This Selected Issues paper attempts to uncover the long-term determinants of the demand for foreign exchange reserves in Tunisia. It assesses the adequacy of current and projected reserves holdings in light of the country’s policy choices. The paper describes recent trends in foreign exchange reserves in Tunisia. Econometric evidence on the determinants of the demand for foreign reserves in Tunisia is presented. The results are used to forecast the desired level of reserves given Tunisia’s medium-term macroeconomic framework and to draw policy implications.

Abstract

This Selected Issues paper attempts to uncover the long-term determinants of the demand for foreign exchange reserves in Tunisia. It assesses the adequacy of current and projected reserves holdings in light of the country’s policy choices. The paper describes recent trends in foreign exchange reserves in Tunisia. Econometric evidence on the determinants of the demand for foreign reserves in Tunisia is presented. The results are used to forecast the desired level of reserves given Tunisia’s medium-term macroeconomic framework and to draw policy implications.

III. How Does Employment Protection Legislation Affect Unemployment in Tunisia? A Search Equilibrium Approach20

A. Introduction

31. Reducing unemployment is at the heart of Tunisia’s economic challenges. In the face of major policy reforms required for Tunisia’s economic liberalization and integration into the world economy, employment generation and unemployment reduction are major concerns of the Tunisian government.

32. The government is well aware that promoting employment and facilitating the reallocation of labor in response to structural changes in the economy require simultaneously sound macroeconomic policies and flexible labor markets. The literature and other countries’ experiences show that high unemployment rates are not only associated with slow economic activity, but can also reflect strict labor market regulations that hamper growth and impede labor market adjustment to shocks. Bentolila and Bertola (1990) and Bertola (1990) argue that high firing costs in Europe and differences in employment protection legislation between countries may explain differences in the dynamics of employment. Similarly, Garibaldi (1998) shows that different dynamics in the US and European labor markets are partly due to the existence of employment protection legislation in Europe. Therefore, tackling unemployment requires action on two fronts: improving economic activity and increasing labor market flexibility. In addition to its direct impact on employment dynamics, a flexible labor market contributes to job creation through its positive effect on growth.

33. Tunisia’s labor regulations are complex and protective, hindering labor market flexibility by not allowing firms to fully adjust employment in response to changing economic circumstances. Although the authorities have recently introduced temporary employment contracts by increasing flexibility on hiring, job termination regulations remain rigid and protective: dismissals for economic reasons are heavily regulated, and there is strong government interference.21 So far, the literature has paid little attention to the relation between employment protection regulation and unemployment in Tunisia, focusing instead on the role of economic growth in promoting employment in Tunisia.

34. Growth has only partially contributed to employment generation in Tunisia. Despite a relatively strong economic performance and progress in structural reforms, unemployment remains above 14 percent (2003) of the labor force. While average economic growth was about 5 percent per year over the last decade, the unemployment rate was down by only 2 percentage points over the period. Moreover, 1 percentage point of this fall occurred in the last two years after the authorities increased labor market flexibility by introducing temporary employment contracts. Therefore, high growth rates have only partially contributed to generate additional employment in Tunisia.22 The insufficient flexibility of the labor market appears a good candidate to explain the low elasticity of employment to growth.

35. Against this background, the purpose of this paper is to examine the role played by employment protection legislation (EPL) in keeping the unemployment rate high in Tunisia. The paper uses a search equilibrium model with firing restrictions following Garibaldi (1998). We calibrate the model for Tunisia (i.e. defining the rate at which vacant jobs and unemployed workers meet in Tunisia) by estimating a matching function for the period 1974-2003. Simulating the calibrated model provides us with indications on how the existence of firing restrictions affects the outcome of the matching process and the natural rate of unemployment.

36. The paper concludes that increasing labor market flexibility will not solve Tunisia’s unemployment problem per-se, but would have a favorable impact on steady state unemployment. The simulations show that the removal of firing restrictions could have a limited impact on reducing unemployment. Furthermore, Tunisia’s increased integration into the world economy could translate into a higher volatility of the shocks to which the labor market needs to adjust. The simulations with higher variance of shocks that could result from increased opening show that in the absence of firing restrictions the labor market would be more resilient to adverse shocks. Therefore, Tunisia would benefit from considering alternative approaches that rely less on protecting workers within the firm. However, to generate a substantial reduction in the unemployment rate, lower employment protection needs to be accompanied by continued efforts to implement sound economic policies aiming at enhancing Tunisia’s growth performance and job creation.

