Flexibility at the margin and labor market volatility in OECD countries∗ Hector Sala
José I. Silva
Universitat Autònoma de Barcelona†
Universitat Jaume I de Castelló‡
and IZA
Manuel E. Toledo Universidad Carlos III de Madrid§
January 26, 2009
Abstract We study the business cycle behavior of segmented labor markets with flexibility at the margin (e.g., just affecting fixed-term employees) and ask whether these types of labor markets can display similar volatilities as fully deregulated ones. We present a matching model with temporary and permanent jobs where (i) there is a gap in the firing costs associated with these type of jobs, and (ii) there are restrictions in the creation and duration of fixed-term contracts. We show that the scenario of flexibility at the margin provides an intermediate situation, in terms of unemployment volatility, between fully regulated and fully deregulated labor markets. This analysis yields new insights into the interpretation of the recent volatility changes witnessed in the OECD area. Key Words: Flexibility at the margin, Volatility, Separation costs, Matching model. JEL Classification Numbers: J23, J41, J63.
∗
We would like to thank the participants of the 2008 International Conference on Labor Market Outcomes: A Transatlantic Perspective, held in Paris, and the participants of the 2007 Symposium on Economic Analysis, held in Granada, for valuable insights. We are also grateful to Julián Messina for insightful comments on earlier versions of this paper. Hector Sala is grateful to the Spanish Ministry of Education and Science for financial support through grant SEJ2006-14849/ECON. Constructive comments from two anonymous referees are also acknowledged. † Departament d’Economia Aplicada, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain; tel.: + 34 93 5812779; email:
[email protected]. Correspondence address. ‡ Departament d’Economia, Universitat Jaume I de Castelló, Campus de Riu Sec, 12071 Castelló, Spain; tel: +34 964 728606;. e-mail:
[email protected]. § Departamento de Economía, Universidad Carlos III de Madrid, Calle Madrid 126, 28903 Getafe, Spain; tel: + 34 91 6249791; e-mail:
[email protected]
1
1
Introduction
Flexibility at the margin is achieved in segmented labor markets with high protection of permanent workers and loose regulation of fixed-term employment. The main objective of this paper is to provide further understanding of the business cycle behavior of segmented labor markets with limitations in the use of fixed-term contracts. In particular we explore whether this flexibility at the margin is the reason why labor markets with relatively high degree of employment protection may display similar volatility as fully flexible ones. Figure 1a shows a negative relationship between unemployment volatility (years 19701990) and the EPL index in the late 1980s (with a correlation coefficient of -0.28). This is in line with the negative relationship documented in Thomas (2006) and Veracierto (2008) in terms of output and employment volatility (rather than unemployment). Since the early 1990s, however, several phenomena have affected this relationship, which now reveals a positive correlation between unemployment volatility and the EPL index (figure 1b).
Figure 1. Unemployment volatility vs. EPL a. 1970-1990 18
b. 1991-2006 18
SWE
16
16 AUT 14
GER
US
Stand ard deviation
Stand ard deviation
14 12 UK CAN 10
DNK
8
POR
JPN BEL
6
FRA
ITA
4
NTH
NTH
GER 10
DNK BEL IRE AUT
US 8
UK
4
2
POR
12
6
SPA
IRE
SWE
SPA JPN
CAN
ITA FRA
2 0
1
2
3
4
5
0
EPL index in late 1980s
1
2
3
4
5
EPL index in late 1990s
Among these phenomena, there is a general fall in the business cycle volatility sometimes known as the "Great Moderation", with specific features in the Anglo-Saxon and Euro-zone countries (Stock and Watson, 2005). In addition, unemployment fluctuations have been magnified by the enhanced international capital mobility, especially in the more responsive small economies (Azariadis and Pissarides, 2007). Finally, there is an increasing use of temporary contracts after several waves of labor market reforms introduced in many OECD countries. These labor market reforms have affected the relative strictness of EPL on fixedterm and permanent contracts (see Table 2.A2.6, in OECD, 2004). As a consequence, firms 2
have adapted their hiring and firing policies, and today have different responses to business cycle shocks. In this context, the aim of this paper is to assess whether these reforms may help to explain the change in the relationship between labor market fluctuations and employment protection. How have firms reacted to the changing institutional setting? Have they adapted their workforce management strategy? Our hypothesis is that, following the possibility of hiring on a fixed-term basis, firms are using flexibility at the margin as an important device for workforce adjustment. According to this, job creation and job destruction would be mainly concentrated on the segment of temporary employment, and this would help explain the high labor market volatilities displayed by countries with strict EPL on permanent contracts and a high (or growing) share of fixed-term employment. Of course, this is not the whole story, but our goal is to assess to what extent this increase in volatility is accounted for by flexibility at the margin. This hypothesis is somewhat endorsed by the OECD (2004), where it is shown that countries having undertaken EPL reforms (and thus eased the relative strictness of EPL for fixed-term contracts relative to the one on permanent contracts) have had a more intensive use of temporary work. Along these lines, Table 1 offers crucial information that allows the differentiation of two labor market types that characterize many OECD economies. First, the well-known Anglo-Saxon type, which is characterized by a small degree of EPL in regular contracts, low firing costs on PCs, and no limitations on the renewal and duration of temporary contracts -which we denote by [NL]-. As a consequence of the high flexibility in the regular segment of the market, there is a limited use of temporary contracts, for example in Australia (5.8%), Ireland (6.8%), UK (6.3%) and US (4.5%). Second, the flexibility-at-the-margin type, which combines a high degree of employment protection in the regular segment with a limited flexibility in the use of temporary contracts -which we denote by [L] in Table 1-. Economies such as Portugal, Sweden and Spain are among those with the highest EPL index values (4.3, 2.9 and 2.6, respectively) and display the highest temporary shares (16.1%, 14.6% and 32.9%). It is well known that fixed-term employment contracts have been introduced in a number of European countries as a way of providing labor flexibility to economies with high employment protection levels. Nevertheless, the implementation of temporary contracts has typically included restrictions such as limited renewals or maximum durations.1 For example, the Spanish 1984 labor market reform crucially broadened the scope of fixedterm contracts while, at the same time, restricted to 3 the maximum number of successive contracts with a maximum accumulated duration of 2 years (OECD, 2004). In Portugal, temporary contracts can also be renewed up to 3 times, but with a longer maximum duration of 30 months. These limitations provide a great source of labor turnover and, thus, of 1
For a comprehensive overview of such restrictions see OECD (2004).
