Jump to ContentJump to Main Navigation
Working Hours and Job Sharing in the EU and USAAre Europeans Lazy? Or Americans Crazy?$

Tito Boeri, Michael Burda, and Francis Kramarz

Print publication date: 2008

Print ISBN-13: 9780199231027

Published to Oxford Scholarship Online: May 2008

DOI: 10.1093/acprof:oso/9780199231027.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017



(p.103) Introduction
Working Hours and Job Sharing in the EU and USA

Francis Kramarz (Contributor Webpage)

Pierre Cahuc

Bruno Crépon

Oskar Nordstörm Skans

Thorsten Schank

Gijsbert van Lomwel

André Zylberberg

Oxford University Press

Theory tells us that work‐sharing is rarely good for employment. Or, at least, conditions under which work‐sharing boosts employment are relatively demanding and are rarely met in practice. However, some countries implemented workweek reductions while others did not. One potential answer to this discrepancy is overly simplistic and populist. In those countries that decided to share work, politicians, union members or leaders, or even public opinion were never confronted with the basic elements contained in Econ 101, the first course in economics taught at virtually all universities. They suffered from a miscomprehension of basic market mechanisms. Another answer runs as follows. For complex reasons (historical, religious, and institutional more generally), countries came up with different agreements between the various social forces. These agreements are also the outcomes of collective bargaining, which takes very different routes in continental Europe. Finally, such agreements reflect basic preferences over various areas of each country's citizens' common life (work and family in particular, or the respective importance and role of the private and the public spheres). The outcomes of the social processes at work should be reflected in the multiple ways of sharing work that have been adopted across Europe, as we shall see.

To illustrate these points, we examine specific countries, namely Germany, France, the Netherlands, and Sweden (see Table 1). Why did we select these countries? One possible answer is given here. We use and complement Jan van Ours's Table C2.2, based on his discussion of our book, and presented below. First, all countries in Table 1, even the UK, reduced hours between 1950 and 2004. This decrease differs across countries but is virtually always larger than 20 percent. Then, part‐time work is present in most countries, but never as prevalent as in the Netherlands, the so‐called part‐time economy. Part‐time is one obvious way of sharing work. As indicated above, this Dutch case will be thoroughly examined. The next column tells us that the number of hours per week is different across European countries, again except for the Netherlands, essentially because of the presence of part‐time working as evidenced in the previous column (this share is included in the computation accounting for (p.104)

Labor Market Effects of Work-Sharing Table 1 Some statistics on hours of work in Europe


Change in hours





Employment rate


























































Notes: For all numbers, see Jan van Ours, Table C2.2 in this book. Germany is West Germany for the relevant period.

the respective number of hours for full‐time and part‐time workers). Moving to the next column, Sweden stands out because Swedes work far fewer weeks than citizens of any other countries in the table. We will focus closely on this issue and try to understand the reasons for and the consequences of such a way of sharing work. The number of hours per year are given in the next column. Italy and the UK are clear leaders, Sweden and Holland clear “losers”, and France and Germany in between. In this respect, these last two countries are very similar but, in fact, experienced very different ways of sharing work as will be seen: bargaining traditions are very different in the two countries and it shows. These two contrasting examples will also be thoroughly investigated. Strikingly, the employment rate also differs widely, almost in a similar fashion as in the previous column. There are couple of exceptions, though: Italy and the UK. These countries have not shared work as analyzed in this book, i.e. by manipulating hours. But Italy has “prevented” its youth and women from having jobs. The employment rates are then directly reflected in the number of hours per person (last column), the product of the two previous columns.

The resulting structure of the book is as follows. To help understand the situation that exists now, we will first investigate the determinants of work‐sharing, building on theoretical work. Then, we will examine the employment effects of work‐sharing using Germany and France as leading examples. Finally, we will try to understand some of the institutional and social factors affecting the choice to reduce the workweek, using Sweden and the Netherlands as leading examples.

Chapter 4 examines the theoretical underpinnings of the effects of work‐sharing on employment. Because institutions vary across Europe, different models are presented to capture this variation. These models shed light on institutional factors that may affect the decision to decrease working hours. We describe these two points in turn.

(p.105) First, models can identify deep factors that modify or are responsible for the employment effects of work‐sharing. Let us summarize the main messages of the chapter on this first issue.