37. The remainder of this paper is organized as follows: Section II reviews the literature on the impact of employment protection regulation on unemployment. Section III provides a background on the institutional settings of Tunisia’s labor market. Section IV presents the broad features of a search matching model with firing restrictions, which is explained in detail in the appendix. Section V analyzes the impact of firing restrictions on current unemployment in Tunisia by calibrating and simulating the matching model. It also assesses the impact of these restrictions as Tunisia increases its integration into the world economy. Section VI presents the main conclusions.

B. Review of the Literature

38. Employment protection legislation is a form of employment regulation which relates to employers’ freedom to dismiss workers. According to OECD (1994), employers’ freedom to dismiss workers may be restricted in several ways: penalties on unfair dismissals, restrictions on lay-offs for economic reasons, compulsory severance payments, minimum notice periods, written justifications, administrative authorizations, etc. The effect of this legislation is similar to a tax on dismissals, even though the firm may not always be required to pay money before it dismisses an employee.

39. Much attention has been devoted to the analysis of the consequences of EPL in industrial countries. EPLs inhibit labor market flexibility by reducing the ability of firms to hire or fire workers. The perception that flexible labor markets promote employment and reduce unemployment is widely accepted. Yet, the theoretical and empirical evidence on the net effects of firing restrictions on employment and unemployment are ambiguous. Nevertheless, it has often been suggested that the elevated severance pay and job security requirements in Europe are in part to blame for the high unemployment levels in this continent (Kugler and Pica, 2004).

40. The theory of job creation and job destruction implies that EPL reduces job destruction, but it also reduces job creation (Millard and Mortensen, 1997; Millard, 1996; Nickell, 1982). The reason for the reduction in job destruction is obvious enough. EPL is a firing cost (direct or indirect) imposed on the employer, so job destruction, which necessitates firing the employee, becomes more expensive and difficult. Job creation falls for two reasons: (i) since dismissal is costlier (direct cost or lengthier process), the firm creates a job and recruits an employee only if it expects to need the employee for a longer length of time. Equivalently, jobs that are not expected to have a long life are not created when there are strict job destruction rules (Pissarides, 1999); and (ii) with less job destruction, and less flow into unemployment, there are both fewer job seekers and fewer vacancies, which implies that fewer job matches take place.

41. Theoretical work on the effects of firing costs shows that while reductions (increases) in firing costs are expected to increase (reduce) hiring and firing as well as employment volatility, the net effects of reducing firing costs on employment are ambiguous. While a recent study (Elmeskov, Martin, and Scarpetta, 1998) suggests a somewhat more robust effect on unemployment if changes in OECD EPL over the past two decades are taken into account, OECD (1999) could not find a statistically significant effect of EPL on aggregate employment. Most of the models in the literature provide the same type of prediction: stricter employment protection has an ambiguous impact on the level of overall employment, because it reduces both job creation and destruction. Nevertheless, empirical regularities were found in other areas, mainly: EPL reduces both job destruction (unemployment inflows) and creation (unemployment outflows). EPL has important effects on the composition of employment, since countries with stricter EPL are associated with higher youth unemployment and larger self employment.

42. The multiple dimensions of employment protections are difficult to model in a simple way. The simplest and most widely modeled form of EPL is a fixed firing cost to be incurred by the firm when firing takes place (Bentolila and Bertola (1990) and Bentolila and Saint-Paul (1994) in partial equilibrium models of labor demand, Burda (1992), Millard (1994) and Millard and Mortensen (1994) in search equilibrium models). Another form of job security provision consists of the existence of firing permissions that cannot be quantified such as fixed firing costs (Garibaldi (1998) in search equilibrium model). The way to capture the effects of procedural constraints in an aggregate model is to assume that a firm can accomplish firing only when it is granted an exogenous firing permission.

43. This paper uses a search equilibrium approach with firing restrictions, which has the advantage of being dynamic and able to analyze the impact of legal and not cost-related restrictions on labor market dynamics. By being dynamic, this approach can model both the unemployment stock and its duration. It is also known as the flow approach to the modeling of unemployment, because unemployment flows play a key role in the modeling.23 The natural rate of unemployment equates the flow into unemployment with the flow out of unemployment. This modeling strategy enables the study of questions related to job search and job turnover, which often provide the key link between policy and unemployment. The model explicitly includes legal restrictions on dismissals, thus it describes well the labor market in Tunisia, where severance payments are relatively low but legal restrictions are high (see below).

C. The Tunisian Labor Market: Institutional Settings

44. Tunisia has significant labor market regulations, including employment protection. Government intervention in the labor market has been traditionally substantial. The social dialogue too, operating through a tripartite mechanism, has been important in shaping a wide range of labor market outcomes, including labor reallocation and wage setting. The logic has been to limit the shocks of economic change through strong job security rules.