3
labor market volatility. For example, unemployment volatilities in Portugal and Sweden are 12.0% and 12.9%, well above those of UK and US (5.3% and 8.5%) in the period 19912006. Spain, with 7.2%, falls between these two deregulated countries. In other words, flexibility at the margin may be important in order to explain the quick adjustments and large volatilities observed in some European labor markets that are characteristic of the Anglo-Saxon economies. Table 1. Contract legislation and unemployment volatility in OECD countries Contract legislation Volatility Restrictions on TCs
EPL on PCs
Firing costs
Share
s.d. (u)
[]
[]
[]
[]
[]
[NL] [L] [NL] [L] [L] [L] [L] [NL] [L] [NL] [L] [L] [L] [L] [NL] [NL]
1.5 1.7 1.3 1.5 2.3 2.3 2.7 1.6 1.8 2.4 3.1 4.3 2.6 2.9 0.9 0.2 2.4 1.3
4 16 28 0 26 32 69 24 11 4 17 95 56 26 22 0 13.7 34.8
5.8 7.3 12.3 10.4 17.3 12.7 11.8 6.8 8.9 11.6 12.5 16.1 32.9 14.6 6.3 4.5 14.4 7.9
6.9 8.1 5.9 8.9 10.9 4.8 10.5 8.0 3.9 5.2 12.9 12.0 7.2 12.9 5.3 8.5 9.2 6.6
Australia Belgium Canada Denmark Finland France Germany Ireland Italy Japan Netherlands Portugal Spain Sweden UK US Average [L] countries Average [NL] countries [A] [B] [C] [D] [E] Note:
Refers to limited renewals and a maximum duration of temporary contracts (TCs): OECD, (2004); [NL] stands for No Limitations, [L] for Limitations; EPL index on permanent contracts (PCs) in the late 1990s: OECD (2004); Weeks of salary: Doing Business Database, The World Bank Group (2008); Share of TCs in 1991-2006: Eurostat (2007); except Australia (1997), Canada and Japan (1997-2004), and US (1995-2001): OECD (2006); Standard deviation of the cyclical component of standardized unemployment in 1991:1 - 2006:4: OECD Main Economic Indicators (2007). We detrend unemployment data using the HP filter with a smoothing parameter of 1600.
To further check to what extent this is a promising hypothesis, Table 1 also provides the averages by countries with and without restrictions on temporary contracts. Observe that the first group has substantially higher average EPL index (2.4 versus 1.3) and firing costs
4
(34.8 versus 13.7), and a higher share of temporary contracts (14.4% versus 7.9%), while it displays a larger unemployment volatility (9.2 versus 6.6). Thus, despite the Anglo-Saxon labor markets having less stringent legislation, unemployment in segmented/dual labor markets with restrictive EPL and high firing costs in the regular side is, on average, 33% more volatile. The influence of institutions on the performance of the labor market has been receiving utmost attention in recent years. Within this context, the study of the impact of firing taxes on business cycle fluctuations is in its initial steps, and only recently a number of studies have started to deal with this issue. Veracierto (2008), for example, develops a Real Business Cycle model to show that firing costs are important in reducing business cycle fluctuations. They preclude employment adjustments and lower the response of the economy to aggregate productivity changes. Thomas (2006) reaches the same conclusion using a matching model and considering economies with different firing costs. Zanetti (2007), also within the matching framework with nominal rigidities in the goods market, studies the impact of unemployment benefits and firing taxes on business cycle fluctuations. Higher firing taxes are found to reduce the volatility of output and employment. Here we extend the equilibrium matching model of Mortensen and Pissarides (1994) by introducing the possibility that firms hire workers on a fixed-term basis. We thus differentiate between permanent and temporary employees, where the latter have fixed-term contracts and virtually zero firing costs. This is a standard distinction in the matching literature, but has generally been used to conduct long-term analyses (for example, Wasmer, 1999; Blanchard and Landier, 2002; Kugler et al., 2002; Cahuc and Postel-Vinay, 2002; Osuna, 2005). In contrast,2 this paper aligns with those of Cabrales and Hopenhayn (1997), Thomas (2006), Veracierto (2008), and Zanetti (2007), and focuses on business cycle fluctuations. The first of these studies distinguishes between permanent and temporary work, but calibrates a labor demand rather than a matching model. The latter three abstain from such a distinction and have no possibility of considering how these fluctuations are influenced by flexibility at the margin. This consideration, important in the assessment of the firing costs’ impact on these fluctuations, is among the contributions of this paper. More precisely, this paper contributes to the understanding of the sources of unemployment volatility by assessing the role played by (i) the gap between the separation costs of the fixed-term and permanent employees; and (ii) restrictions in the use of temporary contracts. Although not providing a full account of the facts, we claim that these are important channels enhancing the volatility achieved by segmented labor markets. Our model provides a stylized representation of a labor market with restrictions on the hiring, job conversion and firing processes. Regarding hiring and job conversion, we 2
Our paper also differs from Boeri and Garibaldi (2007), who focus on the transitional dynamics of EPL reforms providing flexibility at the margin.
5
assume regulations preventing (i) all new contracts to be signed on a fixed-term basis, and (ii) unlimited duration of fixed-term contracts. As for firing, temporary and permanent jobs are both subject to firing costs (the former being lower) which result in a gap in separation costs. This set of restrictions generates flexibility at the margin and provides different channels whereby regulated labor markets may display similar volatilities as fully flexible ones. The calibration of our benchmark model to a representative European labor market with restricted flexibility at the margin, and its simulation, allow us to analyze the crosscorrelation of key labor market variables. In particular, the model shows a procyclical behavior of both temporary and permanent jobs and in the share of fixed-term contracts. Vacancies, together with the job finding and job conversion probabilities, are also procyclical, and the expected negative correlation between vacancies and unemployment is reproduced. We show that these results are broadly consistent with the cyclical behavior observed in the Spanish and French labor markets, which are economies with limited flexibility in the use of temporary jobs. With respect to labor market volatility, the main results stem from considering a situation with flexibility at the margin. We show that, in this situation, the gaps in separation costs between temporary and permanent jobs and the use of fixed-term contracts, although restricted, increase the unemployment volatility with respect to a situation of strict EPL and no gap in firing costs. It should be noted that, within each scenario, we find the standard result that a rise in firing costs reduces the volatility of unemployment (Thomas, 2006; Veracierto; 2008; and Zanetti, 2007). Moreover, we show that the scenario of flexibility at the margin provides an intermediate situation, in terms of unemployment volatility, with respect to the fully regulated and fully deregulated labor markets. In short, our simulated scenarios are not able to explain, on their own, why some OECD countries with limited flexibility in the use of temporary contracts have become as volatile as the fully deregulated ones. The main reason behind the higher unemployment volatility observed in the scenario with flexibility at the margin with respect to a fully regulated labor market is simple. To avoid transitions to a permanent status, which entails future costs in case of adjustments, firms’ workforce adjustments take place more intensively and with higher frequency in temporary jobs. In particular, rather than converting fixed-term contracts into permanent ones, firms will let those temporary contracts affected by the legal conversion restriction expire and hire workers with new temporary contracts not yet affected by this restriction. This situation generates large volatilities in the job destruction probabilities associated with temporary contracts, as well as in unemployment. However, in the presence of limitations in the hiring, and duration of fixed-term contracts, these lager volatilities are somewhat reduced. Adjustments on temporary workers become constricted, it is thus more difficult to 6
avoid the firing costs on permanent contracts, and firms respond by reducing the intensity of job destruction. Summing up, higher restrictions in the hiring and conversion process of fixed-term contracts into permanent ones reduces the volatility of job destruction, and therefore, of unemployment. When flexibility at the margin is suppressed in a scenario of employment protection, most of the unemployment volatility in our model vanishes and gives rise to a scenario similar to the one before the explosion of fixed-term employment in many OECD countries. This situation corresponds to the one in the aftermath of the labor market reforms that took place since the 1990s in many of these economies. A final important result is the countercyclical behavior we find between job destruction and the business cycle both in fully deregulated and flexibility-at-the-margin labor market types. This result clarifies a similar empirical finding for Spain in Messina and Vallanti (2007). Moreover, it helps explain why the job turnover rate of some regulated labor markets displays a countercyclical behavior in contrast to the acyclical or even procyclical movements suggested by some studies (Garibaldi, 1998). The remaining of the paper is structured as follows. Section 2 presents the model which, in Section 3, is calibrated and simulated to a benchmark labor market representative of an average European one with flexibility at the margin. In Section 4 we use it to assess the effects of potential labor market reforms affecting key parameters. Section 5 concludes.
