OECD Employment Outlook 2022

Building Back More Inclusive Labour Markets

Report

OECD Employment Outlook

9 September 2022

Available in:

English
français

Download PDF

5. Well-being, productivity and employment: Squaring the working time policy circle

Sandrine Cazes

Clara Krämer

Sebastien Martin

Chloé Touzet

Abstract

Working time is both a key element of workers’ lives and a production factor. Understanding how working time policy relates to well-being and economic outcomes is thus crucial to design measures balancing welfare and efficiency concerns. Evidence so far has largely focused on the use of maximum hours’ regulation to prevent detrimental effects on workers’ health, and the effect of normal hours reductions on employment levels. This chapter brings two new perspectives: first, it accounts for the fact that workers’ well-being is an increasingly central societal objective of working time policies, and therefore considers well-being effects alongside productivity and employment effects. Second, it accounts for the use of flexible hours and the development of teleworking in the aftermath of the COVID‑19 crisis and considers their impact on well-being, productivity and employment. Building on these analyses, the chapter discusses the potential of various working time policies to enhance non-material aspects of workers’ well-being such as health, work-life balance and life satisfaction while preserving employment or productivity.

In Brief

Working time is a key component of people’s working lives. Regulating its duration and its organisation is necessary to correct market failures leading to an inefficient allocation of working time and inadequate workers’ protection and to prevent negative externalities linked to long hours or variable schedules. Further, working time regulation can help − and historically has helped − enhancing non-material aspects of workers’ well-being. At the same time, working time being a production factor, policies affecting it will also impact employment, wages and productivity, and ultimately workers’ material well-being. On that basis, this chapter discusses the potential of various working-time policies to enhance workers’ well-being, while accounting for their possible effects on employment and productivity. Although data availability and heterogeneity across countries prevent generalisations, interesting insights emerge.

Working hours

The empirical literature suggests a close relationship between working long hours and poor health outcomes (particularly when workers have little control on their work schedule), but offers less-clear-cut results for other aspects of workers’ non-material well-being, such as life satisfaction. The literature moreover usually points to beneficial effects of reducing normal weekly hours on non-material aspects of workers’ well-being, if the reduction does not result in higher work intensity.
New empirical evidence for selected OECD countries confirms that working long hours (e.g. more than 45 hours per week) tends to be associated with a lower probability to report good health outcomes in the majority of selected countries. Yet, working a reduced amount of hours (e.g. less than 35‑30 hours per week) is not necessarily associated with a higher probability to report good health outcomes across countries. In fact, an inverse U-shape pattern emerges in Australia, Switzerland, the United Kingdom and pooled European data, where working less than 35‑30 hours is also associated with lower health outcomes. By contrast, the relationship between working hours and other non-material well-being outcomes is generally linear, e.g. working long hours decreases the probability that a worker is satisfied with her life, job, and free time, while working a reduced amount of hours increases these probabilities, except for France.

These results suggest that besides regulating maximum hours and overtime, a reduction of normal hours may also be considered as a possible lever of working time policy to enhance workers’ non-material well-being under certain conditions. In particular, such reductions in normal hours should be considered taking into account their potential impact on employment and productivity. To shed light on this, the chapter next analyses the effects of legislative reforms reducing normal hours on employment and productivity in European countries, as well as the relationship between episodes of reductions of contractual hours at firm level observed in the data and the growth in employment, average wage and productivity in Germany, Korea and Portugal.

Results from the analysis of legislative reforms implemented in Belgium, Italy, France, Portugal and Slovenia between 1995 and 2007 reveal a significant reduction of average yearly working hours for those who were affected by the reform, but no significant effects on employment, and similar −yet still insignificant effects− on wages and productivity. The absence of significant effect on employment may at least in part stem from heterogeneous effects cancelling each other at the aggregate level. Importantly, these reforms took place with constant monthly wages, thus leading to higher hourly wages, but they did not systematically include compensatory measures (such as e.g. public subsidies) for firms to limit possible adverse impacts of rising labour costs.
Firm-level analyses of the relationship between observed contractual hours reductions and economic outcomes in Germany, Korea and Portugal − point to contrasted results, but suggest that virtuous circles might exist in some instances, whereby the reduction in hours generates a productivity increase that limits the rise of unit labour cost and therefore prevents the potentially negative effect on employment growth. Understanding why such virtuous circles manifest in some cases and not in others should be investigated in future research, but could be explained by national differences in the institutional context of the decision-making process, notably well-functioning collective bargaining and strong social dialogue.

These two empirical approaches assess two different types of hours reduction. The first one looks at the employment and productivity effects of national legislative reforms generally applying to all firms and sectors and widely anticipated by employers. The second one focuses on contractual hours reductions at firm-level that might result from legislative reforms, collective bargaining or unilateral decisions from employers. Yet, despite their differences, the results emerging from these approaches are consistent and aligned with the majority of the existing literature. Reducing working hours (at constant monthly wage) might preserve employment on average if the impact on unit labour cost remains limited (either due to sufficient induced productivity gains or to public subsidies to affected firms/sectors). These results may also arise if the reduction of hours takes place in a pre‑existing situation of labour market monopsony (where the existence of a profit rent means that firms can absorb higher labour costs, see Chapter 3).

Outside the case of firms enjoying monopsony power, the results of this chapter point to the need to fully factor in the possible impact on unit labour costs when considering reductions of normal hours. This could be done through dedicated accompanying measures, or by designing the reduction so that it taps into the productivity-enhancing potential of the reform. Careful attention should also be devoted to the timing, conditions of implementation and scope of the reduction which are all likely to influence the effect of the reduction.

Flexible hours and teleworking

Workers’ ability to work flexible hours, i.e. to autonomously decide their starting and finishing times, is associated with better non-material well-being for all outcomes considered – both in the literature and in new individual-level evidence available for Australia, Germany, Korea, Switzerland and the United Kingdom (although to varying degrees between countries). The literature to date also points to positive associations with employment, wages and productivity. New evidence on German firms adopting flexible hours suggests that this arrangement might indeed boost employment without significantly affecting productivity per worker. Firms choosing flexible hours also see a decrease in average wage growth – suggesting a possible trade‑off between wage increases and higher autonomy in determining hours.
In contrast to flexible hours, the link between teleworking and workers’ non-material well-being varies for different outcomes and across countries – both in the literature and in the new empirical evidence presented for Australia, Switzerland and in the United Kingdom. Empirical results show a negative association with self-assessed health, positive associations with life‑ and job satisfaction and contrasting associations with work-life balance, which is particularly high for teleworkers in Australia, but very low for teleworkers in Switzerland. As for productivity and employment, associations with teleworking in the empirical literature to date are generally positive, especially in terms of attracting and retaining workers, as well as increasing female labour force attachment.

Introduction

Working time is a defining aspect of working lives.1 How many hours workers spend at work, how their working hours are scheduled, and how much control they have over them can affect their physical and mental health, work-life balance, job satisfaction and performance. More generally, working time directly affects workers’ allocation of time between work and other activities, such as leisure, which itself is likely to influence their life satisfaction. At the same time, working time is a key production factor that can affect economic outcomes such as employment, productivity and wages, which in turn impact workers’ material well-being. Therefore, understanding how working time policy relates to workers’ well-being and economic outcomes is crucial to identify and carefully design measures balancing welfare and efficiency concerns.

Regulating working time duration and organisation is necessary to correct possible market failures (due e.g. to asymmetry in market power between workers and employers) leading to an inefficient allocation of working time and an inadequate protection of workers’ health and work-life balance, and to prevent negative externalities linked to excessive working hours or variable schedules. Historically, it has also helped enhancing several aspects of workers’ well-being, notably through regulations reducing working time. Yet, this historical trend towards shorter working hours which has been accompanied by productivity gains and could be traced back to the 19th century in most OECD countries has considerably slowed down – if not almost halted in a number of countries (OECD, 2021[1]). While working time is regulated at various levels across OECD countries, statutory regulations on working time have the most effect on actual working time in OECD countries, even where derogations at lower levels of governance are possible (OECD, 2021[1]).

Policy makers interested in using working time measures as a lever to influence workers’ well-being outcomes have several tools at their disposal: they can regulate the maximum number of hours that a worker can legally work in a given period of time and define a premium wage for overtime work; they can regulate the number of normal hours regarded as representing a full-time job; they can allow for greater flexibility in working time arrangements and provide or modify incentives for using various working time arrangements – see OECD (2021[1]). The pros and cons attached to each of these tools, and how they might affect workers’ well-being as well as employment, wages and productivity need to be factored in when designing working time policy.

Policy debates and related empirical evidence on working time policy so far have generally focused on the regulation of maximum hours to prevent any detrimental effects on workers’ health, and on the reduction of normal weekly hours, with a view to increasing employment. This chapter brings in two new perspectives: first it accounts for the fact that workers’ well-being is an increasingly central societal objective of working time policies, and therefore considers well-being effects alongside productivity and employment effects. In particular, in line with the OECD well-being framework, it distinguishes material aspects (earnings, job status, etc.) and non-material aspects (health, work-life balance, life satisfaction, etc.) of workers’ well-being (OECD, 2015[2]). Second, it accounts for the use of flexible hours and the development of teleworking, given its prevalence and relevance in the aftermath of the COVID‑19 crisis, and considers the impact of these schemes on non-material well-being, productivity and employment. Identifying virtuous circles between welfare and efficiency objectives could help square the working time policy circle.

The chapter starts by exploring the relationship between working time (maximum and normal hours, part-time, flexible hours, and teleworking2) and a set of selected measures of non-material well-being, namely health status (both mental and physical), work-life balance and life and job-satisfaction. Drawing on a combination of literature reviews and on analyses of individual-level data, it first investigates how hours worked, flexible hours arrangements, part-time work and teleworking relate to the above‑mentioned non-material well-being outcomes, to identify potential levers of well-being enhancement (Section 5.1). As results suggest that reducing normal weekly hours and fostering the use of flexible hours and teleworking might in some circumstances lead to well-being gains, the chapter next turns to analysing the impact of these policies on employment, wages and productivity (Section 5.2). To shed some light on these key issues, Section 5.2 next analyses the effects of national legislative reforms that reduced normal weekly hours on employment and productivity in various European countries, before studying the relationship between concrete episodes of reductions of contractual hours at firm level and the growth of productivity, average wage and employment, in Germany, Korea and Portugal where data are available. Finally, the chapter concludes by bringing together all the results and discussing policy options while outlining the importance of timing, scope and careful design and implementation.

5.1. Working time and workers’ non-material well-being

The amount of time spent at work, how hours are scheduled and the relative flexibility workers have in determining these schedules – see OECD (2021[1]) – have direct implications for several outcomes of workers’ non-material well-being, such as health status, work-life balance and life and job satisfaction. Working time policy might be able to improve these outcomes. Drawing on a mix of literature reviews and new analyses using individual-level data in OECD countries, this section explores the relationship between working hours (both normal hours and overtime), flexible hours, part-time work and teleworking and a set of non-material well-being measures (health status, work-life balance and job- and life satisfaction), to identify possible levers of well-being gains.

5.1.1. How do working hours relate to workers’ non-material well-being?

The relationship between time spent working and workers’ well-being (both material and non-material) has initially been investigated in the epidemiology and occupational health literature that assesses the effects of working long hours3 on both mental and physical health (Beswick and White, 2003[3]; Sparks et al., 1997[4]). This literature is plagued by the identification problem known as “the healthy worker effect” – a problem of reverse causality when assessing the impact of working time on health, since healthy workers are more likely to be in employment and to be able to work long hours than unhealthy ones – and by the difficulty of dealing with unobserved confounding factors.4 Nonetheless results usually suggest a close relationship between long hours and poor health outcomes. Working long hours and overtime are associated with unhealthy behaviours, such as alcohol consumption, smoking and lack of exercise (Ahn, 2016[5]). Long hours are also directly related to poor physical health outcomes, such as cardiovascular diseases or stroke (Kivimäki et al., 2015[6]) and considered as one of the major risk factors for workplace accidents (Dembe, 2005[7]; Vegso et al., 2007[8]). Beyond physical health and workers’ safety, long working hours are also associated with stress, depression, and suicidal ideation in young Korean employees (Park et al., 2020[9]) and also more generally with negatively impacted cognitive functions (Virtanen et al., 2008[10]). Research has also explored the relationship between long hours and other well-being outcomes, such as life satisfaction, and found less clear-cut results. Hamermesh at al. (2017[11]) find for instance beneficial effects of overtime reduction on workers’ life satisfaction in Japan and Korea; but other studies find that long hours are not necessarily related to lower well-being outcomes, notably for men or fathers (Hewlett and Luce, 2006[12]; Gray et al., 2004[13]). On the other side of the hour spectrum, working a low number of hours also has an impact on workers’ well-being, mainly because it results in insufficient earnings – see for example Friedland and Price (2003[14]) and Heyes and Tomlinson (2021[15]). This idea is also mentioned in the literature on involuntary part-time work, which is further discussed in the next section on working time arrangements.

Beyond the negative well-being impacts of both long and insufficient working hours, authors have also explored the relationship between a reduction of normal weekly working hours and well-being outcomes. While results vary by outcomes considered, the scope of the hours reduction and the extent to which wages are adjusted, studies find that reducing hours tend to positively affect non-material well-being. Lee and Lee (2016[16]) exploit a quasi-natural experiment in Korea, where normal hours were reduced gradually from 44 to 40 hours at different times by industry and establishment size between 2004 and 2011, and find that on average a one‑hour reduction in normal weekly working hours in Korea significantly decreases the injury rate by about 8%. Berniel and Bietenbeck (2020[17]) provide causal evidence on smoking reduction and lower body mass index for France in the context of the reduction of the 35 hours reform. Lepinteur (2019[18]) shows beneficial effects of normal hours reduction on job and leisure satisfaction of workers in France and Portugal, especially for women and workers with heavy family burden. However, other studies point to less clear effects on well-being should the working time reduction result in a higher time pressure on workers (Askenazy, 2004[19]). Rudolf (2013[20]) finds for instance that a reduction of normal hours in Korea did not have the expected positive impact on workers’ job and life satisfaction and suggests that the reduction in hours was offset by greater work intensity.

Further, other factors beyond work intensity are found to interact in the relationship between working time and workers’ non-material well-being, including workers’ control of their schedules and the mismatch between their desired and actual working hours. The extent to which workers can choose or control the number of hours they work is key in determining how detrimental long hours might be for their health (Bassanini and Caroli, 2015[21]; Bell, Otterbach and Sousa-Poza, 2012[22]; Burke et al., 2009[23]; Caruso et al., 2006[24]; Frijters, Johnston and Meng, 2009[25]). Salo et al. (2014[26]) for instance find that for those working 40 hours a week, less control over working time is associated with greater sleep disturbances in Finland (while sleep disturbances were high irrespective of the degree of workers’ control for those working longer hours). Looking then at the link between working hours mismatch (i.e. the difference between workers’ preferred working hours and their actual hours) and job satisfaction, Grund and Tilkes (2021[27]) find a negative link between working time mismatch and a positive − moderating − link between working time autonomy and job satisfaction in Germany. Moreover, Holly and Mohnen (2012[28]) find that the desire to reduce hours has a negative impact on satisfaction although if overtime is appropriately compensated, satisfaction rises and working time mismatch decreases.

In order to shed further light on the results of this literature review, Figure 5.1 and Figure 5.2 present new OECD empirical evidence exploring the relationship between actual weekly hours in the main job and several measures of workers’ non-material well-being, namely self-assessed health outcomes, life and job satisfaction and satisfaction with free time (as a proxy for work-life balance). Pooled results for European countries are based on European Social Survey (ESS) data, while results for Australia, France, Germany, Japan, Korea, Switzerland and the United Kingdom draw on country-specific, individual-level panel data. Results presented in the figures correspond to the marginal effect of working less (or more) than a particular threshold, compared to those working more (or less) than this threshold: for instance, the left light blue bar in the first graph in Figure 5.1 corresponds to the difference in likelihood (in percentage) of being satisfied with one’s health when working less than 20 hours per week, compared to when working more than 20 hours per week in European countries represented in the ESS data.

In terms of workers’ health, (Figure 5.1), results generally confirm the negative relationship found in the literature between working long hours and poor health outcomes in the majority of selected countries. Working more than 45 hours reduces one’s probability to be satisfied with one’s health in Australia, Germany, Switzerland and Japan (for those working more than 50 hours). Those working more than 45 hours are also less likely to report facing no limitations in their work due to health problems in Germany, Switzerland, the United Kingdom, and countries covered by the ESS data (for those working more than 55 hours). At the same time, there is no significant effect of working long hours on workers’ health when it is measured by health satisfaction in the ESS data, France, Korea and the United Kingdom, or by health-related limitations in Australia and Korea. Surprisingly, working more than 45 hours even increases the probability of reporting no health-related limitations in France – which might be due to the self-selection issues discussed in note 4, survey biases, or cultural differences affecting subjective well-being survey items differently in different countries.

The relationship with health outcomes is however less clear-cut on the other side of the hours spectrum, and working a short amount of hours (starting from less than 35 hours in some cases) is not associated with a linear improvement of workers’ health across countries. The probability to be satisfied with one’s health is higher for those working less than 30, 25 or 20 hours in Germany, compared to those working more than these thresholds. Health satisfaction is not significantly related to any of the short hours’ threshold in Australia. By contrast, the probability to be satisfied with one’s health is lower for those working less than 25 hours in France and Korea, 30 hours in the ESS data and Switzerland and less than 25 hours in the United Kingdom, compared to those working more than these respective thresholds. Similarly, workers doing less than 35 hours a week in the ESS data and France, 30 hours a week in Australia and Switzerland and less than 25 hours a week in the United Kingdom are less likely to declare facing no health-related limitations, compared to those working more than these respective thresholds (while the relationship is not significant in Germany). Of course, these results at the bottom of the hours distribution could be due to some form of healthy worker effect: workers in poor health may be more likely to work fewer hours.

Overall, these findings primarily emphasise heterogeneity across OECD countries. Yet they also confirm the existence of a link between long hours and poor health outcomes in the majority of the selected countries. In addition, they reveal that health outcomes are not linearly related to hours, and not always improving for those working shorter hours. Rather, an inverted U-shaped pattern emerges in some countries (ESS data, Australia, Switzerland and the United Kingdom) when considering health outcomes, with a lower likelihood to be satisfied with one’s health, and to declare no health-related limitations at both ends of the spectrum.

Figure 5.1. Self-assessed health by working hours thresholds in selected OECD countries

In terms of other non-material well-being outcomes, Figure 5.2 shows the marginal effect of working less than (or more than) particular thresholds on the likelihood of being satisfied with one’s life, job, and free time, the latter as a proxy for work-life balance (effects on different outcomes are tested separately). Results are more linear than for health outcomes, with long hours reducing the probability to be satisfied with all three outcomes (e.g. job, life or free time), and short hours increasing these probabilities in most countries. In particular, the probability to be satisfied with one’s free time is higher for those working less than 30 hours (in Australia and Japan) and less than 35 hours (in Germany, Switzerland and the United Kingdom), while it is lower for those working more than 45 hours (Australia, France, Germany, Japan, Switzerland and the United Kingdom). As for life‑and job satisfaction, relationships generally follow a similar pattern but the marginal effects of working shorter hours are generally smaller and less significant. France is again an outlier in that regard, since the marginal effects of working shorter hours show a reverse pattern: the probability to be satisfied with one’s job, life or free time is lower for those working less than 30 hours (and less than 35 for job satisfaction) compared to those working more than these thresholds.5 Another outlier is Korea, where people working shorter hours have a lower probability to be life‑satisfied, and those working long hours a higher probability to be job satisfied, which might again be due to cultural differences affecting subjective well-being survey items differently in different countries.

Figure 5.2. Life, job and free time satisfaction by working hours thresholds in OECD countries

Finally and in line with the literature, OECD estimates available for Australia, Germany and Switzerland also reveal a significant negative relationship linking the mismatch between workers’ preferred working time and their actual working time on the one hand, and the selected measures of non-material well-being on the other hand. Interestingly, this negative relationship is mostly driven by those wanting to work less rather than more: evidence shows that the marginal effects of working more hours than one would like to (excessive hours) are negative for all non-material well-being outcomes, while the marginal effects of working less hours than one would like to (insufficient hours) are also negative but smaller for life and health satisfaction, and are positive for job satisfaction and work-life balance (Figure 5.3). While the data for Australia and Germany in this analysis are based on a precise survey question that asks respondents their preference while stating that their income would be unaffected, this precision is missing for Switzerland. This might bias estimations downward for Switzerland compared to Australia and Germany, if most workers assume that working less would come with a pay cut. The limits inherent to a fixed effects regression analysis also apply, which calls for caution in causally interpreting the results, as the analysis cannot address selection effects, e.g. the fact that workers with different life‑ and health satisfaction might select into jobs with different normal hours.