The simplistic view that work‐sharing, through a decrease in standard hours, creates employment is easily expressed in a simple framework in which firms' output is predetermined (i.e. Keynesian and short‐term). In this view, hourly wages do not adjust. And the effect of a reduction in standard hours depends on the natural choice of hours by firms. When this natural level is low vis-à-vis the new standard, work‐sharing has no effect. At the opposite extreme, when this natural level is high vis-à-vis the new standard, the reduction has the counter‐intuitive effect of raising the number of hours worked by all employees and the associated effect of decreasing employment. Finally, when the natural level equals standard hours, hours do indeed drop but the employment effect is ambiguous. Simple simulations show that the positive effect on employment for these firms appears to be of a smaller magnitude than the negative effect on firms that need long hours. Notice though that the positive effect is exactly the one that proponents of work‐sharing have in mind.

However, wages should adjust, mostly because fixed hourly wages as shown just above entail decreasing monthly or weekly wages after a reduction in hours. Hence, workers will ask for wage compensation. Then, the analysis of a 10 percent decrease in standard hours shows that employment, profits, and production never increase, in fact always decrease, unless there is no wage compensation. In particular profits will decrease for most firms and, hence, a reduction in hours will eventually lead to firm closures, the size of this effect depending on the size of the wage compensation. In addition, the chapter shows the size of subsidies that must be given to firms in order to maintain profits and employment.

Hence, in a simple framework, work‐sharing potentially destroys jobs and firms if workers demand wage compensation to maintain monthly wages, and therefore forces the government to distribute payroll tax subsidies, a source of financing that might have a better use given the effect on jobs.

Then, because workweek reductions are likely to destroy jobs when markets are perfect, the chapter examines the effects of work‐sharing when unions and firms bargain over wages and hours. Here, if the union's objective places significant pressure on employment during collective bargaining, the outcome may entail more employment but, at the same time, the bargained solution implies more hours (not less) and lower weekly wages. Hence, to increase employment when collective bargaining takes place, hours must be increased and monthly wages decreased. Not exactly what appears to have been observed in the various experiments that we examine in this book.

However, the chapter also shows that when workers have a very strong preference for leisure, it may be possible to increase employment by decreasing (p.106) weekly hours. Other favorable factors for employment in this bargaining framework are bargaining centralization and firms' market power; when the product market is less competitive, a decrease in standard hours is more likely to create jobs.

Second, models can help us understand the choice of hours, and therefore the forces against or in favor of work‐sharing, as a function of institutions and as a function of the preferences of the collective or individual members of any given society.

This question is examined in Chapter 4 in two opposed, and complementary, situations: perfect competition and collective bargaining. Perfect competition allows us to understand the role of labor supply. Collective bargaining allows us to understand the role played by the unions as well as the role played by the economic environment in which firms operate.

In the case of perfect competition, a worker faces a wage schedule offered by firms that specifies earnings as a function of hours. He or she maximizes utility subject to this wage function. In particular, when workers prefer leisure or home production over work, they will tend to choose shorter hours. Furthermore, because interactions in the family matter, labor supply decisions, i.e. the amount of hours chosen by workers, depend on hours chosen by other family members. Hence, in the so‐called European model, workers may well have “collective” preferences for leisure through externalities: returns to leisure increase when more people take vacations.

In the case of collective bargaining, the chapter shows very deep, and maybe surprising, results that tie various markets together. First, stronger unions tend to demand shorter hours. More interestingly, when negotiations are centralized, bargained hours should also be shorter. And, even more interesting, economies with less competition in the product market should also display shorter negotiated hours. However, the first effect—stronger unions demand shorter hours—is counterbalanced by another effect. When unions attach significant weight to employment in their objective function, they should ask for longer hours (and lower monthly wages).

Our theory chapter delivers very useful lessons on the two aspects that are studied in our country chapters.

Chapter 5 examines the German case. Germany is a perfect illustration of the role of bargaining in determining hours of work. The metal and engineering workers' union (IG Metall), with 2.5 million members in 2005 the second biggest union in Germany, has played a dominant role in postwar bargaining.1 In the metal‐working industry, normal working time was reduced to forty hours in 1967. These cuts were intended to enhance quality of life. Indeed, other industries followed these settlements several years later, and by 1975 the prevailing conditions for full‐time (p.107) workers were six weeks of annual holidays and just above forty hours per week.

Given rising unemployment in the 1970s, in 1978–9 IG Metall again launched a campaign to reduce standard working time below forty hours in order to promote work‐sharing. While their attempts failed in the face of employers' strong resistance, they were more successful a few years later, when, after a seven‐week strike in 1984, normal working time was reduced to 38.5 hours in 1985. This was followed by further agreements between IG Metall and Gesamtmetall (the metal and engineering employers' association) on reductions of standard hours to thirty‐seven hours in 1988, to thirty‐six hours in 1993, and to thirty‐five hours in 1995. Other industries followed, even though standard hours varied (from thirty‐five to thirty‐nine hours). The most prominent firm‐level agreement on working time reductions has been the settlement between Volkswagen AG and IG Metall (down to 28.8 hours). This was mostly defensive, trying to preserve jobs.