2
The model
The economy consists of a continuum of risk-neutral, infinitely-lived workers and firms. We normalize the measure of workers to 1. Workers and firms discount future payoffs at a common rate . Moreover, capital markets are perfect and time is discrete. Workers may be either unemployed or employed. Unemployed individuals enjoy an instantaneous utility each period. Those who are employed can be so either under a temporary or a permanent contract, and earn a wage and , respectively. Each period any worker may be endogenously terminated if the firm chooses to do so, which entails a firing cost if temporary or if permanent. We assume that . Exogenous separations may also occur at no cost with probability for any type of worker. When an employment relationship is broken, the worker becomes unemployed. Moreover, each temporary contract expires each period with probability . This parameter reflects legal restrictions regarding the use of fixed-term contracts such as limited number of renewals and maximum duration. For instance, a higher indicates stringer legal restrictions in the use of temporary contracts. We use this stochastic approach to model the overall duration of fixed-term contracts for simplicity. When this type of contract expires the worker may be either hired under a permanent contract or laid off. That 7
is, firms have the option of offering these workers a permanent contract (which we define as job conversion). Thus, although the expiration probability is constant, the job conversion probability is still an endogenous variable. Also notice that if separation occurs in this case, the firm pays no firing costs. Each firm consists of only one job which is either filled or vacant, and uses only labor as input. If a job is filled, it produces units of output each period , where and represent aggregate and match-specific productivity shocks, respectively. Aggregate productivity follows a Markov process whereas is independent and identically distributed across firms and time, with cumulative distribution function () and support [0 ¯]. Thus, each productive firm yields an instantaneous profit − ( ), if filled with a temporary worker, or − ( ), otherwise. When a job is vacant, the firm searches for an unemployed worker to fill the position at a constant cost per period. Search frictions in the labor market are captured by a constant-returns-to-scale matching function ( ), where denotes the unemployment rate and is the vacancy rate in period . We follow den Haan et al. (2000) and assume ( ) =
(
+ )1
0
(1)
This functional form ensures that the ratios ( ) and ( ) lie between 0 and 1. The former represents the probability that an unemployed worker meets a vacant job, which we write as ( ) = (1 ), where = . The latter denotes the probability at which vacancies meet workers, ( ) = (1 1). This matching function also implies that the higher the number of vacancies with respect to the number of unemployed workers (i.e., larger ), the easier to find a job and the more difficult to fill up vacancies. Job creation takes place when an unemployed worker meets a vacant job and they agree on an employment contract. When there is a meeting, a match-specific productivity is drawn. If the match is profitable, the firm hires the worker. With no restrictions, a firm would always offer a temporary contract due to our assumption about separation costs. Namely, and no separation costs associated with transitory contract expiration. However, following Cahuc and Postel-Vinay (2002), we assume that with probability , the firm is bound to offer a temporary contract if it chooses to hire the worker. This is a policy parameter reflecting restrictions on hiring. Within a period, the timing of events is as follows. At the beginning of each period unemployed workers and vacancies meet. At the same time, all existing matches (i.e., those who produced last period) learn whether they break exogenously with probability . Right after that, surviving temporary matches realize whether their contract expires according to probability . Afterwards, each match (old and new) draws a idiosyncratic productivity. If is sufficiently large and, consequently, the match has positive surplus, production takes
8
place. Otherwise, the firm either breaks the existing employment relationship or does not offer a contract to the new employee. Moreover, it may choose to open a vacancy and search for a new worker.3 Accordingly, the value of vacancies , and filled positions, ( ) and ( ) (after the idiosyncratic productivity is realized), are represented by the following Bellman equations, "
= − + (1 − ( ))+1 + ( ) +( )(1 − ) "
ÃZ
( ) = − ( ) + (1 − ) + (1 − )(1 − )
ÃZ
¯
+1
"
ÃZ
+1
¯
+1
ÃZ
¯
(+1 ())() + (e +1 )+1
(+1 ())() + (e +1 )+1 ¯
+1
!#
+1 ()() + (e +1 )+1
! (2)
!
# ! ¢ ¡ + +1 (3) +1 ()() + (e +1 ) +1 −
( ) = − ( ) + (1 − )
ÃZ
¯
+1
! ¡ ¢ +1 ()() + (e +1 ) +1 −
+ +1 ] (4)
where e , = , are productivity thresholds defined such that nonprofitable matches (i.e., with negative surplus) are severed.4 Thus, the conditions defining these thresholds (also called reservation productivities) for job creation and destruction are:
) − = 0 (e
(5)
(e ) − = 0
(6)
(e ) − + = 0
(7)
(e ) − + = 0
(8)
The first two expressions (5) and (6) are job creation conditions for temporary and permanent jobs, respectively. Note that in those cases the firm is not liable to pay or in the absence of agreement. Thus, unemployed workers find temporary and permanent jobs with probabilities
)) = (−1 )(1 − (e
(9)
= (−1 )(1 − )(1 − (e )) 3
(10)
In fact, in equilibrium, unmatched firms are indifferent between opening a new vacancy and completely withdrawing from the market. 4 Since the value of a match is increasing in , we can prove that there exists a threshold e ∈ [0 ¯] below which matches are no longer profitable.
9
Recall that when a firm and a worker meet they can only agree on a temporary contract with probability . Finally, the (unconditional) job finding probability is simply = +
(11)
Notice that condition (6) also defines the threshold for temporary-to-permanent conversions. Thus, temporary employees become permanent with the following job conversion probability )) (12) = (1 − )(1 − (e Equations (7) and (8) set the reservation productivity for current temporary and permanent workers. Therefore, these two conditions, together with (6), characterize endogenous job destruction.5 It follows that temporary and permanent matches separate with probabilities
h i = + (1 − ) (1 − )(e ) + (e )
= + (1 − )(e )
(13) (14)
and the total job destruction probability is =
−1 + −1 1 − −1
(15)
where and are the mass of temporary and permanent workers in period . That is, those who actually are in productive matches that period. From the worker’s perspective, the value of being unemployed , and employed, ( )
5
Recall that firms may avoid temporary-to-permanent conversions when fixed-term contracts expire. Therefore, job destruction also depends on condition (6).
10
and ( ), are expressed as follows, "
= + (1 − ( ))+1 + ( ) + ( )(1 − )
"
( ) = ( ) + (1 − ) + (1 − )(1 − ) "
( ) = ( ) + (1 − )
ÃZ
ÃZ
+1
ÃZ
¯
+1
¯
ÃZ
ÃZ
+1 ()() + (e +1 ))+1
+1 ()() + (e +1 )+1
+1 ()() + (e )+1
¯
+1
+1
+1 ()() + (e )+1
¯
+1
¯
+1 ()() + (e )+1
!
!
!
