is particularly weak in new market economies, where unemployment is related primarily to macroeconomic factors. Overall, our findings challenge the policy orthodoxy that comprehensive deregulation is the universal solution to unemployment.
Content may be subject to copyright.
Tenuous link: labour market institutions and
unemployment in advanced and new market
economies
Sabina Avdagic1,* and Paola Salardi2
1
Department of Politics, University of Sussex, Brighton, UK;
2
Department of Economics, University of Sussex, Brighton, UK
*
Correspondence: s.avdagic@sussex.ac.uk
International organizations and mainstream economists have consistently pro-
moted the view that labour market rigidities are responsible for high unemploy-
ment, and that wide-ranging institutional deregulation is an appropriate policy
response. Yet, as demonstrated by recent literature, the empirical support for the
deregulatory view is ambiguous. This paper re-assesses this debate by bringing in
new evidence from a larger group of countries, which includes advanced and
new market economies. Using new data and paying special attention to the robust-
ness of estimation results, we find rather thin support for the deregulatory view. The
sensitivity analysis demonstrates that in most cases the adverse effects of institu-
tions disappear with small changes in the sample or the use of alternative estimators
and specifications. The impact of institutions is particularly weak in new market
economies, where unemployment is related primarily to macroeconomic factors.
Overall, our findings challenge the policy orthodoxy that comprehensive deregula-
tion is the universal solution to unemployment.
Keywords: unemployment, labor market institutions, OECD countries, Central
and Eastern Europe
JEL classification: J4 and J48 General Labour Markets and Public Policy, P16 Pol-
itical Economy, P52 Comparative Studies of Economies
Introduction
The view that institutional rigidities in labour markets are at the root of Europe’s
unemployment problems has become the mainstream view in economics and the
public policy discourse. Originally espoused by the OECD Jobs Study (1994),
this view has found support in a large body of literature (Scarpetta, 1996;
Nickell, 1997,2005;Siebert, 1997;Elmeskov et al., 1998;Nunziata, 2002;IMF,
#The Author 2013. Published by Oxford University Press and the Society for the Advancement of Socio-Economics.
All rights reserved. For Permissions, please email: journals.permissions@oup.com
Socio-Economic Review (2013) 11, 739–769 doi:10.1093/ser/mwt009
Advance Access publication May 6, 2013
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
2003;Be
´lot and van Ours, 2004). While the OECD has subsequently toned down its
initial recommendations about across-the-board institutional deregulation, their
empirical studies continue to emphasize a link between certain labour market rigid-
ities and unemployment (Bassanini and Duval, 2006;OECD, 2006). Although
studies that support this view differ in their conclusions about which labour
market institutions have more of an effect, they generally agree that deregulation
is needed to fight unemployment. Despite its popularity, however, the evidence
supporting the deregulatory view is not conclusive. Some recent contributions
challenge the empirical findings that inform this position and argue that there is
little or no convincing evidence that links institutions to unemployment (Baker
et al., 2005;Baccaro and Rei, 2007,Howell, 2005;Vergeer and Kleinknecht, 2012).
This paper re-assesses this debate by bringing in new evidence from a larger
group of countries, which in addition to OECD economies includes the 10 new
European Union (EU) members fromCentral and Eastern Europe (CEE).
1
The in-
clusion of CEE countries promises new insights for at least three reasons. First, these
countries have experienced substantial institutional reforms over the past two
decades, which allows us to establish more clearly the impact of institutional
changes on unemployment. Second, analysing the role of labour market institu-
tions on a larger sample increases data variability and helps to disentangle the
effects of different institutional settings on unemployment. Finally, most CEE
countries have been under a strong policy influence of international organizations
promoting the deregulatory view. Yet, due to data limitations, the empirical evi-
dence supporting this policy advice has been far less compelling than commonly
believed. Indeed, only a few studies examine the link between institutions and un-
employment in CEE, but due to the lack of data they often rely on simple correla-
tions and cross-sectional analysis (see Cazes and Nesporova, 2003;Behar, 2009).
More recent studies employing panel data techniques focus either on a sample of
OECD economies that includes only a couple of CEE countries since the mid- to
late-1990s (Ederveen and Thissen, 2007;Fialova and Schneider, 2009), or alterna-
tively consider longer panels but focus exclusively on post-communist countries
(Schiff et al., 2006;Lehman and Muravyev, 2009). Given the drawbacks associated
with data availability, it is not surprising that the findings of this literature are not
conclusive. While several studies find a positive link between unemployment and
institutions, such as the tax wedge, employment protection and unemployment
benefits, the evidence supporting these findings is rathe r weak. This paper addresses
this lacuna in prior research. Relying on a newly constructed dataset of labour
market institutions in CEE and the standard data sources for OECD economies,
we re-asses the conventional view that institutional rigidities are responsible for
1
CEE countries include Bulgaria, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland,
Romania, Slovakia and Slovenia.
740 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
high unemployment.
2
Our objective is three-fold. First, we estimate a standard
dynamic model of unemployment on data covering most EU and OECD countries
since 1980. We run a battery of sensitivity checks to assess the robustness of these
estimates. Second, we examine whether the determinants of unemployment are dif-
ferent in advanced and new market economies. Finally, we assess the possibilit y that
institutions affect unemployment not only directly, but also through their interac-
tions with macroeconomic shocks.
Overall, we find no compelling support for systematic deregulation. Only a few
institutions, namely union density and to a lesser extent unemployment benefits,
seem to be associated with high unemployment, but we show that these findings
are highly fragile. Sensitivity checks suggest that the adverse effects of these institu-
tions diminish or disappear entirely with small changes in the sample and data, or
the use of alternative estimators. While we do find some indications that simultan-
eous reforms of unemployment benefits and the tax wedge may be beneficial, this
evidence is also not immune to small changes in the sample. Meanwhile, the effects
of macroeconomic controls, such as GDP growth, and the unemployment-
reducing effects of wage bargaining coordination remain fairly robust. When we
consider the two groups of countries separately, it becomes clear that the impact
of institutions in the new market economies is even weaker: Here, most labour
market institutions are already fairly liberal, and unemployment seems to be
related primarily to macroeconomic developments. Evidence that institutions
affect unemployment indirectly, by exacerbating the effects of macroeconomic
shocks, is similarly inconclusive and highly contingent on the statistical properties
of the models. In sum, our analysis demonstrates rather thin support for the de-
regulatory view. A simple labour market rigidity story appears too simplistic to
account for developments in unemployment in the EU and OECD countries
over the last three decades.
The paper is divided in seven sections. Section 1 provides an overview of un-
employment trends and compares the experience of the advanced OECD econ-
omies and the new market economies from CEE. Section 2 presents the main
models and hypotheses. Section 3 offers a brief summary of the data, while
2
Some authors argue that the employment rate is a better indicator of the overall health of the labour
market because joblessness may be masked by low labour force participation, different types of active
labour market policies, or widespread use of early retirement options (Kenworthy, 2008, p. 62). We
focus on the unemployment rate primarily because the literature that supports the deregulatory view
predominantly uses this measure as an indicator of the current labour market performance. In
addition, the unemployment rate arguably carries more political weight: it seems reasonable to
assume that incumbent politicians are punished more by high unemployment than low employment.
Finally, unemployment and employment rates are highly correlated. Given this, it is not surprising
that using the employment rate as the dependent variable does not generate markedly different results
from those presented here.
Labour market institutions and unemployment 741
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Section 4 outlines our estimation strategy. Section 5 provides a detailed discussion
of the results and employs several sensitivity checks to assess their robustness.
Section 6 discusses the possible reasons for the weak effects of institutions and
for the differences in results within this literature. The last section concludes by
summarizing the key findings.
Labour market institutions and unemployment in the EU and OECD
countries
Figure 1plots the evolution of unemployment and key labour market institutions in
continental Western Europe, the USA, and CEE during 1980 –2007 (for summary
statistics, see online appendix). The first panel shows that since the early 1980s un-
employment in continental Europe has been generally hig her than in the USA with a
particularly large gap during the 1990s. Due to the transformational recession, CEE
countries experienced a dramatic jump in unemployment during the early 1990s
that in many cases surpassed unemployment levels in the advanced countries.
The gap even widened during the late 1990s when a second round of
restructuring in CEE triggered a new increase in unemployment. More recently,
however, the differences between countries have narrowed substantially. The
remaining panels reveal some striking differences in the development of the
main labour market institutions. While union density has been generally declining,
this decline has been particularly striking in CEE where the end of compulsory
union membership led to large membership losses. CEE countries also experienced
a dramatic decline in the unemployment benefit replacement ratio. As unemploy-
ment soared in the early transition, CEE governments adopted a series of reforms
reducing the generosity of benefits. By the late 2000s, the average replacement rate
was significantly lower than in continental Europe, and very similar to the USA.
Employment protection is also more liberal in CEE than in continental Europe,
but not as liberal as in the USA. The gap between continental Europe and CEE
has narrowed in the 2000s when the CEE countries were required to adopt a
number of EU directives that define the procedures regarding collective dismissals
and increased protection of temporary employees. CEE also occupies the middle
position in terms of wage bargaining coordination, with a peak in the early- to
mid-1990s and the subsequent decline to a level that is considerably below contin-
ental Europe. The evolution of the tax wedge displays a different trend from other
variables: here CEE has experienced a steady increase so that in the late 1990s the
average tax wedge exceeded the level in continental Europe and remained at a
relatively high level thereafter.
In sum, this figure shows no clear indication that unemployment is a conse-
quence of labour market rigidities. Although union density and the benefit replace-
ment ratio in CEE show a steep decline throughout the period, unemployment has
742 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
been increasing, with minor dips, pretty much until the mid-2000s. Similarly, un-
employment in CEE was declining at the time when employment protection was
strengthened and wage coordination was largely stable. The lack of a clear associ-
ation between institutions and unemployment is also evident in continental
Europe and the USA. The steady decline in unionization and little movement
with respect to employment protection and wage bargaining coordination do
not seem to offer a convincing explanation of unemployment. Similarly, the
decline in unemployment during the 1990s is at odds with the moderate increases
in benefit generosity. Only the tax wedge shows occasionally some association with
unemployment, but apart from the USA, this association appears rather weak.
Figure 1 The evolution of unemployment and labour market institutions.
Source: See section on data below.
Labour market institutions and unemployment 743
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
In a series of separate plots (not shown here), we examined in more detail bivari-
ate associations between unemployment and labour market institutions in differ-
ent groups of countries. Among these, only employment protection legislation
(EPL) in continental Europe shows a mildly positive relationship with unemploy-
ment (the correlation coefficient of 0.14 for regular contracts and 0.36 for tempor-
ary contracts). However, a closer look reveals that this relationship is driven by only
a few countries, most notably the Netherlands, and to a lesser extent Spain, Greece
and Switzerland. In most countries there is no clear relationship between EPL and
unemployment, and in some countries, such as Sweden, Portugal, Belgium and
Germany, there is even a significant negative relationship. In addition, in countries
that display a positive relationship between EPL and unemployment a causal story
underlying this relationship is not quite clear. For example, while changes in EPL
may have contributed to the rise of part-time employment in the Netherlands,
most scholars tend to agree that the key ingredient of the ‘Dutch employment
miracle’ was not these changes, but the willingness of unions to accept wage mod-
eration (Visser and Hemerijck, 1997). In Spain, a moderate decline in unemploy-
ment during 1994 2007 has been often attributed to the increasing use of
temporary contracts. However, recent research shows that a large gap in EPL strict-
ness between regular and temporar y contracts may actually contribute to rising un-
employment by increasing the number of separations (Bentolila et al., 2010).
