ارزیابی کمی از حقوق و دستمزد یارانه های مالیات برای کارگران کم درآمد: روش جستجو تعادل

Phelps [Phelps, E. (1994): “Low-wage employment subsidies versus the welfare state”, American Economic Review 84, 54–58.] presented the case for a low-wage subsidy policy. Since the mid-1990s, France has experimented with this strategy. This paper evaluates the effect of this policy on employment and also on output and welfare. We construct an equilibrium search model incorporating wage posting and specific human capital investment, where unemployment and the distribution of both wages and productivity are endogenous. We estimate this model using French data. Numerical simulations show that the prevailing minimum wage allows a high production level to be reached by increasing training investment, even though the optimal minimum wage is lower. We show that payroll tax subsidies enhance welfare more than a reduction in the minimum wage when they are spread over a large range of wages in order to avoid specialization in low productivity jobs.

High labor costs are typically considered as the primary cause for high unemployment levels in continental European countries (see Blanchard and Wolfers, 2000) where the welfare state has put in place high payroll taxes and minimum wages. Despite a lack of formal evidence evaluating this claim (Katz, 1996), Phelps (1994) presented a case for a low-wage employment subsidy policy as a means to reduce the unemployment of such workers. Phelps (1994) proposed “a system of low-wage employment subsidies be introduced, a subsidy to every qualifying firm based on the stock of low-wage workers on its roll” (p. 56, Phelps, 1994). This policy differs from most of the prevailing subsidy policies2: for instance, the Targeted Jobs Tax Credit in effect in the US during all of the 1979 to 1994 period concerned hirings, as in most developed countries. However, France has implemented the original strategy, which consists of a high minimum wage3 compensated for by large, permanent payroll tax subsidies on low wage employment. It must be noted that the UK, the Netherlands and Belgium, have also experimented with such permanent subsidies for disadvantaged workers, but to a lesser extent than in France. In this paper, we use the French experience to evaluate the performance of the low-wage subsidy policy suggested by Phelps (1994). Relative to a minimum wage reduction policy, a wage subsidy implies a budgetary cost, but it may preserve the welfare of the low-wage workers. On the other hand, this labor cost-reducing policy may be implemented with two different options for a given budgetary cost: either it concentrates subsidies at the minimum wage or it spreads them over a large range of wages. The former clearly aims at dampening the negative effect on employment of the minimum wage, but risks introducing severe distortions. For instance, Katz (1996) emphasizes the risk of stigmatization against a narrowly targeted population of workers which could explain the low employment impact of subsidy policies. In this paper, as firms are likely to respond to wage subsidies by increasing their utilization of workers in the targeted population, we focus on the risk of distortion in job allocations and its implication in terms of productivity. While several econometric papers have already highlighted the positive impact of this policy on employment in France (for example, Kramarz and Philippon, 2001 and Crépon and Desplatz, 2002), Malinvaud (1998), however, stresses a potential negative impact on productivity due to a bias in job creation at the bottom of the wage distribution. When the wage distribution is strongly interrelated with the productivity distribution, payroll tax subsidies could shrink productivity, which in turn could dampen output. Fig. 1 shows a higher concentration at the bottom of the wage distribution of manual workers after the setting up in 1995 of the subsidy policy.4Crépon and Desplatz (2002) also provide some empirical evidences on the decrease of productivity. In this paper, we propose a wage posting model with specific human capital investments and a bilateral endogenous search including on-the-job search, in the line of Mortensen (2000). This framework generates wage and productivity distributions and an equilibrium unemployment rate that are consistent with the data.5 In our framework, the expected job duration, which depends on the degree of labor market tightness, determines the extent to which firms invest in specific human capital. Beyond a pure distribution effect due to the policy bias towards the low-paid workers, productivity could also be decreased by less investment in specific human capital from firms because lower labor costs reduce the expected job duration. Moreover, allowing heterogeneity in the search intensity of employees and unemployed workers creates an interplay between labor supply and labor demand which we view as necessary to identify the overall impact of a decrease in labor costs. Searcher population is composed of employed workers, short-term unemployed workers (with a higher search intensity than that of employed workers) and long-term unemployed workers (with a lower search intensity than that of employed workers). The average search efficiency thus depends on the degree of labor market tightness, which is in turn influenced by the former. Contrary to Flinn and Mabli (2006) who assume only one type of unemployed workers with high search efficiency, the average job-search efficiency is not necessarily improved when employment decreases. We therefore propose a structural model of the French low-skilled workers labor market that enables us to evaluate quantitatively the employment and productivity effects of the French labor cost reducing policy. This structural strategy differs from recent econometric exercises (see for instance Kramarz and Philippon, 2001 or Crépon and Desplatz, 2002). It allows us to conduct counterfactual policy experiments, and, in particular, study the optimal range of wage levels that should be covered by such subsidies. In contrast to the equilibrium search literature which seeks to match the existing stationary wage distribution (Ridder and Van den Berg, 1998, Bontemps et al., 1999 and Postel-Vinay and Robin, 2002), we propose an equilibrium search model that is also suitable for policy analysis. Despite its empirical performance, the equilibrium search literature has been rarely used to conduct policy evaluations. From an empirical point of view, Ridder