Do Labor Market Rigidities Fulfill Distributive Objectives?: Searching for the Virtues of the European Model
  • 1 0000000404811396 Monetary Fund

The distributional effects of the minimum wage are analyzed in a model where skilled and unskilled labor are inputs into the production function. It is argued that distributional goals are best achieved by letting the labor market clear itself and achieving redistribution through taxes and transfers. The results stand up to the imposition of the additional constraint of political viability, although skilled workers may be harmed by excessive equilibrium tax rates under the second system. [JEL E24, E6, H21, H23, H55, J3]


The distributional effects of the minimum wage are analyzed in a model where skilled and unskilled labor are inputs into the production function. It is argued that distributional goals are best achieved by letting the labor market clear itself and achieving redistribution through taxes and transfers. The results stand up to the imposition of the additional constraint of political viability, although skilled workers may be harmed by excessive equilibrium tax rates under the second system. [JEL E24, E6, H21, H23, H55, J3]

With unemployment in Europe having remained at over 10 percent for more than 10 years and now sharply rising, the issue of labor market flexibility is back in the political debate. Yet few measures—such as marginally broadening the tax base for social security contributions or introducing determined duration contracts subject to many restrictions— have actually been undertaken. Caution and sluggishness are the hallmarks of most of the labor market reforms on the continent. Behind the paralysis lies the fear that labor market flexibility would require adopting an “American model,” with full employment achieved only at the cost of large inequalities, high crime, and a decomposition of the social fabric. Such an argument is presented in the recent proposals of the Organization for Economic Cooperation and Development (OECD) and the European Union. These proposals essentially focus on reducing nonwage labor costs with considerably less emphasis on more fundamental reforms of labor market regulations and unemployment benefit regimes. They maintain that a desirable feature of the status quo is that the current set of labor market rigidities achieve socially desirable redistributive goals.1

This paper is a critique of this view. The bottom line is that these redistributive goals can be achieved at a much lower cost using more traditional tax and transfer instruments, even taking into account the distortionary impact of these instruments. Only for implausibly high values of these distortions can we find some redistributive virtues to labor market rigidities.

The institution on which we focus is the minimum wage. It is unimportant in some countries—Germany, for example—but pervasive in others, such as France. Section I details the distortions introduced by the minimum wage and argues that, even abstracting from alternative forms of redistribution, an increase in the minimum wage may well have adverse impacts on inequality. This is because, while it redistributes income from skilled to unskilled workers, an increase in the minimum wage also redistributes income from the poorest to the lower-middle class by creating unemployment.

Section II develops a simple model to analyze the impact on output and inequality of two alternative systems. One incorporates a minimum wage together with unemployment benefits that are funded using payroll taxes. The second is a simple system in which there is no minimum wage and redistribution is achieved by taxing labor and giving the proceeds to the unskilled. It is shown that the second system typically achieves a higher level of output than the first, with the same level of inequality. Also, if the labor market is “rigid” (in the sense of lower turnover), then it is more likely that raising the minimum wage actually increases inequality, and the gains from shifting to the “tax” system are larger. These results are supported by the model’s calibration, which is developed in Section III.

Some political economy issues are discussed in Section IV. While it is true that shifting to the tax system is a more efficient way to reach a global inequality target, it may well be that employed unskilled workers suffer from such a shift. If the balance of political power is such that unskilled workers can block a change that is detrimental to them, the reform will not be politically feasible. It is shown that this is more likely to happen when there is low turnover, in which case the tax system is much more egalitarian than the minimum wage system. This is because the employed are so secure in their jobs that they do not take into account the increased welfare of the unemployed when computing their own gains from shifting to the new system. The analysis suggests that “politico-economic complementarities” may arise across labor market rigidities: there will be more political support for the minimum wage if turnover is lower.2 The simulations, however, suggest that when the turnover rate is very low, a moderate rise of the tax rate above the one that achieves the same equality level as the minimum wage system is enough to induce the employed to favor the reform. Therefore, while political viability may have to be taken into account in the design of the new system, the conclusion that the tax/transfer system dominates the minimum wage system is left unaltered.