45. Hiring rules are flexible but termination regulations are rigid and protective. Labor regulation reforms in 1994 and 1996 introduced flexibility in hiring through fixed-term contracts and part-time work.24 However, termination regulations are substantial: dismissals for economic reasons are still heavily regulated, and there is strong government interference. Moreover, reforms to the Labor Code in 1994 and 1996 accelerated the administrative procedures and clarified the definition of “licenciement abusif”.25 In the end, however, firms wanting to adjust their workforces for economic or technological reasons still must engage in a heavily bureaucratic process where government and a tripartite mechanism have substantial powers to intervene. Moreover, firms wanting to downsize must first notify the Inspection du Travail (IT) in writing, with at least one month advance notice, indicating the reasons and listing the workers to be affected. The Inspection du Travail then has 15 days to review the request. If this proposal is not accepted by the employer, then within 3 days the case goes to the central or regional Commission du contrôle des licenciements (CCL), which has 15 days to decide on the downsizing application and, where layoffs are involved, on the severance payments owed to the workers.

46. As a result, temporary jobs have increased and small firms often found solutions outside the legal framework to circumvent those regulations. In 2001, 13 percent of the labor force had a nonpermanent contract as employers seeking flexibility found it more attractive to offer nonpermanent positions and to avoid the obligations that still exist regarding layoff rights. With an employed labor force of over 2.5 million during 1998-2001, the total number of proposed layoffs represented less than 1 percent of total employment, compared to 10 percent in OECD (OECD 1994).26 Furthermore, actual retrenchments decreased as temporary unemployment and part-time work became much more important adjustment mechanisms: in 2001, only 14 percent of all proposed layoffs wound up as retrenchments, compared to 30 percent in 1998. Finally, cases going to the CCL (33 percent in 2001 compared to 60 percent in 1998) declined, suggesting that solutions are being found outside the CCL.

47. Severance payments, although not high in general, can be generous in specific cases. In the case of retrenchment, minimum severance requirements are established in the Labor Code. Conventions collectives can set levels above these established rates. The base formula is one day of pay for each month of service, with a 3-month maximum. This level is not excessive by international standards; however, severance requirements increase a great deal—and are high by international standards—in the case of licenciements abusifs. According to the 1994 Labor Code revisions, layoffs are considered abusif when (i) there is no just or serious cause, or (ii) the legal procedures, rules, and conventions have not been respected. If the CCL determines that there was no just or serious cause, the severance guideline is 1-2 months of salary per year of service, with a maximum payment of 3 years’ salary.27 While according to international experience, in cases where only the procedures have been violated, the severance payment should be 1–4 months of salary.

48. Overall, rules to protect job security—mainly Tunisia’s retrenchment procedures—create a duality in the labor market. They increase the stability of existing jobs, though at a price: more long-term unemployment and nonparticipation in the labor force and less opportunity for regular employment in the formal sector.

D. The Model

49. We use the matching framework developed by Garibaldi (1998).28 It is a model à la Mortensen and Pissarides (MP) with firing constraints. The features of the model are as follows:

50. The model considers an economy populated by a continuum of risk-neutral workers of fixed quantity. Workers can be in two states, employed or unemployed. Each firm has one job that can be either filled or vacant. A filled job can be either fully operational or idle, depending on whether the firm is actually waiting for firing permissions. Firms with a vacant position search for filling it. Job creation takes place when a firm with a vacant job and a worker meet. Job destruction takes place when a filled job gets a firing permission, separates and leaves the market.

51. The rate at which vacant jobs and unemployed workers meet is determined by the simple matching function m(v,u), where m is a first degree homogeneous matching function and v and u represent the number of vacancies and unemployed workers respectively, normalized by the fixed labor force size. Vacancies are filled at the rate

q(θ)=m(v,u)/v=m(1,u/v); θ=v/u and q’(θ)<0

where θ is an index of market tightness from the firm’s point of view. The smaller the number of vacancies in relation to the number of unemployed workers, the easier for the firm is to fill vacant jobs. The rate at which workers find jobs is

γ(θ) = m(v,u)/u = m(v/u,1) = θq(θ); γ’(θ)>0

thus, the larger the number of vacancies in relation to the number of unemployed workers, the easier for the worker is to find a vacant job.

52. Job creation is defined by the number of matches

m(v,u)=vq(θ).