Clearly, Figure 1provides only a very crude picture of the links between institu-
tions and unemployment and it conceals considerable differences that exist within
the groups. Within continental Europe, unemployment in the ‘Big Four’—Spain,
France, Italy and to a lesser extent Germany—was hovering around 10% for
most of the period under consideration, and occasionally even around 20% in
Spain. Greece and Belgium also struggled with unemployment for most of this
period, while Finland and Denmark experienced episodes of high unemployment
at different points in time. Other countries, however, had less of a problem with un-
employment, and some of them (Austria, Norway and the Netherlands) even per-
formed better than the USA during the late 1990s and early 2000s. Differences in
unemployment emerged also within CEE countries, despite the fact that they
embarked upon transition with generally similar labour market conditions.
While Poland, Bulgaria and Slovakia had problems with high unemployment for
most of the period, and the Baltic countries were affected from the mid-1990s,
the Czech Republic and Slovenia managed to keep unemployment at relatively
low levels. As documented by previous research, there are also notable cross-
country differences in labour market institutions within these groups (Baker
et al., 2005;Howell, 2005;Nickell et al., 2005;Schiff et al., 2006), and they do not
allow straightforward conclusions about the link between institutions and un-
employment. To examine this variation across countries and over time, in the re-
mainder of this paper we turn to time-series cross-section analysis.
744 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Models and hypotheses
We estimate a dynamic model of unemployment that has been used widely in the
literature (Nickell et al., 2005; see also Layard et al., 1991;Nunziata, 2002;IMF,
2003;Amable et al., 2006;Baccaro and Rei, 2007). In this model the unemployment
rate depends on a set of labour market institutions and macroeconomic controls.
The former determine the equilibrium level of unemployment, while the latter
account for short-term deviations from the equilibrium level. The model has the
following form:
ui,t=
b
0+
b
1ui,t1+Sj
g
jxj,it +Sk
h
kzk,it +Sn
d
nvn,it +
a
i+
l
t+1i,t,(1)
where ui,trepresents the unemployment rate in country iat time t,ui,t1is the
lagged unemployment rate, xj,it are jinstitutional variables, zk,it represent kmacro-
economic controls, vn,it are ninteractions between labour market institutions and
1i,tis the stochastic residual. The model also includes country dummies,
a
i, which
account for unmeasurable time-invariant country-specific characteristics that may
influence unemployment, and year dummies,
l
t, which denote time-varying
shocks affecting all countries. The lagged-dependent variable is included among
the predictors to capture the persistence of unemployment and hysteresis effects
(Nickell et al., 2005).
The vector of institutional variables includes as follows:
Sj
g
jxj,it =
g
1EPi,t+
g
2BRRi,t+
g
3TWi,t+
g
4UDi,t+
g
5BCi,t,(2)
where EPi,tis employment protection legislation, BRRi,tis the unemployment
benefit replacement rate, TWi,tis the tax wedge, UDi,tis union density and BCi,t
is wage bargaining coordination. In the standard competitive model, employment
protection legislation dampens job creation because employers are reluctant to hire
new workers for fear of not being able to fire them easily when the need arises
(Addisson and Texeira, 2003). However, strict EPL also increases job retention as
employers make fewer layoffs during downturns. In addition, strong job protection
may encourage investments in training and enhance overall productivity perform-
ance (Estevez-Abe et al., 2001). Thus, the overall effect of EPL on unemployment, as
Bertola (1992) has argued, is theoretically ambiguous, and it may depend on issues
such as the functional form of labour demand functions, the discount rate, labour
turnover and wage flexibility. The impact of unemployment benefits is generally
less ambiguous. Generous benefits are thought to increase unemployment
because they indicate a high reservation wage, which makes unemployed indivi-
duals both more reluctant to seek actively for jobs and to accept available jobs
(Nickell, 1997;Holmlund, 1998). In addition, generous benefits may contribute
to unemployment by making unions more resolute in pushing for higher wages
Labour market institutions and unemployment 745
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
(Layard et al., 1991).
3
The tax wedge, the difference between the labour cost to
employers and the take-home wage for employees, is generally expected to influ-
ence labour market performance negatively by reducing the demand for labour.
However, theoretically the distribution of taxes between employers and labour
determines the actual impact of the tax wedge (Nickell, 1997). If employees carry
most of the tax burden, this variable alone is not likely to reduce labour demand.
At the same time, the impact on labour supply is indeterminate since a low take-
home pay may either reduce workers’ incentive to accept jobs and keep the existing
ones, or it may motivate them to seek additional jobs. Union density indicates
union bargaining power. In the orthodox view, unions tend to raise wages, and
therefore a high share of workers belonging to unions is expected to increase un-
employment, particularly in contexts of a highly elastic labour supply. Strong
unions are also associated with compressed wage structures, which may reduce
the prospects for employment of low-skill workers (Rueda and Pontusson,
2000). In contrast, the effect of wage bargaining coordination is generally consid-
ered to be beneficial for labour market performance. Because unions in coordinated
systems internalize the externalities of their wage policies, it is expected that real
wages, and thus unemployment, will be lower than in systems characterized by un-
coordinated bargaining (Hall and Franzese, 1998).
The macroeconomic controls include as follows:
Sk
h
kzk,it =
h
1CPIi,t+
h
2GDPi,t+
h
3TOTi,t+
h
4RIRi,t,(3)
where CPI is the change in inflation, GDP is GDP growth, TOT is the terms of trade
and RIR is the real interest rate. Change in inflation captures the influence of eco-
nomic cycles (Nickell, 1997). Following the logic of the Phillips curve, this variable
should be negatively related to unemployment in the short run. Because there are
some concerns about the suitability of this variable in the context of transition
economies (Cazes and Nesporova, 2003), we include GDP growth as an additional
control. A fall in output should be associated with higher unemployment values.
The terms of trade variable should have a negative relationship with unemploy-
ment. A deterioration of the terms of trade requires a downward adjustment of
real wages. If wages do not respond accordingly,unemploymen t is likely to increase.
The real interest rate affects capital accumulation and can cause shifts in labour
demand. This variable should be positively associated with unemployment,
because an increase in real interest rates is likely to reduce aggregate demand,
thereby generating higher unemployment rates (Baker et al., 2005).
Finally, we include three types of interactions among institutional variables that
allow us to examine possible complementarities across labour market reforms (see
3
However,if generous benefits increase the effectiveness of the job matching process, their impact will be
less clear and theoretically indeterminate.
746 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Amable et al., 2006;Bassanini and Duval, 2006). The first is the interaction between
the tax wedge and the generosity of unemployment benefits. Be
´lot and van Ours
(2004) argue that simultaneous reductions in the tax burden and unemployment
benefits have been important ingredients in reforms in countries that managed
to reduce unemployment during the 1990s (see also Bassanini and Duval, 2006).
The theoretical rationale is that if workers shoulder most of the labour taxes, the
incentives of job seekers to invest heavily in job search will be lower provided
that unemployment benefits are generous. The other two interactions include em-
ployment protection on the one hand, and the tax wedge and unemployment ben-
efits on the other. These interactions help to examine whether the impact of
employment protection, which is theoretically ambiguous, may be high when asso-
ciated with another institutional rigidity. In theory, one channel through which
these interactions work is the interdependence of the search intensities of
workers and employers. High labour taxes may discourage vacancy posting
because they reduce either the demand for or supply of labour. By increasing the
costs of hiring and firing, strict employment protection also discourages vacancy
posting. Consequently, the search intensity of workers may be reduced because
the likelihood of finding a job is smaller. The adverse effects of these two institutions
therefore may amplify each other. The interaction between employment protection
and unemployment benefits follows a similar logic. Strict employment protection
may reduce vacancy postings, and this effect may be amplified by generous benefits.
The latter institution reduces workers’ incentives to look for jobs, which conse-
quently may also discourage vacancy posting (see OECD, 1999;IMF, 2003;
Amable et al., 2006). In addition, we also estimate a number of alternative
models that include additional interactions and variables. Section five outlines in
detail the rationale for and the results of these models. All interactions are specified
as deviations from cross-country and over-time sample means. Using such formu-
lation, the coefficients of these institutions in levels can be interpreted as the coeffi-
cients of a country that has the average level of a given institution (Nunziata, 2002,
p. 9). A negative and significant interaction coefficient between two variables that
increase unemployment would suggest reform complementarity (see Bassanini and
Duval, 2006, p. 21).
Finally, we supplement this analysis with an examination of interactions between
institutions and macroeconomic shocks. As argued by Blanchard and Wolfers (2000,
p. C17), labour market institutions may affect the impact of shocks on unemploy-
ment as well as the persistence of unemployment in response to shocks. For
example, with respect to the first channel, a slowdown in productivity growth may
result in unemployment unless wages are adjusted downwards, and this adjustment
may be more difficult in systems with strictemployment protection or generous un-
employment benefits. Similarly, once the adverse shocks generate an increase in un-
employment, the institutions may prolong the time needed for unemployment to
Labour market institutions and unemployment 747
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
return to its normal level. To examine this hypothesis, we re-estimate the Blanchard
and Wolfers model. This model captures the interaction between institutions and
common unobservable shocks, which are treated as time effects:
ui,t=
l
t(1+Sj
g
jxj,it)+
a
i+1i,t,(4)
where ui,tisunemployment in country iat time t,
a
iis the countryeffect for country i,
l
tis the country-unvarying time effect for year tand xjis the same set of institutions
considered in the baseline linear model. The effects of common shocks depend on
labour market institutions, and the coefficients
g
icapture this indirect effect of insti-
tutions on unemployment.
Data
Our data cover 32 EU and OECD countries, including all current EU member states
(apart from Cyprus and Malta), Norway, Switzerland, the USA, Canada, Australia,
New Zealand and Japan during 1980 2009. The series for CEE countries are
shorter, starting roughly at the beginning of their democratic transitions.
4
This is
a significantly larger sample than commonly used in the literature. While recent
analyses by Feldmann (2009) and Bernal-Verdugo et al. (2012) include 73 and 97
countries, respectively, their time series are rather short. The former focuses on 3
years only, while the latter uses series that vary from 3 to 12 years.
Our dependent variable captures the number of unemployed persons as a per-
centage of the labour force and is based on labour force surveys (IMF World Eco-
nomic Outlook and EBRD). Among our independent variables, some are newly
constructed. An important contribution of our analysis is that it includes the
longest and previously unavailable series that measure the strictness of EPL in
CEE countries on a yearly basis during 1990 2009. This is in contrast to the
series provided by the OECD, which are interpolated from a few data points. As
such, these data reflect more accurately the differences in the timing and the
extent of EPL reforms.
5
We combine our data for CEE countries with Allard’s
(2010) EPL index for advanced economies, which also captures annual changes
in legislation and is based on the same methodology. The series on the unemploy-
ment benefit replacement rates for CEE countries is also newly constructed based
on the scheme used by the OECD. These data capture the gross replacement rates
in the first year of unemployment across two levels of earnings (67 and 100% of
4
For reasons of comparability, the analysis excludes the first 3 years of post-communist economic
transformation when these countries experienced profound macroeconomic shocks.