Section V discusses the effects of other types of labor market rigidities, that is, unemployment benefits and job protection laws. Section VI concludes.

I. The Effect of Minimum Wages

In this section we enumerate various channels through which the minimum wage may have perverse effects. These effects arise either because the minimum wage lowers the efficiency of the economy or because it does not achieve the redistributive role it is assigned.

First, the minimum wage creates unemployment among the unskilled, thus reducing output. While it raises the welfare of those who keep their jobs, it lowers the welfare of those who lose jobs. Whether the minimum wage does in fact reduce inequality depends on the level of unemployment compensation, which in Europe guarantees a replacement ratio from 60 to 80 percent, and on what is meant by inequality. If one takes an Atkinson (1970) measure of inequality as representative of society’s preferences, the minimum wage will actually increase inequality for a high index of inequality aversion. This is because such a high index puts a lot of weight on the effects on the poorest, namely, the unemployed, On the other hand, a moderate index of inequality aversion will imply that an increase in the minimum wage will reduce inequality.

Second, the minimum wage reduces the income of skilled workers. This is because, as long as there is (even a small amount of) complementarity between skilled and unskilled workers, a lowering of the unskilled input in production reduces the marginal product of the skilled.

Third, the minimum wage increases the distortion of labor taxation in the unskilled labor market. Figure 1 illustrates this point. In the absence of the minimum wage, equilibrium would be determined by the intersection of a labor supply and a labor demand schedule. The labor supply schedule is likely to be quite inelastic, that is, not very far from vertical. Introducing a tax would only reduce employment by a small amount. The economy would move from point E to point A along the labor demand curve. Wages would be determined by point B’s vertical coordinate, and would fall almost one-for-one to match the increase in the tax. Introducing a minimum wage is equivalent to replacing, from the employer’s viewpoint, the labor supply curve Ls with a horizontal line MM. Employment then reacts to the tax wedge as if its supply were infinitely elastic at the minimum wage level. Introducing a tax would move the economy from E′ to A′, with a large decline in employment.

Labor taxes such as social security contributions are often blamed for unemployment of unskilled workers. But they have this effect only because other rigidities are present in the labor market. In the model developed in Section II, this effect magnifies the direct distortionary impact of the regulation. To prevent inequality from rising too much with unemployment, one must have unemployment benefits, which are financed by means of a labor tax. This tax further reduces output and employment because the very existence of the minimum wage makes it distortionary.3

Figure 1.
Figure 1.

Effect of Taxes on Labor Demand

Citation: IMF Staff Papers 1994, 004; 10.5089/9781451930887.024.A004

Fourth, the redistributive properties of the minimum wage are also affected by other forms of labor market rigidities. In particular, labor turnover, which depends on firing/hiring costs and job protection legislation, is crucially related to the distributional impact of the minimum wage. In a high turnover world, income is on average shared equally among members of the unskilled labor force. While the employed earn more than the unemployed, they will soon switch positions. Therefore, the minimum wage has a small impact on inequality among the unskilled. By contrast, if turnover is very low, the employed expect to keep their jobs for a long time and reap most of the benefits of an increase in the minimum wage. Those unlucky enough to join the larger pool of unemployed workers will lose, and stay in that pool for a long time. The minimum wage is then more likely to be nonegalitarian.

Ironically, the minimum wage is often associated with high job protection and low mobility. This suggests that labor market rigidities may be associated with less philanthropic goals than income redistribution, a point to which we return in Section IV.

II. A Simple Model

In this section we develop a simple model to analyze the effect of the minimum wage on output, income distribution, unemployment, and welfare. There are two types of labor: skilled and unskilled. These are the only inputs to the production function. In the long run, there is no loss of generality associated with ignoring capital as long as the long-run capital stock is restricted by the condition that its marginal product must be equal to some exogenous required return.4

The production function is assumed to be CES in both inputs:


where subscripts 1 and 2 denote skilled and unskilled labor, respectively. α, which is between - ∞ and 1, is an index of complementarity between the two types. If α = 1, they are perfect substitutes and unskilled labor cannot artificially raise its price above market-clearing levels. If α is less than 1, unskilled labor can achieve welfare gains by increasing its price at the expense of skilled labor’s earnings. The more the two inputs are complementary, the less the demand for unskilled labor will fall when its price rises. We also typically assume that ρ is low enough to ensure that the skilled earn more than the unskilled.5 The total labor force is normalized to 1, with χ skilled workers and (1 - χ) unskilled workers. Each worker is endowed with one unit of labor, which has no disutility.