53. Each job is characterized by a fixed irreversible technology and produces a unit of a differentiated product whose productivity is p + σε. The productivity is made up of an aggregate component P, common to every job, and a job specific component ε which differs across jobs.29 The process that changes the idiosyncratic component of prices ε follows Poisson distribution with arrival rate equal to λ When there is a change in the new value of the job specific productivity ε is a drawing from the fixed distribution F(ε), which has finite upper support εu, lower support ε1 and no mass point other than at the upper support εu. This way of modeling implies a memoryless but persistent idiosyncratic productivity. The persistence of any given productivity ε is 1/λ.

54. Filled jobs are said to be fully operative if the idiosyncratic productivity is above some critical value εd, while they are said to be idle if the job specific productivity is below εd. Therefore, the rate at which jobs turn idle is λF(εd), while workers in idle jobs can be dismissed and leave the market at a rate s. The parameter s summarizes EPL in the model: as s →∞, EPL are eliminated. Finally, idle jobs are subject to idiosyncratic uncertainty and can return fully operational at rate λ(1-F(ed)).

55. The model departs from the standard Mortensen-Pissarides (1994) framework in the wage-setting behavior. It assumes that employers capture all the rents associated with a job-worker match by paying workers the common alternative value of their time, b.30 It is well known that in search equilibrium models, wages clear the market since there is no supply and demand to equate. The matching process generates economic rents that need to be shared between the employer and employee according to some exogenous bargaining rule. While often the literature assumes a Nash-symmetric solution for the wage bargaining game, in this model we follow Fanizza (1996) and Garibaldi (1998) assuming that all the rents generated by a match accrue to the employer. This rule can be interpreted as the one which maximizes the flow out of unemployment.

56. The unknowns of the model are the number of job vacancies v and unemployment u, which determine, through the matching technology, job creation, and the critical value for the idiosyncratic component of productivity, εd, that induces idle job. Steady state unemployment is defined as the level at which the flow in and out unemployment are equal.

57. The solution of the model in steady state results in the following:

58. If dismissal of idle workers is unrestricted (s →∞), the idle rate tends to zero and equilibrium unemployment coincides with equilibrium unemployment in more standard matching models (Mortensen & Pissarides, 1994; Pissarides, 1990). As the average waiting time increases, EPL affect, both job creation and job destruction decisions, and have an ambiguous impact on unemployment.

59. On the one hand, stricter EPL has a favorable impact on unemployment since it reduces the flow out of employment by obliging firms to keep workers occupied in idle jobs. On the other hand, stricter EPL reduces the rate at which workers escape unemployment q(θ)) by directly increasing the number of vacant jobs and indirectly by increasing the expected returns from employing a worker. However, the final result of changes in EPL on the steady state unemployment rate will depend upon the values at the parameters of the model, in particular α and λ.

E. Simulation Results

60. To implement the general stochastic model of the previous sections, we first specify the matching elasticity α by estimating a log-linear Cobb-Douglas matching function m(v,u) = uα vβ with a time trend for annual data covering the period 1974-2003:

Log mt = c + α log ut + β log vt + δt + εt

61. The results of the estimation show that the parameter α is equal to 0.14 in Tunisia (Table 1). Compared to coefficient of the matching function in industrial countries (0.25), the matching process appears to be relatively inefficient in Tunisia.

Table 1.

The Matching Function in Tunisia (1974–2003)

article image

62. Calibrating the model for the Tunisian market results in λ=0.05. The real interest rate is set at 0.03. The other parameters values are similar to Mortensen and Pissarides (1994) and to Garibaldi (1996). They are summarized in Table 2.

Table 2.

Baseline parameter value

article image

63. The results of the simulation are summarized in Table 3. The average waiting time being 1/s, a high s is equivalent to low firing restrictions. The results show that easing the current firing restrictions (s=0.2) will reduce the equilibrium unemployment by about 1 percentage point and increase job creation (JC) by 15 percent. Furthermore, firing restrictions affect labor market dynamics through their effect on the relative volatility of job creation and job destruction. Thus, the relative variance of job destruction to job creation σ2 JD/σ2 JC increases dramatically as firing restrictions weaken. Other statistics of Table 4 show that as EPL eases (or as s increases from 0.2 to 1.2), both the average duration and the persistence of unemployment fall. EPLs has obviously a strong effect on average idle capacity, which is the average fraction of jobs waiting for dismissed permission.

Table 3.

Simulation statistics

α=0.14, λ=0.053

article image
Table 4.