5
These data can be accessed at http://store.data-archive.ac.uk/store/collaborativeCollectionEdit
.jsp?collectionPID=archive%3A598&tabbedContext=allCollection.
748 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
average wage). As a check, we also re-ran all the models with the recently released
data on the net replacement rate (van Vliet and Caminada, 2012). The results
were not markedly different from those presented below.
In addition, the analysis uses new data on the tax wedge provided by Labartino
(2010). This database provides longer and more complete series for this sample of
countries than the OECD and Eurostat data. Data on union density and wage co-
ordination are taken from Visser (2009). Data on macroeconomic controls come
from the International Monetary Fund’s International Financial Statistics
(GDP), the World Bank’s World Development Indicators (real interest rate),
OECD National Accounts data files (inflation) and the European Commission’s
AMECO database (terms of trade). Data on central bank independence, used in
one model, come from Crowe and Meade (2007). The number of countries
covered in the final models was governed by data availability. The main specifica-
tions outlined above are based on a sample of 26 countries. The lower number of
observations in these models is primarily a consequence of missing data on the
tax wedge for Bulgaria, the Czech Republic, Slovenia, New Zealand and Switzer-
land, and on employment protection for Luxembourg.
Estimation strategy
We compare the results from two different estimators. The first is a panel weighted
least squares estimator (PWLS), which is the most commonly used estimator in the
literature that supports the deregulatory view. This model assumes country-
specific heteroskedasticity and employs a Prais Winsten transformation to
address a first order (AR1) autoregressive structure in the errors (a common esti-
mated rho). In comparison to the Parks estimator, which produces severely under-
estimated standard errors in analyses where T is not significantly larger than N, this
feasible generalized least squares (FGLS) estimator has better properties. However,
this estimator is not designed to correct for contemporaneously correlated errors,
which characterize our data. In such cases, PWLS may also suffer from overoptimis-
tic errors, which is why we prefer the ordinary least squares procedure with panel
corrected standard errors (OLS-PCSE) (Beck and Katz, 1995). Used widely in com-
parative political economy, this estimator applies OLS with corrected standard
errors to control for common properties of this type of data, including panel het-
eroskedasticity and contemporaneous correlation of the error terms. All models
include country and year effects, as indicated by the F-test for their inclusion.
Given that the inclusion of the lagged-dependent variable can make the fixed-effect
estimator biased due to the correlation between the demeaned-lagged-dependent
variable and the error term (Nickell, 1981), we also estimated the least squares
dummy variable model corrected for the so-called Nickell bias, as suggested by
Kiviet (1995). The results of models that use the Kiviet estimator (obtainable
Labour market institutions and unemployment 749
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
upon request) do not differ appreciably from the results reported below. This sup-
ports the conclusions of Beck and Katz (2011) that this bias is small in panels where
T is around twenty or more. Finally, Equation (4), which captures the interaction
between shocks and institutions, is estimated by non-linear least squares, as pro-
posed by Blanchard and Wolfers (2000).
Empirical results
Table 1reports results for four different specifications across the two estimators.
Models (1) and (2) are our baseline specification. Models (3) and (4) add the inter-
actions between the tax wedge, unemployment benefits and employment protec-
tion.
The coefficients of the lagged-dependent variable are high (0.81 –0.83), indicat-
ing considerable persistence of unemployment, but also potential problems with
stationarity. Unit root tests, however, suggest that most series are stationary.
6
The Augmented Dickey Fuller (204.28) and Philips-Perron (83.34) tests
(Maddala and Wu, 1999) reject the null hypothesis of non-cointegration at the
1% level. The dynamic specification does not eliminate fully serial correlation in
the residuals, but Monte Carlo evidence indicates that the associated bias is unlikely
to be substantial (Beck and Katz, 2011).
At first glance, columns 1 4 suggest that labour market institutions have a role
in explaining unemployment. At the same time, however, it is evident that support
for the deregulatory view is far from strong. Among the institutions, only union
density and to a lesser extent unemployment benefits are consistently positively
associated with unemployment. While the finding about union density seems to
hold across the estimators and specifications, the significance of the benefit replace-
ment rate is more sporadic when FGLS is used. The tax wedge has a noticeable effect
only in interactions with other institutions. Specifically, Models (3) and (4) signal
reform complementarity, suggesting that a reduction in the tax wedge may
augment unemployment-reducing effects of replacement rate cuts. Employment
protection legislation does not have any discernable impact on unemployment, re-
gardless of the choice of estimators and specifications. Finally, wage coordination
shows unemployment-reducing properties, and this finding is robust to changes
in specifications and estimators. The coefficients for all macroeconomic controls
are signed as expected, but only growth is robustly significant across the models.
The results are similar if we use the output gap instead of GDP growth.
6
We ran unit root tests with one lag and two lags, with and without drift, with and w ithout trendand with
and without demean option. Central bank independence, used in one model, is the only variable for
which it is clear that the null of a unit root cannot be rejected.
750 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Table 1 Determinants of unemployment in the EU and OECD countries, 1980 –2006
(1) (2) (3) (4) (5) (6) (7) (8)
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
L.ur 0.821***
(0.028)
0.813***
(0.026)
0.827***
(0.028)
0.818***
(0.026)
0.825***
(0.028)
0.825***
(0.024)
0.816***
(0.028)
0.814***
(0.029)
BRR 0.481
(0.601)
1.306***
(0.428)
0.136
(0.642)
0.810**
(0.405)
0.491
(0.616)
1.248***
(0.432)
0.707
(0.644)
1.354**
(0.576)
EP 20.024
(0.156)
0.166
(0.220)
0.066
(0.169)
0.186
(0.233)
20.009
(0.159)
0.150
(0.240)
TW 0.821
(0.661)
0.981
(0.698)
1.150
(0.699)
1.582**
(0.747)
0.888
(0.668)
1.324*
(0.706)
0.836
(0.669)
1.050
(0.720)
BC 20.146***
(0.055)
20.236***
(0.068)
20.117**
(0.057)
20.206***
(0.073)
20.111*
(0.064)
20.155**
(0.073)
20.162***
(0.056)
20.254***
(0.064)
UD 4.393***
(1.330)
2.855***
(0.755)
4.305***
(1.312)
3.542***
(0.713)
4.417***
(1.368)
2.470***
(0.574)
4.288***
(1.364)
3.014***
(0.781)
GDP 20.262***
(0.022)
20.284***
(0.019)
20.259***
(0.022)
20.287***
(0.019)
20.264***
(0.023)
20.285***
(0.018)
20.247***
(0.023)
20.279***
(0.019)
CPI 20.012
(0.022)
20.003
(0.036)
20.019
(0.022)
20.009
(0.038)
20.012
(0.022)
0.001
(0.035)
20.013
(0.022)
20.002
(0.036)
TOT 20.003
(0.007)
20.019***
(0.007)
20.002
(0.007)
20.016**
(0.007)
20.004
(0.007)
20.020***
(0.008)
20.003
(0.007)
20.019**
(0.008)
RIR 0.018
(0.014)
0.028
(0.020)
0.018
(0.014)
0.029
(0.019)
0.021
(0.014)
0.032*
(0.020)
0.019
(0.014)
0.028
(0.020)
EP*BRR 0.176
(0.532)
0.562
(0.523)
EP*TW 0.557
(0.836)
0.091
(0.931)
BRR*TW 212.873***
(3.990)
212.228***
(4.037)
Continued
Labour market institutions and unemployment 751
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Table 1 Continued
(1) (2) (3) (4) (5) (6) (7) (8)
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
FGLS het
(AR1) OLS-PCSE
CBI 0.025
(0.290)
20.604*
(0.343)
BC*CBI 0.187
(0.203)
0.554***
(0.151)
EPR 0.023
(0.157)
0.099
(0.118)
EPT 0.024
(0.087)
0.087
(0.120)
EPR*EPT 20.050
(0.091)
0.011
(0.069)
Constant 22.610
(3.254)
2.502***
(0.850)
22.690
(3.274)
1.908**
(0.822)
22.696
(3.099)
2.602***
(0.878)
22.597
(3.249)
2.532***
(0.821)
Country and year effects Yes Yes Yes Yes Yes Yes Yes Yes
No. of obs. 451 451 451 451 451 451 448 448
R
2
0.949 0.949 0.950 0.949
Estimated rho 0.353 0.357 0.348 0.359
Durbin M test for
remaining serial
correlation of residuals
Coeff. 0.092
P-value ¼0.026
Coeff. 0.092
P-value ¼0.026
Coeff. 0.155
P-value ¼0.000
Coeff. 0.155
P-value ¼0.000
Coeff. 0.130
P-value ¼0.002
Coeff. 0.130
P-value ¼0.002
Coeff. 0.113
P-value ¼0.008
Coeff. 0.113
P-value ¼0.008
Multicollinearity test:
mean VIF
1.25 1.25 1.43 1.43 1.30 1.30 1.37 1.37
Note: BRR, unemployment benefit replacement rate; EP,employment protection; TW, tax wedge; BC, wage bargaining coordination; UD, union density; GDP,GDP growth; CPI, change
in inflation; TOT, terms of trade; RIR, real interest rate; CBI, central bank independence. Standard errors in parentheses: *P,0.10, **P,0.05, ***P,0.01. Waldtests on country and
time dummies, macro controls and interactions reject the null hypothesis at 1%. The Breusch –Pagan test for cross-sectional independence [Chi
2
(325) ¼650] and group-wise het-
eroskedasticity [Chi
2
(26) ¼214.05] reject the null. Augmented Dickey Fuller
(204.28) and Phillips Perron
(83.34) tests imply rejection of null hypothesis of cointegration at 1%.
752 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Columns 5 8 augment the baseline model with further interactions that the
previous literature found to be significant in explaining unemployment. Models
(5) and (6) include the interaction between wage coordination and central bank in-
dependence. The latter variable, when observed in isolation is generally expected to
increase unemployment, but it has been shown that its adverse effect tends to be
lower when wage setting is highly coordinated (Hall and Franzese, 1998). In such
contexts, the bargaining actors are more sensitive to the likely response of monetary
policy regarding wage settlements than in non-coordinated systems. This inter-
action appears significant only in the OLS-PCSE model, but its positive coefficient
is at odds with the standard view in the literature. However, we do not place much
credence in this finding given that the significance of this coefficient is evidently
sensitive to the choice of estimators as well as specifications. In models that
include only inflation or GDP growth as the only control, this interaction is not sig-
nificant. Models (7) and (8) examine in more detail the impact of employment pro-
tection legislation. Here, we disaggregate this variable on rules for regular and
temporary contracts. In this way, we consider the possibility that the insignificant
coefficient of employment protection in previous columns may mask two opposite
effects, namely that EPL on regular contracts increases unemployment, while EPL
on temporary contracts pushes in the opposite direction (Bassanini and Duval,
2006). We find no support for this hypothesis.
In separate regressions, we also estimated models that include the duration of
unemployment benefits, which (following the FRDB Social Reforms database)
reflects the number of months during which benefits are payable. We found no evi-
dence that this variable increases unemployment either independently or through
the interaction with the replacement rate. We also experimented by including the
minimum wage, but this variable was not consistently significant in any models.