We shall consider two systems. In system A, there is a minimum wage, which is above the market-clearing level for unskilled workers. Labor is taxed at a flat rate, and the proceeds are used to finance unemployment benefits. In system B, there is no minimum wage and both markets clear. Labor is taxed at a flat rate to finance a transfer to the unskilled. We shall compare both systems from three points of view: efficiency, income distribution, and political incentives.

If taxation were nondistortionary, then system A would clearly dominate system A, since full employment would prevail and a high enough tax rate would achieve whatever distributional goal is set by society. Therefore, it is only when taxation is distortionary that the comparison is not obvious.

We introduce tax distortions by assuming that a dollar of tax levied generates less than a dollar of transfers. More specifically, we assume that a dollar of taxes generates [1 - Φ(τ)] dollars of transfers, where τ is the tax rate and ω is an increasing, convex function of τ. This loss can be thought of as a tax collection/processing cost. An alternative approach would be to introduce an elastic labor supply curve. The approach taken here gives an advantage to the minimum wage: it redistributes income directly without having to incur the collection cost. This advantage will however be insufficient to justify it as an appropriate redistributive tool.

System A

Let us first consider the outcome under system A. Let ¯w be the (post-tax) minimum wage. The skilled labor market clears, implying


The cost of employing an unskilled worker is equal to ¯w/(l - τ), where τ is the tax rate. This must be equal to the marginal product of an unskilled worker, implying


Using equation (2), equation (3) can be inverted to compute L2:


The unemployment rate for unskilled workers u2 is then computed as


The (pre-tax) skilled wage is determined as the marginal product of skilled labor at full employment:


This essentially closes the model. In order to make some welfare comparisons, let us define the consumption level of each group. That of skilled labor is simply equal to the post-tax income:


We now turn to unskilled labor. Those who are employed have a net income of ¯w. Those who are unemployed have a net income equal to unemployment benefits b, These are determined by the condition that the government budget is balanced, which is the following:


There are several ways to convert these incomes into consumption levels for the unskilled. The way it is done in this paper allows one to bring an interesting parameter into the discussion, namely, the extent to which the labor market is “flexible.”At one extreme, one can assume that the employed consume ¯w and the unemployed consume b. At the other extreme, one can envisage a perfect redistribution scheme among the unskilled, all of them consuming bu2 + ¯w(1 - u2) regardless of whether or not they are employed.

In practice, such redistribution between the employed and the unemployed is achieved through labor market turnover. We thus specify the employed and unemployed consumption levels in a way that allows us to take into account mobility between the two states. If mobility is very high, then the initial status of an unskilled worker (employed or not) will have very little impact on his/her permanent income, which will be close to bu2 + ¯w(1 - u2) If mobility is very low, then the unemployed will practically never find a job, and their permanent income will be close to b; similarly, the employed will almost never lose their jobs, and their permanent income will be close to ¯w. More generally, let σ be the flow probability of losing one’s job and becoming unemployed. Let μ be the flow probability of an unemployed finding a job. Then if δ is the discount rate, the permanent income of an employed unskilled worker is


The permanent income of an unemployed worker is, similarly,


In a steady state with unemployment μ2, inflows must be equal to outflows, implying μ = σ(l - μ2)/μ2. Plugging this into equations (9) and (10) and assuming that consumption equals permanent income yields the following rule for computing the consumption levels of employed and unemployed unskilled workers:


where s = σ/δ. Equations (11) and (12) tell us that the consumption levels of the employed and unemployed unskilled workers are a weighted average of the minimum wage and unemployment benefits. The weight on ¯w is decreasing in μ2 and higher for the employed. When s increases, C2c falls and C2u rises. The parameter s may be thought of as an index of labor market “flexibility.” It will tend to be lower in economies with regulations that prevent labor reallocation and turnover, such as firing restrictions, rent controls, or subsidies to declining industries. When s goes to zero, C2e goes to ¯w and C2u goes to b. When s goes to infinity, both C2c and C2u go to bu2 + ¯w(1 - u2). The specification therefore encompasses the two extreme cases described above. As will become clear, s will play a key role in the analysis of the distributional and political impact of a shift from system A to system B.