!#
! (16)
#
+ +1 (17) #
+ +1 (18)
We also assume that there is free entry for firms. Hence firms open vacancies until the expected value of doing so becomes zero. Therefore, in equilibrium: = 0
(19)
Furthermore, because neither workers nor employers can instantaneously find an alternative match partner in the labor market, and because hiring and firing decisions are costly, a match surplus exists. To divide this surplus we assume wages to be the result of bilateral Nash bargaining between workers and firms. They are revised every period upon the occurrence of new shocks, and the Nash solution is the wage that maximizes the weighted product of the workers’ and the firms’ net return from the job match. The first-order conditions for the temporary and permanent employees yield the following two equations: (1 − )( ( ) − ) = ( ( ) − + )
(1 − )( ( ) − ) = ( ( ) − + )
(20) (21)
where ∈ (0 1) denotes the workers’ bargaining power relative to firms. Note that the Nash conditions present terms depending on and . Because separation costs are operational they are explicitly considered in the wage negotiation. This implies that the firms’ threat point when negotiating with a worker is no longer the value of a vacancy but ( − ) or ( − ) depending on the type of worker. To fully characterize the dynamics of the model economy, we need to define the law of motion for the unemployment rate , and the mass of temporary and permanent workers,
11
and . These evolve according to the following difference equations, = −1 + −1 + −1 − −1 − −1
= −1 + −1 − −1 − −1
= −1 + −1 + −1 − −1
(22) (23) (24)
with + + = 1. Finally, aggregate output is equal to h i ¡ ¢ = ¯ (1 − )−1 + ¯ −1 + −1 + ¯ (1 − )−1 + ¯ −1 − (25) where ¯ = [| ≥ e ].
3
3.1
Calibration and simulation of the model economy Calibration
We calibrate the model at quarterly frequencies in order to match key empirical facts and steady-state values of a representative European labor market with limited flexibility in the use of fixed-term contracts. Our parametrization is summarized in Table 2. The hiring probability of a new temporary job , and the exogenous separation probability are set to match: (i) the actual average share of temporary workers on total dependent employment which, as shown in Table 1, amounts to 14.4%; and (ii) the 10 years average job tenure of a permanent contract estimated by the OECD (2004). This yields = 09 and = 0025. Also according to the OECD (2004), six months is the average duration of a fixed-term contract in the OECD economies. These contracts are thus assumed to expire stochastically with probability = 05. Regarding the firing costs we rely on several sources. First, on the World Bank’s Doing Business survey and its detailed study of the EPL in many economies. In particular, we take the average firing costs of a permanent contract in the countries with limited flexibility, which is equivalent to 35 weeks of weekly wages (see Table 1). Second, on Garibaldi and Violante (2005), who place the ratio of firing tax over total firing costs between 0.34 (when worker and firm reach no off-court agreement) and 0.20 (when there is a 50% probability of reaching such agreement). We consider the first of these scenarios and set this ratio to 0.34. Thus, the firing tax component of permanent jobs amounts to = (3512) × ∗ × 034 ≈ ∗ , which is about 1 in the steady-state equilibrium. Moreover, we assume no effective firing tax on temporary jobs and set = 0. Following den Haan et al. (2000), the idiosyncratic productivity is assumed to be
12
log-normally distributed with mean and standard deviation , whose values are set to 0 and 0.2, respectively. For the literature provides a range of values between 0.1 (den Haan et al., 2000) and 0.4 (Trigari, 2005) and we choose an intermediate case (see also Burgess and Turon, 2005; and Walsh, 2005, who use values within this range). In turn, the steady-state aggregate labor productivity ∗ is normalized to one. The logarithm of this variable follows an AR(1) process log = log −1 + , with ∼ (0 ). The values of the autoregressive parameter and the standard deviation of the white noise process, = 095 and = 00095, are calibrated to match the average cyclical volatility (1.3 percent) and persistence (0.75) of the labor productivity, (1 − ).6 We assume a plausible quarterly interest rate of nearly 1 percent in the steady state and, accordingly, set the discount factor at 099. The workers’ bargaining power is set to 05. Petrongolo and Pissarides (2001) identify an elasticity of unemployment with respect to the matching function in the range 0.5-0.7. We take 0.7 as reference and thus set the matching parameter at 072.
Table 2. Calibrated parameters for the benchmark economy Parameters Value Source ∗ Aggregate labor productivity 1 Normalized Mean for the distribution of 0 Normalized Discount rate 099 [A] Standard deviation for the distribution of 02 [A] Workers’ bargaining power 05 [A] Duration probability of a temporary contract 05 [A] Parameter of the matching function 072 [A] Firing tax of permanent contracts 1 [A] Firing tax of temporary contracts 0 [A] Exogenous separation probability 0025 [B] Hiring probability of a temporary contract 09 [B] Persistence parameter of 095 [C] Standard deviation of 00095 [C] Hiring costs 003 [D] Leisure parameter 05 [D] Unemployment benefits 04 [D] [A] Other studies, data or own assumptions as explained in main text. [B] Set to match the average share of temporary workers and the average job tenure of permanent contracts. [C] Set to match the cyclical volatility and persistence of labor productivity. [D] Set to match the elasticity of unemployment duration with respect to unemployment benefits (UB) and the ratio of UB to average wage.
6
Data on labor productivity for each country is taken from the OECD’s Main Economic Indicators Database, and reported as deviations from an HP trend with smoothing parameter 1,600.
13
The instantaneous utility of unemployment is assumed to have two components. Thus, = + , where represents the utility associated to leisure and home production, and denotes unemployment benefits. These parameters, together with the vacancy costs , are set to ensure simultaneously that in the steady state (i) job creation is equal to job destruction; (ii) the elasticity of unemployment duration with respect to unemployment benefits is equal to 2.0, as suggested by Addison, Centeno and Portugal (2004); and (iii) the ratio of unemployment benefits to the average wage is equal to 0.4 as reported by Nickell and Nunziata (2001). In this way we obtain = 05, = 04, and = 003.7
3.2
Simulation
Table 3 summarizes the steady state and business cycle results from our benchmark simulation. In this simulation we (i) create 10,000 sample paths of 1,052 quarters, throw away the first 1,000 and keep the 52 quarters corresponding to 1991-2003; (ii) detrend the generated data using the HP filter with the smoothing parameter equal to 1 600; and (iii) calculate the standard deviations and correlation coefficients of the relevant variables.
Table 3. Simulated results for a representative economy with limited flexibility ( = 076; = 000; = 100; = 050) St. stt.
0.138
0.374
2.719
0.124
0.738
0.144
0.503
0.080
1.009
0.132
1.030
St. dv. Autoc.
0.063 0.897
0.083 0.566
0.124 0.713
0.043 0.861
0.012 0.961
0.041 0.889
0.059 0.832
0.034 0.544
0.010 0.711
0.101 0.713
0.013 0.751
1
-0.450 1
-0.804 0.893 1
-0.320 0.845 0.724 1
-0.807 -0.075 0.358 -0.302 1
-0.093 0.779 0.565 0.973 -0.514 1
-0.919 0.614 0.873 0.642 0.525 0.450 1
0.720 -0.528 -0.714 -0.090 -0.668 0.081 -0.522 1
-0.820 0.880 0.999 0.699 0.389 0.535 0.875 -0.735 1
-0.804 0.893 1.000 0.724 0.358 0.565 0.873 -0.714 0.999 1
-0.812 0.883 0.998 0.747 0.351 0.587 0.896 -0.675 0.996 0.998 1
Correl. matrix
How do our results for a stylized economy compare with actual figures? Some key figures for France, Spain, and the US are presented in Table 4. The first two are salient cases, in Europe, of regulated labor markets with flexibility at the margin but different 7
Note that the steady-state average wage shown in Table 3 below is nearly 1.
