Reducing the Working Week: A Free Lunch or Irrational ideology?

The French government's decision to reduce the working week from 39 to 35 hours became law on 1 February 2000. Martine Aubry, France's employment minister, has stated that the objective of the policy is to reduce unemployment by sharing out the available work. The appeal of this proposal lies in its promise to reduce unemployment without affecting the Welfare State. Yet many economists have labelled the policy an 'irrational ideology' and have argued that imposing further restrictions on an already highly regulated labour market will result in even greater inefficiencies and may actually worsen France's unemployment problem.

A Discussion Paper by Ramon Marimon and Fabrizio Zilibotti addresses the questions of both the employment and the redistributional effects of a policy that reduces working time. Much of the previous literature on this subject has cautioned that government action to reduce the working week may not result in the desired reduction in unemployment. The main contribution of Marimon and Zilibotti's work is that it is able to trace back the possible employment effects to basic parameters such as workers' preferences for consumption and leisure, and the degree of capital mobility. Hence, their model can be calibrated and solved numerically to obtain a quantitative assessment of the effects of the policy. It also provides a clear rationale as to why workers often lobby for legislative restrictions on working time.

The proponents of 'Restrictions on Working Hours' (RWH) argue that the policy will induce employers to substitute some of the labour provided by their current employees with new hirings. But opponents point out that some of the firms' labour costs are fixed per employee (e.g. screening, hiring, training, etc) and, as such, are independent of the number of hours worked. Hence hiring more workers will increase the costs of production, which in turn will reduce the incentives for firms to generate employment.

In the paper, the authors develop a general equilibrium model where unemployment originates from search matching frictions; wages are set endogenously via a standard Nash bargaining solution; and capital is assumed to be fixed. The main forces stressed by advocates and opponents of the policy are both present. First, in the tradition of the search-matching literature, fixed costs are included in the form of a vacancy creation cost associated with the hiring of new workers. The existence of fixed costs means that the simple 'lump of labour' argument does not apply. Second, the model assumes diminishing returns to labour, and workers and hours are assumed to be perfect substitutes. This means that the marginal product of labour increases and firms therefore have an incentive to post new vacancies when the number of hours of labour per employee falls. The presence of forces with opposite signs makes the employment effects of the policy a priori ambiguous.  

In order to study the employment and welfare effects of RWH, the paper first characterizes the equilibrium in a laissez-faire environment (wages and hours are endogenous) and then constrains the maximum number of hours worked (wages are endogenous). The initial result is that workers and employers have largely differing (endogenous) preferences on working time. In general, RWH benefits workers, both unemployed and employed, but reduces profits and output. This is because restricting working hours below the laissez-faire solution increases the workers' bargaining power. Although the bargaining process under regulation gives a socially inefficient outcome, the distributional gains for the workers more than outweigh the efficiency loss.

The employment effects of RWH crucially depend on the response of wages. If hours were reduced but the total wage per employee remained constant, employment would unambiguously fall. In the model, however, wages adjust endogenously and the final effect on employment depends on the extent to which the reduction in hours affects both the workers' marginal utility of consumption and the marginal productivity of labour. The net employment effect will therefore depend on both technology and the workers' preferences for consumption and leisure.

Although a standard Cobb-Douglas production technology is maintained throughout, the paper does offer results for different classes of preferences. The benchmark utility function (GHH) is a generalized version of quasi-linear utility, first introduced into the real business cycle literature by Greenwood, Hercowitz and Huffman (1988). GHH preferences have the property that the marginal rate of substitution between consumption and leisure is independent of the consumption level within the period. The second utility function used is one that exhibits a Constant Elasticity of Substitution (CES) between consumption and leisure.