Indeed, companies are often allowed to deviate from hours negotiated at the industry level. In addition, schedules are often quite flexible at the company level, in particular in those firms without a bargaining contract.

Many analysts have tried to evaluate the employment effect of these work‐sharing agreements. Chapter 5 rapidly reviews their findings. None has found positive effects. One main reason, outlined in Chapter 4, is that full wage compensation was almost always negotiated in the agreements (Hunt, 1999, finds almost full wage compensation). And, as we have seen, there is little hope of creating employment when wages do not decrease. And, as pointed out in Chapter 4 when analyzing collective bargaining and hours determination, German unions appear to have been able to protect their members, certainly because employment was accorded low priority in their objectives.

However, as mentioned above, firms and unions in recent years negotiated agreements to make hours more flexible. For instance, opening clauses allow workers to work longer hours than in the collective agreement. They also allow firms to implement working‐time corridors within which firms freely set their working time. Finally, firms may decrease working time with no wage compensation (for a limited period of time). On the workers side, employees are endowed with working‐time accounts. These accounts measure the amount of extra time worked, which can later be exchanged for free time. In 2004, 42 percent of workers had such accounts. Notice that this chapter provides the first analysis of these new developments, using firm‐level data sources. In particular, this new flexibility has been used to increase working time in many firms (Siemens and Daimler‐Chrysler are two prominent examples).

The effects of such increases are analyzed using the IAB data source. Roughly 2.5 percent of plants had increased working time over twelve months (during 2004, approximately). In most such firms, monthly wages stayed put (hourly (p.108) wage decreased). Results show that firms that increase standard hours also have decreasing employment while firms that decrease standard hours have stable employment. In particular, when standard hours increase, firms use fewer part‐time workers as theory predicts (full‐time workers become less costly). However, and more encouraging, in western Germany increasing standard hours are associated with increased productivity, measured by value added per worker, the coefficients being large but only marginally significant, while decreasing standard hours are associated with essentially unchanged productivity. In eastern Germany, decreasing standard hours generate the expected result: productivity decreases.

Chapter 6 examines the French case. Many complementary lessons can be derived from the two workweek reductions that took place in France.

The first episode of hours' reduction took place in 1982. Less than one year after FranÇois Mitterrand was elected president, the workweek fell from forty hours to thirty‐nine. The stated objective was to reduce employment. The policy was part of the program of the left‐wing parties, the so‐called “programme commun”, and was implemented without negotiation with business unions or workers, or unions.

The employment effects of this policy were evaluated in Crépon and Kramarz (2002). They confirm the main message of Chapter 4. Because monthly pay was not allowed to decrease, hence hourly wages increased, firms reacted according to the textbook: employment decreased. More troubling, the workers primarily affected were those paid around the minimum wage (the SMIC) who “benefited” the year before from large hikes.

Potentially informed by the previous episode, the Jospin government, with Martine Aubry as Minister of Labor, decided to fulfill another electoral promise and move to thirty‐five hours. Discussions between the government, which included the Green Party, and business unions were tense. Negotiations started within various industries and firms. But, at some point, Martine Aubry enacted a law essentially forcing firms with over twenty employees to come to some agreement with their workers' unions or delegates. In addition, various incentives and subsidies were proposed at different moments in time.

In June 1998, the so‐called Aubry I laws gave establishments incentives to reduce their workweek and create or preserve employment in exchange for large subsidies. In order to receive these subsidies, firms had to reduce hours by at least 10 percent in order to attain an average weekly duration of thirty‐five hours. In such a case, employment creation had to amount to 6 percent of total employment. A “defensive” aspect also allowed firms to receive subsidies to avoid economic separations or collective dismissals.

The 2000 law, Aubry II, offered payroll tax subsidies for all firms that decided to move to thirty‐five hours per week.

Hence, among firms with more than twenty employees, at the beginning of the twenty‐first century, various agreements prevailed. Some firms were (p.109) still at thirty‐nine hours and had to pay overtime, others went to thirty‐five hours between June 1998 and January 2000 and received incentives and subsidies; others refused the incentives (but received some “structural” subsidies) even though they went to thirty‐five at similar dates (the so‐called Aubry II forerunners). Firms also went to thirty‐five hours after January 2000, receiving only the “structural” subsidies. Indeed, some firms decided to “not ask” subsidies that they were entitled to receive.