Figure 5.3. Working time mismatch and non-material well-being in selected OECD countries

5.1.2. Flexible working time arrangements and workers’ non-material well-being

In contrast, the literature generally6 points to positive effects on non-material well-being of working time arrangements that provide employee‑oriented flexibility, namely flexible hours (e.g. an arrangement whereby workers decide their starting and finishing times), teleworking and, to a lesser extent, part-time work – highlighting again the importance of workers’ control over their schedules as an important factor for their well-being. The underlying mechanism is twofold: on the one hand, flexible working time arrangements help reconcile work with private life and, in the case of flexible hours and teleworking, also coping with job demands and increasing autonomy. Teleworking additionally reduces commuting time. On the other hand, flexible working time arrangements may increase work intensity, (unpaid) overtime hours and work-life conflict (Tucker and Folkard, 2012[29]; Hurtado et al., 2015[30]; Tavares, 2017[31]; Charalampous et al., 2019[32]; Samek Lodovici et al., 2021[33]). Which of these mechanisms outweighs the other likely differs between groups of workers and work contexts, but some patterns emerge from the – mainly correlational – empirical evidence to date.

Overall, the non-material well-being effects of flexible hours, tend to be largely positive − see for example the review by Tucker and Folkard (2012[29]). Moen et al. (2011[34]) for instance find that the introduction of flexible working hours in an experimental setting in the United States improved workers’ health, because it enabled them to get more and better sleep, reduced the postponement of doctors’ appointments and increased the time workers spent on physical activity. Measures of life‑ and job satisfaction are also reportedly higher for workers with flexible hours in Europe and the United States (Atkinson and Hall, 2011[35]; Golden, Henly and Lambert, 2012[36]; De Menezes and Kelliher, 2017[37]; Angelici and Profeta, 2020[38]; Kröll and Nüesch, 2019[39]). At the same time, some studies report none or negative effects, mainly because they find that flexible hours are linked to increases in working hours, particularly for men (Lott and Chung, 2016[40]; Krug, Kemna and Hartosch, 2019[41]), and increases in work-life conflict, particularly for women (Kim et al., 2020[42]). Importantly however, such negative side effects may diminish when analysing flexible hours in connection with supporting policies such as parental leave (see for example, Wanger and Zapf (2021[43])).

Contrary to flexible hours, the use of teleworking spread only recently because of COVID‑19‑induced lockdown measures in most OECD countries − for a detailed overview, see OECD (2021[1]) – but is often linked to flexible hours as a package deal. Since hybrid arrangements mixing teleworking and work in the office are likely to stay,7 research increasingly investigates the effects of teleworking on well-being during the COVID‑19 pandemic. Yet, drawing on resulting evidence would be ambivalent as a number of confounding factors are at play (see Box 5.1). Pre‑pandemic evidence suggests that the impact of teleworking on workers’ non-material well-being is generally positive but more mixed than for flexible hours − see for example the reviews by Tavares (2017[31]) and Charalampous et al. (2019[32]). Henke et al. (2016[44]) find for instance that teleworking improves a number of health outcomes in the United States, such as lower risks of obesity, alcohol abuse, physical inactivity, tobacco use and depression. Teleworking has also positive effects on work-life balance, but mainly if it is occasional and home‑based (instead of highly mobile) (Kim et al., 2020[42]; Rodríguez-Modroño and López-Igual, 2021[45]; Pabilonia and Vernon, 2022[46]). This is because the resulting regularity mitigates some of the negative consequences of teleworking on work-life balance through increased working hours and intensity – as found for instance by Felstead and Henseke (2017[47]) and Song and Gao (2020[48]). The beneficial effects of teleworking also appear to be at least partially mediated by workers’ attitude towards teleworking (Adamovic, 2022[49]) and perceived autonomy (Gajendran and Harrison, 2007[50]), which is found to decrease stress and buffer teleworking-induced increases in work intensification (Curzi, Pistoresi and Fabbri, 2020[51]). The moderating effect of autonomy on teleworkers’ non-material well-being should be contrasted with the risks of new supervision mechanisms, for instance in the form of surveillance software, being deployed to compensate for the lack of physical supervision, and their possible adverse effect on privacy, autonomy and ultimately well-being.

In terms of commuting time, Frazis (2020[52]) and Pabilonia and Vernon (2022[46]) estimate that teleworking saves workers in the United States an hour to 75 minutes per day of commuting and grooming time, which they instead spend on leisure. While objective health measures (e.g. diagnosed health problems) are barely affected by commuting, subjective health measures (e.g. self-perceived health satisfaction and status) are clearly higher for those commuting less, particularly for women and those commuting by car (Künn-Nelen, 2016[53]). Giménez-Nadal et al. (2019[54]) find that the saving in commuting time also improves life satisfaction, but with larger increase for men than for women, one potential reason being that the former use their saved time primarily on leisure, while women also increase their household production on a workday (but not over the entire work week) – at least according to time‑use data from the United States (Pabilonia and Vernon, 2022[46]). This is in line with findings from Arntz et al. (2019[55]) in Germany and Song and Gao (2020[48]) in the United States, who find positive and non-negative teleworking effects on life satisfaction only for men and women without children.

Box 5.1. Mixed effects of teleworking on well-being during the COVID‑19 pandemic

The outbreak of the COVID‑19 pandemic in spring 2020 led to a massive shift to teleworking, and an increasing number of studies make use of this exogenous shock to analyse the link between teleworking and workers’ well-being. Yet, COVID-induced restrictions significantly affected both the experience of teleworking and workers’ well-being, thus results from these studies cannot simply be extrapolated to post-pandemic teleworking arrangements. One important issue is that teleworking during COVID‑19 was a forced experiment. Yet, pre‑pandemic evidence suggests that teleworkers’ well-being is higher in occasional and voluntary arrangements (Rodríguez-Modroño and López-Igual, 2021[45]; Adamovic, 2022[49]). Moreover, COVID‑19‑induced teleworking was widespread, concerning also occupations for which it is feasible but suboptimal – see e.g. Eurofound (2021[56]) while the support of colleagues physically co-located in the office can be important to reap the well-being benefits of teleworking (Raghuram et al., 2019[57]), such support was often lacking during the pandemic. The full-time and widespread nature of teleworking during the pandemic also exacerbated risks of work-life conflict, as some had to telework in limited physical space, with insufficient technical equipment, and with other household members also teleworking or following distance schooling (DeFilippis et al., 2020[58]; Bertoni et al., 2021[59]). Finally, the shift to teleworking happened abruptly in many workplaces, without much consideration for health and safety requirements that would otherwise apply (ILO, 2020[60]). Because of this, workers also faced an unprecedented challenge in quickly adapting to teleworking, for example by learning new IT skills, which is a source of mental distress particularly for senior workers (Bertoni et al., 2021[59]).

Figure 5.4. Confounders of teleworking and well-being during COVID‑19

Against this backdrop, a few studies have already attempted to isolate the effect of teleworking on worker’s well-being from that of other confounders, finding mixed and heterogeneous results for different groups of workers. Sasaki et al. (2020[61]) find positive effects of teleworking on workers’ psychological distress in Japan, but their cross-sectional data is very limited. Using email meta-data from over 3 million workers worldwide, DeFilippis et al. (2020[58]) find an increase in the average workday span, but their analysis is subject to aggregation bias and has unclear implications for workers’ well-being. This is in line with a recent online Eurofound survey (2021[56]), in which over one‑fifth of teleworkers reportedly worked during their free time every or every other day during the pandemic, but at the same time appreciated the absence of commuting to the office; spending more time with their children and spouses; and the flexibility of working hours. Using longitudinal European data, Bertoni et al. (2021[59]) find positive effects of teleworking on mental health only for men and women with no co-residing children.

While flexible hours and teleworking are compatible with full-time employment, part-time work by definition is not. In this respect, part-time jobs in most OECD countries tend to be associated with many labour market disadvantages including lower income, lower job security and reduced access to unemployment benefits, training and promotion (OECD, 2020[62]), which are important factors for job-quality and well-being (Cazes, Hijzen and Saint-Martin, 2015[63]). On the one hand, the disadvantages associated with working part-time appear to be compensated by better health and work-life balance – see for instance the OECD Employment Outlook (2010[64]). On the other hand, part-time workers tend to work more unpaid overtime hours relative to full-time workers (Fernández-Kranz and Rodríguez-Planas, 2011[65]; Chung and van der Horst, 2020[66]), which may hamper some of the non-material well-being effects associated with part-time. More recent evidence confirms positive effects of part-time work on both objective and subjective health measures in the United States and the United Kingdom (Benson et al., 2017[67]; Cho, 2018[68]), and on workers’ satisfaction with work-life balance, but primarily in more gender egalitarian countries (Beham et al., 2019[69]) or where part-time work is more likely to be the norm (Nikolova and Graham, 2014[70]).8 Yet in practice, part-time work is not the norm in most OECD countries, where women make up the vast majority of part-time workers (OECD, 2021[1]) and experience negative impacts on their career progression as a result (OECD, 2018[71]).

Finally, a crucial factor ensuring positive well-being effects of flexible hours, teleworking and part-time work is that they are adopted voluntarily (Joyce et al., 2010[72]; Nikolova and Graham, 2014[70]; Pirani, 2015[73]; Bell and Blanchflower, 2019[74]; Adamovic, 2022[49]). Moreover, workers may have different reasons as to why they voluntarily take up flexible working time arrangements, which can impact well-being differently and be shaped by employers’ reasons to offer these arrangements in the first place. Scholars have pointed out for instance that flexible arrangements lead to more negative side effects like increased overtime hours if they are primarily offered to cut costs or incentivise workers to increase their performance (Chung and van der Horst, 2020[66]). Along these lines and beyond the firm level, promoting part-time work for instance is not only part of countries’ efforts to help workers reconcile work with private life, but also to reduce unemployment and increase labour market flexibility in low-paid occupations (Carrillo-Tudela, Launov and Robin, 2018[75]; Biewen, Fitzenberger and de Lazzer, 2018[76]; Barbieri et al., 2019[77]). Such and other forms of involuntary part-time work can be problematic, because they not only hamper well-being through the lower living standards resulting from income losses associated with part-time work (Bell and Blanchflower, 2019[74]), but also prevent any of the offsetting effects on health and work-life balance discussed above. Those who take‑up a part-time job but would prefer to work more are especially likely to experience negative well-being effects, as insufficient working hours negatively affects their material well-being as discussed in the previous section. Moreover and related to the gendered nature of part-time jobs, women tend to be more constrained in their adoption of flexible working time arrangements, having to opt most often for part-time work, while men tend to be able to use flexible working time arrangements with a greater degree of choice, and to opt most often for flexible hours (Wheatley, 2017[78]).

New OECD individual-level evidence presented here (Figure 5.5 and Figure 5.6) explores the relationship between three flexible working time arrangements that promote employee‑oriented flexibility (part-time, flexible hours and teleworking) and the same aspects of workers’ non-material well-being than above (e.g. health, work-life balance and job-and life satisfaction). As data are only available for three to seven OECD countries depending on the working arrangement considered (Australia, France, Germany, Japan, Korea, Switzerland and the United Kingdom), caution is needed in generalising the results. Nonetheless, they point to interesting results. First, the results confirm the general patterns in the literature: out of the three working time arrangements considered, flexible hours are positively associated with all non-material well-being outcomes, namely self-assessed health, life and job satisfaction, and work-life balance (proxied by satisfaction with free time in Japan and the United Kingdom). Second, the relationship between teleworking and non-material well-being is more mixed, indicating a negative association with self-assessed health, small but positive associations with life‑ and job satisfaction and contrasting associations with work-life balance: while work-life balance is particularly high for teleworkers in Australia, it is particularly low in Switzerland. Finally, both voluntary and involuntary part-time work are negatively associated with all non-material well-being indicators. Interestingly though, distinguishing voluntary part-time workers into those who simply prefer it over full-time work and those who (have to) opt for it because of caring reasons reveals that the latter is associated with negative impacts on well-being, while truly voluntarily adopted part-time work is associated with high well-being. Such granular information is not (yet) available in many surveys and in any case not regarding teleworking and flexible hours, but points to a very important avenue of future research.

Figure 5.5. Self-assessed health by flexible working time arrangements in selected OECD countries

Figure 5.6. Life‑job satisfaction, work-life balance and satisfaction with free time by flexible working time arrangements in selected OECD countries

5.1.3. Scope for improvement? Working time policy levers to enhance workers’ non-material well-being

Results from the literature and new OECD empirical evidence on working time and workers’ non-material well-being presented in the previous paragraphs suggest that some levers of working time policies exist that might enhance workers’ non-material well-being, such as policies regulating working hours (maximum and normal). While limits on maximum hours and overtime are already in place in most OECD countries to prevent their detrimental effect on workers’ health (OECD, 2021[1]), the regulation of normal weekly hours has less often been considered as a potential instrument to foster workers’ well-being. Yet, available evidence on the link between actual working hours and various non-material well-being outcomes presented above cautiously suggests that a reduction of normal weekly hours could enhance workers’ non-material well-being. Other options to improve workers’ non-material well-being discussed above include flexible hours, teleworking and part-time work. Yet, as shown in Figure 5.5 and Figure 5.6, part-time, even when voluntary, might be associated with negative well-being outcomes, in cases where it is chosen for caring reasons – which is likely to be the case for a large proportion of female workers in particular. In addition, the already existing extensive research on part-time work also suggests that even voluntary forms have limited potential for increasing workers’ non-material, let alone material well-being. By contrast, results in Figure 5.5 and Figure 5.6 suggest that flexible hours might be a more promising means of improving workers’ non-material well-being – and one that has been less researched so far.

Beyond assessing their impact on non-material well-being, the effect of these policy options on employment and productivity should also be evaluated, since these two outcomes have ripple effects on workers’ material well-being. A crucial element to consider in this analysis is the extent to which a reduction in normal working hours would maintain the same monthly/weekly income for workers, thus inducing an increase of hourly pay and potentially on labour cost if increases in hourly productivity do not offset increases in pay. Effects on employment levels should also be carefully assessed.

The remainder of this chapter sets out to investigate the effect of normal hour reductions and flexible hours on employment and productivity. While the effect of teleworking on non-material well-being outcomes is less clear-cut, its effect on employment and productivity are also evaluated, on account of its increased prevalence and relevance in the aftermath of the COVID‑19 crisis – and since teleworking and flexible hours often come as a package deal.

5.2. Working time, productivity and employment

In order to carefully discuss the feasibility of the policies identified above as potentially well-being enhancing, this section starts by presenting comprehensive literature reviews on the employment and productivity effects of changes in normal hours. This assessment of the literature is complemented by new evidence analysing the effects of national legislative reforms reducing normal hours in European Union countries and of firm-level episodes of contractual hours reductions in Germany, Korea and Portugal. This two‑pronged empirical approach helps understanding the effect of concrete episodes of hours reductions implemented in different ways. Finally, the section reviews the literature on the employment and productivity effects of flexible hours and teleworking (the latter, as explained above, on account of its increase prevalence in the aftermath of the COVID‑19 crisis), and presents new evidence on the productivity and employment effect of flexible hours in German firms.

5.2.1. The economic impact of reducing normal hours: A literature review

The employment effect

This section presents a summary of the most salient theoretical arguments and of the most robust empirical findings – a more comprehensive literature review is available in Annex Table 5.C.1. Theoretical predictions on the effect of reducing normal hours on employment depend on the underlying mechanisms and assumptions at play on the labour demand side. In this respect, two factors are of particular importance: whether the reduction of hours takes place at constant monthly (or annual) pay – which would lead to a rise in hourly labour cost, and could have adverse effect on employment − or not, and whether hourly productivity gains may be generated and mitigate this potential detrimental employment effect.

Theoretical papers for instance generally assume that working time reductions take place at constant monthly (or annual wage).9 Under this assumption, a reduction of normal hours has an ambiguous effect on employment.10 Simplified versions of the main arguments are as follows (see e.g. Kapteyn et al. (2004[79]) for a more thorough review). Following a simple logic, one could assume that in firms not usually resorting to overtime (i.e. firms where the pre‑reform normal working time was equivalent to the optimal working time), reducing normal hours could incentivise firms to hire more workers in order to meet orders, thus leading to a positive effect on employment. Yet, this logic11 assumes that the optimal working time remains the same after the change, and that hours and workers are substitutable (notably ignoring the fixed costs associated with each additional worker). In firms already using overtime before the reduction in normal hours, the marginal cost of hiring an additional worker goes up after the change (since a larger proportion of her time now has to be paid the overtime premium), while the marginal cost of an additional hour is left unchanged: to compensate for the reduction in normal hours these firms might then choose to pay for more overtime rather than hiring new workers, leading thus to a negative effect on employment (Cahuc et al., 2014[80]; Calmfors and Hoel, 1988[81]).12 More generally, the increase in the hourly labour cost following a normal hours reduction could lead firms to substitute capital for labour, leading to a reduction in employment. However, higher hourly pay could be compensated by gains in hourly productivity induced by the reduction in hours – for instance through productivity-enhancing organisational changes, higher investment, the recruitment of more productive workers, or through labour supply responses (more rested workers could have a higher hourly productivity). Gains in hourly productivity would at the same time limit the negative effect on employment, but also suppress the incentives to hire more workers, therefore preserving employment.

Turning to empirical results, purely correlational studies (i.e. studies that do not account for any possible endogeneity, and that focus on measuring the statistical significance of covariations13) tend to yield mixed results, ranging from studies finding a negative impact of hours reduction on employment (Steiner, Peters and Steiner, 2000[82]; Sagyndykova and Oaxaca, 2019[83]), to the majority of correlational studies finding non-significant effects (Andrews, Schank and Simmons, 2005[84]; Hunt, 1999[85]; Trejo et al., 2016[86]; Kramarz et al., 2008[87]; Brown and Hamermesh, 2019[88]),14 to those finding a positive effect (Fiole, Roger and Rouilleault, 2002[89]; Husson, 2002[90]; Kapteyn, Kalwij and Zaidi, 2004[79]). Among authors using a quasi-causal research design (which, by contrast to purely correlational ones, aim to account for some forms of endogeneity, although they do not correct for all of it), Crépon and Kramarz (2002[91]) find a negative effect of the 1996 statutory reduction of working time from 40 to 39 hours in France on employment. Raposo and van Ours (2010[92]) find that the reduction of working hours in Portugal decreased the separation rate of workers affected by the working time reduction. Crépon et al. (2004[93]) find that employment increased in firms reducing their hours in France (they argue that at least part of this increase is likely to be driven by a concomitant reduction in social security contributions and to wage restraint, rather than by the hours reduction – although on this issue, the meta‑analysis by Gubian et al. (2004[94]) attributes a larger positive effect to the reduction itself). Finally, a majority of quasi-causal studies finds non-significant results – see e.g. (Estevão and Sá, 2006[95]; Costa, 2000[96]; Skuterud, 2007[97]; Sánchez, 2013[98]; Chemin and Wasmer, 2009[99]; Kawaguchi, Naito and Yokoyama, 2017[100]).

Of course, different studies are based on the analysis of different reforms and/or contexts. Hence, differences in results might be due to differences in the parameters of the reforms analysed, such as their size and starting point, and their implementation. Similarly, non-significant results in country-specific analyses could stem from heterogeneous effects in the pool of firms observed. Hence, while the review of existing literature presented above suggests that in most cases, there were no significant effect on employment, it does imply that a reduction of normal hours should not be considered without paying careful attention to its design and implementation.

As explained above, the theoretical prediction that reducing normal hours might have adverse effects on employment rests on two assumptions: first, that monthly (or annual) wages are kept constant; second, that hourly productivity does not increase sufficiently to keep unit labour cost approximately constant. The non-significant results observed in many empirical papers could be explained by the fact that either of these assumptions does not hold in practice.15 Regarding the first assumption, two of the papers reviewed in Annex Table 5.C.1 that use a quasi-causal research design and consider wages as an outcome indeed find evidence of wage cuts or wage restraint (meaning that wage growth was slowed down): Sanchez (2013[98]) in the case of Chile, and Crépon, Leclair and Roux (2004[93]) in the case of France. However, all other papers find that reducing working hours increased hourly wages, but without negatively affecting employment (Estevão and Sá, 2006[95]; Raposo and van Ours, 2010[92]; Kawaguchi, Naito and Yokoyama, 2017[100]). One possible explanation for the results of this second group of papers is that the second assumption does not actually hold and that hourly productivity may have increased sufficiently to maintain unit labour cost approximately constant. This possibility is considered in the literature review on productivity effects below (Sections 5.2.2 and 5.2.3 then present new evidence on this issue).

Another potential explanation for studies finding no negative effect on employment despite an increase in hourly labour cost is that the hours reduction takes place in a context where wages have not fully adjusted to past productivity growth: in that situation, firms can absorb higher labour costs while preserving employment thanks to their accumulated rent. Such rents can typically exist in monopsonistic labour markets. In these contexts, characterised by an asymmetry in market power between employers and workers leading to an inefficient allocation of working time, or a suboptimal wage growth, a reduction in hours inducing a rise in hourly wage can in fact have a similar impact as a minimum wage increase in standard monopsony models, e.g. counteract excessive employers’ market power without creating additional unemployment – see e.g. Manning (2020[101]) and Chapter 3. The possibility that working hours reduction might preserve employment in monopsonistic labour markets is in fact acknowledged and discussed in the literature16.

The productivity effect

Compared with employment, the link between working hours and productivity remains understudied in the empirical literature. From a theoretical point of view, reducing normal hours could result in an increase in hourly productivity per worker, sustaining total productivity per worker17 through at least two channels. First, reducing working hours could reduce workers’ fatigue and increase their work engagement, hence resulting in an increase in hourly productivity. Second, reducing working hours could prompt firms to rethink their production processes and implement productivity-enhancing investments as well as organisational and managerial innovations – including potentially through replacing less productive workers with more productive ones to compensate for reduced hours. Beyond these two channels, productivity could also be enhanced at a more aggregate level if the time freed from work helps spark innovation and new firms creation (Gomes, 2021[102]).