Finally, we examined the impact on unemployment of the coverage by collective
agreements. This is because union density may underestimate the strength of
unions in countries where a low proportion of the labour force belongs to
unions, but a large share is covered by collective agreements (e.g. France). Given
the lack of full time series for bargaining coverage for all countries we could only
use the average values of this variable for the whole period. Following Bassanini
and Duval (2006), we created a dummy for high bargaining coverage where coun-
tries with coverage exceeding 50% were assigned score 1. Since this variable is time-
invariant, its impact can be gauged only through interactions with other institu-
tions. No interaction, however, turned out to be consistently significant across
the models.
In sum, the main conclusions from Table 1remain unaffected in the alternative
specifications. Among institutions union density remains consistently associated
with unemployment, while wage coordination helps to reduce unemployment.
The results also suggest that generous benefits may play a role, but this finding is
Labour market institutions and unemployment 753
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
less stable and specification-dependent. It needs to be noted though that although
union density and benefit generosity seem to be associated with higher unemploy-
ment, the magnitude of these effects is relatively small. The estimates of our baseline
OLS-PCSE model imply that the impact of a 20 percentage point increase (equal to
one standard deviation) in union density yields an increase in the unemployment
rate of 0.57 percentage points. Similarly, a 19.4 percentage increase in the benefit
replacement rate is associated with an increase in the unemployment rate of only
0.25 percentage points.
Before exploring in more detail the sensitivity of these results, the potential
endogeneity of institutions needs to be addressed. Although the mainstream litera-
ture focuses on the impact of institutions on unemployment, it is reasonable to
think that the state of the labour market and employment prospects influence deci-
sions of policy makers about reforms. For example, Howell and Rehm (2009) have
shown that the causality may run from unemployment to benefit generosity rather
than vice versa. Although this hypothesis is plausible, Granger causality tests on our
data show no clear evidence of reverse causality.
7
One could, however, argue that
endogeneity problems may be still present if omitted variables influence simultan-
eously institutions and unemployment. In separate regressions, we checked this
possibility by re-estimating our baseline model using a difference GMM estimator
(Arellano and Bond, 1991), where institutions are instrumented with their lagged
values. The results are not very different from our baseline point estimates.
8
Sensitivity analysis
The finding that union density and benefit replacement rates are positively asso-
ciated with unemployment is in line with the mainstream literature on the
subject. But how robust are these findings? As we saw, the results are generally
not sensitive to changes in specifications. In addition, the inclusion of fixed
effects is supposed to capture possible country- and year-specific influences.
7
We performed Granger causality tests by estimating models with two lags of the unemployment rate and
labour market institutions. Benefit generosity and union density are of particular interest, because their
baseline point coefficients are positive and significant. The F-statistic of the two lagged term of the
explanatory variables is not significant, implying that causation does not run from unemployment to
institutions. Results obtainable upon request.
8
Because this estimator is designed for small T panels, we follow the approach used by Bassanini and
Duval (2009) and estimate these models on 5-year averaged data. Levels of endogenous variables
dated t22 and earlier are used as instruments. In the first model, all institutions are treated as
endogenous variables, while in the second model only those institutions that have a significant
impact on unemployment in the baseline models are treated as endogenous. The results of these
models are very similar, although the first one is more fragile due to a higher number of instruments.
In both models benefit generosity and wage bargaining coordination retain their significance.
754 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Nonetheless, given the relatively small sample size, it is still possible that some coun-
tries or even individual observations greatly influence the coefficient estimates.
Thus, we perform two additional checks. First, we run a jackknife analysis on the
models presented in Table 1. This procedure re-estimates repeatedly the same
models by omitting from the sample one country each time. Second, we re-estimate
the same models by removing a small number of potentially influential observa-
tions that were identified through a combination of a visual inspection and dffits
and covratio cut-offs. This exercise revealed six observations that may have a dis-
proportional influence on the results: Lithuania in 1995 and 2002, Spain in 2001,
Latvia in 1996, Romania in 2004 and Finland in 1993. Table 2shows the outcomes
of these checks for the baseline model. Column 3 reports the original estimates.
Column 6 shows the results when the six outliers are excluded from the sample.
The remaining columns report the maximum and minimum value of coefficients
obtained by jackknifing, and the country that was omitted when those coefficients
were obtained.
As evident, the sensitivity analysis implies significant differences in the substan-
tive conclusions about the impact of institutions. Although none of the originally
significant coefficients changes the sign in the jackknife analysis, the results are
clearly not robust as the key coefficients become insignificant upon exclusion of
a single country. In particular, the results regarding union density, which showed
up as the main culprit of unemployment, are fragile and hinge entirely on the pres-
ence of one country in the sample—omitting Lithuania reduces the coefficient con-
siderably and makes it statistically indistinguishable from zero. Similarly, the
benefit replacement ratio also becomes insignificant if Lithuania, Portugal or
Austria is excluded from the sample. Among initially significant institutions only
wage bargaining coordination remains robust, but this institution is associated
with lower unemployment. Turning to macroeconomic controls, GDP growth
retains its negative sign and significance irrespective of the changes in the
sample. The main findings of the jackknife analysis are corroborated when we
exclude the six country-years identified as outliers. As column 6 shows, labour
market institutions do not seem to have a direct detrimental effect. Again bargain-
ing coordination remains robustly associated with lower unemployment.
The results (not shown) of the sensitivity anal ysis for the main interaction model
(Table 1, column 4) are less conclusive. While the jackknife analysis shows that the
coefficient of the interaction between the tax wedge and the benefit replacement
becomes insignificant when Spain is omitted, this coefficient retains its significance
when the six outliers are removed from the sample. In sum, contrary to the initial
findings, the sensitivity analysis indicates that individual institutions do not have a
clear detrimental effect on unemployment. At the same time, evidence about
reform complementarity between the tax wedge and unemployment benefits
reforms remains inconclusive.
Labour market institutions and unemployment 755
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Table 2 Jackknife analysis and the exclusion of outliers for the baseline model
Minimum Country Estimate Maximum Country Estimate w/o outliers
L.ur 0.786*** (0.028) SP 0.813*** (0.026) 0.910*** *0.029) LT 0.904*** (0.032)
BRR 0.169 (0.388) LT (PT, AUT)
a
1.307*** (0.428) 2.538*** (0.347) IT 0.321 (0.401)
EP 20.160 (0.246) SP 0.166 (0.220) 0.385 (0.238) SWE 0.110 (0.239)
TW 20.086 (0.663) LT 0.981 (0.698) 2.444*** (0.793) SP (GR, IT)
b
0.319 (0.723)
BC 20.280*** (0.076) AUS 20.236*** (0.068) 20.177*** (0.068) LT 20.144** (0.071)
UD 0.437 (0.861) LT 2.855*** (0.755) 4.889*** (0.878) SP 0.774 (0.878)
CPI 20.032 (0.030) IRL 20.003 (0.036) 0.019 (0.0423) CAN 20.290*** (0.017)
GDP 20.341*** (0.019) RO 20.284*** (0.019) 20.262*** (0.018) PL 20.003 (0.034)
TOT 20.024*** (0.007) NO 20.019*** (0.007) 20.007 (0.006) LT (SP, SWE)
c
20.008* (0.005)
RIR 0.002 (0.024) RO 0.028 (0.020) 0.061*** (0.019) LT (SK)
d
0.056*** (0.017)
Notes: Entries in the first five columns are OLS-PCSE coefficient estimates from the baseline model (Table 1, column 2), together with minimum and maximum coefficient estimates
obtained by re-estimating the model so that each country is omitted one at a time. The last column presents the estimates of the model where six outliers (Lithuania 1995 and 2002, Spain
2001, Latvia 1996, Romania 2004 and Finland 1993) are excluded from the sample.
a
Exclusion of Portugal or Austria also makes the coefficient of the replacement rate insignificant.
b
Apart from Spain, the tax wedge becomes significant when Greece or Italy are dropped from the sample, albeit with a smaller coefficient (1.346 and 1.597, respectively).
c
The terms of trade coefficient becomes also insignificant with the exclusion of either Spain or Portugal.
d
The coefficient of the real interest rate turns marginally significant when Slovakia is omitted (0.026).
756 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Do the determinants of unemployment differ between advanced and new market
economies?
The previous section has shown that the estimates are highly sensitive to changes in
the sample. The overall results, therefore, can conceal potentially different effects of
institutions in different groups of countries. In this section, we re-assess this finding
by re-estimating the main models for the advanced and new market economies sep-
arately. This additional check is also warranted on substantive grounds. Although
we control for the initial transition shock in CEE countries by excluding the first
three years of the transition from the analysis, there may be still good reasons to
believe that the key culprits of unemployment are different in the two regions.
The literature suggests that trade unions in CEE are weaker than in continental
Europe and that wage bargaining is largely uncoordinated (Cazes and Nesporova,
2003). The welfare states, at the same time, do not appear very generous and
support for the unemployed has been generally much more modest (Schiff et al.,
2006). Our data also indicate that on the whole employment protection is fairly
liberal in CEE. Thus, apart from the tax wedge, which remains high, this general evi-
dence suggests that CEE countries do not seem to suffer from excessive labour
market rigidities. Given that unemployment in CEE has been generally higher
and labour market rigidities less pronounced, the effects of institutions on un-
employment should be even weaker than in the advanced economies.
Table 3reports th e estimates of th e fully dummy-int eractive model , which allows
us to see the results of the main models from Table 1for the advanced and new
market economies separately. The first four columns show the results for the base-
line model and the main interaction model across the two estimators. Columns 5 – 8
repeat this exercise for the sample that excludes the six outliers identified earlier.
The results suggest that labour market institutions appear to be more influential
in the advanced economies. Union density and to a lesser extent benefit replace-
ment ratios are both consistently associated with high unemployment. At the
same time, wage bargaining coordination remains negative and significant in
most models. As in all previous models, GDP growth remains robustly and nega-
tively associated with unemployment. The interaction between the tax wedge
and the benefit replacement rate suggests the potential for reform complementar-
ity, but this result is evidently no longer so robust when we exclude the outliers.