Simulation Statistics, Higher Frequency and Variance of Idiosyncratic Shocks

α = 0.14, λ = 0.056

article image

64. Tunisia’s increased integration into the world economy is likely to increase the frequency and variance of idiosyncratic shocks to the labor market (i.e. higher values of λ). However, increased integration will push up job creation. The results of the simulation with higher values of λ show that more frequent and stronger shocks will have an adverse effect on average unemployment in Tunisia (Table 4). However, this effect will be more pronounced if Tunisia maintains its current firing restrictions than if it eases them. Thus, when λ increases to 0.056, unemployment could increase up to 17.7 percent if firing restrictions are in place, while it might go up to only 16.9 percent if restrictions are abolished. EPL will raise job creation (JC) slightly more than under the assumption of a lower λ Furthermore, as Tunisia further opens to the world economy (higher λ), the duration and persistence of unemployment will also continue to be negatively affected by a stricter EPL.

F. Conclusion

65. Increased labor market flexibility in Tunisia will contribute to reducing unemployment. The removal of firing restrictions is likely to produce a positive but limited impact on unemployment of not more than 1 percent. Furthermore, while Tunisia’s increased integration into the world economy could expose the labor market to adverse shocks and increase the rate of unemployment, the latter would increase less in case firing restrictions are removed.

66. Increased labor market above flexibility will not solve Tunisia’s unemployment problem. Given the limited impact of removing firing restrictions on unemployment, increased flexibility cannot solely address the unemployment problem in Tunisia. This result suggests that the existence of other factors that are preventing unemployment to fall rapidly. One of these factors could be the skill mismatch (i.e. gaps between skills in demand by employers and skills offered by job seekers).

67. The findings of this study suggest that a three-pronged approach to improve employment performance is needed:

  • Continue to implement sound economic policies aiming at achieving high growth rates and promoting private activity.

  • Consider introducing greater flexibility in the labor market to facilitate the reallocation of labor in response to structural changes in the economy. As Tunisia becomes more engaged in opening and restructuring its economy, based on international experience, it would be favorable if employment policy moves away from a model where the government played a central role through large public sector employment, and labor market was tightly regulated. In particular, Tunisia would benefit from considering alternative approaches that rely less on protecting workers through within the firm and more on offering opportunities and protection outside the firm.

  • Review the education and vocational training systems to improve the skill matching: These systems should be tailored in a way to equip new labor force entrants with the skills in demand by private employers.

APPENDIX Development of the Model

The rate at which vacant jobs and unemployed workers meet is determined by the simple matching function m(v,u), where m is a first degree homogeneous matching function and v and u represent the number of vacancies and unemployed workers respectively, normalized by the fixed labor force size.

Vacancies are filled at the rate: q(θ)=m(v, u)/v=m(1, u/v); θ=v/u and q’(θ) < 0

If Cobb-Douglas m(v, u) = v1-α uα and q(θ)=(v1-α uα)/v = (u/v) = θ

The rate at which workers find job is: γ (θ) = m(v,u)/u = m(v/u, 1) = θq(θ); γ’(θ) > 0

Job creation is defined by the number of matches: m(v,u)=vq(θ).

Each job is characterized by a fixed irreversible technology and produces a unit of a differentiated product whose productivity is p + σε. The productivity is made up of an aggregate component p, common to every job, and a job specific component ε.31 The parameter σ reflects dispersion, and increases in σ representing a symmetric mean preserving spread in the job-specific shock distribution or equivalently an increase in productivity variance.

The process that changes the idiosyncratic component of prices ε is Poisson with arrival rate equal to λ. When there is a change in ε, the new value of the job specific productivity ε is a drawing from the fixed distribution F(ε), which has finite upper support εu, lower support ε1 and no mass point other than at the upper support εu. This way of modeling implies a memoryless but persistent idiosyncratic productivity. The persistence of any given productivity ε is 1 /λ.

The model assumes that firms have the option to select the best productivity in the market, and create jobs at the upper support p + σεu. Once a job is created, however, the firm has no choice over its productivity. Filled jobs are said to be fully operative if the idiosyncratic productivity is above some critical value εd, while they are said to be idle if the job specific productivity is below εd. Therefore, the rate at which jobs turn idle is λF(εd), while idle jobs get firing permissions and leave the market at rate s. The parameter s summarizes EPL in the model as s →∞ EPLs are eliminated. Finally, idle jobs are subject to idiosyncratic uncertainty and can return fully operational at rate λ(1-F(εd)).

The model assumes that employers capture all the rents associated with a job-worker match by paying workers the common alternative value of their time, b.

The unknowns of the model are the number of job vacancies v and unemployment u, which determine, through the matching technology, job creation, and the critical value for the idiosyncratic component of productivity, εd, that induces idle job.