However, the limited number of existing studies on working hours and productivity focuses almost exclusively on the potential productivity effect of reducing workers’ fatigue through regulation on maximum hours and overtime. On the latter, the evidence in the literature is rather unanimous:18 productivity decreases with long hours. The evidence on the productivity effect of reducing normal hours is scarcer.19 Delmez and Vandenberghe (2017[103])’s analysis on total hours (which therefore linearly averages effects of normal hours and overtime) shows clear evidence of a declining productivity of hours in Belgian firms (with a 1% increase in firm-level hours leading to a 0.8% increase in firm-level value added). Crépon et al. (2004[93]), however, observe a slight decrease in total factor productivity following the reduction in normal hours from 39 to 35 hours in France in the early 2000s. By contrast, Park and Park (2019[104]) exploit the stepwise reduction in normal hours from 44 to 40 hours between 2004 and 2011 in Korean manufacturing firms, and find that it even increased total output per worker (i.e. not only hourly productivity). Evidence of decreasing marginal returns to normal working hours has been found in cross-country (Cette, Chang and Konte, 2011[105]) as well as micro-level analyses (Collewet and Sauermann, 2017[106]). This last study, based on an experiment – and therefore with particularly robust results – with Dutch call-centre workers in the 2010s, is particularly enlightening. Indeed, it exploits variation in the effective working time (i.e. excluding breaks, slack or training hours) due to random changes in weekly schedules, of workers paid by the hour and employed on average for 6 hours per day, 4 days a week (and effectively working 17.7 hours per week). Using these precise data, Collewet and Sauermann find strong evidence of a fatigue effect, with hourly productivity decreasing with hours, even for workers in intensive part-time jobs.

All of the above suggests that there could be some potential for working time policy to be productivity-enhancing over and above reducing long hours and overtime and also focusing on the reduction in normal working hours. Quests for the “optimal” length of the workday are therefore not over, and answers are likely to vary with job characteristics (Pencavel, 2016[107]; Dolton, Howorth and Abouaziza, 2016[108]).

5.2.2. Evidence on the effects of European national legislative reforms reducing normal hours

As outlined above (and see also Annex Table 5.C.1), much of the empirical literature on the impact of working time reforms, and in particular on working time reduction, concentrates on the employment effect. When productivity effects are considered, this is often done in isolation from employment effects, so that the broader economic impact of working time reforms (and the potential interaction between employment and productivity effects) remains poorly understood. To overcome these limitations, this section draws on results from Batut, Garnero and Tondini (2022[109]) to consider the employment and productivity effect of several working time reforms that took place in Europe between 1995 and 2007 allowing for general equilibrium effects.

The analysis focuses on national working time reductions reforms that were implemented in five European OECD countries; while these reforms kept monthly wages constant, thus leading to higher hourly wages, they did not all include compensatory measures for firms to buffer the impact on labour cost (see Table 5.1 for an overview of the reforms). By lumping several reforms together in a relatively short time period, in countries with a similar legislative framework (the EU Working Time Directive) and relatively similar societal preferences, this analysis allows presenting average effects and minimise the idiosyncrasies linked to specific national reforms. The causal effect of working time reductions on the outcomes of interest (hours worked, employment, hourly wage and hourly productivity) is identified via a difference‑in-difference approach that exploits the initial differences in the share of workers exposed to the reforms across sectors.20 The treatment group is composed of sectors in reforming countries above the median of the share of affected workers before the reform, i.e. those previously working more hours than the new threshold specified in the reform (see Box 5.2 for a discussion of the specification). The analysis uses information from multiple sources to document working time reforms in European Union countries.21 It relies on sectoral data in 22 countries for hours worked, employment, wages and productivity from EU KLEMS, since they are among the most reliable cross-country comparable sources for industry-level data. Out of 22 countries, 17 serve as full control.

Results are presented in Figure 5.7 for a discrete treatment variable (as in Equation 5.1 in Box 5.2) and for both a discrete and a continuous measures of exposure (Panel A and Panel B in Annex 5.A, as defined in Equation 5.1 and Equation 5.2 in Box 5.2). They show that the reforms examined appear to reduce significantly the share of workers who were working more than the new threshold introduced by the reform (by around 5 percentage points with the specification with the discrete treatment variable i.e. a reduction of one‑third compared to the pre‑reform difference between more and less exposed sectors) and the yearly number of hours worked on average by workers (by 1.3%, relative to sectors below the median, with the discrete treatment variable i.e. a reduction of two‑thirds compared to the pre‑reform difference22). However, reforms had no significant effects on employment, on workers’ compensation nor on hourly productivity (Figure 5.7). Although insignificant, the evidence displayed on employment reduction suggests that effects varied a lot across industries, reflecting perhaps different degrees of monopsonistic labour market situations; so overall, the absence of significant effect for employment is likely to be the average of heterogeneous positive and negative effects.

Results do not vary when the estimation is run only on the sample of countries implementing a reform (i.e. Belgium, France, Italy, Portugal and Slovenia, thus exploiting sectoral differences in the exposition to reforms in these countries only) and are robust to extended checks of alternative specifications, samples and estimators.23

Country	Year	Implementation	Reduction of weekly working time	Monthly wage	Compensations for firms
Portugal	1996	1997‑98	44h -> 40h	Constant	None
Italy	1997	1998	48h->40h	No specific adjustment.	None
France	1998	2000	39h->35h	Constant	Reduction in Social Security contributions
Belgium	2001	2002	40h->38h	Constant	Reduction in Social Security contributions
Slovenia	2002	2003	42‑>40h	Constant	None

Note: Adoption refers to the year of adoption of the legislation, while implementation refers to the year in which the legislation was actually implemented.

In 1997 and 2002, Poland also reduced weekly working time but the LFS data for Poland do not cover these years and therefore these reforms are not part of the analysis in this section.

Source: Batut C., Garnero A., and Tondini A. (2022[109]) “The Employment Effects of Working Time Reductions: Sector-Level Evidence from European Reforms”, FBK-IRVAPP Working Papers Series.

Box 5.2. Estimating the impact of working time reductions in Europe

Batut, Garnero and Tondini (2022[109]) estimate the effect of reductions in working hours on value‑added per hour worked, employment and wages, using the following specification:

where $Y$ stands for the dependent variable (e.g. productivity, employment, etc.), $X'$ is a vector of sector and time‑varying controls at the country level (share of self-employed, gender, part-time, temporary contract, occupation, education and age), $γ$ and $Ɵ$ are fixed effects (respectively sector × country, sector × year and country × year fixed effects), $ε$ is the error term, $i$ indexes the sector, $c$ the country and $t$ is the year. $T r e a t e d$ is a binary variable indicating whether a sector is above the median of the share of affected workers before the reform (e.g. those working more hours than the threshold specified by the reform) interacted with $P o s t$ which indicates the staggered implementation of the reform across countries. The coefficient of interest, $β$ , is identified by the evolution of more affected sectors relative to less-affected sectors in reforming countries at the moment of the reform.

There are two important caveats to point out about the β coefficient: first, it is identified only through variation within reforming countries, hence non-reforming countries play a role only in the estimation of the set of sector × year fixed effects; second, it only identifies a relative effect, i.e. the effect of more treated sectors relative to less treated sectors.

Moreover, a second specification is tested that introduces a continuous measure of sectoral exposure to the reform (and not a discrete one as in Equation 5.1. This also allows to recover a relative effect, leveraging the full variation in exposure to the reform, at the price of assuming a linear relation between the effect and the measure of exposure. Equation 5.1 is rewritten as follows:

where exposure indicates the share of workers above the reform level in each sector.

Source: Batut, Garnero and Tondini (2022[109]), “The Employment effects of Working Time Reductions in Europe”, FBK-IRVAPP Working Papers Series.

Several potential explanations could be behind these results, echoing the theoretical arguments discussed in Section 5.2.1. First, between 1995 and 2007, all European countries (with the exception of Italy) experienced relatively robust growth, together with productivity and wage growth (although with a lot of heterogeneity across sectors/countries) and stable, low inflation. It is therefore possible that, even in the context of a standard competitive model, the reduction of working time and the increase in labour cost per hour worked might have been quickly absorbed with no effect on employment (in line with the observed results of insignificant but positive effect on productivity). Second, an alternative partial explanation would be that the classical hypotheses do not hold, and the reductions in working time with constant monthly wage act like an increase in the minimum wage in a monopsony model (e.g. the increase in hourly labour cost induced by the reduction in hours counteracts pre‑existing excessive employers’ market power as described in Section 5.2.1). A third potential explanation could be that some mechanisms limited the rise of labour costs in practice, such as a decrease in social security contributions (as in the French and the Belgian reforms24) or some voluntary wage restraint by social partners in wage negotiations. Finally, as outlined before, even if statistically insignificant, the average estimated employment effect is negative and not small: employment is estimated to have decreased by 2.3% in more exposed industries with respect to less exposed industries. These results suggests that the average estimated effect could result from the aggregation of heterogeneous positive and negative effects in different industries and local labour markets, for example because certain local labour markets are more monopsonistic while others are more competitive (see Chapter 3).

Figure 5.7. Average differential impact of normal hour’s reductions between sectors with above and below median intensity of long hours, 1995‑2007

5.2.3. Going to the firm-level: How do contractual hours reductions affect productivity, wage and employment?

In order to get additional evidence on the relationships between normal hours reduction, employment and productivity, this section looks at how these relationships materialise at the firm level. Exploiting firm-level panel data, it explores the effect of observed episodes of reductions in average contractual hours on the growth of productivity per worker, employment, and average wage, in three countries where data are available, namely Germany, Korea and Portugal. The analysis adopts a difference‑in-difference framework, comparing log changes in productivity per worker (using information on value added and number of workers in the data), in number of employees, and in average wage, between firms that reduced their contractual working hours and similar firms that did not,25 around the time of the change. Treated firms are matched to control firms based on a series of firm-level descriptive variables, including their pre‑change trajectories in terms of value added per worker, total employment and average wage. The detailed identification strategy is presented in Box 5.3.

Box 5.3. How do reductions in contractual hours relate to productivity and employment? A firm-level identification strategy

The identification strategy implemented in this analysis requires access to firm-level panel data with information on contractual normal hours (as opposed to effective hours, which take into account overtime and sick leave, and therefore are not a good means of measuring the impact of a change in normal hours). This information is available in three countries: Germany, Portugal and Korea.

For Germany, the analysis uses data from the IAB Establishment Panel, a nationwide representative survey of employers conducted by the German Institute for Employment Research (IAB). Data on individual establishment characteristics as well as on many employment policy related topics are collected annually from employers in 15 500 German firms, from all industries and firm sizes categories. The longitudinal dataset goes back to 1993 in Western Germany and 1996 in Eastern Germany and allows deriving information on year-on-year changes in value added per worker (from information on business volume and medium-term inputs), number of employees, and average annual wage (which is the total wage bill divided by the number of workers in a given year).

The dataset used in the Portuguese analysis is a merge from two sources, the Quadros de Pessoal (QP) and Sistema de Contas Integrado das Empresas (SCIE). The QP dataset is a matched employer-employee administrative dataset covering all Portuguese firms with at least one wage earner in the private sector. Individual level data on firms’ employees, as well as some data on firms (e.g. industry, sales, ownership, size, legal form…) have been collected annually since 1985. Firm-level information is completed with data from SCIE, a dataset compiled by Statistics Portugal (INE) from the online Simplified Business Information (IES) system used by the tax authority, Ministry of Justice, Banco de Portugal and Statistics Portugal. All non-financial firms are included in the dataset, which has existed since 20041. The SCIE dataset covers detailed information on firms’ annual balance sheet and income statement, and includes variables on annual value added, annual total employment, and annual gross staff expenses which allow deriving the three dependent variables used in the analysis (year-on-year changes in value added per worker, average annual wage, and number of employees). The final sample therefore covers all private‑sector, non-financial firms between 2004 and 2019.

Finally, data on Korea comes from the Koran Workplace Panel Survey (WPS), a longitudinal survey of 4 300 firms with more than 30 employees in all industries except agriculture and mining, conducted every two years by the Korean Labour Institute since 2005. The WPS collects information on the various characteristics of individual workplaces, and covers a wide‑range of employment related topics, including business volume, employment, and wage bill, which allows deriving information on wave‑on-wave change in business volume per worker, employment, and average wage (total wage bill divided by the number of workers).

Treatment is defined as a firm-level reduction in contractual hours and identified by spells: a treated spell is made of a 4‑year period2 around the year when the reduction in contractual hours is observed, with one pre‑year and two post-years without changes in contractual hours. Several treatments spells can therefore be identified for the same firm. Spells during which hours are increased are excluded from the sample, but the possibility that firms increase their use of overtime as a result of the contractual hours reduction is taken care of by matching firms according to their use of overtime before the change as well as adding a dedicated control in the regression below – see Equation 5.3.

To estimate the effect of treatment on productivity and employment, a control group is identified through the following matching procedure. Spells are grouped in clusters by set of four years3, industry and firm size. Each cluster contains treated and non-treated spells. Event dummies (t‑1, t, t+1, t+2)4 allow for a common identification of time across clusters. Within clusters, a nearest-neighbour algorithm is used to match treated spells with the five closest non-treated spells. The matching algorithm uses the following firm characteristics at t‑15 in all three countries: the year-on-year percentage change6 in the number of employees, in value added per employee7, and in average wage, a dummy capturing whether the firm is using overtime, and a categorical variable describing the firm’s profit situation. Additional variables include a dummy capturing the presence of worker representation (e.g. a work’s council) in Germany and Korea, the collective bargaining level of the collective agreement covering the firm and the share of exports in business volume in Portugal and Germany, the share of highly-educated employees in Portugal and the average level of education of workers in the largest occupational group in Korea, as well as the share of permanent workers and of full-time workers, the share of investment in value added and the change in business volume in Portugal. This allows obtaining three balanced8 samples made of pooled treated spells and their matched controls – Annex Table 5.D.1 presents descriptive statistics of the balanced sample in each country.

Equation 5.3 is then estimated on each balanced sample (one per country):

Where $α$ is the coefficient of interest, $Y$ represents the outcome variable analysed (i.e. either log change in value added per worker, in number of employees, or in average annual wage per worker) in firm $f$ at time $t$ ; ${T r e a t m e n t}_{f}$ is a dummy variable identifying treated spells; $P o s t$ is a vector of post-treatment time dummies. $X_{f, t}$ is a vector of time‑varying observable firm characteristics and potential confounding factors, namely: annual log change in real wage compensation per worker, change in the firm’s use of overtime, investment in communication technology/data processing, level of the applicable collective agreement, and region. Finally, $γ t$ represents year fixed effects, $δ f$ “spell fixed effects”, $β m a t c h g r o u p$ is a fixed effect for each group of one treated spell and its five matched controls, and $ε_{f, t}$ is an idiosyncratic error term. Errors are clustered at the spell level.

There are three main sources of errors attached to this identification strategy. First, self-selection: firms that already have a higher productivity growth might decide to reduce their contractual hours. However, since treated spells are matched with non-treated spells with comparable productivity trajectories in the year before the change, this source of error should be largely neutralised. Second, reverse causality: an increase in the growth of productivity per worker might cause, rather than follow from, a reduction in working hours. This issue should also be partly dealt with by using pre‑change outcomes in the matching algorithm – although this is insufficient to exclude the possibility that a change in productivity growth simultaneously causes a reduction of working time. Third, unobserved confounding factors: time‑invariant confounding factors are in principle neutralised by the introduction of spell fixed effects, and by matching firms on the outcome variables; however, time‑varying unobserved confounding factors might also be at play. For example, certain firms may introduce working time reductions together with (or just after) a reorganisation process which also makes them more productive (through e.g. more efficient processes or the hiring of more productive workers). This last source of error cannot be solved with this identification strategy.

1. Through merging the two datasets, the Portuguese sample is de facto reduced to the period 2004‑19.

2. Except in the Korean data, for which spells are periods of five years identified over at least three consecutive waves in the four that are available (2007, 2009, 2011 and 2013).

3. Five years in the Korean data.

4. In the Korean data this can only be done for t‑2, t and t+2.

5. t‑2 in the Korean data.

6. Since the analyses focuses on growth rates rather than levels, growth rates variables (rather than levels) are also used in the matching algorithm.

7. In Korea, productivity is measured as log change in business volume per worker, since the variable on value added has too many instances of missing value.

8. At t‑1, there are no statistically significant differences in the dependent variables of interest, namely log changes in value added per worker, in number of employees and in average wage between the pool of treated observations and that of control observations in any of the three countries, meaning that the pre‑change trends in independent variable between t‑2 and t‑1 (t‑4 and t‑2 in Korea) are parallel. As shown in Annex Table 5.D.1, samples for the three countries are balanced when considering levels and percentage changes in total number of employees, firm size, industry, change in business volume, profit situation, share of export in business volume, share of full-time employees, share of permanent workers, use of overtime and change in the use of overtime, education level, level of the applicable collective agreement, and, for Germany and Korea only, coverage by a collective agreement on wage and presence of a works council, and for Germany and Portugal only, investment growth, share of investment in value added or business volume, and investment in technology. The Korean and German samples are also balanced regarding levels and percentage change in value added per worker and average wage. In Portugal, the sample is balanced for percentage and log changes, but not when considering levels of value added per worker and average wage: firms that reduce their hours have a significantly higher level of value added per worker and pay a higher average wage at t‑1. This does not affect the identification strategy, since the analysis is based on growth rather than levels. Yet, to correct for this imbalance, controls for pre‑change levels of value added per worker and average wage are added in the baseline analysis for Portugal and therefore reflected in the results presented in Figure 5.8 below.

Results for all three countries are presented in Figure 5.8. They show positive and significant associations with productivity growth in two countries out of three (Germany and Korea, although the cumulative effects at t+2 26 disappears in Germany; results are positive but insignificant in Portugal). On employment growth, results show insignificant associations in two countries out of three (Germany and Korea), but a negative significant association in one (Portugal). Finally there are insignificant associations with wage growth in Korea and Portugal, and positive significant results on wage growth in Germany.

Looking at country-specific results, Figure 5.8 suggests that in Germany, episodes of contractual hours reduction observed in the data on average led to an increase in productivity and wage growth, while they did not significantly affect employment growth. The analysis exploits the variation in employment, productivity per worker and average annual wage observed in 204 spells (3.7% of total spells in the sample)27 of hours reduction (on average amounting to a 2.1 hours reduction per week), compared to the variation observed in matched control spells. The association between contractual hours reduction and log change in employment is insignificant in both post-change years. Change in productivity per worker, by contrast, is significantly and positively related to contractual hours reduction at t+1 – but the association becomes insignificant at t+2. Log change in real average wage is positively and significantly related to contractual hours reduction at t+1, and this association remains statistically significant at t+2. Spells of hours reductions are more concentrated in 2002, 2004 and 2006;28 results hold when excluding 2004 and 2006 from the analysis as a robustness test, however when excluding 2002, the effect on average wage growth becomes insignificant in both years, and the positive productivity effect observed at t+1 still holds at t+2. Results are also robust to adding a control for the presence of a works council, for organisational change, and for total investment.

To test for the hypothesis that the positive association with productivity is mediated by an increase in investment prompted by the change in contractual hours, Equation 5.3 is also estimated with growth in total investment as outcome variable. The association between contractual hours reduction and the growth of total investments is positive and significant at t+1 (and loses significance at t+2), which lends support to the idea that the positive effect on productivity might be mediated by a spike in firms’ investment following the reduction in hours.

Figure 5.8. How do reductions in contractual hours relate to changes in productivity, employment and wages in Germany, Korea and Portugal?

The Korean story emerging from Figure 5.8 is aligned with the German one: episodes of contractual hours reduction observed in the Korean data on average led to an increase in the growth of productivity per worker, while it did not significantly affect the growth of employment. In contrast with the German case, wage growth was also insignificantly affected. The sample behind these results contains 31 spells (5.5% of total spells in the sample) of hours reduction – by 4 hours per week on average. The structure of the Korean data (survey waves of limited sample size) only allows looking for spells of contractual hours reductions in two years, 2009 and 2011. This corresponds to the implementation period of a reform reducing normal working hours in Korea: between 2004 and 2011, normal weekly hours were reduced from 44 to 40 hours per week – which is in line with the average reduction observed in our sample. The reform was implemented gradually to give small firms more time to adjust (Hijzen and Thewissen, 2020[110]). Since treated and controls are matched within similar firm-size and industry groups, this staggered implementation does not invalidate the identification strategy detailed in Box 5.3.

The association between contractual hours reduction and log change in productivity per worker in Korea is positive and significant at t+2. By contrast, the associations with log change in number of workers, and log change in average real wage are insignificant. The small sample sizes for the Korean analysis should be kept in mind when interpreting results, however results are robust to using an alternative specification of the collective bargaining variable and adding a control for organisational change.