The link between the labour market institutions and unemployment is less
evident in CEE. The coefficients on the interactions with the CEE dummy
capture the difference between the two groups of countries. In other words, the co-
efficient of a given variable for CEE is equal to the sum of its interaction with the
CEE dummy and the corresponding coefficient for the advanced economies. As
evident, the terms of trade is the only variable that seems to have a clear impact
on unemployment. The tax wedge seems to have influence only in the models
Labour market institutions and unemployment 757
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Table 3 Differences between the advanced and new market economies
(1) (2) (3) (4) (5) (6) (7) (8)
FGLS het
(AR1) OLS-PCSE FGLS het (AR1) OLS-PCSE FGLS het (AR1) OLS-PCSE FGLShet (AR1) OLS-PCSE
L.ur 0.875***
(0.025)
0.917***
(0.018)
0.888***
(0.025)
0.928***
(0.019)
0.865***
(0.024)
0.900***
(0.022)
0.874***
(0.025)
0.910***
(0.025)
BRR 1.296**
(0.606)
1.279***
(0.300)
1.428*
(0.733)
0.809**
(0.387)
1.216**
(0.586)
1.215***
(0.325)
1.443**
(0.708)
0.911*
(0.484)
EP 20.050
(0.138)
0.260
(0.181)
0.047
(0.148)
0.292
(0.187)
20.097
(0.133)
0.179
(0.209)
20.021
(0.144)
0.224
(0.213)
TW 20.003
(0.571)
21.391***
(0.509)
0.111
(0.602)
20.590
(0.484)
0.170
(0.539)
20.913
(0.579)
0.312
(0.580)
20.203
(0.531)
BC 20.149***
(0.050)
20.155**
(0.065)
20.120**
(0.052)
20.112
(0.071)
20.141***
(0.049)
20.143**
(0.068)
20.113**
(0.051)
20.098
(0.073)
UD 4.455***
(1.282)
1.796***
(0.461)
3.599***
(1.260)
1.736***
(0.448)
4.467***
(1.228)
2.232***
(0.548)
3.838***
(1.232)
1.998***
(0.498)
GDP 20.293***
(0.023)
20.321***
(0.019)
20.292***
(0.023)
20.320***
(0.019)
20.294***
(0.022)
20.309***
(0.021)
20.290***
(0.022)
20.307***
(0.021)
CPI 20.009
(0.022)
20.003
(0.029)
20.012
(0.022)
20.010
(0.029)
20.009
(0.022)
20.004
(0.028)
20.011
(0.022)
20.012
(0.029)
TOT 20.003
(0.007)
20.003
(0.004)
20.001
(0.007)
0.001
(0.005)
20.003
(0.006)
20.003
(0.004)
20.002
(0.006)
0.001
(0.005)
RIR 0.021
(0.016)
0.052**
(0.022)
0.026
(0.016)
0.061***
(0.022)
0.022
(0.015)
0.049**
(0.022)
0.025
(0.016)
0.056**
(0.022)
L.ur*CEE 20.435***
(0.093)
20.470***
(0.064)
20.434***
(0.095)
20.490***
(0.050)
20.073
(0.091)
20.132
(0.092)
20.120
(0.089)
20.211**
(0.103)
BRR*CEE 2.094
(2.229)
0.780
(1.504)
6.969**
(3.148)
7.673***
(1.907)
20.836
(1.830)
21.142
(1.578)
4.265
(3.014)
3.777
(2.605)
758 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
EP*CEE 21.163
(1.017)
21.569
(1.239)
20.885
(1.384)
21.403
(0.883)
20.639
(0.627)
20.742
(1.051)
20.170
(0.790)
21.539
(1.082)
TW*CEE 7.526
(5.545)
10.198*
(6.124)
6.717
(5.470)
8.907*
(4.908)
7.036*
(4.168)
8.599*
(5.053)
9.310**
(4.367)
11.211***
(3.765)
BC*CEE 20.565
(0.692)
20.738
(0.923)
0.089
(0.690)
0.144
(0.823)
0.176
(0.523)
0.174
(0.731)
0.737
(0.566)
0.829
(0.695)
UD*CEE 2.108
(6.603)
4.928
(8.268)
23.042
(7.070)
21.694
(6.783)
24.640
(5.180)
20.758
(6.021)
25.657
(5.345)
24.412
(5.348)
GDP*CEE 0.043
(0.085)
0.011
(0.047)
0.117
(0.084)
0.085*
(0.045)
0.109*
(0.061)
0.124***
(0.042)
0.141**
(0.066)
0.144***
(0.056)
CPI*CEE 20.070
(0.252)
0.001
(0.348)
20.062
(0.256)
0.012
(0.252)
0.189
(0.213)
0.179
(0.396)
0.247
(0.229)
0.360
(0.297)
TOT*CEE 20.150***
(0.056)
20.167***
(0.051)
20.149***
(0.054)
20.172***
(0.031)
20.101**
(0.041)
20.122***
(0.041)
20.143***
(0.043)
20.159***
(0.038)
RIR*CEE 20.005
(0.036)
20.015
(0.046)
20.032
(0.037)
20.057*
(0.029)
20.019
(0.027)
20.025
(0.039)
20.040
(0.029)
20.057*
(0.031)
CEE 24.048***
(7.578)
—, — 21.120***
(7.374)
—, — 12.126**
(5.887)
—, — 10.878*
(5.860)
—, —
EP*BRR 20.407
(0.525)
0.262
(0.461)
20.453
(0.510)
0.103
(0.530)
EP*TW 20.220
(0.670)
20.428
(0.738)
20.304
(0.637)
20.552
(0.737)
BRR*TW 26.214*
(3.558)
210.564***
(3.585)
24.998
(3.463)
29.425***
(3.475)
EP*BRR*CEE 14.420*
(8.100)
14.108***
(3.712)
10.818*
(5.814)
4.402
(5.437)
EP*TW*CEE 21.158
(13.232)
31.627**
(13.618)
17.850*
(10.169)
28.791**
(13.070)
BRR*TW*CEE 68.352
(46.596)
70.027**
(35.626)
24.619
(35.642)
30.800
(36.865)
Continued
Labour market institutions and unemployment 759
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Table 3 Continued
(1) (2) (3) (4) (5) (6) (7) (8)
FGLS het
(AR1) OLS-PCSE FGLS het (AR1) OLS-PCSE FGLS het (AR1) OLS-PCSE FGLShet (AR1) OLS-PCSE
Constant 22.995
(2.005)
1.352***
(0.505)
24.937**
(2.130)
0.876*
(0.525)
22.501*
(1.378)
11.281*
(6.833)
24.327***
(1.542)
10.684**
(4.481)
Country and year
effects
Yes Yes Yes Yes Yes Yes Yes Yes
No. of
observations
451 451 451 451 440 440 440 440
No. of countries 26 26 26 26 26 26 26 26
R
2
0.965 0.969 0.974 0.976
Notes: Estimates in columns 1 –4 are based on the full sample. Columns 5 –8 exclude the outliers. Standard errors in parentheses:* significant at 10%, ** significant at 5%, *** sig-
nificant at 1%.
760 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
that exclude the six outliers. Union density and bargaining coordination, which
matter for unemployment in the advanced economies, have no discernable
impact in CEE and are even ‘wrongly’ signed in most models. Whilethe interactions
between employment protection on the one hand, and benefit replacement rates
and the tax wedge on the other appear as statistically significant in several
models, neither of these interactions is robust to changes in estimators or the exclu-
sion of outliers. On the whole, these findings suggest that institutions in CEE play
even less of a direct role in explaining aggregate unemployment than in the
advanced economies.
As above, we also assessed the robustness of these resultsthrough a jackknife ana-
lysis, which showed that the results are fragile to the exclusion of individual coun-
tries. In the case of advanced economies, the benefit replacement ratio is no longer
significantly associated with unemployment if either Portugal or Austria is omitted.
The only institution that remains robustly associated with higher unemploym ent is
union density, though the magnitude of the coefficient and the significance level
become markedly smaller when the UK or Finland is dropped from the sample.
This finding concurs with Baccaro and Rei (2007) who demonstrate that union
density is the only institutional variable that displays a robust positive association
with unemployment in OECD countries. Apart from union density, only GDP
growth survives the jackknife procedure. Finally, the significance of the interaction
between the benefit replacement rate and the tax wedge turns out to depend entirely
on the presence of Spain. Only when the interactions are added one at a time does
this interaction survive the jackknife procedure, though the exclusion of Spain
reduces the coefficient considerably (from 10.82 to 5.05), making it significant
only at the 10% level.
The results for CEE are even more fragile. The tax wedge is no longer significant if
any of the following countries is excluded from the analysis: Latvia, Lithuania,
Poland, Romania, Italy, Norway, Spain or the UK. The interaction between employ-
ment protection and the benefit replacement rate becomes insignificant when
either Estonia or Latvia is omitted, confirming the results of the analysis that
excludes the outliers. When these interactions are added to the baseline model
one at a time, only the interaction between the tax wedge and employment protec-
tion survives the jackknife procedure, but this interaction becomes insignificant as
we add other interaction terms. The results for other variables are also sensitive to
the exclusion of individual countries, and some coefficients even change their sign
depending on the sample.
Overall, the analysis shows that the results of these models are rather fragile.
Most estimates suggesting an adverse impact of institutions no longer hold when
we exclude particular countries or even a small number of potentially influential
observations.
Labour market institutions and unemployment 761
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Interactions between institutions and macroeconomic shocks
Althoughwe did not find strongevidence that unemployment is a direct consequence
of rigid labour market institutions, it is possible that institutions play a more indirect
role by amplifying the effects of economic shocks. Table 4reports the results of the
model that examines this possibility. Following Blanchard and Wolfers (2000), this
model is estimated via non-linear least squares. Positive coefficients indicate that
institutions exacerbate the effects of shocks, while the negative coefficient suggests
that institutions mitigate the adverse effects of shocks on unemployment.
Institutions in Blanchard and Wolfers’s analysis are expressed as deviations from
the sample means. Following Baccaro and Rei (2007), we extend this formulation
and consider annual data in both levels and deviations. In addition, because the as-
sumption of i.i.d. residuals is unrealistic (cf. Blanchard and Wolfers, 2000, p.20), we
rely on the results obtained using Rogers robust standard errors. These correspond
to White standard errors adjusted to account for the possible correlation within a
cluster (i.e. country) and country-specific heteroskedasticity.
9
When data are in
levels the coefficients on the time dummies (not reported) indicate the impact of
shocks on unemployment in a country in which all institutional variables are set
to zero. In this case, the coefficients of the institutional variables shown in
columns 1 and 2 capture the additional effect of shocks on unemployment when
a given institution increases by one unit. When data are in deviations the coeffi-
cients of the time dummies capture the impact of shocks in a country where all insti-
tutions are at the sample mean, and the coefficients of institutions in columns 3 and
4 capture the additional effect of shocks when the given institution increases one
unit above the sample mean.
As Table 4shows, we cannot draw strong conclusions about the indirect impact
of institutions since the way in which the data are expressed and the choice of stand-
ard errors evidently has a big impact on the results. When the data are expressed in
levels, institutions do not seem to amplify the effects of shocks. When the data are in
deviations, the same institutions identified initially in the linear model appear sig-
nificant: union density and benefit replacement ratio seem to increase the impact of
adverse shocks, while bargaining coordination mitigates it. However, the estimated
time effects are negative, suggesting that they make no significant contribution to
the overall increase in unemployment (see Blanchard and Wolfers, 2000, p. 20).
Discussion
Despite a wide range of models and specifications, we find little support for the
standard argument that unemployment is a consequence of institutional rigidities.
9
See Petersen (2008) for the discussion of the choice of standard errors and simulation evidence that
Rogers standard errors perform best with this type of data structure.
762 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
As demonstrated, most positive associations between institutions and unemploy-
ment disappear with small changes in specifications or the sample. While these
findings concur with recent research that questions the empirical evidence
behind the deregulatory view (Baker et al., 2005;Baccaro and Rei, 2007;Howell
and Rehm, 2009;Vergeer and Kleinknecht, 2012), they are at odds with a
number of studies that report adverse effects of institutions on unemployment
(e.g. OECD, 1994;Scarpetta, 1996;Siebert, 1997;Elmeskov et al., 1998;Blanchard
and Wolfers, 2000;Nickell et al., 2005;Bernal-Verdugo et al., 2012). How can we
explain such different findings? Three possible explanations are worth considering.