14
outcomes in terms of labor market performance. The US is at the other extreme and provides the paradigm of a fully flexible labor market. The unemployment volatilities in France (0.047) and Spain (0.066) are lower than in the US (0.09), especially in France where it is half that of the US. In turn, the volatilities of the job finding rate are similar in Spain and the US (0.063 and 0.059, respectively), and are higher than in France (0.049). Regarding the job destruction rate the differences are narrower, France being the economy with the lowest volatility (0.038). Our calibration of a stylized economy provides close values to these figures and is representative of a labor market in an intermediate position between France and Spain. Specifically, the volatilities of unemployment (0.063) and the job finding rate (0.059) fall in between, whereas that of the job destruction rate (0.034) is the lowest and not too distant from the French one. Note also that the simulated signs of the correlation coefficients between the unemployment, job finding and job destruction rates are the correct ones for the three economies considered.
Table 4. Summary statistics for France, Spain and US. Quarterly data. Spain (1991-2003)
France (1991-2003)
US (1991-2003)
Std. deviation
0.066
0.063
0.045
0.047
0.049
0.038
0.090
0.059
0.042
Autocorrelation
0.928
0.474
0.201
0.921
0.705
0.809
0.928
0.757
0.360
1
-0.510
0.100
1
-0.804
0.414
1
-0.868
0.295
1
-0.020
1
-0.313
1
-0.040
Correlation matrix
1
1
Sources: OECD Main Economic Indicators (2007) for the unemployment rate; Petrongolo and Pissarides (2008) for the job finding and job destruction rates of France and Spain, Shimer (2005) for the US ones. Note: We detrend the data using the HP filter with a smoothing parameter of 1,600.
Another important feature of our stylized economy is that we match the expected behavior of some other key variables such as vacancies, employment (both temporary and permanent), the share of fixed-term employment, and the job conversion probability. For example, vacancies display a higher volatility than unemployment, as in Spain (see Table 10 in Sala and Silva, 2009) but in contrast with the US where these are very similar (see Table 1 in Hornstein et al., 2005, for 1951-2004). Also, the volatilities of temporary and permanent employment in our model (0.043 and 0.012, respectively) are close to those in Spain (as reported in Sala and Silva, 2009, for 1997-2004). Other significant features are the positive correlations between the job conversion probability and labor productivity (0.998), and between the share of fixed-term contracts and labor productivity (0.587). 15
1
Both, the job conversion probability and the share of fixed-term contracts, are clearly procyclical. Although not perfect, our stylized model provides an accurate picture of what can be thought as a representative European labor market with flexibility at the margin. Of course, this labor market cannot be specifically associated with a particular country, but it is in broad terms an appropriate benchmark for assessing the impact of different labor market reforms on the volatility of the unemployment rate. These reforms will consist of changes in legislation that shift the four crucial parameters governing the hiring and firing processes in our model: , , , and . The goal of this assessment is to shed new light on the potential consequences of these reforms for countries currently seeking to enhance their labor market responsiveness through flexibility at the margin.
4
Assessing the impact of selected EPL changes
We distinguish different legislative scenarios: 1. The benchmark scenario presented above, which is characterized by employment protection and temporary contracts. Since most OECD countries present different degrees of flexibility at the margin, this scenario is modified in different ways: (a) For a given level of restriction on fixed-term job creation () and a given gap in separation costs ( − 0), we consider different job conversion probabilities: takes values from 0.1 to 1. (Recall that a higher value of is associated with tougher restrictions on fixed-term employment in terms of duration and number of renewals of these contracts). (b) For a given level of restriction on fixed-term employment () and a given gap in separation costs ( − 0), we consider different levels of hiring restrictions: takes values from 0 to 1. (Recall that higher values of implies fewer restrictions in the creation of temporary contracts). (c) For a given level of restriction on fixed-term job creation () and fixed-term job conversion (), we consider: (i) different firing costs on permanent contracts: takes values from 0 to 1.5. (ii) a lower gap in separation costs: is progressively increased to approach = 1. (Since 0, this implies that firms are bound to pay firing costs on fixed-term contracts provided the employment relationship terminates before they expire). 2. A scenario with no employment protection, which attempts to mimic an economy with essentially no legal restrictions of the type studied in our analysis. This implies the 16
existence of just one type of job since temporary contracts become perfect substitutes. Therefore and become irrelevant, and we set = = 0. 3. A scenario with employment protection and no temporary contracts, which attempts to mimic the situation of several OECD labor markets before the introduction of temporary contracts and the development of fixed-term employment (a paradigmatic case would be Spain before its 1984 labor market reform). In terms of the model, this has two implications. First, firms are no longer able to make use of fixed-term contracts and, as a consequence, we set = 0 and = 0. Second, and become irrelevant. Next we describe the effects of these EPL changes on the model’s steady-state. Then, we simulate each of the above scenarios and assess their impact on business cycle dynamics.
4.1
Steady state results
When modifying each of these key parameters, we hold the rest constant and compute the new steady state values of the endogenous variables. Table 5 shows the results from the conducted simulations by distinguishing five analytical panels. The first noteworthy result is that higher values of , , and , and lower values of , reduce the number of vacancies relative to unemployment () and this, in turn, diminishes the workers’ hiring probability . Intuitively, this is the outcome of the firms’ internalization of stricter hiring restrictions and higher expected firing costs, whose response hinders job creation. Note that this result is especially significant in the scenario without flexibility at the margin (panel 1), because firing costs become immediately operational once the match takes place. For example, in the absence of temporary contracts, when the firing tax is increased from 0% to 100% of the average quarterly wage, falls from 61.5 to 9.9. Similar sensitivities are observed in the case of tougher restrictions in the hiring process of fixed-term workers (panel 3) and when firing costs become operational in temporary jobs (panel 5). In contrast, this probability is reduced from 61.5 to 50.3 in a situation of flexibility at the margin and no firing costs in fixed-term jobs (panel 2). The second main result is the behavioral change in separations across scenarios. On the one hand, they decline when temporary jobs are not allowed (panel 1), and also the tougher the restrictions on both the creation of fixed-term employment and its conversion into permanent employment (panels 3 and 4). On the other hand, they rise the higher the gap in firing costs ( − ) in a labor market with flexibility at the margin (panels 2 and 5). This is due to the additional separations affecting temporary jobs. The first response (in panel 1) is a well known result in the literature. The higher the firing costs, the more expensive it becomes to shed workers, and the lower the job destruction rate . In our case from 4.8% to 2.5% when the gap in firing costs rises from 17
0% to 100%. This response, however, changes dramatically when firms are allowed to have flexibility at the margin (panel 2), in which case the same rise in firing costs increases the job destruction probability from 4.8% to 8.0%. To avoid incurring firing costs, firms make more intensive use of fixed-term contracts the higher the separation tax on permanent workers. Note that this larger response takes place when the gap in firing costs surpasses 50%, and is entirely driven by endogenous job destruction of fixed-term contracts.
Table 5. Simulated steady states under different EPL scenarios (%).