Finally, the story observed in the Portuguese data differs from that emerging from the German and Korean analyses. In Portugal, on average in the data, observed episodes of contractual hours reduction did not significantly affect productivity and wage growth, but they negatively affected employment growth. This is based on data from 668 spells of hours reduction observed in the Portuguese sample (4.2% of total spells in the sample), during which contractual hours were reduced by 3 hours per week on average. The association between reductions of contractual hours and growth in value added per worker is insignificant in both post-change years. Similarly, there are no significant associations between average real wage growth and reductions of contractual hours. By contrast, growth in employment is significantly and negatively associated with reductions of contractual hours when considering both changes between t and t+1, and those between t and t+2).29 Results hold when excluding 2012 and 2013 − which display a higher concentration of cases30 − from the analysis, and when replacing the control for investment in technology with a more precise control for investment in software (to test for the possibility that digital solutions are adopted to compensate for the lost hours of work).

Regarding insignificant results, while those observed in the Korean analysis (on employment growth and average wage growth) might be difficult to interpret due to the limited sample sizes available, the panel data used for the German and Portuguese analyses are rich enough to cautiously interpret the results that are non-significant (i.e. the results on employment growth in Germany, and on the growth of value added per worker and average wage growth in Portugal) as an actual absence of statistical relationship on average (possibly due to heterogeneous effects cancelling out each other) rather than as the effect of a weak statistical power. Overall, these results show that reductions of contractual hours can yield positive results in terms of productivity growth and leave employment growth unaffected on average in some cases (e.g. the German case), while they can leave productivity growth unaffected and yield negative results in terms of employment growth in others (e.g. the Portuguese case). Considered together, these results suggest (although they do not prove) that there could be a virtuous circle in some cases – which does not however materialise in all instances – with productivity increases potentially limiting the rise of unit labour cost and therefore the potentially negative effect on employment growth.

Regarding the causes of increased productivity following a reduction of contractual hours, the analyses above do not allow giving a definitive answer. There are some suggestive evidence that investment growth induced by the hours reduction might be at play (e.g. in Germany) although other mechanisms might be at play as well, e.g. organisational change, workforce re‑composition, or reduced worker fatigue. Independently of the mediating factors behind the increase in productivity, the reason why reducing hours led to an increase in productivity growth in Germany and Korea but not in Portugal should be explored further in the future; tentatively, one can perhaps posit that it might have to do with differences in firm-level institutions of collective representation and negotiation between these countries, and/or to the different institutional contexts in which contractual hours rseduction happened.

Beyond increased productivity growth, other factors (discussed in Section 5.2.1 above) could have limited the impact on unit labour cost and therefore explain the absence of negative effects on employment in Germany and Korea, namely wage restraint or public subsidies compensating the rising hourly wage for workers. While wage restraint can be ruled out in Germany and Korea (since average wage growth is not negatively affected), there is no information on whether public subsidies played a role or not. While this is unlikely in the German case, which exploits episodes of spells reduction scattered over more than 20 years, in Korea, as explained above the majority of reductions spells observed are the result of a legislative reform, which included accompanying measures for firms – although no direct subsidies (Hijzen and Thewissen, 2020[110]).31 Finally, beside cases of limited impact on unit labour cost, the absence of significant effect on employment growth in Korea and Germany might be explained if the increase in hourly pay induced by contractual hours reductions was absorbed by a pre‑existing profit rent in firms, generated for instance if wage growth and working time did not follow previous productivity increases – which would typically be the case in the monopsonistic labour markets described in Chapter 3.

5.2.4. The economic impact of fostering the use of flexible hours and teleworking: A literature review

The employment effect

In the literature, flexible hours and teleworking are theoretically expected to have a positive impact on employment, mainly since they might allow workers to stay in full-time employment when they face schedule constraints or family responsibilities – see for example Chung and van der Horst (2018[111]) and Fuller and Hirsh (2019[112]). Flexible hours and teleworking might notably represent ways to increase female labour force attachment – and may also lead to higher earnings for women, see below. The expected effect of teleworking and flexible hours on career progression is, however, less clear-cut – and might depend on whether their use is exceptional or relatively mainstreamed in a given firm.

The empirical evidence on flexible hours and teleworking to date remains mainly correlational, with overall32 positive effects for employment outcomes. Offering flexible hours and teleworking in particular is consistently shown to have positive effects on worker attraction (He, Neumark and Weng, 2021[113]; Wiswall and Zafar, 2016[114]; Mas and Pallais, 2017[115]; Maestas et al., 2018[116]) and also to partly reduce attrition rates (Bloom et al., 2015[117]; Kröll and Nüesch, 2019[39]). Workers across OECD countries value flexible hours and teleworking – which might be of increasing importance for firms seeking to attract talent in times of labour shortages. In accordance with expectations, flexible hours and teleworking indeed appear to be a successful means of increasing female labour force attachment, especially after childbirth (Chung and van der Horst, 2018[111]; Arntz, Sarra and Berlingieri, 2019[55]). Yet, robust evidence of the (gender-differentiated) effect of teleworking and flexible hours on long-term career progression is still missing today and would be a welcome focus for future research.

The wage effect

The theoretical effect of teleworking and flexible hours on wages is unclear. They may reduce wages if they are costly for employers, but may increase wages if they also increase productivity (Arntz, Sarra and Berlingieri, 2019[55]). In terms of gender, flexible hours and teleworking are ways to increase female labour force attachment and may thus lead to higher earnings for women. Yet, these arrangements may also increase the gender wage gap if women view flexibility as a job amenity and accept lower pay in exchange, while men may view flexibility as a job demand and select into jobs that pay a flexibility premium (Pabilonia and Vernon, 2022[46]).

Empirical studies in Canada and Germany find positive wage effects of both flexible hours and teleworking for women but suggest that these operate largely by reducing barriers to their employment in higher wage establishments: whereas women do seem to receive higher wages when switching into jobs that allow for flexible hours and teleworking, this is less the case when they opt for these arrangements while remaining in the same firm (Fuller and Hirsh, 2019[112]; Arntz, Sarra and Berlingieri, 2019[55]). The authors conjecture that this could be due to a flexibility stigma that adheres more to women or that the bargaining power of women within firms is weaker for re‑negotiating wages than it is for men. Overall however, while generalisable estimates of wage effects are difficult to derive because of the concentration of flexible hours and teleworking in a limited number of jobs, existing estimates for the wage effects of both arrangements tend to be positive in general (Bonacini, Gallo and Scicchitano, 2020[118]; Weeden, 2005[119]; Pabilonia and Vernon, 2022[46]; Oettinger, 2011[120]; White, 2019[121]; Fuller and Hirsh, 2019[112]) – but mixed for women and parents.

On flexible hours alone, while Weeden (2005[119]) finds that wage premia for flexible hours in the United States do not vary by gender, more recent evidence from Germany suggests that only men receive financial rewards for working flexible hours (Lott and Chung, 2016[40]). Giménez- Nadal et al. (2019[54]) find a U-shaped relationship between flexible hours and both mothers’ and fathers’ wage rates in the United States, with wages being highest for parents who work either very flexible or very inflexible hours. In parallel, Yu and Kuo (2017[122]) find that the wage penalties often experienced by women after childbirth (OECD, 2018[71]), is smaller in workplaces with flexible hours.

Regarding teleworking, experimental evidence from the United States suggests that the average worker is willing to give up 8% of wages for the option to telework – but even though women value teleworking up to twice as much more than men, this preference cannot explain a large part of existing gender wage gaps (Mas and Pallais, 2017[115]). Yet, in contrast to exploiting an experimental setting but analysing detailed time use data from the United States instead, Pabilonia and Vernon (2022[46]) do find wage premia from actual teleworking uptake but not for everyone: for fathers regardless of how often they telework and women without children who occasionally telework. In this respect, evidence from Italy shows that without dedicated policies, an increase in teleworking would boost the wage of male, older, high-educated, and high-paid employees but not of others, thereby increasing income inequalities (Bonacini, Gallo and Scicchitano, 2020[118]).

The productivity effect

Both the employment and wage effects of teleworking and flexible hours partly depend on how they affect workers’ productivity. The effect of teleworking and flexible hours on productivity has been the object of more attention in recent years, although evidence is still patchy. The majority of pre‑pandemic33 studies look at the effect of flexible hours and teleworking together (since the two are usually offered together and therefore hard to separate), analysing the effect on productivity of firms allowing workers to choose both where and when to work. Since the integration of teleworking and flexible hours in standard work practices imply transitioning from a system of input control, and working time registration, to a system of output control, in which performance is evaluated through measurable objectives other than the amount of time workers spend at work, these arrangements have sometimes been labelled “trust-based” arrangements. Viete and Erdsiek (2018[123]) for example find that German firms using “trust-based” working practices experienced enhanced productivity returns to mobile ICT equipment as a result. Moreover, a randomised experiment on a sample of workers in a large Italian company causally shows that workers engaging in “smart working” (another term for the combination of teleworking and flexible hours) one day per week have a higher productivity (Angelici and Profeta, 2020[38]). Beckmann (2016[124]) similarly finds a positive effect on firm productivity of introducing “self-managed” working practices, whereby workers have control over the duration, scheduling and location of their work. In a subsequent paper, Beckmann et al. (2017[125]) explain this positive productivity effect by the fact that workers with such flexible working arrangements exert higher effort levels than their peers with fixed working hours. Accordingly, Godart et al. (2017[126]) find that German firms adopting trust-based working practices (regarding hours and place) are more likely to improve products and engage in process innovation.

On teleworking34 alone, Bloom et al. (2015[117]) conducted an experiment in a Chinese call-centre where they randomly assigned workers to telework or work in the office, and found that teleworking led to a 13% performance increase. By contrast, Monteiro et al. (2019[127]) found a small but significantly negative effect on productivity in Portuguese firms allowing teleworking, albeit with a large degree of heterogeneity: the effect was positive for firms undertaking R&D activities, but negative in others, and in particular in small firms in the sheltered sector employing a below-average skill level workforce. The experiment conducted by Dutcher (2012[128]) showed that teleworking could also have heterogeneous productivity effects depending on the nature of tasks affected, with negative effects on productivity when it comes to routine tasks, but positive effects for creative tasks.

When it comes to flexible hours, Boltz et al. (2020[129])’s experiment with routine job workers in Colombia revealed that allowing workers to decide their start and finishing times increased total productivity per worker by as much as 50%. Such productivity gains are likely to be weighed against the organisational costs induced by the transition to flexible working; however, the COVID‑19 pandemic has led businesses to identify tasks that can be performed flexibly and many firms have already paid the fixed costs of that transition (Pabilonia and Vernon, 2020[130]).

To sum up, the majority of empirical studies to date point towards positive or neutral effects of teleworking and flexible hours on employment and productivity, albeit with more heterogeneous results for teleworking. At the same time, there is evidence of wage effects increasing pre‑existing gender differences and pay gaps, if no counter-acting measure (e.g. pay transparency policies and similar mechanisms (OECD, 2021[131])) are in place to strengthen female bargaining power in firms adopting teleworking and flexible hours. Future research should aim to systematically look at gender-differentiated effects where possible. Another crucial aim for future research would be to address two of the main limitations plaguing existing studies, namely the lack of comparability in the definitions of teleworking and flexible hours used in studies, and the fact that many studies only consider firms willingly adopting teleworking, which limits the potential extrapolation of findings to the universe of firms (OECD, forthcoming[132]). Overall, more evidence is still needed on the productivity, employment and wage effect of teleworking and flexible hours. The next section aims to contribute to that effort by looking at how the adoption of flexible hours affect employment and productivity at the firm-level in Germany.

5.2.5. How does the adoption of flexible hours affect firm-level productivity and employment? An analysis on German data

The German IAB Establishment Panel data used for the firm-level analysis on contractual hours reduction above also contains data on whether firms have a system of flexible hours (whereby workers can autonomously determinate their starting and finishing times) in place. This allows replicating the analysis, adapting the identification strategy described in Box 5.3 above. Treatment is now defined as the firm-level adoption of flexible hours, when a firm which previously did not offer flexible hours starts doing so; it is still identified by spells: a treated spell is made of a 4‑year period around the year when flexible hours adoption is observed, with at least one pre‑year and two post-years without changes in hours. Control spells are stable spells of non-adoption (periods of four consecutive years during which a firm which previously did not have flexible hours continues not to do so). The analysis uses the same matching algorithm and regression specification as described in Box 5.3, adding a variable on the share of workers with a university education, and a dummy variable on whether the firms has changed its use of overtime (started using it, or stopped using it) that year, to obtain a balanced sample. Annex Table 5.D.2 presents descriptive statistics of the balanced sample behind this analysis.

Results are presented in Figure 5.9. The adoption of flexible hours is not significantly associated with growth in productivity per worker in either t+1 or t+2. By contrast, flexible hours adoption is positively and significantly related to employment growth at t+1 and t+2. This result is aligned with findings from the literature, which find that flexible hours have a positive impact on worker’s attraction and retention.35 Finally, the adoption of flexible hours is significantly and negatively related to average wage growth in the first year after the change (while the relationship becomes insignificant at t+2). Therefore, a decrease in average wage growth is observed on average in firms adopting flexible hours, in the year after the adoption of flexible hours. One possible interpretation is that increased autonomy in determining hours through flexible hours might indeed be traded against wage increases in when negotiating wages, as suggested by other results in the literature.

Figure 5.9. How does the adoption of flexible hours relate to changes in productivity, employment and wages in Germany?

5.3. Concluding remarks

This chapter discusses the pro‑and cons of various working time policies at the disposal of policy makers interested in enhancing workers’ well-being, while accounting for their potential adverse effects on employment, wages and productivity. Results from the literature, as well as new empirical evidence presented in this chapter, suggest that reducing normal working hours and facilitating the use of flexible working hours might, in some circumstances, help improve workers’ non-material well-being. Reducing normal hours might in particular help improve workers’ satisfaction with their free time, their job, and their life more generally, while fostering the use of flexible hours might also help enhance all these three outcomes, together with health satisfaction. Further analyses on these policies’ effects on productivity and employment suggest that they might in some circumstances be valid options worth reviewing by policy makers, but impacts on productivity and employment should be closely monitored in the aftermath of the reform. Wide‑ranging recommendations are ill-advised as working time policy should always be analysed in their concrete institutional and national contexts.

Reducing normal hours

Provided they are carefully designed and implemented, evidence presented in this chapter suggests that a reduction in normal working hours might enhance workers’ well-being without adverse effects on employment and productivity. Analysis of the effects of a number of national legislative reforms and firm level contractual reductions indicate that reducing normal hours (at constant monthly or annual wage) might preserve employment and enhance workers’ well-being if the impact on unit labour cost remains limited, either due to sufficient induced productivity gains or to public subsidies, or if the reduction takes place in a pre‑existing situation of labour market monopsony. Hence, any foreseen reduction of working hours in the future should be carefully designed to tap into the productivity-enhancing potential of working shorter hours, to generate a positive feedback loop and preserve employment. Moreover, accompanying measures limiting the impact on unit labour costs might also be considered.

Since one key issue here is to identify how to ensure that normal hours reduction generate sufficient productivity gains, a promising way to structure such pre‑policy design analyses would be to look, within each country and possibly sector, at the various channels through which reducing hours might be associated with or stimulate such productivity gains (e.g. increased investment, managerial reorganisation and innovation). Reforms should be designed to provide the right incentives for these channels to be activated. For instance, incentives for investing in IT or organisational innovations could be built in to maximise the productivity enhancing potential of reducing hours.

Beyond measures fostering productivity gains, an important parameter that should also be kept in mind when designing working time policy reforms is negotiated wage progression over the implementation period. In that regard, negotiating working time reduction and wage increases together, as a longer term package deal would allow smoothing the induced increase in hourly wages over a longer period of time, and therefore limiting the rise in unit labour costs.

More generally, careful attention should be devoted to the implementation process, starting with:

The initial level of weekly normal hours applying in the country, as well as the scope of the reduction are also key parameters which are likely to influence the hours reduction effect.
Second, the timing of the policy measure, as robust economic growth – together with productivity and wage growth might provide scope for easing pressure on unit labour costs.
Third, the mode of adoption, e.g. by law or collective bargaining, is also important: legislative reforms on the one hand, can ensure a maximum coverage of the measure, but could on the other be perceiced as a straightjacket for some enterprises or sectors; in that sense, collective bargaining has been shown to be an efficient tool to negotiate reduction in hours at sectoral level in recent years (OECD, 2019[133]). At the same time, if decisions on working hours are too individualised and not influenced by statutory or negotiated rules, working time policy runs the risk of losing its power as a policy lever altogether.
Fourth, policy makers could also consider gradual changes when implementing the reduction of hours, e.g. to give small firms more time to adjust.
Finally, counter-productive effects should be carefully considered: for instance, to prevent heterogeneous effects among firms/workers and avoid that a reduction of hours result in higher work intensity for workers.

Fostering the use of flexible hours

Since flexible hours are also identified as a potential measure to increase workers’ non-material well-being, the chapter considers the effect of adopting such arrangement on employment and productivity. Results from the literature suggest that flexible hours can be a successful means of increasing the labour force attachment of women with children, while also allowing them to remain in their job, occupation and skill-level (by opposition to women opting for part-time who often have to select into lower-paid jobs and occupations allowing for this arrangement). Existing estimates for the wage effects of flexible hours also tend to be positive in the literature but indicate risks of increasing pre‑existing gender differences and pay gaps, if no counter-acting measure (e.g. pay transparency policies and similar mechanisms) are in place to strengthen female bargaining power in firms adopting teleworking and flexible hours.

Results from an analysis in German firms find that firms adopting flexible hours also see a decrease in average wage growth – suggesting that consistent with theoretical assumptions, there can also be a possible trade‑off between wage increases and higher autonomy in determining hours. In this context, the relevant question for firms might increasingly revolve around how they can implement flexible arrangements to remain attractive to workers in the most beneficial way with regards to other outcomes such as productivity and employment. While additional analyses in other contexts are necessary before allowing for further generalisation, results from the analysis in German firms suggest that this is possible: the adoption of flexible hours resulted on average in an increase of employment, while it did not significantly affect productivity per worker in the sample of firms studied.

Monitoring the development of teleworking

Considering teleworkings widespread development in the aftermath of the COVID‑19 crisis, the chapter also considers its relationships with non-material well-being outcomes, employment, wage and productivity. It finds a more ambiguous link between teleworking and workers’ non-material well-being than when considering flexible hours: results vary for different outcomes and across countries, in the literature and in the new empirical evidence. In particular, empirical results show a negative association with self-assessed health, and contrasting associations with work-life balance across countries. Turning to productivity and employment, while associations with teleworking in the empirical literature to date are generally positive, especially in terms of attracting and retaining workers, as well as increasing female labour force attachment, there is heterogeneity across studies and across types of tasks. The literature also suggest that without dedicated counteracting policies, the adoption of teleworking runs the risk of dispropotionately favouring the wage of male, older, high-educated, and high-paid employees but not of others, thereby increasing general income inequalities.

These findings suggest that policy makers should aim to guarantee an enforceable right to access teleworking across groups, to limit disparities linked to differences in legal frameworks between different employees – in that regard, the OECD typology of access to teleworking might be a useful basis (OECD, 2021[131]). Crucially, the limited number of results available to date suggest that the effect of teleworking on various outcomes will still need to be closely monitored in the future. In that regard, one important issue to address is the lack of good quality data. Conceiving appropriate data collection strategies now will be key to ensure good quality research capable of informing policy making and considering the many possible repercussions (on gender disparities, on geographical disparities, etc.) in the future.

References

[49] Adamovic, M. (2022), “How does employee cultural background influence the effects of telework on job stress? The roles of power distance, individualism, and beliefs about telework”, International Journal of Information Management, Vol. 62, p. 102437, https://doi.org/10.1016/j.ijinfomgt.2021.102437.

[156] Adams, Z., L. Bishop and S. Deakin (2010), CBR Labour Regulation Index (dataset of 117 countries), Centre for Business Research, Cambridge.

[5] Ahn, T. (2016), “Reduction of Working Time: Does It Lead to a Healthy Lifestyle?”, Health Economics, Vol. 25/8, pp. 969-983, https://doi.org/10.1002/hec.3198.

[142] Andrews, M. et al. (2015), “More hours, more jobs? The employment effects of longer working hours”, Oxford Economic Papers, Vol. 67/2, pp. 245-268, https://doi.org/10.1093/OEP/GPU026.

[84] Andrews, M., T. Schank and R. Simmons (2005), “Does worksharing work? Some empirical evidence from the iab-establishment panel”, Scottish Journal of Political Economy, Vol. 52/2, pp. 141-176, https://doi.org/10.1111/J.0036-9292.2005.00339.X.

[38] Angelici, M. and P. Profeta (2020), “Smart-Working: Work Flexibility Without Constraints”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3556304.

[55] Arntz, M., B. Sarra and F. Berlingieri (2019), “Working from Home: Heterogeneous Effects on Hours Worked and Wages”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3383408.

[19] Askenazy, P. (2004), “Shorter Work Time, Hours Flexibility, and Labor Intensification”, Eastern Economic Journal, Vol. 30/4, https://www.jstor.org/stable/40326151 (accessed on 24 November 2021).

[35] Atkinson, C. and L. Hall (2011), “Flexible working and happiness in the NHS”, Employee Relations, Vol. 33/2, pp. 88-105, https://doi.org/10.1108/01425451111096659.

[77] Barbieri, P. et al. (2019), “Part-time employment as a way to increase women’s employment: (Where) does it work?”, International Journal of Comparative Sociology, Vol. 60/4, pp. 249-268, https://doi.org/10.1177/0020715219849463.