In order to compare income distributions, we will use an Atkinson (1970) measure of inequality with parameter ε ∈ [−∞, 1]. In system A, it is equal to


where ¯c is aggregate consumption, defined by


System B

We now turn to system B, in which wages are fully flexible, implying that both the skilled and unskilled are fully employed. One therefore has L1 = x and L2 = (1 - x). Pretax wages are then equal to marginal products:


Taxes are used to finance a transfer to the unskilled. Let a be the amount of the transfer per unskilled worker. Then the budget balance equation implies that


The skilled consumption level is then given by c1 = w1(1 - τ), while the unskilled consumption level is given by c2 = w2(1 - τ) + a. The Atkinson inequality index is then given by


where aggregate consumption is ¯c = xc1 + (1 - x)c2.

The points made in the previous section are apparent from the above equations. An increase in ¯w reduces L2 and therefore w1. This tends to reduce the inequality index defined by equation (13). At the same time, it increases the number of unemployed and reduces output. This, for a given tax rate, will reduce unemployment compensation. Both the increase in the unemployment rate and the reduction in unemployment compensation tend to increase the inequality index defined by equation (13), In principle, the government could compensate for that by increasing taxes. But this would lead to further increases in unemployment and output losses.6

The net effect of the minimum wage on inequality is therefore ambiguous, At low levels of unemployment benefits, it will tend to increase inequality. At high levels of unemployment benefits, it will tend to reduce it.

That the minimum wage increases the distortionary impact of taxes is also clear if one compares equations (15)(18) with equations (1)(14). In system B, taxes have no effect on output since there is full employment in both markets. They reduce consumption, however, because of the collection cost. A 1 percent increase in the tax rate would be matched by a 1 percent drop in post-tax wages so as to leave w1 and w2 unaffected (equations (15)(16)). The economy would remain at full employment with the same level of output. Consumption would drop by -ϕ′(τ)/ [1 - ϕ(t)] percentage points. In system A, the unskilled’s post-tax wage cannot drop, so a 1 percent increase in the tax rate has to be matched by an increase in u2 (equations (3)(4)). This introduces an additional distortion, since output drops: everything works as if the unskilled’s labor supply, which was inelastic under system B, were infinitely elastic. Furthermore, a 1 percent increase in τ generates a drop in the skilled aftertax wage that is greater than 1 percent. This is because the pretax wage of skilled workers falls due to the fall in L2 (equation (6)).

III. Model Calibration

The preceding discussion suggests that the minimum wage is a powerful tool to redistribute income from the skilled to those unskilled workers who remain employed. In this section we calibrate the model and answer the following question: can system B achieve the same inequality level as system A at a lower cost in terms of distortions?

Given the model’s simplicity, it is not easy to define what “realistic” parameter values should be. However, this is not very important because the results are robust for a wide range of variations in those values.

Table 1 presents the set of parameter values for the benchmark simulation. The proportion of skilled workers x was chosen to be x = l/2.7 The production function parameters α and ρ were calibrated so as to imply w1w2 = 2 under system B, meaning that the average income of the richer 50 percent is twice the average income of the poorer 50 percent. This leaves one degree of freedom for α, which, given the moderate elasticities of labor demand implied by econometric studies, was not allowed to exceed 0.2. The exponent in the inequality measure ε is fixed to −1, but results are very robust for wide variations in ε. Finally, the index of labor mobility, s, is 1, which may seem on the low side. Recent work on labor turnover (Burda and Wyplosz (1994), Davis and Haltiwanger (1992), Blanchard and Diamond (1990)) suggests a job destruction rate of about 10 percent, which, with a real interest rate of 5 percent, would imply s = 2. However, one has to take into account the fact that the labor market is very heterogeneous: a lot of the high turnover rate is probably due to a “secondary tier” of workers with very low job security and high turnover. In fact, most of the employed workers in the “primary tier” probably have a transition probability to unemployment of less than 10 percent. This is why we prefer s = 1. Concerning tax distortions, a linear specification ϕ(τ) = φτ was chosen. A wide range of values for φ was tried, with little change in results. The value φ = 1 was retained in the benchmark simulations, which implies a Laffer curve peaking at τ = 0.5 in the full employment case, and less under system A. Higher values of φ therefore seem rather implausible.