(1−)
Panel 1: = 000 = 000 (Full flexibility) = 025 = 050 = 075 = 100 = 150
7.18 6.53 10.57 15.21 20.22 31.11
61.54 35.79 21.13 13.93 9.86 5.53
4.76 2.50 2.50 2.50 2.50 2.50
– – – – – –
– – – – – –
– – – – – –
Panel 2: = 090 = 000 (Full flexibility) = 025 = 050 = 075 = 100 (Benchmark case) = 150
7.18 5.19 7.35 10.24 13.76 22.80
61.54 58.20 55.16 52.61 50.33 45.87
4.76 3.18 4.38 6.00 8.00 13.54
– 14.94 26.78 35.22 41.00 48.41
– 2.50 2.50 2.50 2.50 2.50
– 5.54 7.73 10.70 14.37 24.10
Panel 3: = 025 = 050 = 075 = 090 (Benchmark case) = 100
18.79 17.14 15.15 13.76 12.76
19.00 29.64 41.98 50.33 56.30
4.39 6.13 7.50 8.00 8.23
45.09 44.05 42.39 41.00 39.89
2.50 2.50 2.50 2.50 2.50
4.46 8.75 14.37 12.53 15.33
Panel 4: = 010 = 025 = 050 (Benchmark case) = 075 = 100
15.02 14.55 13.76 13.39 13.18
52.01 50.84 50.33 50.13 50.00
9.20 8.66 8.00 7.75 7.60
14.10 24.98 41.00 56.52 71.78
2.50 2.50 2.50 2.50 2.50
57.72 23.39 14.37 9.72 7.34
Panel 5: = 000 (Benchmark case) = 010 = 025 = 050 = 075
13.76 15.77 19.33 23.10 25.05
50.33 40.34 27.64 16.35 11.25
8.00 7.55 6.62 4.90 3.76
41.00 38.43 36.21 30.83 24.33
2.50 2.50 2.50 2.50 2.50
14.37 14.06 12.24 8.48 5.78
18
Moreover, when having to face restrictions in terms of the duration and number of renewals of fixed-term contracts (panel 4), firms will find it optimal to further increase their temporary job destructions (as a way to avoid transitions to a permanent status, which entails future costs in case of adjustments). In particular, before converting fixed-term contracts, they will hire another temporary worker and start the process again. However, given the gap in firing costs, the average separation probability is reduced with higher levels of due to the observed reduction in the share of temporary workers (1 − ). In short, in response to higher firing costs in permanent contracts, a situation of flexibility at the margin with restrictions generates a large job destruction rate in temporary jobs and, eventually, in the aggregate separation probability . Third, it is also important to note that, for 0, firms will choose to retain their permanent workers to avoid incurring in those separation costs. In this situation, separations become exogenous to the firms’ decisions. Finally, the outcome in terms of unemployment depends on the relative strength of the hiring and firing probabilities, which may pull in opposite directions (a lower job hiring probability increases unemployment, whereas a lower firing probability decreases it). The direction and relative strength of these effects varies across scenarios. Without temporary employment (panel 1), the effect of the reduced job finding probability in response to higher overcomes the impact of the lower separation probability. In contrast, with flexibility at the margin (panel 2), both the lower and the higher enhance unemployment. The more restricted the conversion of temporary jobs (panel 4), the more this upward effect is reverted.
4.2
Business cycle results
As in the previous section, when modifying the key parameters (, , , and ) we hold the rest constant and compute the new steady state equilibrium. We then solve and simulate the model around this new steady state, and compute the second moments of the relevant variables. These exercises provide new insights into the effects of various EPL schemes and different degrees of flexibility at the margin on the cyclical behavior of the job finding probability , the separations in temporary and permanent jobs ( and ), and unemployment (). Table 6 summarizes the results from the conducted simulations, again by distinguishing five analytical panels. The first important result, in terms of business cycle fluctuations, is the increased volatility of in response to higher values of , and , and lower values of . This holds within the five scenarios considered, as shown in the second column of Table 6. In the absence of temporary contracts (panel 1), when the firing tax is increased from 0% to 100% of the average wage, the volatility of is multiplied by 3.6 (from 3.3 to 11.9). 19
When this rise affects and there is flexibility at the margin (panel 2), this volatility is increased by 77% (from 3.3 to 5.9), and by an extra 78% when firing costs in temporary jobs are increased from 0% to 0.25% (from the benchmark 5.9 to 10.5 in panel 5). When the hiring probability of a fixed-term contract becomes lower (panel 3), for example from 90% to 25%, this volatility increases by an additional 52% and rises from the benchmark 5.9 to 8.9. Finally, the higher the quarterly exogenous component of the job conversion probability (panel 4), for example 100%, the more the volatility increases (from 5.9 in the benchmark scenario to 6.1). The main intuition behind this result is along the lines of Hornstein et al. (2005), Mortensen and Nagypal (2007), and Silva and Toledo (2009). The presence of firing costs reduces the firms’ surplus, making them more responsive to variations in the level of aggregate labor productivity. The same result applies to the presence of more restrictions in the hiring and conversion processes of temporary workers.8 The fact that the job finding rate becomes more volatile with larger hiring and job conversion probabilities implies that, the more restricted the creation and duration of a temporary contract, the more sensitive is the job creation process to productivity shocks. In other words, a more stringent legislation on temporary jobs makes firms perceive good times as even better times, so that they are more prone to adjust vacancies relative to a labor market without fixed-term contracts. The second main result is the reduced volatility of in response to a more stringent legislation (i.e., higher values of , and ), and lower values of . Again this holds within the five scenarios considered, and note that it is valid for both separation rates (on temporary and permanent jobs). In more detail, with employment protection and no temporary contracts (panel 1), a rise in firing costs from 0% to 100% eliminates the volatility of the job destruction probability (which goes from 17.3 to 0.0). Similarly, when firms are allowed to use fixed-term contracts (panel 2), it falls by 80.3% (from 17.3 to 3.4). Further, when the conversion probability from a temporary to a permanent contract, , moves from 50% to 100% (panel 4), the volatility of falls by 82.0% (from 17.3 to 3.12). Similar results are observed with more stringent restrictions on new temporary contracts (panel 3) and higher firing costs (panel 5). This is a well-known result (Garibaldi, 1998; Thomas, 2007) whose essential intuition is that the higher the firing costs, the more expensive it becomes to shed workers, in which case the firms’ job destruction rate becomes less sensitive to shocks. It is also worth 8
In our benchmark calibration, the total expected surplus of a hired worker with a temporary job, ∗ ∗ ∗ ∗ ∗ [ ()| ≥ e ] = [ ()| ≥ e ] + + [ ()| ≥ e ] − , amounts to 35.5% of the aggregate quarterly output ∗ . This value is somewhat lower to the surplus of 45.2% obtained by Costain and Reiter (2008), but significantly higher with respect to the value of 8.0% from Hagerdon and Manovskii (2008). In other words, we avoid the small surplus approach used by Hagerdon and Manovskii by preventing an excessive sensitivity of unemployment duration to unemployment benefits in the steady state as suggested by Costain and Reiter. ∗
20
remarking that, as in the case of the steady state, it is natural to expect zero volatility of when surpasses some threshold value for which separations on permanent jobs become exogenous and equal to 2.5%.
Table 6. Simulated volatilities under different EPL scenarios (%).