[150] Barrero, J., N. Bloom and S. Davis (2021), Why Working from Home Will Stick, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w28731.

[21] Bassanini and Caroli (2015), “Is Work Bad for Health? The Role of Constraint versus Choice”, Annals of Economics and Statistics 119/120, p. 13, https://doi.org/10.15609/annaeconstat2009.119-120.13.

[109] Batut, C., A. Garnero and A. Tondini (2022), “The Employment Effects of Working Time Reductions: Sector-Level Evidence from European Reforms”, FBK-IRVAPP Working Papers Series, pp. 2022-04.

[124] Beckmann, M. (2016), “Self-managed working time and firm performance: Microeconometric evidence”, WWZ Working Papers, WWZ, University of Basel, https://edoc.unibas.ch/61322/.

[125] Beckmann, M., T. Cornelissen and M. Kräkel (2017), “Self-managed working time and employee effort: Theory and evidence”, Journal of Economic Behavior and Organization, Vol. 133, pp. 285-302, https://doi.org/10.1016/j.jebo.2016.11.013.

[69] Beham, B. et al. (2019), “Part-time work and gender inequality in Europe: a comparative analysis of satisfaction with work–life balance”, European Societies, Vol. 21/3, pp. 378-402, https://doi.org/10.1080/14616696.2018.1473627.

[74] Bell, D. and D. Blanchflower (2019), “The well-being of the overemployed and the underemployed and the rise in depression in the UK”, Journal of Economic Behavior & Organization, Vol. 161, pp. 180-196, https://doi.org/10.1016/j.jebo.2019.03.018.

[22] Bell, D., S. Otterbach and A. Sousa-Poza (2012), “Work Hours Constraints and Health”, Annals of Economics and Statistics 105/106, pp. 35-54, https://www.jstor.org/stable/23646455 (accessed on 9 February 2022).

[67] Benson, R. et al. (2017), “Do work and family care histories predict health in older women?”, European Journal of Public Health, Vol. 27/6, pp. 1010-1015, https://doi.org/10.1093/eurpub/ckx128.

[17] Berniell, I. and J. Bietenbeck (2020), “The effect of working hours on health”, Economics & Human Biology, Vol. 39, p. 100901, https://doi.org/10.1016/j.ehb.2020.100901.

[59] Bertoni, M. et al. (2021), “Remote working and mental health during the first wave of Covid-19 pandemic”, IZA Discussion Paper Series, Vol. No. 14773.

[3] Beswick, J. and J. White (2003), “Working Long Hours”, Health and safety laboratory working papers HSL/2003/02, http://www.si-whp.eu/resources/working-long-hours.pdf (accessed on 9 February 2022).

[76] Biewen, M., B. Fitzenberger and J. de Lazzer (2018), “The role of employment interruptions and part-time work for the rise in wage inequality”, IZA Journal of Labor Economics, Vol. 7/1, p. 10, https://doi.org/10.1186/s40172-018-0070-y.

[117] Bloom, N. et al. (2015), “Does Working from Home Work? Evidence from a Chinese Experiment *”, The Quarterly Journal of Economics, Vol. 130/1, pp. 165-218, https://doi.org/10.1093/qje/qju032.

[129] Boltz, M. et al. (2020), “How Does Working-Time Flexibility Affect Workers’ Productivity in a Routine Job? Evidence from a Field Experiment”, SSRN Electronic Journal, https://doi.org/10.2139/SSRN.3725302.

[118] Bonacini, L., G. Gallo and S. Scicchitano (2020), “Working from home and income inequality: risks of a ‘new normal’ with COVID-19”, Journal of Population Economics, Vol. 34/1, pp. 303-360, https://doi.org/10.1007/s00148-020-00800-7.

[148] Brachet, T., G. David and A. Drechsler (2012), “The effect of shift structure on performance”, American Economic Journal: Applied Economics, Vol. 4/2, pp. 219-246, https://doi.org/10.1257/app.4.2.219.

[88] Brown, C. and D. Hamermesh (2019), “Wages and Hours Laws: What Do We Know? What Can Be Done?”, SSRN, https://doi.org/10.3386/W25942.

[23] Burke, R. et al. (2009), “Work hours, work intensity, satisfactions and psychological well-being among Turkish manufacturing managers”, Europe’s Journal of Psychology, Vol. 5/2, https://doi.org/10.5964/ejop.v5i2.264.

[80] Cahuc, P. et al. (2014), Labor Economics, MIT press.

[143] Cahuc, P. and S. Carcillo (2014), “The Detaxation of overtime hours: Lessons from the French experiment”, Journal of Labor Economics, Vol. 32/2, pp. 361-400, https://doi.org/10.1086/674987/SUPPL_FILE/11256DATA.ZIP.

[81] Calmfors, L. and M. Hoel (1988), “Work Sharing and Overtime”, The Scandinavian Journal of Economics, Vol. 90/1, p. 45, https://doi.org/10.2307/3440149.

[75] Carrillo-Tudela, C., A. Launov and J. Robin (2018), “The Fall in German Unemployment: A Flow Analysis”, IZA Discussion Paper Series No. 11442, https://www.iza.org/publications/dp/11442/the-fall-in-german-unemployment-a-flow-analysis (accessed on 14 December 2021).

[24] Caruso, C. et al. (2006), “Long working hours, safety, and health: Toward a national research agenda”, American Journal of Industrial Medicine, Vol. 49/11, pp. 930-942, https://doi.org/10.1002/ajim.20373.

[63] Cazes, S., A. Hijzen and A. Saint-Martin (2015), “Measuring and Assessing Job Quality: The OECD Job Quality Framework”, OECD Social, Employment and Migration Working Papers, No. 174, OECD Publishing, Paris, https://doi.org/10.1787/5jrp02kjw1mr-en.

[105] Cette, G., S. Chang and M. Konte (2011), “The decreasing returns on working time: an empirical analysis on panel country data”, Applied Economics Letters, Vol. 18/17, pp. 1677-1682, https://doi.org/10.1080/13504851.2011.558473.

[32] Charalampous, M. et al. (2019), “Systematically reviewing remote e-workers’ well-being at work: a multidimensional approach”, European Journal of Work and Organizational Psychology, Vol. 28/1, pp. 51-73, https://doi.org/10.1080/1359432X.2018.1541886.

[99] Chemin, M. and E. Wasmer (2009), “Using Alsace‐Moselle Local Laws to Build a Difference‐in‐Differences Estimation Strategy of the Employment Effects of the 35‐Hour Workweek Regulation in France”, Journal of Labor Economics, Vol. 27/4, pp. 487-524, https://doi.org/10.1086/605426.

[68] Cho, Y. (2018), “Part-time employment and worker health in the United States”, The Social Science Journal, Vol. 55/2, pp. 97-107, https://doi.org/10.1016/j.soscij.2017.09.004.

[66] Chung, H. and M. van der Horst (2020), “Flexible Working and Unpaid Overtime in the UK: The Role of Gender, Parental and Occupational Status”, Social Indicators Research, Vol. 151/2, pp. 495-520, https://doi.org/10.1007/s11205-018-2028-7.

[111] Chung, H. and M. van der Horst (2018), “Women’s employment patterns after childbirth and the perceived access to and use of flexitime and teleworking”, Human Relations, Vol. 71/1, pp. 47-72, https://doi.org/10.1177/0018726717713828.

[106] Collewet, M. and J. Sauermann (2017), “Working hours and productivity”, Labour Economics, Vol. 47, pp. 96-106, https://doi.org/10.1016/j.labeco.2017.03.006.

[139] Contensou, F. and R. Vranceanu (2000), “A Model of Working Time under Utility Competition in the Labor Market on JSTOR”, Journal of Economics, Vol. 67/2, pp. 145-166, https://www.jstor.org/stable/41770445.

[96] Costa, D. (2000), “Hours of Work and the Fair Labor Standards Act: A Study of Retail and Wholesale Trade, 1938-1950”, Industrial and Labor Relations Review, Vol. 53/4, p. 648, https://doi.org/10.2307/2696141.

[91] Crépon, B. and F. Kramarz (2002), “Employed 40 Hours or Not Employed 39: Lessons from the 1982 Mandatory Reduction of the Workweek”, Journal of Political Economy, Vol. 110/6, pp. 1355-1389, https://doi.org/10.1086/342807.

[93] Crépon, B., M. Leclair and S. Roux (2004), “RTT, productivité et emploi : nouvelles estimations sur données d’entreprises”, Economie et Statistique, Vol. 376/1, pp. 55-89, https://doi.org/10.3406/ESTAT.2004.7234.

[155] Criscuolo, C. et al. (2021), “The role of telework for productivity during and post-COVID-19: Results from an OECD survey among managers and workers”, OECD Productivity Working Papers, No. 31, OECD Publishing, Paris, https://doi.org/10.1787/7fe47de2-en.

[51] Curzi, Y., B. Pistoresi and T. Fabbri (2020), “Understanding the stressful implications of remote e-working: Evidence from Europe”, DEMB Working Paper Series No. 165, https://doi.org/10.25431/11380_1196359.

[37] De Menezes, L. and C. Kelliher (2017), “Flexible Working, Individual Performance, and Employee Attitudes: Comparing Formal and Informal Arrangements”, Human Resource Management, Vol. 56/6, pp. 1051-1070, https://doi.org/10.1002/hrm.21822.

[58] DeFilippis, E. et al. (2020), Collaborating During Coronavirus: The Impact of COVID-19 on the Nature of Work, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w27612.

[103] Delmez, F. and V. Vandenberghe (2017), Working long hours: less productive but less costly? Firm-level evidence from Belgium.

[7] Dembe, A. (2005), “The impact of overtime and long work hours on occupational injuries and illnesses: new evidence from the United States”, Occupational and Environmental Medicine, Vol. 62/9, pp. 588-597, https://doi.org/10.1136/oem.2004.016667.

[108] Dolton, P., C. Howorth and M. Abouaziza (2016), The Optimal Length of the Working Day : Evidence from Hawthorne Experiments., https://www.ucl.ac.uk/ioe/events/2017/oct/optimal-length-working-day-evidence-hawthorne-experiments.

[153] Elridge, L. and S. Pabilonia (2010), “Bringing work home: implications for BLS productivity measures”, Monthly Labor Review, Vol. 133/12, pp. 18-35, https://www.bls.gov/opub/mlr/2010/article/bringing-work-home-implications-for-bls-productivity-measures.htm (accessed on 2 May 2022).

[95] Estevão, M. and F. Sá (2006), “Are the French Happy with the 35-Hour Workweek?”, IMF Working papers, https://www.imf.org/-/media/Websites/IMF/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp06251.ashx.

[56] Eurofound and European Commission Joint Research Centre (2021), What just happened? COVID-19 lockdowns and change in the labour market, Publications Office of the European Union, Luxembourg, https://www.eurofound.europa.eu/publications/report/2021/what-just-happened-covid-19-lockdowns-and-change-in-the-labour-market (accessed on 22 November 2021).

[135] Fagnart, J., M. Germain and B. Van der Linden (2020), “Working Time Reduction and Employment in a Finite World | IZA - Institute of Labor Economics”, IZA DP, No. 13880, IZA, Bonn, https://www.iza.org/publications/dp/13880/working-time-reduction-and-employment-in-a-finite-world (accessed on 5 January 2022).

[47] Felstead, A. and G. Henseke (2017), “Assessing the growth of remote working and its consequences for effort, well-being and work-life balance”, New Technology, Work and Employment, Vol. 32/3, pp. 195-212, https://doi.org/10.1111/ntwe.12097.

[65] Fernández-Kranz, D. and N. Rodríguez-Planas (2011), “The part-time pay penalty in a segmented labor market”, Labour Economics, Vol. 18/5, pp. 591-606, https://doi.org/10.1016/j.labeco.2011.01.001.

[89] Fiole, M., M. Roger and H. Rouilleault (2002), “Les effets sur l’emploi de la loi du 11 juin 1996 sur la réduction du temps de travail ; suivi d’un commentaire de Henri Rouilleault”, Economie et Statistique, Vol. 357/1, pp. 3-22, https://www.persee.fr/doc/estat_0336-1454_2002_num_357_1_7661 (accessed on 26 November 2021).

[52] Frazis, H. (2020), “Who Telecommutes? Where is the Time Saved Spent?”, BLS Working Papers No. 523, https://www.bls.gov/osmr/research-papers/2020/ec200050.htm (accessed on 29 November 2021).

[14] Friedland, D. and R. Price (2003), “Underemployment: Consequences for the Health and Well-Being of Workers”, American Journal of Community Psychology, Vol. 32/1-2, pp. 33-45, https://doi.org/10.1023/A:1025638705649.

[140] Friesen, J. (2001), “Overtime pay regulation and weekly hours of work in Canada”, Labour Economics, Vol. 8/6, pp. 691-720, https://doi.org/10.1016/S0927-5371(01)00053-7.

[25] Frijters, P., D. Johnston and X. Meng (2009), “The mental health cost of long working hours: the case of rural Chinese migrants”, mimeo, https://conference.iza.org/conference_files/LabEco2009/frijters_p931.pdf (accessed on 9 February 2022).

[112] Fuller, S. and C. Hirsh (2019), ““Family-Friendly” Jobs and Motherhood Pay Penalties: The Impact of Flexible Work Arrangements Across the Educational Spectrum”, Work and Occupations, Vol. 46/1, pp. 3-44, https://doi.org/10.1177/0730888418771116.

[50] Gajendran, R. and D. Harrison (2007), “The good, the bad, and the unknown about telecommuting: Meta-analysis of psychological mediators and individual consequences.”, Journal of Applied Psychology, Vol. 92/6, pp. 1524-1541, https://doi.org/10.1037/0021-9010.92.6.1524.

[54] Giménez-Nadal, J., J. Molina and J. Velilla (2019), “Work time and well-being for workers at home: evidence from the American Time Use Survey”, International Journal of Manpower, Vol. 41/2, pp. 184-206, https://doi.org/10.1108/IJM-04-2018-0134.

[128] Glenn Dutcher, E. (2012), “The effects of telecommuting on productivity: An experimental examination. The role of dull and creative tasks”, Journal of Economic Behavior & Organization, Vol. 84/1, pp. 355-363, https://doi.org/10.1016/j.jebo.2012.04.009.

[126] Godart, O., H. Görg and A. Hanley (2017), “Trust-Based Work Time and Innovation: Evidence from Firm-Level Data”, ILR Review, Vol. 70/4, pp. 894-918, https://doi.org/10.1177/0019793916676259.

[36] Golden, L., J. Henly and S. Lambert (2012), “Work Schedule Flexibility for Workers: A Path to Employee Happiness?”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.2129520.

[102] Gomes, P. (2021), Friday is the New Saturday: How a Four-Day Working Week Will Save the Economy, The History Press Ltd, London.

[13] Gray, M. et al. (2004), “Long Work Hours and the Wellbeing of Fathers and their Families | Australian Journal of Labour Economics”, Australian Journal of Labour Economics, Vol. 7/4, pp. 255-273, https://search.informit.org/doi/abs/10.3316/ielapa.072475979217429.

[27] Grund, C. and K. Tilkes (2021), “Working Time Mismatch and Job Satisfaction – the Role of Employees’ Time Autonomy and Gender”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3934749.

[94] Gubian, A. et al. (2004), “Les effets de la RTT sur l’emploi : des estimations ex ante aux évaluations ex post”, Economie et Statistique, Vol. 376/1, pp. 25-54, https://doi.org/10.3406/ESTAT.2004.7233.

[11] Hamermesh, D., D. Kawaguchi and J. Lee (2017), “Does labor legislation benefit workers? Well-being after an hours reduction”, Journal of the Japanese and International Economies, Vol. 44, pp. 1-12, https://doi.org/10.1016/j.jjie.2017.02.003.

[145] Hart, A. and S. Krall (2007), “220: Productivity: Do 8-9 Hour Shifts Make a Difference?”, Annals of Emergency Medicine, Vol. 50/3, pp. S69-S70, https://doi.org/10.1016/j.annemergmed.2007.06.372.

[113] He, H., D. Neumark and Q. Weng (2021), “Do Workers Value Flexible Jobs? A Field Experiment”, Journal of Labor Economics, Vol. 39/3, pp. 709-738, https://doi.org/10.1086/711226.

[44] Henke, R. et al. (2016), “The Effects of Telecommuting Intensity on Employee Health”, American Journal of Health Promotion, Vol. 30/8, pp. 604-612, https://doi.org/10.4278/ajhp.141027-QUAN-544.

[12] Hewlett, S. and C. Luce (2006), “Extreme jobs: the dangerous allure of the 70-hour workweek.”, Harvard Business Review, Vol. 84/12, pp. 49-59, 162, https://europepmc.org/article/med/17183793 (accessed on 9 February 2022).

[15] Heyes, J. and M. Tomlinson (2021), “Underemployment and well-being in Europe”, Human Relations, Vol. 74/8, pp. 1240-1266, https://doi.org/10.1177/0018726720912297.

[110] Hijzen, A. and S. Thewissen (2020), “The 2018-2021 working time reform in Korea: A preliminary assessment”, OECD Social, Employment and Migration Working Papers, No. 248, OECD Publishing, Paris, https://doi.org/10.1787/0e828066-en.

[28] Holly, S. and A. Mohnen (2012), “Impact of Working Hours on Work-Life Balance”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.2135453.

[85] Hunt, J. (1999), “Has Work-Sharing Worked in Germany?”, The Quarterly Journal of Economics, Vol. 114/1, pp. 117-148, https://www.jstor.org/stable/2586949.

[30] Hurtado, D. et al. (2015), “Schedule control and mental health: the relevance of coworkers’ reports”, Community, Work & Family, Vol. 18/4, pp. 416-434, https://doi.org/10.1080/13668803.2015.1080663.

[90] Husson, M. (2002), “Réduction du temps de travail et emploi: une nouvelle évaluation”, Revue de l’IRES, Vol. 38, pp. 80-108, http://www.ires.fr/index.php/publications-de-l-ires/item/2757-reduction-du-temps-de-travail-et-emploi-une-nouvelle-evaluation.

[60] ILO (2020), Teleworking during the COVID-19 pandemic and beyond: A Practical Guide, International Labour Organization, Geneva, https://www.ilo.org/travail/info/publications/WCMS_751232/lang--en/index.htm.

[72] Joyce, K. et al. (2010), “Flexible working conditions and their effects on employee health and wellbeing”, Cochrane Database of Systematic Reviews, https://doi.org/10.1002/14651858.cd008009.pub2.

[79] Kapteyn, A., A. Kalwij and A. Zaidi (2004), “The myth of worksharing”, Labour Economics, Vol. 11/3, pp. 293-313, https://doi.org/10.1016/J.LABECO.2003.08.001.

[100] Kawaguchi, D., H. Naito and I. Yokoyama (2017), “Assessing the effects of reducing standard hours: Regression discontinuity evidence from Japan”, Journal of the Japanese and International Economies, Vol. 43, pp. 59-76, https://doi.org/10.1016/J.JJIE.2016.12.002.

[154] Ker, D., P. Montagnier and V. Spiezia (2021), “Measuring telework in the COVID-19 pandemic”, OECD Digital Economy Papers, No. 314, OECD Publishing, Paris, https://doi.org/10.1787/0a76109f-en.

[42] Kim, J. et al. (2020), “Workplace Flexibility and Worker Well‐Being by Gender”, Journal of Marriage and Family, Vol. 82/3, pp. 892-910, https://doi.org/10.1111/jomf.12633.

[6] Kivimäki, M. et al. (2015), “Long working hours and risk of coronary heart disease and stroke: a systematic review and meta-analysis of published and unpublished data for 603 838 individuals”, The Lancet, Vol. 386/10005, pp. 1739-1746, https://doi.org/10.1016/S0140-6736(15)60295-1.

[87] Kramarz, F. et al. (2008), “Working Time Developments in Germany”, in Boeri, T., M. Burda and F. Kramarz (eds.), Working Hours and Job Sharing in the EU and USA: Are Europeans Lazy? Or Americans Crazy?, Oxford University Press, https://doi.org/10.1093/ACPROF:OSO/9780199231027.003.0009.

[39] Kröll, C. and S. Nüesch (2019), “The effects of flexible work practices on employee attitudes: evidence from a large-scale panel study in Germany”, The International Journal of Human Resource Management, Vol. 30/9, pp. 1505-1525, https://doi.org/10.1080/09585192.2017.1289548.

[41] Krug, G., K. Kemna and K. Hartosch (2019), “Auswirkungen flexibler Arbeitszeiten auf die Gesundheit von Beschäftigten”, IAB-Discussion Paper, Vol. 1, http://hdl.handle.net/10419/204848 (accessed on 6 December 2021).

[53] Künn-Nelen, A. (2016), “Does Commuting Affect Health?”, Health Economics, Vol. 25/8, pp. 984-1004, https://doi.org/10.1002/hec.3199.

[16] Lee, J. and Y. Lee (2016), “Can working hour reduction save workers?”, Labour Economics, Vol. 40, pp. 25-36, https://doi.org/10.1016/j.labeco.2016.02.004.

[18] Lepinteur, A. (2019), “The shorter workweek and worker wellbeing: Evidence from Portugal and France”, Labour Economics, Vol. 58, pp. 204-220, https://doi.org/10.1016/j.labeco.2018.05.010.