The first one is an obvious point that the data used here are different. Given that
the choice of data (both measures of institutions and the sample considered in the
analysis) inevitably has a large impact, findings of most studies in this literature are
not strictly comparable. A related point is that not all studies use the same robust-
ness checks. The fact that the results may be robust to variations in variable speci-
fication or the estimation method does not guarantee that they are robust to small
changes of the sample. But this is unlikely to be the whole story. The second possible
Table 4 Interactions between Shocks and Institutions: EU and OECD countries, 1980 –2006
(1) (2) (3) (4)
NLS levels,
Rogers stand-
ard errors
NLS levels, l.s
standard
errors
NLS deviations,
Rogers stand-
ard errors
NLS devia-
tions, l.s stand-
ard errors
EP
0.033
(0.340)
0.033
(0.120)
0.013
(0.102)
0.0103
(0.037)
UD
3.829
(5.827)
3.829**
(1.721)
1.211**
(0.544)
1.211***
(0.162)
BRR
1.465
(1.383)
1.465**
(0.719)
0.463*
(0.239)
0.463***
(0.140)
TW
1.345
(2.440)
1.345
(0.947)
0.425
(0.381)
0.425**
(0.179)
BC
20.260
(0.208)
20.260***
(0.084)
20.082**
(0.040)
20.082***
(0.021)
No. of observations 500 500 500 500
R
2
0.806 0.806 0.806 0.806
Adj. R
2
0.798 0.783 0.798 0.783
Notes: Time and country dummies omitted.
Labour market institutions and unemployment 763
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
explanation is that labour market institutions simply do not have strong effects on
unemployment. As indicated above, even when certain institutional variables
display a statistically significant impact on unemployment, this impact is rather
small in substantive terms. It should be noted that in a number of studies that
support the deregulatory view, the size of the effect of particular institutions is
also rather small, although this is often not explicitly acknowledged (see Baker
et al., 2005, pp. 101103). This could be the case either because the positive and
negative effects of institutions balance out or, as Freeman has argued, because ‘bar-
gaining settlements and regulations that are truly expensive to an economy’ are ef-
fectively ruled out (2008, p. 25). Finally, an explanation that seems most convincing
is that there is no universal cause and thus solution to unemployment. The same
institution may have different effects in different countries or time periods (Hall,
2003, p. 383). In this line of reasoning, the impact of institutions is not straightfor-
ward and it may depend on the overall institutional configurations and interactions
between labour markets and other spheres, such as social policy, skill regimes, and
product market—characteristics that, given policy changes in these areas, may not
be adequately captured by country dummies. More generally, this interpretation is
in line with Ragin’s work on ‘multiple conjectural causation’ (1987,2000) as it
implies that labour market institutions do not have a consistent causal ef fect on un-
employment that applies universally. The fact that our sensitivity analysis and, in
particular, changes in the sample lead to very different conclusions about the
effects of institutions supports this interpretation. Overall, while our analysis
does not find strong evidence about the adverse effects of labour market institutions
on unemployment in general, we cannot exclude the possibility that institutions
may be responsible for high unemployment in some countries. However, based
on our analysis, there is little merit in recommendations that call for
across-the-board institutional deregulation.
Conclusion
This paper has examined the role of labour market institutions in the determination
of unemployment in the EU and OECD countries over the last three decades. The
analysis entailed a re-examination of a number of specifications that have been used
frequently in the empirical literature on unemployment which underpins the view
that deregulation improves labour market performance. Our analysis pays special
attention to common, but often neglected, problems associated with macro-
comparative time-series cross-section analysis, such as the potential sensitivity of
the results to the choice of estimators and small changes in model specifications
and the sample. Robustness checks reveal that most results are fragile and that
even a small number of observations may exert inordinate leverage on the coeffi-
cient estimates.
764 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
On the whole, we find no systematic support for the conventional view that un-
employment is a consequence of rigid labour market institutions. Among the insti-
tutions only wage bargaining coordination shows a fairly robust association with
unemployment, but this variable has a beneficial rather than a detrimental effect.
Union density and benefit generosity, which showed up as significant in the
initial analysis, do not survive the robustness checks. Turning to the interaction
models, our initial analysis finds some indications that the complementary tax
and benefit reforms may be beneficial, but these findings are less robust to
changes in the sample. A re-estimation of the main models for the advanced and
new market economies separately reveals that institutions play a more important
role in the advanced economies, with several of them showing some association
with unemployment. However, these initial results are also fragile to changes in spe-
cifications and the sample. In the advanced economies, union density is the only
institutional variable that remains robustly associated with unemployment. The
interaction between the benefit replacement ratio and the tax wedge also seems
to play a role, though this association is more specification dependent. At the
same time, in CEE no institution alone is directly and robustly associated with un-
employment, but there are some (albeit not strong) indications that the interaction
between the tax wedge on the one hand, and employment protection and the benefit
replacement rate on the other may play a role. We also do not find sufficiently robust
evidence for the hypothesis that institutions affect unemployment indirectly by
amplifying the adverse effects of economic shocks.
In sum, our models provide no compelling evidence about the adverse effects of
institutions. Meanwhile, GDP growth, and to a lesser extent the terms of trade, seem
to be more consistent predictors of unemployment. Our analysis therefore chal-
lenges the policy orthodoxy that comprehensive labour market deregulation is ne-
cessary to reduce or stabilize unemploym ent. Calls for further deregulation seem to
be especially unwarranted in CEE countries, where the link between institutions
and unemployment appears particularly weak. But even in the advanced economies
the effects of institutions are weak and in most cases depend heavily on which coun-
tries are included in the analysis. Given the lack of robustness, the most plausible
interpretation of our results is that institutions have different effects in different
contexts, and that therefore there is no universal cause of (and remedy for) un-
employment. An implication of this for further research is that in-depth analyses
of individual countries or particular groups of countries may be more fruitful in
offering sound policy recommendations than the continued search for universal
causes of unemployment. Of course, it is possible that the lack of a strong link
between institutions and unemployment that we found may reflect the fact that
institutions have different effects for different groups of the labour force. Clearly,
our analysis of aggregate unemployment cannot identify such effects and further
research on these issues is needed. Nonetheless, our findings suggest that any
Labour market institutions and unemployment 765
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
results from this type of quantitative macro-comparative research must be taken
with a grain of salt and should not be used as unquestionable evidence for reforming
particular institutions in a particular country.
Acknowledgements
We are particularly grateful to Lucio Baccaro, Barry Reilly, Bernhard Kittel, Lane
Kenworthy and three anonymous reviewers for detailed and constructive com-
ments. An earlier version of this paper was presented at the 23rd Annual Conference
of the Society for the Advancement of Socio-Economics (SASE) and a fellow lecture
series at the Hanse-Wissenschaftskolleg in June 2011.
Funding
Research for this paper was supported by the Economic and Social Research
Council (ESRC) grant RES-061-25-0354.
Supplementary material
Supplementary material is available at SOCECO online.
References
Addisson, J. and Texeira, P. (2003) ‘The Economics of Employment Protection, Journal of
Labor Research,24, 85– 127.
Allard, G. (2010) The Employment Protection Dataset for OECD Countries, 1950 2008,
Madrid: IE Business School.
Amable, B., Demmou, L. and Gatti, D. (2006) Institutions, Unemployment and Inactivity in
the OECD Countries, Working Paper 16, Paris-Jourdan Sciences Economiques.
Arellano, M. and Bond, S. (1991) ‘Some Tests of Specification for Panel Data: Monte Carlo
Evidence and an Application to Employment Equations’, Review of Economic Studies,58,
277297.
Baccaro, L. and Rei, D. (2007) ‘Institutional Determinants of Unemployment in OECD
Countries: Does the Deregulatory View Hold Water?’, International Organization,6,
527569.
Baker, D., Glyn, A., Howell, D. and Schmitt, J. (2005) ‘Labour Market Institutions and Un-
employment: A Critical Assessment of the Cross-Country Evidence’. In Howell, D. (ed.)
Fighting Unemployment. The Limits for Free Market Orthodoxy, Oxford, Oxford University
Press, pp. 72– 118.
Bassanini,A. and Duval, R. (2006) ‘The Determinants of UnemploymentAcross OECD Coun-
tries: Reassessing the Role of Policies and Institutions, OECD Economic Studies,42,786.
766 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Bassanini, A. and Duval, R. (2009) ‘Unemployment, Institutions and Reform Complemen-
tarities: Re-Assessin the Aggregate Evidence for OECD Countries’, Oxford Review of Eco-
nomic Policy,25, 40–59.
Beck, N. and Katz, J. (1995) ‘Whatto do (and Not to Do) with Time Series Cross Section Data
in Comparative Politics’, American Political Science Review,89, 634– 647.
Beck, N. and Katz, J. (2011) ‘Modelling Dynamics in Time-Series– Cross-Section Political
Economy Data’, Annual Review of Political Science,14, 331352.
Behar, A. (2009) ‘Tax Wedges, Unemployment Benefits and Labour Market Outcomes in the
New EU Members’, Czech Economic Review,3, 6992.
Be
´lot, M. and van Ours, J. (2004) ‘Does the Recent Success of Some OECD Countries in Low-
ering Their Unemployment Rates Lie in the Clever Design of Their Labour Market
Reforms?’, Oxford Economic Papers,56, 621 642.
Bentolila, S., Cahuc, P., Dolado, J. and Le Barbanchon, T. (2010) Two-Tier Labor Markets in
the Great Recession: France vs. Spain, London, CEPR Discussion Paper 8152.
Bernal-Verdugo, L. E., Furceri,D. and Guillaume, D. (2012) Labor Market Flexibility and Un-
employment: New Empirical Evidence of Static and Dynamic Effects, WorkingPaper No. 12/
64, Washington, DC, IMF.
Bertola, G. (1992) ‘Labor Turnover Costs and Average Labor Demand’, Journal of Labor Eco-
nomics,10, 389411.
Blanchard, O. and Wolfers, J. (2000) ‘The Role of Shocks and Institutions in the Rise of Euro-
pean Unemployment: The Aggregate Evidence’, Economic Journal,110, 133.
Cazes, S. and Nesporova, A. (2003) Labour Markets in Transition: Balancing Flexibility and
Security in Central and Eastern Europe, Geneva, ILO.
Crowe, C. and Meade, E. E. (2007) ‘Evolution of Central Bank Governance around the
World’, Journal of Economic Perspectives,21, 69 90.
Ederveen, S. and Thissen, L. (2007) ‘Can Labour Market Institutions Explain High Un-
employment Rates in the New EU Member States?’, Empirica,34, 299 –317.
Elmeskov, J., Martin, J. and Scarpetta, S. (1998) ‘Key Lessons for Labour Market Reforms:
Evidence from OECD Countries Experience’, Swedish Economic Policy Review,5, 205 –252.
Estevez-Abe, M., Iversen, T. and Soskice, D. (2001) ‘Social Protection and the Formation of
Skills: A Reinterpretation of the Welfare State’. In Hall, P. and Soskice, D. (eds) Varieties of
Capitalism: The Institutional Foundations of Comparative Advantage, Oxford, Oxford Uni-
versity Press, pp. 145– 183.
Feldmann, H. (2009) ‘The Unemployment Effects of Labor Regulation around the World’,
Journal of Comparative Economics,37, 7690.
Fialova, K. and Schneider, O. (2009) ‘Labour Market Institutions and Their Effect on Labour
Market Performance in the New EU Member Countries’, Eastern European Economics,47,
5783.
Freeman, R. B. (2008) Labour Market Institutions Around the World, CEP Discussion Paper
844, London, LSE, Centre for Economic Performance.