Wage compensation schemes and wage moderation agreements were implemented at the same time so that monthly wages stayed constant in the short term and did not increase too rapidly in the longer run. Labor costs for low‐wage workers increased too strongly, thanks to the payroll tax exemptions that were expanded in those years. Until 2005, when all minimum wages were unified, there existed a flurry of SMICs depending on the point at which the firm reduced the working week.

Contrasting the French experience with the German one shows that the French government intervened massively and did not let the so‐called “social partners” come up with an (unlikely, to be honest) agreement.

How can Chapter 4 illuminate these outcomes? If there was bargaining, it was at the central level; something that should favor workweek reductions, as explained in the theoretical chapter. But, because the financial situations of firms in those years were extraordinarily diverse, firms strongly resisted these changes that clearly meant decreased profits for a large fraction of them. Hence, this forced the government to distribute massive subsidies (again see Chapter 4) and suggest wage moderation. One may wonder why unions accepted, and sometimes pushed for, work‐sharing. First, not all industries were highly competitive, a factor favoring hours' reductions. In those relatively protected industries, unions' bargaining power was quite high (see Abowd et al. 2005), another factor favoring unions' push for shorter hours. Potentially, the French preference for leisure was also high in comparison with other countries (even though we do not have any indisputable estimate of this “deep” preference parameter).

On the employment side, because the thirty‐five hours policy was accompanied by massive subsidies, both theoretical and empirical predictions are difficult. And because implementation was gradual, any serious evaluation is fraught with difficulties, in particular those related to the various selection biases due to firms' self‐selection into the program. Chapter 6, however, provides the most up‐to‐date evaluation of the thirty‐five hours policy, not only on employment but also on other firm‐level outcomes.

The main message may be that the thirty‐five hours policy was a massive shock, with lots of unintended and not‐well‐understood consequences. Many firms that decided to stay at thirty‐nine hours failed in the years following the policy. Was it purely due to increased costs? This is difficult to say at this point. But the question remains. Many firms that moved to thirty‐five hours also (p.110) failed after 2000. What was the role of this policy? Once again, it is difficult to be definitive. Some firms gained employment, in particular those that adopted the Aubry I framework. The estimates provided in the chapter show that a fraction was due to work‐sharing, another was due to decreased labor costs, but structural incentives and other types of help surely played a role that is difficult to assess. In addition, the death of competitors might have also played a part. Clearly though, these firms decreased their labor productivity as well as their total factor productivity, which is rarely a good sign in these years of intense international competition.

Another outcome that must be taken into account is the impact of the policy on workers' happiness. Very little information is available at this point. In the chapter, results on changes in work schedules are presented and show that they are now more irregular and more difficult, in particular for low‐wage or mid‐wage workers. Apparently, engineers, managers, and professionals were affected but might have benefited from the increased vacations that went with the thirty‐five hours policy.

At this point, caution is required when evaluating the recent wave of French work‐sharing. Serious analysis is difficult but the effects on the economy have been pervasive, potentially destructive, with large and unintended consequences mostly through an important reallocation of jobs across firms.

Chapters 7 and 8 present two cases that completely differ from those examined in the two previous chapters. Sweden and the Netherlands, respectively, are “small” countries in which there is a longstanding negotiation tradition. Let us start with Sweden.

The Swedish case suggests that strong unions and a labor movement with close ties to political power do not necessarily lead to large reductions in working hours in the way that was observed in France. Even “worse”, there was virtually no debate on work‐sharing in Sweden. Work‐sharing historically was never a prominent feature of the Swedish debate, nor a motivation for the reductions that were implemented over the years. It is also striking how little has happened to working hours in Sweden in recent decades. And the small changes that took place were always very flexible in their implementation. Instead the Swedes chose to give substantial subsidies to promote career interruptions. These interruptions matter. Indeed, even though legislated and contractual working hours are relatively long, such career interruptions make actual hours equal to the European average, but number of weeks per year the lowest in our Table 1.

Since working‐time reductions and leave subsidies both reduce the number of actual hours worked, they are likely to be competing policies. The theory chapter should help us understand this choice. Given the strong labor movement in Sweden, and its high degree of centralization, Chapter 4 tells us that unions should favor shorter hours. But because Sweden is a small open country, with many industries competing on the world market, Chapter 4 also (p.111) tells us that unions should favor longer hours (see the collective bargaining model and its lessons).