[40] Lott, Y. and H. Chung (2016), “Gender Discrepancies in the Outcomes of Schedule Control on Overtime Hours and Income in Germany”, European Sociological Review, Vol. 32/6, pp. 752-765, https://doi.org/10.1093/esr/jcw032.

[144] Lu, S. and L. Lu (2017), “Do mandatory overtime laws improve quality? Staffing decisions and operational flexibility of nursing homes”, Management Science, Vol. 63/11, pp. 3566-3585, https://doi.org/10.1287/mnsc.2016.2523.

[116] Maestas, N. et al. (2018), The Value of Working Conditions in the United States and Implications for the Structure of Wages, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w25204.

[101] Manning, A. (2020), “Monopsony in Labor Markets: A Review”, ILR Review, Vol. 74/1, pp. 3-26, https://doi.org/10.1177/0019793920922499.

[137] Marimon, R. and F. Zilibotti (2000), “Employment and distributional effects of restricting working time”, European Economic Review, Vol. 44/7, pp. 1291-1326, https://doi.org/10.1016/S0014-2921(00)00032-5.

[115] Mas, A. and A. Pallais (2017), “Valuing Alternative Work Arrangements”, American Economic Review, Vol. 107/12, pp. 3722-3759, https://doi.org/10.1257/aer.20161500.

[34] Moen, P. et al. (2011), “Changing Work, Changing Health”, Journal of Health and Social Behavior, Vol. 52/4, pp. 404-429, https://doi.org/10.1177/0022146511418979.

[127] Monteiro, N., O. Straume and M. Valente (2019), “Does Remote Work Improve or Impair Firm Labour Productivity? Longitudinal Evidence from Portugal”, CESifo Working Paper Series, https://ideas.repec.org/p/ces/ceswps/_7991.html (accessed on 28 July 2020).

[70] Nikolova, M. and C. Graham (2014), “Employment, late-life work, retirement, and well-being in Europe and the United States”, IZA Journal of European Labor Studies, Vol. 3/1, https://doi.org/10.1186/2193-9012-3-5.

[131] OECD (2021), Pay Transparency Tools to Close the Gender Wage Gap, OECD Publishing, Paris, https://doi.org/10.1787/eba5b91d-en.

[1] OECD (2021), “Working time and its regulation in OECD countries: How much do we work and how?”, in OECD Employment Outlook 2021: Navigating the COVID-19 Crisis and Recovery, OECD Publishing, Paris, https://doi.org/10.1787/c18a4378-en.

[62] OECD (2020), OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis, OECD Publishing, Paris, https://doi.org/10.1787/1686c758-en.

[133] OECD (2019), Negotiating Our Way Up: Collective Bargaining in a Changing World of Work, OECD Publishing, Paris, https://doi.org/10.1787/1fd2da34-en.

[71] OECD (2018), OECD Employment Outlook 2018, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2018-en.

[2] OECD (2015), How’s Life? 2015: Measuring Well-being, OECD Publishing, Paris, https://doi.org/10.1787/how_life-2015-en.

[64] OECD (2010), “How Good is Part-Time Work?”, in OECD Employment Outlook 2010: Moving beyond the Jobs Crisis, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2010-5-en.

[132] OECD (forthcoming), “Teleworking through the gender looking glass: Facts and gaps”, OECD Social, Employment and Migration Working Papers, OECD Publishing, Paris.

[120] Oettinger, G. (2011), “The Incidence and Wage Consequences of Home-Based Work in the United States, 1980–2000”, Journal of Human Resources, Vol. 46/2, pp. 237-260, https://doi.org/10.3368/jhr.46.2.237.

[151] Olds, D. and S. Clarke (2010), “The effect of work hours on adverse events and errors in health care”, Journal of Safety Research, Vol. 41/2, pp. 153-162, https://doi.org/10.1016/j.jsr.2010.02.002.

[46] Pabilonia, S. and V. Vernon (2022), “Telework, Wages, and Time Use in the United States”, Review of Economics of the Household, https://doi.org/10.1007/s11150-022-09601-1.

[130] Pabilonia, S. and V. Vernon (2020), “Telework and Time Use in the United States”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3601959.

[104] Park, W. and Y. Park (2019), “When Less is More: The Impact of the Regulation on Standard Workweek on Labor Productivity in South Korea”, Journal of Policy Analysis and Management, Vol. 38/3, pp. 681-705, https://doi.org/10.1002/pam.22136.

[107] Pencavel, J. (2016), “Recovery from Work and the Productivity of Working Hours”, Economica, Vol. 83/332, pp. 545-563, https://doi.org/10.1111/ecca.12206.

[149] Pencavel, J. (2015), “The Productivity of Working Hours”, The Economic Journal, Vol. 125/589, pp. 2052-2076, https://doi.org/10.1111/ecoj.12166.

[73] Pirani, E. (2015), “On the Relationship Between Atypical Work(s) and Mental Health: New Insights from the Italian Case”, Social Indicators Research, Vol. 130/1, pp. 233-252, https://doi.org/10.1007/s11205-015-1173-5.

[57] Raghuram, S. et al. (2019), “Virtual Work: Bridging Research Clusters”, Academy of Management Annals, Vol. 13/1, pp. 308-341, https://doi.org/10.5465/annals.2017.0020.

[92] Raposo, P. and J. van Ours (2010), “How working time reduction affects jobs and wages”, Economics Letters, Vol. 106/1, pp. 61-63, https://doi.org/10.1016/j.econlet.2009.10.001.

[138] Rocheteau, G. (2002), “Working time regulation in a search economy with worker moral hazard”, Journal of Public Economics, Vol. 84/3, pp. 387-425, https://doi.org/10.1016/S0047-2727(01)00087-1.

[45] Rodríguez-Modroño, P. and P. López-Igual (2021), “Job Quality and Work—Life Balance of Teleworkers”, International Journal of Environmental Research and Public Health, Vol. 18/6, p. 3239, https://doi.org/10.3390/ijerph18063239.

[152] Rogers, A. et al. (2004), “The Working Hours Of Hospital Staff Nurses And Patient Safety”, Health Affairs, Vol. 23/4, pp. 202-212, https://doi.org/10.1377/hlthaff.23.4.202.

[20] Rudolf, R. (2013), “Work Shorter, Be Happier? Longitudinal Evidence from the Korean Five-Day Working Policy”, Journal of Happiness Studies, Vol. 15/5, pp. 1139-1163, https://doi.org/10.1007/s10902-013-9468-1.

[83] Sagyndykova, G. and R. Oaxaca (2019), “Raising the Overtime Premium and Reducing the Standard Workweek: Short-Run Impacts on U.S. Manufacturing”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3445823.

[26] Salo, P. et al. (2014), “Work Time Control and Sleep Disturbances: Prospective Cohort Study of Finnish Public Sector Employees”, Sleep, Vol. 37/7, pp. 1217-1225, https://doi.org/10.5665/sleep.3842.

[33] Samek Lodovici, M. (ed.) (2021), The impact of teleworking and digital work on workers and society, Publication for the committee on Employment and Social Affairs, Policy Department for Economic, Scientific and Quality of Life Policies, European Parliament, https://www.europarl.europa.eu/thinktank/en/document/IPOL_STU(2021)662904 (accessed on 17 May 2022).

[98] Sánchez, R. (2013), “Do reductions of standard hours affect employment transitions?: Evidence from Chile”, Labour Economics, Vol. 20, pp. 24-37, https://doi.org/10.1016/j.labeco.2012.10.001.

[61] Sasaki, N. et al. (2020), “Workplace responses to COVID‐19 associated with mental health and work performance of employees in Japan”, Journal of Occupational Health, Vol. 62/1, https://doi.org/10.1002/1348-9585.12134.

[136] Schank, T. (2015), “Employment effects of longer working hours”, IZA World of Labor, https://doi.org/10.15185/izawol.216.

[147] Schank, T. (2005), “Are overtime plants more efficient than standard-time plants? A stochastic production frontier analysis using the IAB Establishment Panel”, Empirical Economics 30, pp. 693-710, https://doi.org/10.1007/s00181-005-0257-5.

[146] Shepard, E. and T. Clifton (2000), “Are longer hours reducing productivity in manufacturing?”, International Journal of Manpower, Vol. 21/7, pp. 540-552, https://doi.org/10.1108/01437720010378999.

[97] Skuterud, M. (2007), “Identifying the Potential of Work‐Sharing as a Job‐Creation Strategy”, Journal of Labor Economics, Vol. 25/2, pp. 265-287, https://doi.org/10.1086/511379.

[48] Song, Y. and J. Gao (2020), “Does Telework Stress Employees Out? A Study on Working at Home and Subjective Well-Being for Wage/Salary Workers”, Journal of Happiness Studies, Vol. 21/7, pp. 2649-2668, https://doi.org/10.1007/s10902-019-00196-6.

[4] Sparks, K. et al. (1997), “The effects of hours of work on health: a meta analytic review”, Journal of occupational and environmental medicine.

[82] Steiner, V., R. Peters and V. Steiner (2000), “Employment effects of work sharing: an econometric analysis for West Germany”, https://www.econstor.eu/handle/10419/24367 (accessed on 7 January 2022).

[31] Tavares, A. (2017), “Telework and health effects review”, International Journal of Healthcare, Vol. 3/2, p. 30, https://doi.org/10.5430/ijh.v3n2p30.

[86] Trejo, S. et al. (2016), “Does the Statutory Overtime Premium Discourage Long Workweeks?”, ILR Review, Vol. 56/3, pp. 530-551, https://doi.org/10.1177/001979390305600310.

[29] Tucker, P. and S. Folkard (2012), “Working Time, Health and Safety: a Research Synthesis Paper”, Conditions of Work and Employment Series, International Labour Office No. 31, https://www.ilo.org/travail/info/publications/WCMS_181673/lang--en/index.htm.

[9] Useche, S. (ed.) (2020), “The negative impact of long working hours on mental health in young Korean workers”, PLOS ONE, Vol. 15/8, p. e0236931, https://doi.org/10.1371/journal.pone.0236931.

[8] Vegso, S. et al. (2007), “Extended work hours and risk of acute occupational injury: A case-crossover study of workers in manufacturing”, American Journal of Industrial Medicine, Vol. 50/8, pp. 597-603, https://doi.org/10.1002/ajim.20486.

[123] Viete, S. and D. Erdsiek (2018), Trust-Based Work Time and the Productivity Effects of Mobile Information Technologies in the Workplace, Elsevier BV, https://doi.org/10.2139/ssrn.3159890.

[10] Virtanen, M. et al. (2008), “Long Working Hours and Cognitive Function: The Whitehall II Study”, American Journal of Epidemiology, Vol. 169/5, pp. 596-605, https://doi.org/10.1093/aje/kwn382.

[43] Wanger, S. and I. Zapf (2021), “For better or worse: How more flexibility in working time arrangements and parental leave experiences affect fathers’ working and childcare hours in Germany”, Journal of Family Research, https://doi.org/10.20377/jfr-644.

[119] Weeden, K. (2005), “Is there a flexiglass ceiling? Flexible work arrangements and wages in the United States”, Social Science Research, Vol. 34/2, pp. 454-482, https://doi.org/10.1016/j.ssresearch.2004.04.006.

[78] Wheatley, D. (2017), “Employee satisfaction and use of flexible working arrangements”, Work, Employment and Society, Vol. 31/4, pp. 567-585, https://doi.org/10.1177/0950017016631447.

[121] White, D. (2019), “Agency Theory and Work from Home”, LABOUR, Vol. 33/1, pp. 1-25, https://doi.org/10.1111/labr.12135.

[114] Wiswall, M. and B. Zafar (2016), Preference for the Workplace, Human Capital, and Gender, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w22173.

[141] Yu, S. and D. Peetz (2018), “Non-Standard Time Wage Premiums and Employment Effects: Evidence from an Australian Natural Experiment”, British Journal of Industrial Relations, Vol. 57/1, pp. 33-61, https://doi.org/10.1111/bjir.12412.

[122] Yu, W. and J. Kuo (2017), “The Motherhood Wage Penalty by Work Conditions: How Do Occupational Characteristics Hinder or Empower Mothers?”, American Sociological Review, Vol. 82/4, pp. 744-769, https://doi.org/10.1177/0003122417712729.

[134] Zwickl, K., F. Disslbacher and S. Stagl (2016), “Work-sharing for a sustainable economy”, Ecological Economics, Vol. 121, pp. 246-253, https://doi.org/10.1016/J.ECOLECON.2015.06.009.

Annex 5.A. Additional information on the impact of working hours and working time arrangements on workers’ well-being

Regression specification

Effects of working time dimensions on the workers’ non-material well-being discussed in Section 5.1 are based on marginal effects derived from individual probit regressions calculated using repeated cross-section data with standard errors clustered at individual level (country fixed effect for estimates based on the European Social Survey) and year fixed effect controlling for demographic characteristics, household composition and income, job characteristics (including contract duration) and life events:

$P r (W i t = 1 | H i t, X i t, δ t) = Φ (H ’ i t α + X ’ i t β + δ t)$

Where i and t are individual and time suffices, $δ$ are year fixed-effects. W is the worker’s well-being outcome, H, the working hours or working time arrangement indicator and, X, the control variables (demographic, household and job characteristics of the individual and life events variables). The marginal effect of the working time indicator on worker’s well-being is then computed as:

$μ (H, α) = \frac{\partial Φ (H, α)}{\partial H}$

Data

The analysis was carried out on two types of data source (See Annex Table 5.A.1):

national panel data, which have the advantage of containing a wide range of information about individual and household characteristics, working time and life events, and,
cross-sectional social surveys data, which makes it possible to cover a large number of countries, but for a more limited set of variables.

As most of the indicators and data have only been available since the early 2000s, the analysis has been conducted, as far as possible, over the last two decades, and due to the obvious impact of the COVID‑19 pandemic on working time and working time arrangements, the years beyond 2019 have not been considered. Finally, the samples were reduced to keep only those common years with all indicators required for the analysis.

Survey	Country covered	Type of data	Years	Obs.
Household Income and Labour Dynamics in Australia (HILDA)	Australia	Panel data	2005‑19	132 189
Statistiques sur les ressources et conditions de vie (SRCV);	France	Panel data	2010‑19	66 216
Socio-Economic Panel (SOEP)	Germany	Panel data	2002, 2004, 2006, 2008, 2010, 2012, 2014, 2016 and 2018	114 038
Japan Household Panel Survey (KHPS/JHPS)	Japan	Panel data	2010‑17	8047
Korean Labor and Income Panel Survey (KLIPS)	Korea	Panel data	2005‑19	79 764
Swiss household Panel (SHP)	Switzerland	Panel data	2004‑19	69 822
UK Household Longitudinal Study or “Understanding Society” (UKHLS)	United Kingdom	Panel data	2010‑11, 2012‑13, 2014‑15, 2016‑17 and 2018‑19	98 162
European Social Survey (ESS)	Austria, Belgium, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Ireland, Israel, Italy, Lithuania, the Netherlands, Norway, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland and the United Kingdom	Cross-sectional social survey data	2010, 2012, 2014, 2016 and 2018	91 608

Self-assessed health and other non-material well-being outcomes

Workers’ health is analysed through two self-assessed indicators: health satisfaction and limitations due to physical and/or mental health problems (see Annex Table 5.A.2).

Health satisfaction is generally measured in the different surveys on a scale from 0 (strongly dissatisfied) to 10 (strongly satisfied) and grouped into dummy variables with satisfied employees defined as those with a score between 6 and 10. However for countries covered by the European Social Social Survey (ESS), France and Korea, health satisfaction refers to self-assessed health conditions with employees satisfied by their health as those in good or very health condition.

Limitations due to health problem are defined as the occurrence of these limitations on work activity (yes or no) for Australia and Korea, and, as the frequency of these limitations on daily activity and work on a scale ranging from 1 (always) to 5 (never) for countries covered by the European Social Social Survey (ESS), France, Germany and the United Kingdom. For Switzerland, the question refers to the intensity of limitations coded on a scale of 1 (not at all) to 10 (a great deal). The reference period to which the employee’s assessment of these limitations refers varies from one survey to another: over the last four weeks for Australia, Germany and the United Kingdom, over the last six months for France and Korea and a general assessment without a reference period for the European Social Survey and Switzerland.

Country	Health satisfaction	No limitations due to health problem
Australia (HILDA)	Q: “All things considered, how satisfied are you with your health?” A: Scale from 0: “Totally dissatisfied” to 10: “Totally satisfied” Recoded as: 0 = 0‑5 Not satisfied 1 = 6‑10 Satisfied	Q: “During the past four weeks, have you had any of the following problems with your work or other regular daily activities as a result of your physical health?” A: “Had difficulty performing the work or other activities (for example, it took extra effort)” Q: “During the past four weeks, have you had any of the following problems with your work or other regular daily activities as a result of any emotional problems (such as feeling depressed or anxious)?” A: “Didn’t do work or other activities as carefully as usual” Recoded as: 0= if answer is yes to one of these questions: Limitations due to physical health problem 1= if answers to both question is no: No limitations due to physical health problem
Europe and Israel (ESS)	Q: “How is your health (physical and mental health) in general?” A: 1: “very good”; 2: “good”; 3: “fair”; 4: “poor”; 5: “very poor” Recoded as: 0=4 or 5: poor health condition 1=1 to 3: good health condition	Q: “Are you hampered (i.e. limited, restricted) in your daily activities in any way by any longstanding illness, or disability, infirmity or mental health problem?” A: 1 “Yes a lot”; 2 “Yes to some extent”; and 3 “No” Recoded as: 0 = 1 or 2: Limitations due to health problem 1 = 3: No limitations due to health problem
France (SRCV)	Q: “How is your health in general?” A: 1: “very good”; 2: “good”; 3: “fair”; 4: “poor”; 5: “very poor” Recoded as: 0=4 or 5: poor health condition 1=1 to 3: good health condition	Q: “Have you been limited for at least six months by a health problem (i.e. discomfort, difficulties, after-effects of accidents) in the activities that people usually do?” A: 1 “Yes, strongly limited”; 2 “Yes, limited”; and 3 “No, not limited” Recoded as: 0 = 1 or 2: Limitations due to health problem 1 = 3: No limitations due to health problem
Germany (SOEP)	Q: “How satisfied are you today with your health?” A: Scale from 0: “completely dissatisfied” to 10: “completely satisfied” Recoded as: 0 = 0‑5 Not satisfied 1 = 6‑10 Satisfied	Q: “How often in the last four weeks, due to health problems of a physical nature, have you been restricted in the type of tasks you can perform in your work or everyday activities?” A: Scale from 1: Always to 5: Never Recoded as: Q: “How often in the last four weeks, due to psychological or emotional problems, did you achieve less in your work or everyday activities than you actually intended?” A: Scale from 1: Always to 5: Never Recoded as: 0= if answers 1 or 2 to one of these questions: Limitations due to physical health problem 1= if answers 3 to 5 to both question: No or few limitations due to physical health problem
Japan (KHPS/ JHPS)	Q: “How do you feel about the present situation regarding your health?” A: Scale from 0: “not at all satisfied” to 10: “fully satisfied” Recoded as: 0 = 0‑5 Not satisfied 1 = 6‑10 Satisfied
Korea (KLIPS)	Q: “How would you rate your overall health?” A: 1: “excellent”; 2 “good”; 3 “fair”; 4 “poor”; 5 “very poor” Recoded as: 0=4 or 5: poor health condition 1=1 to 3: good health condition	Q: “Have you had any persistent – i.e. six months or longer – difficulties in the following activities by physical, mental or emotional conditions? “ A: “Difficulties in working (economic activity)” 0= Limitations due to physical health problem 1= No limitations due to physical health problem
Switzerland (SHP)	Q: “How satisfied are you with your state of health, if 0 means “not at all satisfied” and 10 “completely satisfied”?” Recoded as: 0 = 0‑5 Not satisfied 1 = 6‑10 Satisfied	Q: “Please tell me to what extent, generally, your health is an impediment in your everyday activities, in your housework, your work or leisure activities? 0 means not at all and 10 a great deal” A: Scale from 0: Not at all to 10: Very strongly Recoded as: 0 = 5‑10 Limitations due to physical health problem 1 = 0‑4 No or few limitations due to physical health problem
United Kingdom (UKHLS)	Q: “On a scale of 1 to 7 where 1 means ‘Completely dissatisfied’ and 7 means ‘Completely satisfied’, how dissatisfied or satisfied are you with your health?” Recoded as: 0 = 1‑4 Not satisfied 1 = 5‑7 Satisfied	Q: “During the past four weeks, how much of the time have you had any of the following problems with your work or other regular daily activities as a result of your physical health? A: “Accomplished less than you would like”. Scale from 1: “All of the time” to 5: “None of the time” Q: “During the past four weeks, how much of the time have you had any of the following problems with your work or other regular daily activities as a result of any emotional problems (such as feeling depressed or anxious)? A: “Accomplished less than you would like”. Scale from 1: All of the time to 5: None of the time Recoded as: 0= if answers 1 or 2 to one of these questions: Limitations due to physical health problem 1= if answers 3 to 5 to both question: No or few limitations due to physical health problem

Country

Health satisfaction

No limitations due to health problem

Australia

(HILDA)

Q: “All things considered, how satisfied are you with your health?”