The asset valuation of a filled job, conditional on an idiosyncratic productivity ε is:

rJ(ε)=p+σεb+λ[εIεuJ(x)dF(x)J(ε)]+s[max{0,J(ε)}J(ε)],(1)

where J(.) is the value of a job, r is the exogenous interest rate, p + σε - b are operational profits at idiosyncratic productivity ε. Apart from the flow-term p + σε - b, Eq. (1) involves two capital gain terms. At rate λ the firm loses its current asset value J(e) and draws a new ε from the productivity distribution. At rate s firing permissions arrive and the firm gets an option to destroy the job. Since a destroyed job has zero value, the max operator in Eq. (1) captures the idea that a firm will keep running a job as long as its value is positive. It follows that an operational job is a positively valued job that ignores firing permissions while an idle job is a negatively valued job that is destroyed when permissions arrive. Differentiating Eq. (1) with respect to e, it shows that J(.) is a piece-wise increasing function of ε and its derivative reads:

J(ɛ)=σ/(r+λ)J(ɛ)0,(2)

and,

J(ɛ)=σ/(r+λ+s)J(ɛ)<0.(3)

If we define the reservation productivity εd as:

J(εd) = 0,

making use of Eqs. (1) and (3), after an integration by parts, the expected value of a job in Eq. (1) reads:

ɛ1ɛuJ(x)dF(x)J(ɛ)]=σ/(r+λ)ɛdɛu(1F(z))dzσ/(r+λ+s)ɛ1ɛdF(z))dz(4)

The last term of Eq.(4) is the value (negative)of an idle job and is a measure of expected firing costs. As the average waiting time goes to zero (s → ∞), the second term on the right hand side of Eq.(4) vanishes, firing is always possible and it is accomplished as soon as the value of the job is negative. To obtain the cut off value εd, below which the firm will accept firing permission, we make use of Eq.(4) and we evaluate Eq.(1) at J(.)=0. The reservation productivity solves:

p+σɛb=σ/(r+λ)ɛdɛd(1F(z))dzσ/(r+λ+s)ɛ1ɛdF(z))dz(5)

Eq.(5) is one of the key equations of the model and uniquely determines the reservation productivity as a function of the parameters r, λ, p, s, b, σ and the productivity distribution F(e). The left hand side of Eq.(5) is the profit from the marginal operational job. In an economy with no firing constraints (s → ∞), the second term of the right hand side vanishes, the marginal profit is negative and there is voluntary labor hoarding in equilibrium. When firing is instantaneous (s → ∞) but hiring is costly, the firm will hoard labor up to the level in which current losses compensate savings of hiring costs if conditions improve. The presence of firing delays increases, through the last term in Eq. (5), the value of the marginal profits. As the average waiting time for firing permissions increase, a job will be kept running in bad times for a longer period of time because of exogenous constraints and there will be institutional labor hoarding. Since the firm anticipates firing restrictions when conditions are bad, in Eq.(5) the firm reduces the extent of voluntary labor hoarding. As s falls, it is possible that firing restrictions become so high that the firm will accept firing permissions at a positive profit per period.

Differentiating Eq.(5) with respect to s and rearranging, yields:

(ɛd/s)[s(r+λF(ɛd))+(r+λ)]/(r+λ)(r+λ+s)=λ/(r+λ+s)2ɛ1ɛdF(z))dz(6)

Thus ∂∊d/∂s ≤ 0: an increase in the average waiting time of permission (fall in s) increases the productivity at which the firm takes advantage of firing permissions. This is consistent with the firm anticipating long waiting time when conditions worsen.

The reservation productivity falls with p, the common productivity. Differentiating Eq.(5) with respect to (p-b) and rearranging, yields:

σ[ɛd/(pb)][s(r+λF(ɛd))+r(r+λ)]/(r+λ)(r+λ+s)=1(7)

Thus ∂∊d/∂p ≤ 0: as the productivity increases the firm will find it profitable to keep a job operational for a higher range of productivities. The effect of other parameters on the reservation productivity is ambiguous. Higher discount rate r reduces the flow of income from the job and makes labor hoarding less profitable. This would reduce εd. But simultaneously, the higher discount rate reduces expected firing costs and makes autonomous labor hoarding profitable. Similar arguments hold for changes in the arrival rate of idiosyncratic shocks. Higher λ corresponds to an increase in the arrival rate of productivity shocks. On the one hand the reservation productivity tends to decrease since the firm expects the duration of adverse conditions to be shorter. At the same time, the probability of facing a firing procedure is higher and the net effect depends mainly on the distribution F(.).