The assumption of a GHH utility function produces a number of interesting results. First, the relationship between working time and employment is non-monotonic. Specifically, starting from a laissez-faire equilibrium, there exists a range of reductions in working hours that will increase employment. Second, there exists a range of reductions in working hours that increase the welfare of all workers. Firms, however, lose since profits fall. Third, reducing working time below the laissez-faire equilibrium reduces employment when capital is perfectly mobile and there is no fixed factor of production. This finding suggests that at least part of the positive employment effects which may materialize in the short run are likely to vanish as firms adjust their productive capacity. It might also explain why some proponents of the policy would like it implemented on the widest possible area – e.g. the EU.

In order to assess the quantitative importance of these results, the authors construct calibrated economies and simulate the effects of reductions in working time. Using conventional estimates of parameter values and GHH preferences the paper presents results for three different risk aversion parameters, ranging from risk neutrality (RRA=0) to almost unit relative risk aversion (RRA=1). The above graphs plot the unemployment rate, the welfare of the employed, the welfare of the unemployed and the firms' profits as functions of the number of hours worked for the case of RRA=0.8. The dashed line corresponds to the laissez-faire equilibrium (i.e. 44 hours).

The laissez-faire equilibrium in the calibrated model is confirmed by the data: 44 hours is the average working week in the UK, the only European country with virtually no restrictions on working time. While the length of the working week that maximizes workers' utility is approximately 29 hours. The effect on employment varies with the level of risk aversion, since this affects the wage response. However, for all three values of risk aversion, moving from the laissez-faire to the utility maximizing working week (i.e. from 44 to 29 hours) results in a fall in unemployment, with the decrease in the unemployment rate ranging from 0.5 to 0.9 percentage points – the number of unemployed falling between 6.25% and 11.25%. These small employment effects imply that the total number of working hours in the economy is reduced by almost the full amount of the reduction in hours per worker. Subsequently, GDP falls by about a quarter.

Relating the model to the current policy debate, the paper compares two regulated economies with working weeks of 40 and 35 hours, respectively. As is evident from the graph, the employment differences between these two economies are small. These results do not significantly change if the model is recalibrated with a higher structural unemployment rate. If parameters were set so the unemployment rate in the 40 hours economy was 11% (about the average for continental Western Europe), then the unemployment rate in the 35 hours economy would be 10.7%.

When the analysis is extended to include CES preferences the results are even less positive. Specifically, unless consumption and leisure are better substitutes than in the Cobb-Douglas case, reductions of working time cannot increase employment. The more complementary are consumption and leisure, the more negative are the employment effects of restricting working time. Yet even when these restrictions cause unemployment, the welfare of the workers, both unemployed and employed, is still maximized when a relatively large restriction on working time is imposed.

Marimon and Zilibotti's results are consistent with history: since the industrial revolution workers have supported reductions in the working week and employers have opposed them. Policies that reduce working hours are nothing new, but there is an important difference in the current reasoning for legislation: what was intended as a policy for alleviating the conditions of the employed has now become a policy for alleviating the conditions of the unemployed. Hence, there are very few examples of policies that are motivated by this new rationale. One such example is again that of France, which in 1981 decided to shorten the working week from 40 to 39 hours. And it is this situation that Bruno Crépon and Francis Kramarz study in their Discussion Paper.

A few months after the election of François Mitterrand in May 1981, the French government announced a mandatory reduction in the working week from 40 to 39 hours and a 5% increase in the French minimum wage, the SMIC. This was coupled with mandatory stability of monthly earnings for minimum-wage workers – effectively a further increase in their hourly wage – and a strong recommendation for the stability of monthly earnings for those earning above this level. The law became effective on 1 February 1982.

In order to evaluate the effects of the policy, Crépon and Kramarz compare workers who, in March 1981, were identical in all but one respect: some were working 40 hours a week (i.e. those directly affected by the policy) whereas others were working between 36-39 hours (i.e. those unaffected by the policy). This second group can be viewed as a control group. The paper uses panel data from the French labour force survey. The sample used contains individuals who were surveyed for the years 1981, 1982 and 1983, thus allowing the authors to follow the same individuals and characterize their situation before, during and after the implementation of the policy.