Although Swedish bargaining institutions changed substantially with the breakdown of central agreements in the early 1980s, the Swedish model was set up in order to let industries facing intense international competition govern wage rate increases with the explicit desire to preserve Sweden's international competitiveness. The expressed goal was to increase overall productivity, even at the cost of shutting down low‐productivity firms. For the labor movement, high productivity and high international competitiveness were always viewed as instrumental for building and keeping the Swedish welfare state.

A compromise had therefore to be found. Unions wanted shorter hours but not at the expense of productivity. Work‐sharing, in its simplest form, was not an option. The career‐break policies, on the other hand, were set up as part of the welfare state. When set up in the 1970s they were aimed at facilitating female labor force participation, and making it easier to combine work and family life. In particular, career breaks tried to facilitate sickness absence, childcare, or education. Indeed, they expanded alongside a growing female participation rate.

Chapter 7 shows that career breaks essentially have little impact on those who leave or on those who replace them. For the former, breaks essentially represent leisure and not an education opportunity. And the replacements appear to be unemployed workers with a relatively strong labor force attachment and reasonable reemployment opportunities. To summarize the chapter's conclusions, this policy had negative effects on participants' subsequent wages and is not likely to have contributed to the employability of the long‐term unemployed, in contrast to the initial intention.

In the Netherlands, another strategy was followed. This is described in Chapter 8. As mentioned above, standard hours fall within the realm of collective bargaining between workers and business unions. Hours for full‐time workers fell, as in many countries over the 1970s and 1980s, but remained roughly constant from the 1990s onward. The Wassenaar Agreement in 1982 constitutes the key moment in Dutch labor relations. During the years of recession, firms were looking for more flexibility when workers' unions, to preserve employment, agreed to wage moderation (no automatic adjustment of wages to inflation) in exchange for some work‐sharing. Work‐sharing was defined loosely but clearly included a shorter workweek, increased holidays, early retirement, and part‐time work. Between 1982 and 1985 hours were reduced by increasing holidays. But then the push for shorter hours appears to have stopped until the mid-1990s when unions and firms agreed on a thirty‐six hour week with full flexibility of hours and a reduced overtime premium.

As with all other countries examined, the impact of “straight” work‐sharing (i.e. going from, say, forty to thirty‐five hours) on employment appears (p.112) to have been essentially zero. All evaluations concur on this outcome, as described in the chapter. One reason is the wage compensation that took place: a decrease of 1 percent in working time being associated with an increase of 0.5 percent in hourly wages. This outcome did not escape the attention of unions, apparently. Hence, work‐sharing was mostly implemented through part‐time working.

The move to part‐time work was massive. For instance, between 1980 and 1984 the number of part‐time jobs soared from 132,000 to 829,000. Unions at first opposed part‐time working. But new female membership was also seen positively, most particularly in industries with a large proportion of women workers. And, in 2000, a law enacted the right to work part‐time for all (except for “compelling business reasons”). Now the Netherlands is the first part‐time economy in the world.

Women started to participate in the labor market relatively late, at least compared to other European countries. Factors explaining this late start are described in the chapter. The labor supply considerations involved are the main causes for specific Dutch developments. Once again, our theory chapter shows how supply can also play a role in the appeal and the shape of work‐sharing.

A central fact of Dutch life and values revolves around married women. For instance, the chapter reports that in the 1970s more than 60 percent of the Dutch found that it was troublesome for women to work outside the house (with children in daycare) until the beginning of the 1980s, when this proportion dropped to a relatively high level of 35 percent in 1997. The optimal number of children is four or more for 12 percent of Dutch while only 6 percent of Germans, French, Swedes, or Danes report this number. In their attitude to sharing of roles in the family, the Dutch resemble the Germans more than the Swedes or even the French. In addition, roughly 50 percent of women with young children stopped working after giving birth. Informal childcare arrangements appear to be favored by the Dutch (66 percent use grandparents, family, friends, babysitters, while 13 percent use daycare centers). This stands in stark contrast to countries like France, for instance, where formal daycare has long been favored by the government. And now for my favorite statistics: 70 percent of Dutch couples disagree that both men and women should contribute to household income. The equivalent proportions in West Germany, France, Sweden, and Denmark are respectively 31, 19, 12, and 27 percent. Preferences are clearly expressed. Clearly they matter to understand work‐sharing arrangements in the Netherlands.

Finally, a brief conclusion helps draw policy‐oriented lessons from these four countries' experience.


(1) IG Metall was by far the biggest union until 2001.