Table 1.

Parameter Values of the “Benchmark Simulation”

article image

Table 2 reports the basic results from the benchmark simulation. The minimum wage and tax rates under system A were picked so as to generate an unskilled unemployment rate of about 20 percent, with a replacement ratio bw of about 50 percent. This implies a minimum wage about 10 percent above its Walrasian level and a tax rate of about 4.5 percent.

Table 2 shows that system B clearly dominates system A. While it is true that in the absence of taxes system B would be more unequal than system A, the tax rate needed to achieve the same level of inequality is 2.7 percent, substantially smaller than under system A. The output gain from shifting to system B is more than 8 percent, a large number. What makes a lower tax rate possible under system B is that the same level of inequality can be achieved with a lower level of the unskilled/skilled wage ratio. This is because shifting to system B automatically removes the contribution of unemployment to inequality.

Given the wide margin by which system B dominates system A in the benchmark simulation, the results are quite robust to alterations in parameter values. For example, increasing labor market mobility would reduce inequality under system A, which should be matched by an increase in τ under system B. However, τ stays well below 4.5 percent even for implausibly large values of s, so that system B still dominates unambiguously. Similarly, an increase in e, which would give less weight to the poorest under system A, has only a small impact on the value of τ that achieves the same level of inequality under system B. An increase in the level of distortion φ increases system B’s dominance over system A since the tax rate is lower under system B.

Table 2.

Impact of a Shift from a Minimum Wage System to a Transfer System

article image

System B with τ computed so as to achieve same level of I as in system A.

I = Atkinson’s inequality measure with ε= −1.

y = output.

¯c = aggregate consumption.

Table 3 presents a simulation that was engineered in such a way as to be very favorable to system A. First, a high level of mobility was chosen under system A (s = 2). Second, the value of α = −0.2 implies a low demand elasticity for unskilled workers, thus lowering the impact of the minimum wage on unemployment. These two assumptions tend to make τ larger under system B. To strengthen system A’s case even further, an implausibly high value of φ = 5 is assumed. This implies that for a 10 percent tax rate, half of the proceeds are lost due to distortions. While it is true that the tax rate is now higher under system B, this is insufficient to make it inferior to system A. For the same level of inequality, total consumption is still 5 percent higher under system B than under system A. It is, however, possible to set up even more extreme examples. For example, for α = −0.5, s = 5, φ = 5, and ¯ w = 0.175, it is impossible to achieve system A’s inequality level under system B for any tax rate.

Table 3.

Impact of a Shift from a Minimum Wage System to a Transfer System: Searching for the Virtues of System Aa

article image

See Table 2 for definitions. Parameters as in benchmark simulation except for α = −0.2, s = 2, and ϕ = 5.

Aggregate consumption is still higher, for any tax rate on the left side of the Laffer curve, under system B than under system A.

The main conclusion of this section is therefore that it is very unlikely that distributional considerations could justify the kind of rigidities that plague European labor markets.

IV. Political Aspects

So far, we have assumed the existence of a “benevolent” policymaker who cares only about output and some aggregate measure of inequality. In this section, a slightly more cynical view is adopted, namely, that unskilled employed workers support the minimum wage simply because it increases their own welfare. We first consider what happens if these workers can block any reform from which they lose, and then ask the question: under what conditions would a shift from system A to system B that achieves the same level of inequality be “politically viable,” that is, not be blocked by the unskilled employed workers? Even if a reform is politically viable, it may be the case that relative to the skilled, the unskilled are worse off. This does not imply that they would block the reform, but it does suggest that once the reform is passed, they will have incentives to lobby for a higher tax rate. In principle, this could make system B less efficient than system A. To tackle this issue, we also consider what happens when the tax rate is set in both systems so as to maximize the welfare of the unskilled employed workers.