()
()
()
( )
( )
( (1−) )
Panel 1: = 0 = 000 (Full flexibility) = 025 = 050 = 075 = 100 = 150
16.92 5.74 5.42 4.64 3.92 2.89
3.30 7.81 9.61 10.79 11.90 14.47
17.29 0.00 0.00 0.00 0.00 0.00
– – – – – –
– – – – – –
– – – – – –
Panel 2: = 090 = 000 (Full flexibility) = 025 = 050 = 075 = 100 (Benchmark case) = 150
16.92 6.43 6.95 6.67 6.28 5.80
3.30 4.28 4.82 5.30 5.85 7.22
17.29 3.48 3.95 3.71 3.39 3.04
– 13.91 8.87 6.57 5.42 4.44
– 0.00 0.00 0.00 0.00 0.00
– 3.23 3.93 4.09 4.12 4.20
Panel 3: = 025 = 050 = 075 = 090 (Benchmark case) = 100
4.37 5.00 5.78 6.28 6.59
8.90 7.45 6.43 5.85 5.42
1.82 2.33 2.90 3.39 3.77
3.47 4.22 4.89 5.42 5.78
0.00 0.00 0.00 0.00 0.00
5.42 4.73 4.28 4.12 4.04
Panel 4: = 010 = 025 = 050 (Benchmark case) = 075 = 100
9.02 7.40 6.28 5.90 5.76
4.61 5.44 5.85 6.01 6.09
7.73 4.96 3.39 3.10 3.12
9.30 7.51 5.42 4.24 3.50
0.00 0.00 0.00 0.00 0.00
1.22 2.90 4.12 4.67 4.99
Panel 5: = 000 (Benchmark case) = 010 = 025 = 050 = 075
6.28 5.77 5.25 4.39 3.84
5.85 8.22 10.54 12.51 13.46
3.39 2.28 3.00 3.66 3.21
5.42 2.99 2.15 2.08 2.35
0.00 0.00 0.00 0.00 0.00
4.12 4.39 6.03 8.75 11.03
A third result concerns the response of unemployment, which depends on whether or not the fall in the volatility of separations exceeds the increase in the volatility of the firing 21
probability. Note that in our analysis it does and, therefore, the unemployment volatility declines in the five scenarios considered. Given this third result, by comparing panels 1 and 2, we can also look at the change in unemployment volatility in response to changes in just the volatility of the separation rate. For example, departing from a scenario with strict employment protection (firing costs at 100% of the average wage) and no segmentation (panel 1), when firms are allowed to use fixed-term contracts not subject to firing costs (panel 2), the unemployment volatility increases by 62% (from 3.9 to 6.3). This result contributes to the understanding of the actual rise experienced by countries which, in the 1990s, undertook partial labor market reforms introducing flexibility at the margin. Finally, it is important to highlight the intermediate position, in terms of business cycle volatilities, of the scenario with flexibility at the margin (panels 2 to 5). A natural place, it seems, between fully deregulated (first row of panel 1 or panel 2) and fully regulated labor markets (the other rows in panel 1). Therefore, it is clear that our simulated scenarios are not enough to explain why some OECD countries with limited flexibility in the use of temporary contracts have become as volatile as the fully flexible ones.
4.3
Employment protection and simulated correlations
Messina and Vallanti (2007) have recently examined the impact of firing restrictions on job flow dynamics. They provide evidence that firms with tight firing restrictions smooth job destruction over the business cycle so that job turnover becomes less countercyclical. This result is in line with previous studies that suggest acyclical behavior of the labor flows in continental Europe in contrast to their countercyclical pattern in Anglo-Saxon countries (see Garibaldi, 1998). However, Messina and Vallanti also find some empirical evidence suggesting that the presence of temporary contracts may revert the acyclical behavior on the job destruction rate. This possibility is what we explore next. In particular, we ask our model to what extent the coexistence of EPL in permanent contracts with flexibility at the margin has a relevant incidence on the average separation rate () as well as on the job finding probability . We measure this incidence through the correlation between these variables and the business cycle. The answer is provided in Table 7 where, again, we distinguish three stylized cases: (i) a pure deregulated market, where = = 0 (corresponding to the first row of panel 2); (ii) a regulated market with no temporary contracts, where = 0 and = 1 (fifth row of panel 1); and (iii) a regulated market with restricted flexibility at the margin, where = 09, = 05, = 0 and = 1 (benchmark case).
22
Table 7. Simulated correlations and the business cycle.
( ) ( )
Deregulated market
Regulated market with no temporality
Regulated market with restrictions in temporality
=0 = 000
=0 = 100
= 090 = 05 = 100 = 000
0.854 -0.999
0.942 0.163
0.896. -0.675
Note: As before, we detrend the generated data using the HP filter with the smoothing parameter equal to 1,600.
It is interesting to observe that in the first case, which we associate with the AngloSaxon type labor market, there is an almost perfect negative correlation between the job separation rate and the business cycle (-0.99), which is somewhat higher than that observed in the US data (see Hornstein et al., 2005). Further, also noteworthy is the acyclical relationship (with a correlation of 0.16) obtained in a regulated market with no flexibility at the margin. This is traditionally associated with some continental European labor markets, as in Garibaldi (1998). The added value of this exercise, however, lies in the third case, where the use of fixed-term contracts is restricted. We associate this case with the flexibility-at-the-margin type of labor market defined in the introduction. As shown in Table 7, the large negative correlation between the job destruction rate and the business cycle (-0.68) resembles very much that of the pure deregulated labor market (see Sala and Silva, 2009, for the Spanish case).
5
Conclusions
Flexibility at the margin is achieved in segmented labor markets with high protection of permanent workers and loose regulation of fixed-term employment. Does this type of flexibility warrant the same volatility as fully flexible labor markets? Does it explain the recent change witnessed in the OECD area in the correlation between employment protection legislation (EPL) and unemployment? In this paper we have explored whether, and to what extent, flexibility at the margin is helpful in answering these questions. We have developed a matching model with heterogenous workers (regular and fixed-term employees) and focused our analysis on a twofold dimension of segmented labor markets. First, on the effects that the gap in firing costs among these two type of workers has on the volatility of the labor market. Second, on the additional effects that arise from restricting the creation and duration of fixed-term contracts. This is the first paper that studies the business cycle behavior dealing simultaneously with several of these limitations in the form of hiring restrictions (not all vacancies can be matched with fixed-term contracts), renewal restrictions (not all fixed-term contracts can be renewed, some have to be converted into 23
permanent), and firing restrictions (firing costs affect fixed-term contracts which terminate before their expiring date). This model is calibrated to a stylized economy representative of a European labor market with flexibility at the margin and limitations in the use of temporary employment. This benchmark economy is then used to examine the consequences of EPL changes affecting the main parameters governing the firm’s hiring and firing processes. We find that the gaps in separation costs between temporary and permanent jobs, on the one hand, and the restricted use of fixed-term contracts, on the other hand, increase the unemployment volatility with respect to a situation of strict EPL and no gap in firing costs. Another important result is the almost perfect negative correlation we find between job destruction and the business cycle both in the Anglo-Saxon and the flexibility at the margin labor market types. This result clarifies the analogous finding for Spain in Messina and Vallanti (2007), which could not be confirmed for the rest of the countries in the context of their analysis. Our paper, in fact, provides the rationale for such a finding. Finally, and somewhat in contrast to our hypothesis, our simulated scenarios clearly show that a labor market with limited flexibility in the use of temporary contracts cannot achieve the same unemployment volatility as a fully flexible labor market. In short, on its own our model is not able to explain why some OECD countries with limited flexibility in the use of temporary contracts have become as volatile as the fully flexible ones. This leads us to the conclusion that this is not the whole story and other forces may be at work. For example, as is by now well documented, flexibility at the margin not only produces a gap in separation costs between temporary and permanent workers but also leads to a productivity gap, due to high turnover and lack of on-thejob training of temporary employees (see, among others, OECD. 2002; Aguirregabiria and Alonso-Borrego, 2004; and Albert, García-Serrano and Hernanz, 2005). As suggested by Sala and Silva (2009), this productivity gap makes temporary workers more vulnerable to productivity shocks, thereby increasing the volatility of unemployment. This channel constitutes the next step in our objective to explain why some OECD countries with limited flexibility in the use of temporary contracts have become as volatile as the fully flexible ones.