Labour market institutions and unemployment 767
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Hall, P. A. (2003) ‘Aligning Ontology and Methodology in Comparative Politics’. In
Mahoney, J. and Rueschemeyer, D. (eds) Comparative Historical Analysis in the Social
Sciences, New York, Cambridge University Press, pp. 373404.
Hall, P. A. and Franzese, R. J. (1998) ‘Mixed Signals: CBI, Coordinated Wage-Bargaining, and
European Monetary Union’, International Organization,52, 505535.
Holmlund, B. (1998) ‘Unemployment Insurance in Theory and Practice’, Scandinavian
Journal of Economics,100, 113141.
Howell, D. (ed.) (2005) Fighting Unemployment. The Limits for Free Market Orthodoxy,
Oxford, Oxford University Press.
Howell, D.R. and Rehm, M. (2009) ‘Unemployment Compensation and High EuropeanUn-
employment: A Reassessment with New Benefit Indicators’, Oxford Review of Economic
Policy,25, 60– 93.
IMF. (2003) World Economic Outlook, Washington, DC, International Monetary Fund.
Kenworthy, L. (2008) Jobs With Equality, Oxford, Oxford University Press.
Kiviet, J. F. (1995) ‘On Bias, Inconsistency and Efficiencyof Various Estimators in Dynamic
Panel Data Models’, Journal of Econometrics,68, 53 68.
Labartino, G. (2010) ‘Essays in Labour Economics’, PhD dissertation, Milan, Bocconi Uni-
versity.
Layard, R., Nickell, S. and Jackman, R. (1991) Unemployment: Macroeconomic Performance
and the Labour Market, Oxford, Oxford University Press.
Lehmann, H. and Muravyev, A. (2009) How Important Are Labour Market Institutions for
Labour Market Performance in Transition Countries?, IZA Discussion Paper 4673, Bonn,
Institute for the Study of Labour.
Maddala, G. S. and Wu, S. (1999) ‘A Comparative Study of Unit Root Tests with Panel Data
and a New Simple Test’, Oxford Bulletin of Economics and Statistics, Special Issue,61, 631 –
652.
Nickell, S. (1981) ‘Biases in Dynamic Models with Fixed Effects’, Econometrica,49, 1417
1426.
Nickell, S. (1997) ‘Unemployment and Labour Market Rigidities: Europe versus North
America’, Journal of Economic Perspectives,11, 5574.
Nickell, S., Nunziata, L. and Ochel, W. (2005) ‘Unemployment in the OECD since the 1960s.
What Do We Know?’, Economic Journal,115, 1 27.
Nunziata, L. (2002) Unemployment, Labour Market Institutions and Shocks, University of
Oxford Economics Group, Nuffield College, Working Paper No. 16.
OECD. (1994) OECD Jobs Study, Paris, OECD.
OECD. (1999) Implementing the JobsStrateg y: Assessing Performanceand Policy, Paris, OECD.
OECD. (2006) OECD Employment Outlook, OECD, Paris.
Petersen, M. A. (2008) ‘Estimating Standard Errors in Finance Panel Data Sets: Comparing
Approaches’, Review of Financial Studies,22, 435– 480.
768 S. Avdagic and P. Salardi
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
Ragin, C. (1987) The Comparative Method: Moving Beyond Qualitative and Quantitative
Strategies, Berkeley, University of California Press.
Ragin, C. (2000) Fuzzy Set Social Science, Chicago, University of Chicago Press.
Rueda, D.and Pontusson, J. (2000) ‘WageInequalities and Varieties of Capitalism’, WorldPol-
itics,52, 350383.
Scarpetta, S. (1996) ‘Assessing the Role of Labour Market Policies and Institutional Settings
on Unemployment: A Cross-Country Study’, OECD Economic Studies,26, 43– 98.
Schiff, J., Egoume-Bossogo, P., Ihara, M., Konuki, T. and Krajnyak, K. (2006) Labour Market
Performance in Transition: The Experience of Central and Eastern European Countries, Oc-
casional Paper 248, Washington, DC, IMF.
Siebert, H. (1997) ‘Labor Market Rigidities: At the Root of Unemployment in Europe’,
Journal of Economic Perspectives,11, 37– 54.
Van Vliet, O. and Caminada, K. (2012) ‘Unemployment Replacement Rates Dataset Among
34 Welfare States, 1971– 2009’, v.1. Leiden University, accessed at http://www.law
.leidenuniv.nl/org/fisceco/economie/hervormingsz/on November 15, 2012.
Vergeer, R. and Kleinknecht, A. (2012) ‘Do Flexible Labor Markets Indeed Reduce Un-
employment? A Robustness Check’, Review of Social Economy,70, 451– 467.
Visser, J. (2009) ICTWSS Database. v.2. University of Amsterdam, accessed at http://www
.uva-aias.net/208 on May 25, 2010.
Visser, J. and Hemerijck, A. (1997) A Dutch Miracle: Job Growth, Welfare Reform and Corpor-
atism in the Netherlands, Amsterdam, Amsterdam University Press.
Labour market institutions and unemployment 769
at University of Sussex on October 20, 2014http://ser.oxfordjournals.org/Downloaded from
... The degree of centralization shows the level (firms, industry, or national) at which collective agreements are negotiated (Visser, 2019). According to the major literature, higher bargaining coordination may be particularly "employment friendly" (Scarpetta, 1996;Elmeskov et al., 1998;Bassanini and Duval, 2006;Vergeer and Kleinknecht, 2012;Avdagic and Salardi, 2013;Pesliakaite, 2016). Similarly, bargaining centralization may have a beneficial effect on unemployment (Belot and Van Ours, 2004;Flaig and Rottmann, 2013), although several authors have pointed out the existence of a nonmonotonic relationship with the unemployment rate, the so-called hump-shaped hypothesis (Scarpetta, 1996;Alesina and Perotti, 1997;Elmeskov et al., 1998;Boeri et al., 2000;Daveri and Tabellini, 2000;Pastore and Shorman, 2018). ...
... Using these estimators is consistent with the related literature (IMF, 2003;Bassanini and Duval, 2006;Baccaro and Rei, 2007;Avdagic and Salardi, 2013;Avdagic, 2015;Escudero, 2018;Heimberger, 2019). ...
... 29 The lagged coefficient of unemployment assumes a positive value, and it is significant at the 1% level for the FD-FGLS, PW-PCSE, and FE-DK models; hence, there is firm evidence of persistence in unemployment, with a "transmission coefficient" of 0.46 in the FD-FGLS models and 0.65 in the PW-PCSE and FE-DK specifications. This output is consistent with other studies (IMF, 2003;Nickell et al., 2005;Bernal-Verdugo et al., 2012;Avdagic and Salardi, 2013;Heimberger, 2019). ...
Article
This paper investigates the long-run effect of a wide set of labor market institutions (LMIs) and macroeconomic variables on aggregate unemployment for a panel of 22 European countries over the period 1990-2019. First-difference feasible generalized least squares model, Prais-Winsten regression with panel-corrected standard errors, two-step generalized method of moments estimation of the fixed effects, and fixed-effects regression with Driscoll and Kraay standard errors are estimated. The results suggest that employment protection legislation, wage bargaining coordination and centralization, minimum wage, and immigration inflows are significantly and negatively associated with the aggregate unemployment rate. Conversely, union density, product market regulation (PMR), and tax wedge have a positive and significant correlation with unemployment rate. The impact of corporate tax rate and government size is mostly positive. Moreover, the interaction between LMIs does matter and may sometimes change the interpretation of some reforms taken in isolation. Stronger wage-setting institutions may offset the negative impact of PMR and the tax wedge. Macroeconomic variables are generally consistent with the major literature and do not change LMIs interpretation. Among macroeconomic factors, capital accumulation plays the most important role in reducing the unemployment rate. Finally, my findings suggest the implementation of economic policies consistent with Keynesian theory and all those economists-such as Solow (1990)-who look at the labor market as a social institution. Paper available upon request.
... First, while the previous literature has focused on temporary employment and low wages, this article looks at another form of labour market flexibility that has proliferated in recent years: unstable jobs with variable hours and pay. Secondly, with a few exceptions (Gebel and Giesecke, 2011;Barbieri and Cutuli, 2016), the majority of existing studies have relied on country-level time series to tease out the impact of institutions on labour market outcomes (Iversen and Wren, 1998;Howell, 2002;Kenworthy, 2003;Nickell et al., 2005;Avdagic and Salardi, 2013). While informative, studies relying on country-level variation have a number of well-known limitations such as insufficient institutional variation, the confounding potential of national institutional, cultural or economic characteristics, and drawing conclusions about micro-level processes from aggregate data. ...
... A few studies using country-level time series do find a significant positive effect of labour market rigidities on unemployment, especially youth and long-term unemployment (Di Tella and MacCulloch, 2005;Nickell et al., 2005;Bertola et al., 2007;Bernal-Verdugo et al., 2012;Agnello et al., 2014). Yet, subsequent studies have shown that these results are not robust to variations in model specification, the inclusion of additional countries and/or the use of slightly different institutional indicators (Kenworthy, 2003;Heyes, 2011;Vergeer and Kleinknecht, 2012;Avdagic and Salardi, 2013). A recent meta-analysis of 75 studies concluded there is no robust evidence to support the assertion that employment protection legislation increases unemployment rates (Heimberger, 2021). ...
Article
Full-text available
This article tests the hypothesis that unstable jobs with variable hours or pay enhance the job-finding chances of the working-age non-employed in the UK, by using a combination of the UK Household Longitudinal Study and the Labour Force Survey data and a discrete time model. We find no evidence on the share of unstable jobs in the non-employed person’s local labour market impacts on the probability to move into employment. This result holds both for men and women and for groups with low employability such as the low educated and the long-term unemployed. It is robust to alternative ways of defining unstable jobs and to the inclusion of unobserved heterogeneity. Overall, findings cast doubt on the importance of unstable jobs for employment creation in the UK.
... Following these recommendations, since the 1980s many countries have approved reforms to enhance labour market flexibility acting on unemployment protection schemes, collective bargaining or employment protection legislation (Brancaccio, Garbellini, and Giammetti 2018;Tridico and Pariboni 2017). However, empirical studies are not conclusive and have not conclusively proved that these reforms had a positive impact on labour markets (Avdagic and Salardi 2013;Bertola 2017b). ...
... However, the empirical evidence about the impact of high employment protection on unemployment is not conclusive (Bertola 2017a;Boeri, Cahuc, and Zylberberg 2015;Ferreiro and Gomez 2020;Heyes and Lewis 2015;Myant and Brandhuber 2016). Many studies conclude that high employment protection has no negative impact on unemployment (Avdagic 2015;Avdagic and Salardi 2013;Bertola 2017a;Flaschel et al. 2012) and, consequently, that labour market flexibilization has not reduced unemployment rates. ...
Article
This paper analyses the impact of employment protection legislation on the evolution of employment and unemployment in European Union economies during the Great Recession. The results show that employment protection did not have a significant impact on employment growth. Regarding unemployment rates, we obtain contrasting results: high employment protection for temporary workers was associated with larger increases in unemployment rates, whereas high protection for permanent workers against individual dismissal was associated with lower increases in unemployment rates. Nonetheless, employment protection for permanent in conjunction with that for temporary workers had a positive impact on unemployment rates.