Can Reform Be Blocked?

The unskilled employed workers may lose from the reforms if system A is very rigid (i.e., there is a very low value of s). In such a case, they are not very exposed to unemployment, so they do not value the increase in the welfare of unemployed workers generated by the reform at all. The inequality-preserving tax rate will then not be high enough to compensate them; a higher tax rate is necessary. The question is then whether the excess distortions created by this higher tax rate can be large enough to make the shift to system B undesirable from an efficiency point of view.

Table 4 presents some elements of an answer to this question. In panel A of Table 4, the parameters are the same as for the benchmark simulation. For low values of s, the tax rate that equalizes inequality across systems-is insufficient (column 3). The politically viable tax rate (column 5) is substantially higher, but distortions are not large enough to make the shift welfare reducing (column 6). Panel B of Table 4 presents a similar simulation in a world where tax distortions are larger (φ = 5), while distortions from the minimum wage are smaller (α = −0.2). In that case, for low values of s there exists no tax rate that compensates the employed for a shift to system B: the tax system is politically unviable.

Table 4.

Political Viability of Reform

article image

τ(I) = tax rate that matches inequality under system B.

P.V. = political viability index equal to the relative gain to the unskilled employed of shifting from system A to system B with tax rate τ(I).

¯ cBcA(I) = welfare gain, in terms of aggregate consumption, of shifting from system A to system B with tax rate τ(I).

τ(P.V.) = politically viable tax rate, i.e., that which makes the unskilled employed indifferent between system A and system B.

¯ cBcA(P, V.) = welfare gain, in terms of aggregate consumption, of a shift from A to B with τ = τ(P.V.).

0 means inequality lower under system B with τ = 0.

NN means no tax rate politically viable.

The value of s above which it becomes viable is, however, quite small (0.3), In that case, the shift to system B is, again, welfare-improving by a wide margin.

While these simulations convey the qualitative message that Europe’s low labor mobility may generate political opposition to system B, Table 4 nevertheless suggests that this will only arise for very low values of the mobility parameter. Furthermore, for the benchmark parameter values, which are more reasonable than those in panel B, an increase in the tax rate can compensate the unskilled employed and leave the ranking of the two systems in terms of welfare unaffected.

Can Reform Lead to Too Much Expropriation?

We now consider what happens if the tax rate is in both systems endogenously determined by the unskilled employed so as to maximize their welfare. This will typically lead to much higher tax rates than the ones considered above.

In system B, the tax rate that maximizes the welfare of the unskilled employed is readily computed as the solution to


Given that w1 and w2 are unaffected by τ, the first order condition with ϕ(τ) = φτ yields


With w1 = 2w2 and χ = 0.5, this is equivalent to τ = l/(3φ), which for φ = 1 yields a tax rate of 33 percent.

In system A, the tax rate benefits the employed only indirectly since it is used to finance unemployment insurance. As a result, the equilibrium tax rate that prevails under system A is much lower. However, taxes are much more distortionary under system A because they reduce employment. Simulations suggest that this effect is likely to dominate. For the benchmark simulation, the equilibrium tax rate under system A is 17 percent, but this generates an unemployment rate among the unskilled of 40 percent. This large distortionary impact makes aggregate consumption lower under system A than under system B.

However, an interesting new problem arises: skilled workers now object to the reform. This is not surprising, since they are taxed under system B at a rate twice that under system A, The fact that their gross marginal product is also higher is not strong enough to outweigh this tax effect. In principle, the skilled and unskilled could agree on a future tax rate before shifting to system B. Such a Pareto-improving tax rate would exist for a wide range of parameter values, as suggested in Table 4. However, if there is no commitment device to enforce this agreement, the skilled will block the reform.

Now, this result is in some sense “cooked” because under system A we have ruled out any transfers to the unskilled employed, who decide on the tax rate. It may therefore be useful to run a more symmetrical experiment, assuming that under system A taxes are redistributed as transfers to both the employed and unemployed unskilled workers. The simulations (not reported here) suggest that the tax rate will only be slightly lower under system A, and that system B unambiguously dominates system A.