68. Job creation comes through the posting of vacancies. When creating a job, we assume the existing technology is fully flexible and the productivity distribution is common knowledge. This implies that new firms have the option to select the best productivity in the market and job creation takes place at the upper support of the distribution (εu). A posted vacancy yields an asset return of -c per period, c being the constant cost of hiring, and a probability q(θ) of being filled with a job created at the upper support of the distribution. The vacancy asset valuation is:

rV=c+q(θ)[J(ɛu)V].(8)

With free entry into the job market, there are, in equilibrium, zero expected profits (V=0) (Pissarides, 1990) and the value of a job equals the expected searching costs:

J(ɛu)=c/q(θ),(9)

where the value of a job at the upper support of the distribution is obtained subtracting Eq.(5) from Eq.(1) and reads:

J(ɛu)=(ɛuɛd)/(r+λ),(10)

Eq.(9) is the job creation condition and uniquely determines the vacancy unemployment ratio θ as a function of the parameters r, λ, c, the matching function q(.), the upper support of the distribution εu and the reservation productivity εd.

(ɛuɛd)/(r+λ)=c/q(θ),(11)

Differentiating Eq.(11) with respect to common productivity p, yields

[ɛd/p][1/(r+λ)]=[q(θ)c/q(θ)2][θ/p],(12)

and, making use of the facts that ∂∊d/∂p < 0 and q’ (.) < 0, ∂θ/∂p > 0. Higher common productivity, increasing the flow of future profits, increases job creation at given unemployment Conversely, higher job security provisions reduce the expected value of a job and reduce the profitability of new jobs. Job creation at given unemployment falls. Differentiating Eq.(11) with respect to s,

[ɛd/s][1/(r+λ)]=[cq(θ)/q(θ)2][θ/s],(13)

making use of ∂∊d/∂s < 0 Eq.(13) implies that ∂θ/∂s>0.

To close the model, we need to introduce unemployment. With a fixed labor force, a worker can be either unemployed or employed. If employed, a worker can be attached to a fully operational (ε ≥ εd) or to an idle job ε< εd. Normalizing variables in terms of a constant labor force, the relationship among different labor force status is:

u+nj+ni=1,(14)

where u is the unemployment rate, ni is the employed idle capacity, and nj is the employed operational rate. In an interval dt, the outflow rate (job creation) corresponds to the number of matches per unemployed times the number of unemployed, while the inflow rate (job destruction) corresponds to the fraction of workers in the idle state whose employers obtained firing permission.

Δu(t)=sni(t)θq(θ)u(t),(15)

where 6 q(θ) is the job finding rate. Eq. (15) defines unemployment variation as the difference between job destruction and job creation. Simultaneously, there are a number of fully operational jobs that are hit by a shock below the reservation productivity and enter the idle state. The outflow from the idle state corresponds to the idle jobs that have obtained firing permissions plus those idle jobs that, hit by a positive productivity shock, return to be fully operational. The inflow into the idle state is given by the operational jobs hit by a shock below the reservation productivity. The change in the idle rate is:

Δni(t)=λF(ɛd)nj(t)[sλ(1F(ɛd))]ni(t).(16)

In steady state equilibrium, the unemployment rate and the employment composition between idle and operational jobs is constant. From Eqs. (15) and (16) it follows that unemployment and the idle rate are constant if the inflow rate is equal to the outflow rate. Steady state idle rate is:

ni*=[θq(θ)u*]/s,(17)

and steady state equilibrium unemployment is:

u*=λF(ɛd)/[λF(ɛd)+θq(θ)(s+λ)/s].(18)

In steady state, the system is recursive and it reduces down to four equations. Eq.(5) uniquely determines the reservation productivity εd, while Eq.(11), given εd, uniquely determines the vacancy/unemployment ratio θ. Given θ and εd, Eq.(17) and (18) simultaneously determine unemployment and the idle rate. Finally, given the unemployment rate, θ determines vacancies.

References

  • Bentolila, S., and G. Saint-Paul, 1994. “The Macroeconomic Impact of Flexible Labor Contracts, with an Application to Spain”. European Economic Review, Vol.36, No. 5.

    • Search Google Scholar
    • Export Citation
  • Bertola, G., 1990. “Job Security, Employment, and Wages”, European Economic Review, Vol. 54, No. 4.

  • Bertola, G., and S. Bentolila, 1990. “Firing Costs and Labor Demand: How Bad is Eurosclerosis?”. Review of Economic Studies, No. 57.