and Van den Berg (1998), Bontemps et al. (1999), or Postel-Vinay and Robin (2002) show that it is necessary to add an exogenous productivity distribution in order to match the observed data. With exogenous heterogeneity in productivity, a decrease in the minimum wage makes additional productivity segments profitable and increases the number of firms in the economy (Ridder and Van den Berg, 1998). Due to composition effects, labor cost changes influence average productivity. However, in models with exogenous productivity, the productivity level in any firm is given, which limits the effect on average productivity. On the other hand, the labor demand side in this literature is not well-suited to policy analysis as the firms' hirings do not depend on labor costs. Flinn (2006) extends the equilibrium search approach in this direction by considering a stochastic matching model with Nash bargaining. However, Flinn (2006) assumes an exogenous productivity distribution and does not allow on-the-job search, thus restricting search to unemployed workers along the lines of Albrecht and Axell (1984) and Eckstein and Wolpin (1990). Our modeling strategy relies on three key assumptions: the wage posting hypothesis, the absence of counter-offers and the fact that productivity is governed by specific human capital investments.6Manning (2003) argues that bargaining cost considerations make the wage setting assumption not restrictive for anonymous markets with low-skilled workers. While Shimer (2006) points out that the equilibrium wage distribution is no longer unique in wage bargaining models with on-the-job search, Cahuc et al. (2006) show that the introduction of renegotiation circumvents the multiple equilibria outcome. Moreover, they estimate that the bargaining power of the unskilled workers is small: the wage posting assumption cannot be rejected. Our assumption regarding the absence of counter-offers7 is motivated by Postel-Vinay and Robin (2004) who argue that it is optimal for low productive firms not to commit to matching outside wage offers.8 Concerning the third assumption, the specific human capital, Postel-Vinay and Robin (2002) show that the productivity differential across firms explains about half of the French low-skilled wage variance. The remaining part is due entirely to search friction, leaving no room for individual fixed effects. We interpret this as the absence of a significant general human capital effect for the low-skilled workers.9 We also exclude physical capital from the analysis: Robin and Roux (2002) show that the introduction of physical capital in a model à la Burdett and Mortensen (1998) does not help match the observed French wage distribution. Hence, we assume that the firms' heterogeneity observed in the data comes only from their various investments in specific human capital. A key implication of our theoretical framework is that the wage distribution will not display a point mass at the minimum wage. As can be seen from Fig. 1, this implication is validated on French data if one considers only full-time workers (see also Postel-Vinay and Robin, 2002).10 In contrast, part-time workers exhibit a point mass at the minimum wage (DARES, 2006). While such workers account for 40% of the density at the minimum wage, taking into account these part time jobs would have introduced too many complexities and heterogeneities into our theoretical framework. Moreover, part-time jobs in France benefit from specific policy arrangements, especially concerning the subsidy policy in question. We therefore prefer to focus on a consistent worker population. We estimate key parameters of the model on French data using an indirect inference method. Based on statistical tests, we cannot reject the hypothesis that the theoretical wage distribution is generated by the same law as the observed one. In particular, because the productivity distribution plays a central role in the replication of the observed wage density, it provides a powerful identification strategy to estimate the elasticity of productivity relative to human capital investment. We show that this model is able to better replicate the deformation of the wage distribution following the low-wage subsidy policy than a traditional equilibrium search model with exogenous productivity. Thus, while search models based on exogenous productivity differences can replicate existing wage distribution, they are unable to explain variations in such distributions owing to policy changes. This is a key result which strongly supports the need to adopt a framework with endogenous human capital investments when considering the impact of labor cost-reducing policies. Taking into account a costly human capital investment creates a potential trade-off between employment and productivity. The investment channel competes with the employment channel to determine the output effect of the labor cost-reduction policies. As a benchmark case, we investigate the various implications of a minimum wage change on output. The optimal level for a minimum wage seems to be slightly lower (by 12%) than the existing one: a decrease in the minimum wage leads to an employment boost, partially compensated by a decline in labor productivity. The investment channel explains why the optimal minimum wage is relatively high which contrasts with the result of Ridder and Van den Berg (1998) and to a lesser extent of Flinn (2006). We then evaluate the performance of the subsidy policy. Given that the goal of the payroll tax subsidies is to reduce labor costs (without removing the minimum wage legislation), we show that this policy leads to an employment boost which is offset in part by a deterioration in productivity. We show that the current implementation of this policy in France is close to be optimal owing to the fact that subsidies are allocated over a broad range of wages and not only at the minimum wage level. Removing the investment channel from the analysis leads to an opposite recommendation, namely to concentrate exemptions at the minimum wage level. Our results thus imply that the French policy of labor subsidies provides an efficient way of conciliating a higher employment level with maintaining the welfare of the low-wage workers. This paper is organized in three sections. Section 1 is devoted to the presentation of the theoretical model. Section 2 presents the calibration and empirical performances of the model. Quantitative results for different policy experiments are discussed in Section 3.