References [1] Addison, J.T., M. Centeno, and P. Portugal (2004): “Key elasticities in job search theory: international evidence”, IZA Discussion Paper Series No. 1314, IZA, Bonn.
[2] Aguirregabiria, V. and C. Alonso-Borrego (2004): “Labor contracts and flexibility: evidence from a labor market reform in Spain”, mimeo, Boston University and Universidad Carlos III de Madrid.
24
[3] Albert, C., C. García-Serrano, and V. Hernanz (2005): “Firm-provided training and temporary contracts”, Spanish Economic Review, vol. 7, pp. 67-88.
[4] Azariadis, C. and C.A. Pissarides (2007): “Unemployment dynamics with international capital mobility”, European Economic Review, vol. 51, 1, pp. 27-48.
[5] Blanchard, O.J. and A. Landier (2000): “The Perverse Effects of Partial Labor Market Reform: Fixed Term Contracts in France”, The Economic Journal, vol. 112, 480, pp. F214F244.
[6] Boeri, T. and P. Garibaldi (2007): “Two Tier Reforms of Employment Protection: A Honeymoon Effect?”, The Economic Journal, vol. 117 (521), pp. F357-F385. [7] Burgess, S. and H. Turon (2005): “Worker flows, job flows and unemployment in a matching model”, Department of Economics Discussion Paper 05/572, University of Bristol.
[8] Cabrales, A. and H.A. Hopenhayn (1997), “Labor-market flexibility and aggregate employment volatility”, Carnegie-Rochester Conference Series on Public Policy 46, pp. 189-228.
[9] Cahuc, P. and F. Postel-Vinay (2002): “Temporary jobs, employment protection and labor market performance”, Labour Economics, vol. 9, 1, pp. 63-91.
[10] Costain, J.S. and M. Reiter (2008): “Business cycles, unemployment insurance, and the calibration of matching models”, Journal of Economic Dynamics & Control, vol. 32, pp. 1120-1155.
[11] den Haan, W.J., G. Ramey and J. Watson, (2000): “Job Destruction and Propagation of Shocks”, The American Economic Review, vol. 90, 3, pp. 482-498.
[12] Dolado, J.J., C. García-Serrano and J.F. Jimeno (2002): “Drawing lessons from the boom of temporary jobs in Spain”, The Economic Journal, vol. 112, 480, pp. F270-F295.
[13] Garibaldi, P. (1998): “Job Flow Dynamics and Firing Restrictions”, European Economic Review, vol. 42, pp. 245-275.
[14] Garibaldi, P. and G.L. Violante (2005): “The employment effects of severance payments with wage rigidities”, The Economic Journal, vol. 115, 506, pp. 799-832.
[15] Hagedorn, M. and I. Manovski (2008): The cyclical behavior of equilibrium unemployment and vacancies revisited”. American Economic Review, vol. 98, 4, pp. 1692-1706.
[16] Hornstein, A., P. Krusell, and G.L. Violante (2005): “Unemployment and vacancy fluctuations in the matching model: Inspecting the mechanism”, Federal Reserve Bank of Richmond Economic Quarterly 91 (3), pp. 19-51.
[17] Kugler, A., J.F. Jimeno and V. Hernanz (2002): “Employment consequences of restrictive permanent contracts: evidence from Spanish labor market reforms”, IZA Discussion Paper Series No. 657, IZA, Bonn.
[18] Messina, J. and G. Vallanti (2007): “Job Flow Dynamics and Firing Restrictions: Evidence from Europe”, The Economic Journal, vol. 117, 521, pp. 279-301.
[19] Mortensen, D.T and C.A. Pissarides (1994): “Job Creation and Job Destruction in the Theory of Unemployment”, Review of Economic Studies, vol. 61, pp. 397-415.
[20] Mortensen, D.T. and E. Nagypal (2007): “More on unemployment and vacancy fluctuations”, Review of Economic Dynamics, vol. 10, 3, pp. 327-347.
[21] Nickell, S. and L. Nunziata (2001): “Labour market institutions database”, mimeo, Centre of Economic Performance, London School of Economics, London.
25
[22] OECD (2002): “Taking the measure of temporary employment”, OECD Employment Outlook 2002, Paris, pp. 127-185.
[23] OECD (2004): “Employment Protection Regulation and Labour Market Performance”, OECD Employment Outlook 2004, Paris, pp. 61-123.
[24] OECD (2006): “Employment and Labour Market Statistics”, vol. 2006, release 01. [25] Osuna, V. (2005): “The Effects of Reducing Firing Costs in Spain: A Lost Opportunity?”, Contributions to Macroeconomics, vol. 5, 1, Article 5.
[26] Petrongolo, B., and C. Pissarides (2001): “Looking into the Black Box: A survey of the Matching function”, Journal of Economic Literature, vol. 39, 2, pp 390-431.
[27] Petrongolo, B. and C. Pissarides (2008): “The Ins and Outs of European Unemployment ”, The American Economic Review, vol. 98, 2, pp 256-62.
[28] Sala, H. and J.I. Silva (2009): “Flexibility at the margin and labour market volatility: The case of Spain”, Investigaciones Económicas, vol. 33, forthcoming.
[29] Shimer, R. (2005): “The Cyclical Behavior of Equilibrium Unemployment and Vacancies”, The American Economic Review, vol. 95, 1, pp. 25-49.
[30] Silva, J.I. and M. Toledo (2009): “Labor Turnover Costs and the Cyclical Behavior of Vacancies and Unemployment”, Macroeconomic Dynamics, forthcoming.
[31] Stock, J.H. and M.W. Watson (2005): “Understanding changes in international business cycle dynamics”, Journal of the European Economic Association, vol. 3, 5, pp. 968-1006.
[32] Thomas, C. (2006): “Firing costs, labor market search and the business cycle”, mimeo. [33] Trigari, A. (2005): “Equilibrium unemployment, job flows and inflation dynamics”, Bocconi University Working Paper 268, Milano.
[34] Veracierto, M. (2008): “Firing costs and business cycle fluctuations”, International Economic Review, vol. 49, 1, pp. 1-39.
[35] Walsh, C. (2005): “Labor market search, sticky prices, and interest rate policies”, Review of Economics Dynamics, vol. 8, pp. 829-849.
[36] Wasmer, E. (1999): “Competition for jobs in a growing economy and the emergence of dualism in employment”, The Economic Journal, vol. 109, 457, pp. 349— 371.
[37] Zanetti, F. (2007): “Labor market institutions and aggregate fluctuations in a search and matching model”, Bank of England Working Paper 333.
26