A: Scale from 0: “Totally dissatisfied” to 10: “Totally satisfied”

Recoded as:

0 = 0‑5 Not satisfied

1 = 6‑10 Satisfied

Q: “During the past four weeks, have you had any of the following problems with your work or other regular daily activities as a result of your physical health?”

A: “Had difficulty performing the work or other activities (for example, it took extra effort)”

Q: “During the past four weeks, have you had any of the following problems with your work or other regular daily activities as a result of any emotional problems (such as feeling depressed or anxious)?”

A: “Didn’t do work or other activities as carefully as usual”

Recoded as:

0= if answer is yes to one of these questions: Limitations due to physical health problem

1= if answers to both question is no: No limitations due to physical health problem

Europe and Israel

(ESS)

Q: “How is your health (physical and mental health) in general?”

A: 1: “very good”; 2: “good”; 3: “fair”; 4: “poor”; 5: “very poor”

Recoded as:

0=4 or 5: poor health condition

1=1 to 3: good health condition

Q: “Are you hampered (i.e. limited, restricted) in your daily activities in any way by any longstanding illness, or disability, infirmity or mental health problem?”

A: 1 “Yes a lot”; 2 “Yes to some extent”; and 3 “No”

Recoded as:

0 = 1 or 2: Limitations due to health problem

1 = 3: No limitations due to health problem

France

(SRCV)

Q: “How is your health in general?”

A: 1: “very good”; 2: “good”; 3: “fair”; 4: “poor”; 5: “very poor”

Recoded as:

0=4 or 5: poor health condition

1=1 to 3: good health condition

Q: “Have you been limited for at least six months by a health problem (i.e. discomfort, difficulties, after-effects of accidents) in the activities that people usually do?”

A: 1 “Yes, strongly limited”; 2 “Yes, limited”; and 3 “No, not limited”

Recoded as:

0 = 1 or 2: Limitations due to health problem

1 = 3: No limitations due to health problem

Germany

(SOEP)

Q: “How satisfied are you today with your health?”

A: Scale from 0: “completely dissatisfied” to 10: “completely satisfied”

Recoded as:

0 = 0‑5 Not satisfied

1 = 6‑10 Satisfied

Q: “How often in the last four weeks, due to health problems of a physical nature, have you been restricted in the type of tasks you can perform in your work or everyday activities?”

A: Scale from 1: Always to 5: Never

Recoded as:

Q: “How often in the last four weeks, due to psychological or emotional problems, did you achieve less in your work or everyday activities than you actually intended?”

A: Scale from 1: Always to 5: Never

Recoded as:

0= if answers 1 or 2 to one of these questions: Limitations due to physical health problem

1= if answers 3 to 5 to both question: No or few limitations due to physical health problem

Japan

(KHPS/

JHPS)

Q: “How do you feel about the present situation regarding your health?”

A: Scale from 0: “not at all satisfied” to 10: “fully satisfied”

Recoded as:

0 = 0‑5 Not satisfied

1 = 6‑10 Satisfied

Korea

(KLIPS)

Q: “How would you rate your overall health?”

A: 1: “excellent”; 2 “good”; 3 “fair”; 4 “poor”; 5 “very poor”

Recoded as:

0=4 or 5: poor health condition

1=1 to 3: good health condition

Q: “Have you had any persistent – i.e. six months or longer – difficulties in the following activities by physical, mental or emotional conditions? “

A: “Difficulties in working (economic activity)”

0= Limitations due to physical health problem

1= No limitations due to physical health problem

Switzerland

(SHP)

Q: “How satisfied are you with your state of health, if 0 means “not at all satisfied” and 10 “completely satisfied”?”

Recoded as:

0 = 0‑5 Not satisfied

1 = 6‑10 Satisfied

Q: “Please tell me to what extent, generally, your health is an impediment in your everyday activities, in your housework, your work or leisure activities? 0 means not at all and 10 a great deal”

A: Scale from 0: Not at all to 10: Very strongly

Recoded as:

0 = 5‑10 Limitations due to physical health problem

1 = 0‑4 No or few limitations due to physical health problem

United Kingdom

(UKHLS)

Q: “On a scale of 1 to 7 where 1 means ‘Completely dissatisfied’ and 7 means ‘Completely satisfied’, how dissatisfied or satisfied are you with your health?”

Recoded as:

0 = 1‑4 Not satisfied

1 = 5‑7 Satisfied

Q: “During the past four weeks, how much of the time have you had any of the following problems with your work or other regular daily activities as a result of your physical health?

A: “Accomplished less than you would like”. Scale from 1: “All of the time” to 5: “None of the time”

Q: “During the past four weeks, how much of the time have you had any of the following problems with your work or other regular daily activities as a result of any emotional problems (such as feeling depressed or anxious)?

A: “Accomplished less than you would like”. Scale from 1: All of the time to 5: None of the time

Recoded as:

0= if answers 1 or 2 to one of these questions: Limitations due to physical health problem

1= if answers 3 to 5 to both question: No or few limitations due to physical health problem

Note: Q: question asked; A: Answers.

Other non-material well-being outcomes (see Annex Table 5.A.3) refer to the evaluation of employees’ satisfaction with life in general, health, current job, work-life balance and satisfaction with free time. Satisfaction is generally measured in the different surveys on a scale from 0 (strongly dissatisfied) to 10 (strongly satisfied). In the case of the United Kingdom, satisfaction is measured on a scale from 1 (completely dissatisfied) to 7 (completely satisfied) and for Korea on a scale from 1 (very satisfied) to 5: (very dissatisfied). Work-life balance indicators are generally based on questions about the difficulties to reconcile work and family obligations excepted for Germany where this variable refers to the satisfaction with housework. Satisfaction with free time refers to satisfaction with leisure without any clear reference to time spent for France.

Outcome	Country	Question	Answer	Recoding
1. Life satisfaction	Australia (HILDA)	“All things considered, how satisfied are you with your life?”	Scale from 0: Totally dissatisfied to 10: Totally satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Europe and Israel (ESS)	“All things considered, how satisfied are you with your life as a whole nowadays?”	Scale from 0: Extremely dissatisfied to 10: Extremely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	France (SRCV)	“On a scale from 0 (not at all satisfied) to 10 (completely satisfied), indicate your overall life satisfaction”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Germany (SOEP)	“How satisfied are you today with your life?”	Scale from 0: completely dissatisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Japan (KHPS/JHPS)	“How do you feel about the present situation regarding overall life?”	Scale from 0: not at all satisfied to 10: fully satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Korea (KLIPS)	“Overall, how satisfied or dissatisfied are you with your life?”	1: very satisfied; 2: satisfied; 3: neither satisfied nor dissatisfied; 4: dissatisfied; 5: very dissatisfied	Not satisfied: answers 3 to 5 Satisfied: answers 1 to 2
	Switzerland (SHP)	“In general, how satisfied are you with your life if 0 means “not at all satisfied” and 10 means “completely satisfied”?”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	United Kingdom (UKHLS)	“On a scale of 1 to 7 where 1 means ‘Completely dissatisfied’ and 7 means ‘Completely satisfied’, how dissatisfied or satisfied are you with your life overall?”	Scale from 1: Completely dissatisfied to 7: Completely satisfied	Not satisfied: answers 1 to 4 Satisfied: answers 5 to 7
2. Job satisfaction	Australia (HILDA)	“All things considered, how satisfied are you with your job?”	Scale from 0: Totally dissatisfied to 10: Totally satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	France (SRCV)	“On a scale from 0 (not at all satisfied) to 10 (completely satisfied), indicate your satisfaction with main job”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Germany (SOEP)	“How satisfied are you today with your job?”	Scale from 0: completely dissatisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Japan (KHPS/JHPS)	“How do you feel about the present situation regarding your employment?”	Scale from 0: not at all satisfied to 10: fully satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Korea (KLIPS)	“What is your feelings regarding your current job (work, tasks)?” I am satisfied with my current job	1: strongly disagree, 2: disagree 3: neutral; 4: agree; 5: strongly agree	Not satisfied: answers 1 to 3 Satisfied: answers 4 or 5
	Switzerland (SHP)	“On a scale from 0 “not at all satisfied” to 10 “completely satisfied” can you indicate your degree of satisfaction for each of the following points?”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	United Kingdom (UKHLS)	“On a scale of 1 to 7 where 1 means ‘Completely dissatisfied’ and 7 means ‘Completely satisfied’, how dissatisfied or satisfied are you with your present job?”	Scale from 1: Completely dissatisfied to 7: Completely satisfied	Not satisfied: answers 1 to 4 Satisfied: answers 5 to 7
3. Work-life balance	Australia (HILDA)	“All things considered, how satisfied are you with the flexibility available to balance work and non-work commitments (in your main job)?”	Scale from 0: Totally dissatisfied to 10: Totally satisfied	Bad work-life balance: answers 0 to 5 Good work-life balance: answers 6 to 10
	France (SRCV)	“Do you find it difficult to reconcile work and family obligations?”	1 “Always”; 2 “Often”; 3 “Sometimes”; 4 “Never”	Bad work-life balance: answers 1 or 2 Good work-life balance: answers 3 or 4
	Germany (SOEP)	“How satisfied are you today with housework?”	Scale from 0: completely dissatisfied to 10: completely satisfied	Bad work-life balance: answers 0 to 5 Good work-life balance: answers 6 to 10
	Switzerland (SHP)	“How strongly does your work interfere with your private activities and family obligations, more than you would want this to be, if 0 means “not at all” and 10 “very strongly”?”	Scale from 0: not at all to 10: very strongly	Bad work-life balance: answers 5 to 10 Good work-life balance: answers 0 to 4
4. Satisfaction with free time	Australia (HILDA)	“All things considered, how satisfied are you with the amount of free time you have?”	Scale from 0: Totally dissatisfied to 10: Totally satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	France (SRCV)	“On a scale from 0 (not at all satisfied) to 10 (completely satisfied), indicate your satisfaction with leisure”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Germany (SOEP)	“How satisfied are you today with your leisure time?”	Scale from 0: completely dissatisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Japan (KHPS/JHPS)	“How do you feel about the present situation regarding your amount of leisure time?”	Scale from 0: not at all satisfied to 10: fully satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	Switzerland (SHP)	“How satisfied are you with the amount of free time you have, if 0 means “not at all satisfied” and 10 “completely satisfied”?”	Scale from 0: not at all satisfied to 10: completely satisfied	Not satisfied: answers 0 to 5 Satisfied: answers 6 to 10
	United Kingdom (UKHLS)	“On a scale of 1 to 7 where 1 means ‘Completely dissatisfied’ and 7 means ‘Completely satisfied’, how dissatisfied or satisfied are you with the amount of leisure time you have?”	Scale from 1: Completely dissatisfied to 7: Completely satisfied	Not satisfied: answers 1 to 4 Satisfied: answers 5 to 7

Working hours and working-time arrangements

The impact of working time and working time arrangement (see Annex Table 5.A.4) is assessed through actual weekly working hours, reasons for part-time work, hour’s mismatches, telework and flexitime. Actual weekly working hours include overtime hours and exclude commuting hours (and when specified, excluding meal time) and have been recoded into dummy variables according to an increasing threshold of hours worked (from 20 to 55 hours) and then included one by one in the regressions in order to test the impact of increasing hours on workers’ well-being. Involuntary part-time, voluntary part-time job (caring reasons) and voluntary part-time job (free choice) are based on employees’ stated reasons for working part-time and are available for Australia, France, Germany (only for involuntary part time), Japan and Switzerland. Involuntary part-time refers to employees who could not find a full-time job. In the case of Germany, this indicator refers to part-time employees for whom the hours usually worked are not sufficient. A voluntary part-time job by choice refers to part-time employees who prefer part-time job (Australia and Japan) or are not interested in full-time job (France and Switzerland); voluntary part-time for caring reasons refers to employees with a part-time job due to own illness or disability, cares for children, disabled or elderly relatives or other personal or family responsibilities. For Japan, voluntary part-time for caring reason refers to employees holding a part-time job because they cannot work full-time due to personal or other reasons.

Hours mismatches (wish to work more or less) are based on preferred weekly hours worked that employees wish to work taking into account that this change may affect their income. This indicator is only available for Australia, Germany and Switzerland. For the latter, the question asked does not explicitly take into account how income may be affected (“How many hours a week would you like to work as regards your main activity?”).

The information on telework is very close to the usual definition for Australia and the United Kingdom. Indeed, the HILDA and Understanding Society surveys identify teleworking employees by asking, firstly, whether this form of organisation is agreed or authorised by the employer and, secondly, whether the employee works at least one hour at telework. For Germany, telework refers to employees working at home, and, for Switzerland, to employees working at home for overtime, always or sometimes and using a computer at work.

Information on employees working under flexitime arrangement refers to a specific question on this form of working time arrangement for the United Kingdom while for Australia, Germany, Korea and the United Kingdom, flexitime is derived from question about work scheduling and the ability of employees to determine their working hours. Flexitime refers to employees who consider that their working hours can be flexible for Australia (score of 6 to 10 on a scale of 0: strongly disagree to 10: strongly agree), to employees who decide their own working hours for Germany (“self-determined working time”) and Japan (“Flex time system, i.e. self-starting and ending time self-adjustment within certain hours”), Korea (“working hours determined at employee discretion”) and Switzerland (Hours varied from day to day and decided by the employee).

Survey	Hours mismatches	Reasons for PT job	Telework	Flexitime
Australia (HILDA)	Q: “If you could choose the number of hours you work each week, and taking into account how that would affect your income, would you prefer to work”	Q: “You have said that (currently) you usually work fewer than 35 hours per week. What is the main reason for your working part-time hours rather than full-time hours?” Recoded as: Involuntary PT job: 6 “Could not find full-time work” or 12 “Prefer job & part-time hours are a requirement of the job” Voluntary PT job (choice): 7 “Prefer part-time work” Voluntary PT job (caring reasons): 1 “Own illness or disability”, 2 “Caring for children”, 3 “Caring for disabled or elderly relatives (not children)” and 4 “Other personal or family responsibilities”	Q: “Are the hours worked at home the result of a formal arrangement with your employer?” A: 1 “Yes” Q: “In your main job, are any of your usual working hours worked at your home (that is, the address of your usual place of residence)? A: 1 “Yes”	Q: “My working times can be flexible” A: Scale from 1: strongly disagree to 7: Strongly agree Recoded as: 1 = 6‑7 Flexitime 0 = 1‑5 No flexitime
France (SRCV)		A: Which of the following is the main reason why you work, on average, less than 30 hours per week (in all jobs)? Recoded as: Involuntary PT job: 4. “Although you would like to work more, you cannot find a job with more hours” or 6. “The combined hours of all your jobs are equivalent to full time” Voluntary PT job (choice): 5. “You do not want to work more” Voluntary PT job (caring reasons): 2. “You have health problems (illness or disability)” or 3. “You are involved in housekeeping, childcare or other of children or other people”
Germany (SOEP)	Q: “If you could choose your own working hours, taking into account that your income would change according to the number of hours: How many hours would you want to work?”	Involuntary PT job proxied using PT status and wish to work less.	Q: “Does it happen that you do your job at home?” A: 1 “Yes”	Q: “There are very different working arrangements nowadays. Which of the following applies to your work best?” A: 3 “Self-determined working time”
Japan (KHPS/ JHPS)	-	Q: “Why do you work under that work status?” Recoded as: Involuntary PT job: 1 “I wanted to work as a regular employee but no company would hire me” Voluntary PT job (choice): 2 “The wages and working terms and conditions are good” Voluntary PT job (caring reasons): 3 “I cannot work as a regular employee due to personal reasons” or 4 “Other”	-	Q: “Which of the following is closest to your work system (working hours system)?” A: 2 “Flex time system (self-starting and ending time self-adjustment within certain hours)”
Korea (KLIPS)				Q: “How work hours are determined?” A: 4 “Own discretion”
Switzerland (SHP)	Q: “How many hours a week would you like to work as regards your main activity?”	Q: “Why do you work part-time?” Recoded as: Involuntary PT job: 4 “because you could not find a full-time job” Voluntary PT job (choice): 5 “because you are not interested in working full-time” Voluntary PT job (caring reasons): 1 “for family reasons/caring for children or relatives” or 3 “because of a disability or illness”	Q: “Do you sometimes work at home?” A: 1 Yes, overtime, 2 Yes, occasionally, 3 Yes, always Q: “Do you personally use a computer in your job?” A: 1 “Yes”	Q: “Are your working hours…” (type of working hours) A: 4 “Varies from day to day, you decide”
United Kingdom (UKHLS)			Q: I would like to ask about working arrangements at the place where you work. Which of the following arrangements are available at your workplace? Work from home on a regular basis” A: 1 “Yes” Q: “Do you currently work in any of these ways? Work from home on a regular basis” A: 1 “Yes”	Q: I would like to ask about working arrangements at the place where you work. Which of the following arrangements are available at your workplace? Flexi-time” A: 1 “Yes” Q: “Do you currently work in any of these ways? Flexi-time” A: 1 “Yes”

Survey

Hours mismatches

Reasons for PT job

Telework

Flexitime

Australia

(HILDA)

Q: “If you could choose the number of hours you work each week, and taking into account how that would affect your income, would you prefer to work”

Q: “You have said that (currently) you usually work fewer than 35 hours per week. What is the main reason for your working part-time hours rather than full-time hours?”

Recoded as:

Involuntary PT job: 6 “Could not find full-time work” or 12 “Prefer job & part-time hours are a requirement of the job”

Voluntary PT job (choice): 7 “Prefer part-time work”

Voluntary PT job (caring reasons): 1 “Own illness or disability”, 2 “Caring for children”, 3 “Caring for disabled or elderly relatives (not children)” and 4 “Other personal or family responsibilities”

Q: “Are the hours worked at home the result of a formal arrangement with your employer?”

A: 1 “Yes”

Q: “In your main job, are any of your usual working hours worked at your home (that is, the address of your usual place of residence)?

A: 1 “Yes”

Q: “My working times can be flexible”

A: Scale from 1: strongly disagree to 7: Strongly agree

Recoded as:

1 = 6‑7 Flexitime

0 = 1‑5 No flexitime

France

(SRCV)

A: Which of the following is the main reason why you work, on average, less than 30 hours per week (in all jobs)?

Recoded as:

Involuntary PT job: 4. “Although you would like to work more, you cannot find a job with more hours” or 6. “The combined hours of all your jobs are equivalent to full time”

Voluntary PT job (choice): 5. “You do not want to work more”

Voluntary PT job (caring reasons): 2. “You have health problems (illness or disability)” or 3. “You are involved in housekeeping, childcare or other

of children or other people”

Germany

(SOEP)

Q: “If you could choose your own working hours, taking into account that your income would change according to the number of hours: How many hours would you want to work?”

Involuntary PT job proxied using PT status and wish to work less.

Q: “Does it happen that you do your job at home?”

A: 1 “Yes”

Q: “There are very different working arrangements nowadays. Which of the following applies to your work best?”

A: 3 “Self-determined working time”

Japan

(KHPS/

JHPS)

Q: “Why do you work under that work status?”

Recoded as:

Involuntary PT job: 1 “I wanted to work as a regular employee but no company would hire me”

Voluntary PT job (choice): 2 “The wages and working terms and conditions are good”

Voluntary PT job (caring reasons): 3 “I cannot work as a regular employee due to personal reasons” or 4 “Other”

Q: “Which of the following is closest to your work system (working hours system)?”

A: 2 “Flex time system (self-starting and ending time self-adjustment within certain hours)”

Korea

(KLIPS)

Q: “How work hours are determined?”

A: 4 “Own discretion”

Switzerland

(SHP)

Q: “How many hours a week would you like to work as regards your main activity?”

Q: “Why do you work part-time?”

Recoded as:

Involuntary PT job: 4 “because you could not find a full-time job”

Voluntary PT job (choice): 5 “because you are not interested in working full-time”

Voluntary PT job (caring reasons): 1 “for family reasons/caring for children or relatives” or 3 “because of a disability or illness”

Q: “Do you sometimes work at home?”

A: 1 Yes, overtime, 2 Yes, occasionally, 3 Yes, always

Q: “Do you personally use a computer in your job?”

A: 1 “Yes”

Q: “Are your working hours…” (type of working hours)

A: 4 “Varies from day to day, you decide”

United Kingdom

(UKHLS)

Q: I would like to ask about working arrangements at the place where you work. Which of the following arrangements are available at your workplace? Work from home on a regular basis”

A: 1 “Yes”

Q: “Do you currently work in any of these ways? Work from home on a regular basis”

A: 1 “Yes”

Q: I would like to ask about working arrangements at the place where you work. Which of the following arrangements are available at your workplace? Flexi-time”

A: 1 “Yes”

Q: “Do you currently work in any of these ways? Flexi-time”

A: 1 “Yes”

Note: Q: question asked; A: Answers.

Additional control variables

To address composition effects and factors that may affect the perception of workers’ non-material well-being, regressions include four types of variable controls:

Socio-demographic characteristics of the employee: sex, age group, marital status and migration status:

Household characteristics: household composition (number of members and number of children aged 0‑4, 5‑9 and 10‑14) and deciles of the gross household income;

Job characteristics: job autonomy, contract duration (permanent vs temporary or regular vs irregular contract for Korea), type of contract (full-time / part-time job), existence of other jobs, job tenure, occupation, supervisory responsibilities, hourly earnings deciles, industry, sector (public / private), firm size and, where possible, overtime rules and compensation; and life events last year: pregnancy and/or birth, death of close relatives or friends, change of residence, change in marital status, change in working life (promotion, separation etc.), own illness or illness of household member, other serious life events (violence, conflict, jail etc.).