V. Other Types of Rigidities

The above discussion has focused on a peculiar type of rigidity, namely, the minimum wage. More generally, it is applicable to any institution (such as unemployment benefits) that allows unskilled workers to raise their compensation beyond the market-clearing level.

However, it may be argued that the “European model” consists of a wide array of interventions in the labor market and that some of them may have beneficial effects. It is of course impossible to review this in detail here. Let us, however, discuss two common arguments possibly justifying job protection and unemployment insurance.

The first argument is that job protection may be good for productivity and long-run growth because long-term jobs are associated with more job-specific human capital than short-term jobs. Several criticisms of this argument can be made. First, there is no reason to believe that the market cannot internalize the effect of tenure on productivity. Nothing prevents firms from voluntarily providing implicit or explicit job protection, and this is indeed observed. As pointed out by Hall (1982), in the United States, where no job protection is granted by the state, most jobs last a long time. Second, the supposed productivity effects of firing costs are likely to prevail in declining sectors where they are not much needed. At the same time, firing costs slow reallocation toward new productive sectors. If one takes the Schumpeterian view that growth essentially comes from new goods rendering old goods obsolete, then the negative effect of firing costs on growth through slower reallocation is likely to dominate their positive effects through on-the-job training.8 Indeed, while the latter effect is likely to be a growth effect, the former is only a level effect: a higher mean tenure is associated with a higher productivity level. Third, the productivity gains from higher job protection have to be weighed against the productivity losses from higher unemployment duration. The view that higher unemployment duration generates skill loss is as plausible as the idea that higher job duration generates skill gains. Fourth, the empirical evidence on the effect of tenure on wages is not strong, particularly since there may be a bias because a positive association between the two variables may just be due to “good news” about match quality rather than an effect of tenure per se.9

The second common argument in favor of job protection and unemployment insurance is that unemployment benefits allow workers to search longer for a better match, thus improving aggregate productivity through the distribution of matches in society. This argument is not likely to be very relevant, for the following reasons. First, unemployment is not a promising position from which to find a job. As Tobin (1972) argued, the employed are better informed about jobs than the unemployed. Moylan, Millar, and Davies (1982), cited by Layard, Nickell, and Jackman (1991), report that in the United Kingdom, more than half the unemployed spend fewer than five hours a week looking for a job. Similarly, the mean rate of job applications by the unemployed is, in the United Kingdom, on the order of one a month.10 Second, if waiting for better matches were that important, one would not observe the monotonously decreasing exit rates from unemployment as unemployment duration increases.

VI. Conclusion

We have developed a model that analyzes the effect of alternative labor market regulations on the welfare of individuals according to their skill level and employment status. The model facilitates discussion of the desirability of labor market reform from the point of view of efficiency, income distribution, and political feasibility. It also allows us to relate this discussion to other characteristics of the economy, such as labor mobility and the distortionary impact of taxation.

The model could be adapted to analyze other sorts of regulations that have a similar impact on the demand for unskilled labor—for example, an unemployment benefit system that raises the workers’ reservation wage.

The model points to the conclusion that there is little justification for the set of labor market regulations that now prevails in Europe. Alternative instruments, such as direct taxes and transfers, would be a more efficient way to achieve distributional goals. While there may be some political resistance to reform, a properly designed package should be able to overcome this problem while still increasing the economy’s efficiency.

It therefore remains a puzzle why reform has been so timid. One may blame irrationality or shortsightedness, failure to recognize gains from general equilibrium effects, or uncertainty about the pace of events once reform has been passed. Analyzing these factors would be a useful subject for future research.


  • Abraham Katherine G., and Henry S. Farber, “Job Duration, Seniority, and Earnings,” American Economic Review, Vol. 77 (June 1987), pp. 278 –l97.

    • Search Google Scholar
    • Export Citation
  • Atkinson, Anthony B., “On the Measurement of Inequality,” Journal of Economic Theory, Vol. 2 (September 1970), pp. 244 –63.

  • Bettola, Giuseppe, “Flexibility, Investment, and Growth,” (unpublished; Princeton, New Jersey: Princeton University, 1991).