... This body of research finds that shocks increased unemployment rates more in contexts with strong EPL compared to other contexts (Bertola, Blau, and Kahn 2001;Blanchard and Wolfers 2000). More recent research, however, has disputed these findings and showed that they are very sensitive to model specification (Avdagic and Salardi 2013). Related studies on labor market flows, which examine mobility rates and typical length of employment and unemployment, also considered the interaction between labor market institutions and macroeconomic environment, finding that market flows are generally lower in contexts with high EPL and less sensitive to macroeconomic shocks (DiPrete and Nonnemaker 1997;. ...
Article
Full-text available
The Great Recession raised the concern that employment protective institutions that are effective during macroeconomic stability might become counterproductive under growing macroeconomic volatility. We study this question by examining the relationship between employment protection legislation (EPL) and unemployment scars on earnings in 21 countries during the period surrounding the Great Recession. We use harmonized work history data for 21 countries from 2004 to 2014 and combine propensity score matching and multilevel-regression to estimate how earnings losses due to unemployment vary with the strength of labor market regulation and over changing macroeconomic conditions. We find that unemployment scarring is lower in contexts with robust employment protection, both under positive and negative macroeconomic environments. We also show that economic downturns intensify unemployment scarring significantly more in countries with weak EPL, largely because long-term unemployment is more strongly penalized. Taken together, our study finds that the positive effects of employment protection for workers remain robust during economic downturns.
Conference Paper
Full-text available
The paper examines the degree of connection and the impact of negative demographic trends on the movement of unemployment rates in Central and Southeast Europe (CSEE countries) through correlation and regression analysis. The variable number of working population (15-64) as % of population was taken as an indicator of negative demographic trends. Negative demographic trends have a significant impact on the movement of unemployment rates for most of the observed CSEE countries (Czech, Hungary, Bulgaria and Romania), as indicated by statistically significant coefficients of correlation and regression analysis. Additional confirmation of the strength of the relationship of the observed variables are the high coefficients of determination. Effective demographic policies need to be established and implemented to prevent negative demographic trends in the future. Due to the declining number of working population, it will also be necessary to increase the labour force participation rate through active labour market measures. This is especially true for those who are long-term unemployed. It would increase the supply of labour that will meet the demand for the necessary professionals. The results of the research indicate the need to include demographic factors when studying the determinants of unemployment in European countries, given that all countries face negative demographic trends that can have a significant impact on reducing the unemployment rate.
Conference Paper
Full-text available
In 2017 was approved in Brazil a series of changes on the Brazilian Employment Protection Legislation, the so-called Consolidação das Leis do Trabalho (CLT), regulating temporary work and the Insurance Severance Guarantee Fund for workers. A series of changes on collective negotiations, non-compulsory union dues, economic groups, liability of the departing shareholder, remote work, intermittent work, working hours, labor proceedings and arbitration were conducted. By that time, the generation between two million (in the short run) and ten million formal jobs (in the long run) was expected by policy makers. The main objective of this work is to analyze the effects of these changes on Brazilian labor market in the short run and, based on identified trends on its economy on the last decades, to verify how these recent changes affected job quality and the economic structure over time in the long run. Using the Employment Quality Index (EQI), developed by Oreiro et al. (2022), we can observe that there was not a considerable increase in the quality of employment after implementation of the changes in the Brazilian Labor Legislation (CLT). On the contrary, it was verified a greater precariousness of the new jobs, increased unemployment, and the decreasing of real wages due to several shocks on price indexes. After a great fall observed in the EQI in the years of 1995-1996, the Brazilian economy did not recover the employment quality that was observed in the previous periods, remaining below 0.40. Part of this effect is cyclical and related to exogenous shocks, such as the pandemic effects over Brazilian economy, which has been in falling behind trajectory over the last four decades. However, there are causes related with economic reforms adopted in Brazil in the period of (2016-2019) that did not manage to boost economic growth in the recent period, such as the fiscal ceiling for primary government expenditures, which depressed public investments in infrastructure, thereby reducing the profit rate for private investment due to the pecuniary and technological externalities of infrastructure over incentives for investment by the private sector (Ros, 2013). Furthermore, according to microdata analyzed, informality rate rose, as well as demand for low skilled and low technological intensive jobs in the service sector. These changes promote a regressive structural change in Brazil, transferring labor from modern and high productivity sectors (manufacturing sector) to subsistence and low productivity activities (low-tech services), increasing the duality in the sense of Lewis (1954) of the Brazilian economy. Moreover, according to econometric analysis performed in the paper, unemployment rate in Brazil is more responsive to investment rates and business environment than unit labor cost.
Article
This article analyses from a Keynesian approach the effect of wage devaluation on the Spanish labour market during the Great Recession post-2008. It challenges the pro-flexibility literature, which attributes to labour relations reforms the prevention of larger job destruction in the recession and a larger reduction in unemployment during the subsequent expansion. Instead, we examine the role of wage devaluation in the operation of Okun’s law and gross domestic product, using an extended version of the Bhaduri–Marglin model. We find that wage devaluation has not significantly modified Okun’s law and that through its impact on income distribution, the unemployment rate rose by 1.9 percentage points. We therefore provide evidence for the negative effect of wage devaluation on gross domestic product and the positive effect on the unemployment rate. JEL Codes: C22, E11, E24
Article
Full-text available
This paper analyses the role played by the flexibilization of labour markets on functional income distribution. Specifically, we analyse whether employment protection legislation affects the evolution of labour income share, measured by the size of compensation of employees as a percentage of GDP, the sum of wages and salaries as a percentage of GDP and the size of the adjusted wage share, in twenty European economies. Our study’s results show that the evolution of labour income share is explained by the economic growth, the growth of employment and unemployment rates, and the growth of real wages. Regarding the role played by the flexibility of the labour market, and specifically of the employment protection legislation, only employment protection for temporary workers has a significant impact on the evolution of labour shares. Our results show that stricter provisions on the use of fixed-term and temporary agency contracts have a positive impact on the growth of labour shares.
Article
Full-text available
The empirical analysis is conducted from two perspectives. Firstly, it examines the role that different policy and institutional settings have played in determining the marked differences in the level of structural or 'equilibrium' unmemployment across the OECD countries during the past decade. Secondly, it looks at the role of these same policy and institutional factors in determining the persistence of unemployment. The results encompass most of the previous cross-country studies comparing labour market performance and, in particular, those of Layart et al. (1991) and Bean and Symons (1989). They also offer new insights as to how policies and the mechanisms of wage determination may affect aggregate unemployment and other users of labour market slack, such as youth and long-term unemployment rates and non-employment rates. The use of these other measures of labour market slack gives a better understanding of the mechanisms through which distortions in the labour market affect unemployment and gives a better identification of potential beneficiaries of reforms. The broad empirical conclusions suggest that policy variables and the institutional mechanisms of wage determination do matter for the level of structural unemployment as well as for the speed of labour market adjustment in OECD countries.
Book
Much of Europe remains plagued by high levels of unemployment. Fighting Unemployment critically assesses the widely accepted view that the culprit is excessive labor market regulation and overly generous welfare state benefits. The chapters include both cross-country statistical analyses and country case studies and are authored by economists from seven North American and European countries. They challenge the standard free market prescription that lower wages for less skilled workers, weaker labor unions, greater decentralization in bargaining, less generous unemployment benefits, and much less job security are necessary for good employment performance. There is little or no evidence of an equality-employment tradeoff: more wage equality is not associated with higher unemployment (or lower employment) rates. And while some recent statistical tests of the role of protective labor market institutions across the most affluent countries have been interpreted to lend support to the orthodox view and have been highly influential in both professional and policy circles, these results are shown to vary significantly across studies and to be highly sensitive to minor changes in the way the tests are run. The case study chapters suggest that good employment performance has been achieved less by shrinking the welfare state and deregulating the labor market than by effectively coordinating macroeconomic and social policies with the wage bargaining system -an achievement that requires both strong employer and union associations and a relatively stable and consensual political environment. The larger message of this book is that fundamentally different labor market models are compatible with low unemployment, ranging from the free market "American Model" to the much more regulated and coordinated Scandinavian systems.
Chapter
Some of the liveliest debates about methodology in the social sciences center on comparative research. This essay concentrates on comparative politics, a field often defined by reference to the use of a particular “comparative method,” but it also bears on sociology, where there is active controversy about methodological issues. I use the term “methodology” to refer to the means scholars employ to increase confidence that the inferences they make about the social and political world are valid. The most important of these are inferences about causal relationships, where the object of a methodology is to increase confidence in assertions that one variable or event (x) exerts a causal effect on another (y). One of the curious features of contemporary debates is that they pay more attention to methodology than to issues of ontology. “Ontology” refers to the character of the world as it actually is. Accordingly, I use the term to refer to the fundamental assumptions scholars make about the nature of the social and political world and especially about the nature of causal relationships within that world. If a methodology consists of techniques for making observations about causal relations, an ontology consists of premises about the deep causal structures of the world from which analysis begins and without which theories about the social world would not make sense. At a fundamental level, it is how we imagine the social world to be.
Chapter
Chapter 3 takes up the question of the robustness of the cross-country evidence for the orthodox claim that labor market institutions explain the pattern of unemployment across the affluent countries. A detailed survey of the most influential cross-country statistical studies finds a wide range of results that are highly sensitive to the nature of the variables, the time period, and the econometric method employed. Simple scatter plots of unemployment against six standard measures of labor market institutions for five-year periods between 1980 and 1999 show no significant relationships. In their multivariate tests, which follow standard approaches, the authors find weak and even perverse effects of the standard institutional variables. They conclude that "the empirical case has not been made that could justify the sweeping and unconditional prescriptions for labor market deregulation which pervade much of the policy discussion."
Article
When a model for panel data includes lagged dependent explanatory variables, then the habitual estimation procedures are asymptotically valid only when the number of observations in the time dimension (T) gets large. Usually, however, such datasets have substantial sample size in the cross-section dimension (N), whereas T is often a single-digit number. Results on the asymptotic bias (N → ∞) in this situation have been published a decade ago, but, hence far, analytic small sample assessments of the actual bias have not been presented. Here we derive a formula for the bias of the Least-Squares Dummy Variable (LSDV) estimator which has a approximation error. In a simulation study this is found to be remarkably accurate. Due to the small variance of the LSDV estimator, which is usually much smaller than the variance of consistent (Generalized) Method of Moments estimators, a very efficient procedure results when we remove the bias from the LSDV estimator. The simulations contain results for a particular operational corrected LSDV estimation procedure which in many situations proves to be (much) more efficient than various instrumental variable type estimators.
Article
Nickell et al. (200522. Nickell , S. , Nunziata , L. and Ochel , W. 2005. “Unemployment in the OECD Since the 1960s. What do We Know?”. Economic Journal, 115 January: 1–27. [CrossRef], [Web of Science ®]View all references) have frequently been cited as empirical evidence that labor market rigidities cause high unemployment. We find that their model is not robust. Leaving their database unchanged and changing three details in their estimation procedure, it turns out that several policy-relevant coefficients change sign or significance. We conclude that their claim from Non Accelerating Inflation Rate of Unemployment (NAIRU) theory that labor market rigidities cause unemployment is rather shaky. There is a remarkable discrepancy between weak empirical results and sweeping conclusions by policy practitioners with respect to the call for deregulation of labor markets.