    • Search Google Scholar
    • Export Citation
  • Burda, M., 1992. “A Note on Firing Costs and Severance Benefits in Equilibrium UnemploymentScandinavian Journal of Economics, No. 94.

    • Search Google Scholar
    • Export Citation
  • Elmeskov, J., J.P. Martin, and S. Scarpetta, 1998. “Key Lessons for Labor Market Reforms: Evidence from OECD Countries’ Experiences”. Swedish Economic Policy Review, Vol. 5, No. 2.

    • Search Google Scholar
    • Export Citation
  • Fanizza, D., 1996. “Employment Cycles in Search Equilibrium”. Journal of Economic Dynamics and Control, No. 20, Elsevier Science.

  • Garibaldi, P. 1998. “Job Flow Dynamics and Firing Restrictions”. European Economic Review, Vol. 42.

  • Kugler, A.D., and G. Pica, 2004. “Effects of Employment Protection and Product Market Regulations on the Italian Labor Market”. Center for Economic Policy Research, Discussion paper No.4216.

    • Search Google Scholar
    • Export Citation
  • Millard, S., 1996. The Effect of Employment Protection Legislation on Labor Market Activity: A Search ApproachBank of England Working Paper.

    • Search Google Scholar
    • Export Citation
  • Millard, S., and D. Mortensen, 1997. “The Unemployment and Welfare Effects of Labor Market Policy: A Comparison of the US and UK”. In Unemployment Policy: Government Options for the Labor Market, Edited by Dennis J. Snower and Guillermo de la Dehesa. Cambridge University Press.

    • Search Google Scholar
    • Export Citation
  • Mortensen, D. and C. Pissarides, 1994. “Job Creation and Job Destruction in the Theory of Unemployment”. Review of Economic Studies, Vol. 61.

    • Search Google Scholar
    • Export Citation
  • Nickell, S. J., 1982. “The determinants of Equilibrium Unemployment in BritainEconomic Journal, No. 92.

  • OECD, 1994. The OECD Job Study, OECD Publications.

  • OECD, 1999. Employment Outlook, OECD Publications.

  • Pissarides, C., 1990. Equilibrium Unemployment Theory. Basil Blackwell, Oxford.

  • Pissarides, C., 1999. “Policy Influences on Unemployment: The European Experience”. Scottish Journal of Political Economy, Vol. 46, No. 4.

    • Search Google Scholar
    • Export Citation
  • The World Bank, 2003. Republic of Tunisia, Employment Strategy. World Bank Publications.

20

The authors of this paper are Taline Koranchelian and Domenico Fanizza.

21

The law does not require tax payments when firing employees, neither severance payments are high by international standards.

22

The World Bank estimated the elasticity of employment to GDP growth to 0.5 during 1994-2001.

24

The main reforms included:

  • Introduction of two categories of fixed-term contracts, determinate and indeterminate period of time. Fixed-term contracts are permitted for a maximum of four years, subject to the agreement of the parties.

  • Introduction of part-time work. Part time is defined as less than 70 percent of the normal hours. It is based on two principles: freedom of choice for employees and equal treatment with full-time employees.

25

These reforms stipulated that the complete process, from initial application for downsizing to a final decision of the Commission du controle des licenciements, should take a maximum of 33 days, unless the parties agree to an extension.

26

This includes job turnover from closures and firm contraction. Data cover the 1980s and early 1990s for 16 OECD countries.

27

In the case of licienciement abusif for fixed-term (CDD) employees, payment should equal the remaining part of the contract.

28

See appendix for the description of the complete model and the characterization of equilibrium.

29

σ reflects dispersion and is common to every job. It is a normalizing parameter useful for the simulations.

30

As Diamond (1971) has shown, this outcome is an equilibrium in a wage setting game played among employers when workers have only the power to accept or reject offers and workers search sequentially at some positive costs. Given this outcome, workers have no incentive to search on the job and their parameters, other than b, do not affect the equilibrium. Alternatively, if we allowed a continuously renegotiated Nash bargain between the firm and the worker, the wage would be higher than the worker reservation utility in operational jobs, where the surplus from the match is positive. But the presence of firing restrictions, would force the firm to pay the worker even when the job is idle and the worker’s participation constraint is binding. This would force idle firms to offer the worker his reservation utility b, exactly as in the present model. Thus, a continuously renegotiated bargain would only affect the wage of operational jobs, leaving unchanged the behavior of idle jobs, the distinctive feature of this model. To keep track of such bargains would be analytically tedious and would not change the qualitative results of the paper.

31

σreflects dispersion and is common to every job. It is a normalizing parameter useful for the simulations.