Marimon and Zilibotti's paper emphasized the importance of the response of wages to the result of the policy. In 1981 minimum-wage earners were guaranteed the same take-home pay as before the implementation of the policy, but for other workers this was only a recommendation. However, a survey carried out in 1982 shows that more than 90% of these workers had their monthly pay unchanged after the introduction of the law. Marimon and Zilibotti's work suggests that this scenario would lead to a fall in employment. And this is confirmed by the results of Crépon and Kramarz.

They found that for observationally identical workers, those affected by the policy were twice as likely to lose their jobs compared with those unaffected by the policy: 3.2% of workers employed 36-39 hours in March 1981 were unemployed in 1982, whereas 6.2% of workers employed 40 hours in March 1981 were unemployed in 1982. As such, these results do not constitute definitive evidence that the policy had an adverse impact. They may stem from different unobservable characteristics of those working 40 hours compared to those working 36-39 hours or from the simultaneous increase in the SMIC. Hence the authors examined the employment to unemployment transition before the announcement of the policy, which was unexpected as Mitterrand's election is widely regarded as being unforeseen. The results of this show that between 1979-81 those working 40 hours appeared less likely to become unemployed than those working 36-39. In addition, removing the relatively small proportion of minimum wage workers had little effect on the results.

For those working 40 hours the most severely affected were workers who were also affected by the rise in the SMIC. For workers earning above the SMIC, those most affected worked in the service sector, had long job tenure and a low level of education. Hence, it seems that the institutions that were designed to protect such workers have harmed them disproportionately: it is the workers that have the greatest difficulties in finding a job once unemployed (i.e. minimum wage workers, the poorly educated and senior workers) that have been most severely affected by the policy.

Not all firms were able to implement the working time restrictions immediately. Hence, the authors extend their analysis by examining the employment to unemployment probabilities of those workers who were still working 40 hours after February 1982. Using a completely different data set, they found that workers employed 40 hours in 1982, 1983 and 1984 were significantly more likely to lose their jobs that those employed 39 hours in the same years. This second analysis reinforces the authors' initial result and they conclude that the reduction of the workweek from 40 to 39 hours is directly responsible for increased unemployment. 

The central message from both of the papers is that the response of wages is crucial. Crépon and Kramarz show that if wages remain unchanged then unemployment will rise. But France's current policy is slightly different in structure from the 1982 version. As in 1982, minimum-wage workers are guaranteed the same take-home pay, but this is now partly subsidized by the government. Employers of non-minimum-wage workers will be 'expected' to maintain the monthly earnings of their employees, although no legal arrangements have been made to enforce this. However, there are reasons to believe that employers may be less likey to comply with this recommendation than they were in 1982 (when 90% of workers had their pay unaltered): a number of firms have negotiated a commitment to wage moderation, many having frozen wages for the next 18 months. Hence, if wages are allowed to adjust then, according to Marimon and Zilibotti, unemployment could initially fall.

So is reducing the working week a free lunch or an irrational ideology? The answer would seem to be neither. Marimon and Zilibotti's paper suggests that there may be nothing irrational behind workers' support for working time reductions. Their model suggests that even when restrictions on working hours cause unemployment, the welfare of both unemployed and employed workers is maximized when a relatively large restriction is imposed. But the conditions for obtaining even small increases in employment are rather restrictive. In particular, employment increases can only occur if firms have some fixed factor of production: if capital can freely adjust then reducing working hours unambiguously reduces employment. In addition, the output loss associated with the policy will have negative effects on both the tax base and government expenditure. Therefore what little benefits there are may well be short-lived.

Discussion Paper No. 2127: 'Employment and Distributional Effects of Restricting Working Time' by Ramon Marimon (European University Institute, Firenze, and CEPR) and Fabrizio Zilibotti (Institute for International Economic Studies, Stockholm University, and CEPR).

Discussion Paper No. 2358: 'Employed 40 Hours or Not-Employed 39: Lessons from the 1982 Mandatory Reduction of the Workweek' by Bruno Crépon (CREST-INSEE, Paris) and Francis Kramarz (CREST-INSEE, Paris, and CEPR).