  • Blanchard, Olivier J., and P. A. Diamond, “The Cyclical Behavior of Gross Flows of U.S. Workers,” Brookings Papers on Economic Activity 2 (1990), The Brookings Institution, pp. 144 –53.

    • Search Google Scholar
    • Export Citation
  • Blanchard, Olivier J., and Lawrence H. Summers, “Fiscal Increasing Returns, Hysteresis, Real Wages and Unemployment,” European Economic Review. Vol. 31 (April 1987), pp. 543 –50.

    • Search Google Scholar
    • Export Citation
  • Burda, Michael, and Charles Wyplosz. “Gross Workers and Job Flows in Europe,” European Economic Review, Vol. 38 (June 1994), pp. 1287 –1315.

    • Search Google Scholar
    • Export Citation
  • Davis, Steven J., and John Haltiwanger, “Gross Job Creation, Gross Job Destruction, and Employment Reallocation,” Quarterly Journal of Economics, Vol. 107 (August 1992), pp. 819 –64.

    • Search Google Scholar
    • Export Citation
  • Emerson, Michael, with André Dramais, What Model for Europe? (Cambridge, Massachusetts: MIT Press, 1988).

  • Hall, Robert E., “The Importance of Lifetime Jobs in the U.S. Economy,” American Economic Review, Vol. 72 (April 1982), pp. 716 –24.

    • Search Google Scholar
    • Export Citation
  • Jovanović, Boyan, “Job Matching and the Theory of Turnover,” Journal of Political Economy, Vol. 87 (October 1979), pp. 972 –90.

  • Layard, P.R.G., S.J. Nickell, and Richard Jackman, Unemployment: Macroeconomic Performance and the Labor Market (Oxford: Oxford University Press, 1991).

    • Search Google Scholar
    • Export Citation
  • Moylan, Sue, Jane Millar, and Bob Davies, “Unemployment—The Year After,” Employment Gazette, Vol. 90 (August 1982), pp. 334 –40.

  • Ortega, Javier, “Loss of Skill During Unemployment, Complementarities in Production, and Persistence of Shocks,” DELTA Working Paper No. 93–26 (Paris: DELTA, Ecole Normale Supérieure, December 1993).

    • Search Google Scholar
    • Export Citation
  • Saint-Paul, Gilles, “On the Political Economy of Labor Market Flexibility,” in NBER Macroeconomics Annual 1993, ed. by Olivier Blanchard and Stanley Fischer (Cambridge, Massachusetts: MIT Press, 1993).

    • Search Google Scholar
    • Export Citation
  • Tobin, James, “Inflation and Unemployment,” American Economic Review, Vol. 62 (March 1972), pp. 1 –18.


Gilles Saint-Paul is a researcher with CERAS and DELTA and a fellow of the Center for Economic Policy Research. DELTA is a joint research unit of ENS-CNRS-EHESS. This paper was written while he was a visiting scholar at the IMF’s Research Department. Its main theme was suggested by David Coe and Robert Ford, and it has benefited from discussions with José De Gregorio.


See Emerson (1988) for a discussion on the welfare state and labor market reform.


For an analysis of the political issues in increasing labor market flexibility with an emphasis on complementarities, see Saint-Paul (1993).


In principle, the problem could be alleviated by raising the tax rate in the skilled market, in accordance with the prescriptions of optimal second-best taxation.


This would be the case in an open economy with capital mobility. The marginal product of capital would then be equal to the world rate of interest.


As emphasized by Ortega (1993), there is no canonical way of labeling workers as “skilled” or “unskilled” as long as they are not perfect substitutes.


Note that the model exhibits “fiscal increasing returns” as analyzed in Blanchard and Summers (1987): an increase in the unemployment rate lowers the tax base, which for a given target for unemployment benefits calls for an increase in taxes, which leads to further increases in unemployment.


Given that there are only two skill levels in the economy, there is no clear-cut way to calibrate x. If one takes the number of people earning the minimum wage as a guideline, then 0.5 is too low for x, since the minimum wage will then Be binding for 50 percent of the labor force. However, if one takes the proportion of workers with higher education as an alternative criterion, then 0.5 is too high.


See Layard, Nickell, and Jackman (1991).