In the case of the European Social Survey (ESS), control variables on job characteristics and life event are more limited in number than for the other data sources and the results from this survey should be compared with caution (see Annex Table 5.A.5). For the other countries, the availability of the first three sets of controls is relatively complete except for job autonomy (only available for Australia and the United Kingdom), the existence of side job (Germany and Korea), job tenure (France and Switzerland but this variable is approximated by work experience and the United Kingdom due to lack of reliable information on the start dates of current employment), supervisory responsibilities (Germany, Japan and Korea) and overtime rules and compensation (only available for Germany and Korea). Life event contain a heterogeneous amount of information that was not always possible to find or derive in most surveys. The Australian data contains by far the most information as specific questions on life events in the past year were asked. For the other panel data (France, Germany, Japan, Korea, Switzerland and the United Kingdom) information was derived from information on household members and information for the same individual last year. For the European Social Survey (ESS), information is scarce and relate mainly on violence and discrimination.

Control variable	ESS	GWP	HILDA	SRCV	SOEP	KHPS /JHPS	KLIPS	SHP	UKHLS
1. Demographic characteristics
Sex, age groups and education	●	●	●	●	●	●	●	●	●
Marital status	●	●	●	●	●	●	●	●	●
Migration status	●	●	●	●	●			●	●
2. Household characteristics
Region of residence	●	●	●	●	●	●	●	●	●
household income deciles	●	●	●	●	●	●	●	●	●
Number of HH members	●	●	●	●	●	●	●	●	●
Number of child(ren)	●	●	●	●	●	●	●	●	●
3. Job characteristics
Job autonomy	●	●[1]	●						●
Contract duration (permanent vs temporary)	●		●	●	●	●[2]	●[2]	●	●
Contract (FT vs PT)		●	●	●	●	●	●	●	●
Side jobs			●	●		●		●	●
Job tenure			●	●[3]	●	●	●	●[3]
Occupation	●		●	●	●	●	●	●	●
Supervisory responsibilities			●	●				●	●
Hourly earnings deciles			●	●	●	●	●	●	●
Industry	●		●	●	●	●	●	●	●
Sector (public vs private)	●		●	●	●	●	●	●	●
Firm size	●		●	●	●	●	●	●	●
Overtime rules and compensation					●		●
4. Life events since last year
Pregnancy and/or birth			●	●	●	●	●	●	●
Death of close relatives or friends			●	●	●	●	●	●
Change of residence			●	●	●	●	●	●	●
Change in marital status			●	●	●	●	●	●	●
Change in working life			●	●		●			●
Own illness or illness of household member			●					●
Other serious life events (violence, conflict, jail etc.)	●	●	●					●

Note: [1]: Employees engaged in their job. [2] Regular versus irregular contracts according to the national definition. [3] Total work experience.in regular paid job. ESS: European Social Survey; GWP : Gallup World poll; HILDA: Household Income and Labour Dynamics in Australia; SRCV: Statistiques sur les ressources et conditions de vie (France); SOEP: Socio-Economic Panel (Germany); KHPS / JHPS: Japan Household Panel Survey (Japan); KLIPS: Korean Labor and Income Panel Survey (Korea); SHP: Swiss household Panel (Switzerland); UKHLS: UK Household Longitudinal Study or “Understanding Society” (United Kingdom).

Annex 5.B. Additional information on the impact of working time reductions in Europe

	A. Discrete treatment variable	B. Continuous measure of initial exposure
Percentage share of workers affected by reform	‑4.863*** (1.369)	‑4.773*** (1.381)	‑34.124*** (10.939)	‑33.909*** (10.933)
Log of average annual hours per worker	‑0.014*** (0.004)	‑0.013*** (0.004)	‑0.063*** (0.018)	‑0.059*** (0.019)
Log of total hours worked within a sector	‑0.040** (0.018)	‑0.036** (0.017)	‑0.184** (0.093)	‑0.172** (0.088)
Log employment	‑0.026 (0.017)	‑0.023 (0.027)	‑0.120 (0.086)	‑0.113 (0.080)
Log of compensation per worker	0.001 (0.011)	0.004 (0.010)	0.008 (0.062)	0.007 (0.057)
Log of compensation per hour worked	0.015 (0.012)	0.018 (0.011)	0.071 (0.062)	0.066 (0.055)

A. Discrete treatment variable

B. Continuous measure of initial exposure

Without controls

With controls

Without controls

With controls

Percentage share of workers affected by reform

‑4.863***

(1.369)

‑4.773***

(1.381)

‑34.124***

(10.939)

‑33.909***

(10.933)

Log of average annual hours per worker

‑0.014***

(0.004)

‑0.013***

(0.004)

‑0.063***

(0.018)

‑0.059***

(0.019)

Log of total hours worked within a sector

‑0.040**

(0.018)

‑0.036**

(0.017)

‑0.184**

(0.093)

‑0.172**

(0.088)

Log employment

‑0.026

(0.017)

‑0.023

(0.027)

‑0.120

(0.086)

‑0.113

(0.080)

Log labour productivity (VA per worker)

‑0.003

(0.023)

‑0.002

(0.022)

0.102

(0.120)

0.110

(0.115)

Log hourly labour productivity (VA per hour worked)

0.011

(0.023)

0.012

(0.022)

0.165

(0.112)

0.169

(0.119)

Log of compensation per worker

0.001

(0.011)

0.004

(0.010)

0.008

(0.062)

0.007

(0.057)

Log of compensation per hour worked

0.015

(0.012)

0.018

(0.011)

0.071

(0.062)

0.066

(0.055)

Note: This table gives the estimates of Equation 5.1 and Equation 5.2 presented in Box 5.2 on the share of workers above the threshold, and the log of average hours per worker, employment, value added per hour and compensation per hour. Share of workers affected by reform indicates the share of workers working more than the value specified by the existing legislation (for countries without a reform) or introduced by the reform (for countries with reform). Sectors are weighted by the within – country share of employment in the pre‑reform period. Standard errors are clustered at the country × sector level. Panel A gives the results of Equation 5.1 with a discrete treatment variable. Panel B presents the results of Equation 5.2) with a continuous measure of initial exposure (the share of workers above the threshold). Panel A shows the effect of being in a sector above the median of exposed workers before the reform; Panel B, the effect of going from 0 to 100% of workers exposed to the reform. Controls included are at the 2‑digits Nace Rev.1.1. from an ad hoc extraction by EUROSTAT, and include age, education, gender, type of contracts, tenure and occupation. Regressions are calculated with 7 345 observations.

Source: Batut, Garnero and Tondini (2022[109]), “The Employment Effects of Working Time Reductions: Sector-Level Evidence from European Reforms”, FBK-IRVAPP Working Papers Series.

Annex 5.C. Comprehensive literature review on the employment effect of change in hours worked

Robustness level	Author and year	Title	Method	Data	Type of change	Outcomes	Empirical scope	Results
Theoretical	(Zwickl, Disslbacher and Stagl, 2016[134])	Work sharing for a sustainable economy	Theoretical discussion	N/A	Hours reduction	Unemployment	N/A	Little expected employment effect of working time reduction reforms and consider it promising to mitigate unemployment in context of low growth.
Theoretical	(Fagnart, Germain and Van der Linden, 2020[135])	Working Time Reduction and Employment in a Finite World	Theoretical modelling	N/A	Hours reduction	Hours worked, earnings per worker, employment, unemployment, hourly wage.	N/A	The impact of working time reductions (WTRs) on (un)employment and the hourly wage is expected to depend on the relative scarcity of natural resources used in the economy. If the resource inflow was unlimited, a WTR would lower the employment and wage levels in the long run. When the resource inflow is finite the economy tends toward a stationary state with a finite output level. If the resource is scarce enough, notably if the technical progress on human factors (labour and capital) is unbounded, a WTR has a favourable effect on employment and on the hourly wage.
Theoretical	(Calmfors and Hoel, 1988[81])	Work Sharing and Overtime	Theoretical modelling	N/A	Statutory hours reduction	Employment, overtime hours.	N/A	A reduction in statutory hours is expected to increase the cost per worker in relation to the cost of overtime, with the consequence that firms substitute overtime for workers. When output is fixed by demand, this substitution effect may reduce employment. Second, when firms choose a profit-maximising level of output, the cost increase due to a reduction in normal working time produces, in addition, a negative scale effect on employment. With a fixed output level, an employment increase can always be achieved, however, through the combination of an increased overtime premium and reduced normal working time that produces a substitution effect in the right direction.
Theoretical	(Schank, 2015[136])	Employment effects of longer working hours	Theoretical modelling	N/A	Negotiated hours increase	Employment, wages.	N/A	Extending work hours may reduce employment in the short term but may increase it in the long term if hourly pay remains constant (which means a welfare decline for workers). Extending standard hours could also safeguard jobs in firms under competitive pressure.
Theoretical	(Marimon and Zilibotti, 2000[137])	Employment and distributional effects of restricting working time	Theoretical modelling	N/A	Hours reduction	Employment, profits, output.	N/A	Small reductions in working time starting from the laissez-faire equilibrium solution, always result in a small increase in the equilibrium employment, while larger reductions reduce employment. The regulation benefits workers, both unemployed and employed (even if wages decrease and even in cases where employment falls), but reduces profits and output.
Theoretical	(Rocheteau, 2002[138])	Working time regulation in a search economy with worker moral hazard	Theoretical modelling	N/A	Hours reduction	Employment, worker well-being	N/A	When unemployment is high, reducing working hours increases aggregate employment. At the opposite, for low unemployment countries, a working time reduction worsens the labour market situation. If the working time reduction takes place with no wage loss, the model predicts a non-ambiguous increase in the equilibrium unemployment rate.
Theoretical	(Contensou and Vranceanu, 2000[139])	A model of working time under utility competition in the labour market	Theoretical modelling	N/A	Hours reduction	Employment.	N/A	If a limit on the duration of work is imposed to an initially free system, at first, a favourable effect on employment might be achieved for a constant utility level of workers. A too “strong” working-time constraint would have a perverse effect on the demand for workers.
Correlational	(Andrews, Schank and Simmons, 2005[84])	Does worksharing work? Some empirical evidence from the IAB establishment panel	Fixed effects regression analysis	Firm-level	Contractual hours reduction.	Employment, overtime hours.	Germany	A regression analysis of hours reduction on employment level shows non-significant results in most cases, except in small plants in the non-service sector in East Germany, where effects are strongly positive.
Correlational	(Brown and Hamermesh, 2019[88])	Wages and Hours Laws: What Do We Know? What Can Be Done?	Literature review	N/A	Applicability of overtime pay	Demand for overtime, weekly hours, employment	United States	Overtime provisions have only small effects on labour-market outcomes: they reduce employers’ demand for overtime hours, and weekly hours of work slightly. The law probably spreads employment among a few more labour-force participants, although total labour input – hours per worker times employment – probably decreases because hours drop more than employment increases. In the long run it has no impact on unemployment rates.
Correlational	(Sagyndykova and Oaxaca, 2019[83])	Raising the overtime premium and reducing the standard workweek: short-run impacts on US manufacturing	Model estimation	Individual-level	Overtime premium increase and statutory hours reduction	Employment, wages, capital use, weekly hours.	Unites States	The simulation results suggest that raising the overtime premium to double‑time would have a modest negative impact on employment and aggregate earnings growth and a negligible effect on the growth rate of weekly hours and earnings per worker. Lowering the standard workweek from 40 hours to 35 hours would reduce the industry-wide employment growth rate by a substantial ‑1.54 percentage points. Overall, the growth rate effects for capital, aggregate hours, total earnings, and weekly hours and earnings per worker would also be substantially negative.
Correlational	(Fiole, Roger and Rouilleault, 2002[89])	Les effets sur l’emploi de la loi du 11 juin 1996 sur la réduction du temps de travail	Difference‑in-difference (DiD) with self- selection issues	Firm-level	Negotiated hours reduction	Employment growth	France	The analysis finds a positive effect of working hours reduction on employment growth.
Correlational	(Friesen, 2001[140])	Overtime pay regulation and weekly hours of work in Canada	Fixed effects regression analysis	Individual-level	Statutory hours reduction.	Hours, moonlighting, wages.	Canada	Coverage by overtime pay regulation is associated with an increase in the straight-wage rate. The constraints created by overtime pay regulation appear to induce a considerable number of workers to take up a second, moonlighting job.
Correlational	(Husson, 2002[90])	Réduction du temps de travail et emploi: une nouvelle évaluation	Regression analysis	Country level	Statutory hours reduction.	Job creation, productivity	France	The estimation of the impact of working time in the production function suggests that 500 000 jobs were created between 1997 et 2001 thanks to the reduction of working time in France.
Correlational	(Hunt, 1999[85])	Has Work-Sharing Worked in Germany?	Fixed effects regression analysis	Industry-level & individual-level	Negotiated hours reduction.	Employment, actual hours worked, wages.	Germany	Results are insignificant when including industry-specific trends (which is essential with this identification strategy) and when considering the whole sample (only particular specification concentrating on men only sometime yield significant negative results.
Correlational	(Trejo et al., 2016[86])	Does the statutory overtime premium discourage long workweek?	Regression analysis	Industry-level	Applicability of overtime pay.	Overtime hours, total hours, employment (derived).	United States	Increases in overtime pay coverage did not reduce overtime incidence and hours. Authors suggest that employment is therefore unlikely to have been affected (although no direct analysis).
Correlational	(Kramarz et al., 2008[87])	Working time developments in Germany	Fixed effects regression analysis	Firm-level	Contractual hours increase.	Employment, productivity, wages, rate of female employment.	Germany	Firms that increase standard hours also have decreasing employment when firms that decrease standard hours have stable employment. In particular, when standard hours increase, firms use less part-time workers as theory predicts (full time workers become less costly). In Western Germany increasing standard hours is marginally significantly associated with an increase in productivity; decreasing standard hours is associated with unchanged productivity. Decreasing hours does not affect employment growth.
Correlational	(Steiner, Peters and Steiner, 2000[82])	Employment effects of work sharing: an econometric analysis for West Germany	Regression analysis	Industrial level	Contractual hours reduction.	Employment, unemployment, wages.	Germany	Labour demand elasticities with respect to real wages differ significantly between unskilled, skilled and high-skilled workers. Given wages, the direct employment effect of a reduction in weekly normal hours is negligible for all three groups. However, taking the adjustment of wages into account, the net employment effect becomes negative on average. This negative effect is particularly strong for the unskilled.
Correlational	(Kapteyn, Kalwij and Zaidi, 2004[79])	The myth of worksharing	Regression analysis	Country-level	Statutory hours reduction	Employment and wages	16 OECD countries	The results show a positive direct effect on employment of a reduction in working hours. However, taking into account indirect effects, in particular the upward effects on wages, we find that the long-run effect becomes small and insignificant.
Quasi-causal	(Estevão and Sá, 2006[95])	Are the French Happy with the 35‑Hour Workweek?	Difference‑in-difference (DiD) with imperfectly comparable control group	Individual level	Statutory hours reduction.	Hours distributions, wages, dual job holdings, transition from large to small firms, employment, satisfaction with hours.	France	Employment of persons directly affected by the law declined, although the net effect on aggregate employment was not significant. The law constrained the choice of a significant number of individuals: dual-job holdings increased, some workers in large firms went to small firms where hours were not constrained, and others were replaced by cheaper, unemployed individuals as relative hourly wages increased in large firms.
Quasi-causal	(Costa, 2000[96])	Hours of Work and the Fair Labor Standards Act: A Study of Retail and Wholesale Trade, 1938‑50	Difference in difference (DiD) with imperfectly comparable control group.	Industry-level and individual level	Applicability of overtime pay.	Employment, overtime hours, total hours.	Unites States	The paper finds no clear effect on employment (and supposes that the positive effect was offset by a negative effect -unproven – of the parallel increase in the minimum wage).
Quasi-causal	(Yu and Peetz, 2018[141])	Non-Standard Time Wage Premiums and Employment Effects: Evidence from an Australian Natural Experiment	Difference‑in-difference (DiD) with imperfectly comparable control group	Country-level and individual level	Applicability of overtime pay	Labour force participation, wages	Australia	Results show that the introduction of a Sunday overtime premium in Australia did not have negative employment effects, but resulted in more flexible hours (though in an industry dominated by casual employment).
Quasi-causal	(Crépon and Kramarz, 2002[91])	Employed 40 Hours or Not Employed 39: Lessons from the 1982 Mandatory Reduction of the Workweek	Difference‑in-difference (DiD) with imperfectly comparable control group	Individual-level	Statutory hours reduction.	Job losses, hours change.	France	Workers more affected by the reform were less likely to be employed after it than observationally identical workers who were not affected by it. Workers more affected lost their jobs more often than less affected – especially minimum wage workers; better compensated workers were less directly affected by the reduction of the workweek.
Quasi-causal	(Skuterud, 2007[97])	Identifying the Potential of Work-Sharing as a Job-Creation Strategy	Difference‑in-difference (DiD) with imperfectly comparable control group	Individual-level	Statutory hours reduction.	Employment.	Quebec	Results on employment are non-significant: the point estimate is negative, but non-significant throughout, and, crucially, it is not increasing with the share of supposedly affected workers in the industry. These negative – insignificant – coefficients are observed only for men, while coefficients are positive for women.
Quasi-causal	(Sánchez, 2013[98])	Do reductions of standard hours affect employment transitions?: Evidence from Chile	Difference‑in-Difference (DiD) comparing workers with different likelihood to be affected	Individual-level	Statutory hours reduction	Wages, Employment transitions	Chile	The reduction of standard hours had no significant effects on employment transitions (i.e. no effect on excess job destruction), but had a significant effect on hourly wages (i.e. evidence of wage compensation).
Quasi-causal	(Raposo and van Ours, 2010[92])	How working time reduction affects jobs and wages	Difference in Difference (DiD) comparing workers with different likelihood to be affected	Matched employee‑employer	Statutory hours reduction	Wages, Employee retention	Portugal	For workers affected the reduction reduced the job separation rate and increased hourly wages, keeping monthly earnings approximately constant. The working hours reduction also affected workers who already worked less than the new norm, who were more likely to lose their job.
Quasi-causal	(Chemin and Wasmer, 2009[99])	Using Alsace‑Moselle Local Laws to Build a Difference‑in-Differences Estimation Strategy of the Employment Effects of the 35‑Hour Workweek Regulation in France	Difference‑in-difference‑in-difference (DDD) with two imperfectly comparable treatment groups	Individual-level	Statutory hours reduction.	Employment, unemployment.	France	No significant effect on employment or unemployment.
Quasi-causal	(Andrews et al., 2015[142])	More hours, more jobs? The employment effects of longer working hours	Difference‑in-Difference (DiD) with matched control group	Firm-level	Contractual hours increase.	Employment, wages.	Germany	Significant positive employment response in firms offering overtime (for which the increase in hours with wage concession corresponds to a fall in labour cost), while they find no effect in firms offering no overtime.
Quasi-causal	(Cahuc and Carcillo, 2014[143])	The detaxation of overtime hours: lessons from the French experiment	Difference‑in-difference (DiD) with imperfectly comparable control group, complemented by convincing robustness tests.	Individual-level	Overtime pay reduction.	Overtime hours, total hours, employment (derived).	France	Overtime hours of highly qualified employees working in France rose, relative to those of the transborder employees, following the overtime pay reduction. There were no difference in the evolution of hours worked, whatever category of employee is considered. The fact that hours worked do not increase suggests that the measure must have had a very limited effect on employment.
Quasi-causal	(Crépon, Leclair and Roux, 2004[93])	RTT, productivité et emploi: nouvelles estimations sur données d’entreprises	Difference‑in-Difference (DiD) with self-selection issues but numerous convincing test of parallel trends hypothesis.	Firm-level	Statutory hours reduction	Employment, productivity, wages.	France	Firms adopting the 35 hours saw a slight reduction in total factor productivity, less than expected based on the hours change; employment in these firms increased. Authors posit that this is largely due to wage restraint and social security cuts, rather than to the reduction in hours per se.
Quasi-causal	(Kawaguchi, Naito and Yokoyama, 2017[100])	Assessing the effects of reducing standard hours: Regression discontinuity evidence from Japan	Regression Discontinuity Design (RDD)	Firm-level	Statutory hours reduction	Hours worked, monthly wages, annual bonuses, and hires.	Japan	The results of the RD analyses show that the reduction of standard hours from 44 to 40 in the manufacturing industry decreased hours worked, but this effect is not statistically significant when we estimate the average treatment effect. Overall, on average, the reduction of standard hours did not change hours worked, monthly wages, annual bonuses, and employment in statistically significantly ways. Results on new hires are still insignificant when considering heterogeneous establishment types.

Annex 5.D. Balanced samples of treated and control groups for the firm-level analyses in Germany, Korea and Portugal

Country	Variable	Control group mean	Treated group mean	Difference of the means	Means difference 95% CI lower bound	Means difference 95% CI upper bound
Germany	Value added per worker (level)	58 117	57 679	438	‑7 983	8 860
	Change in value added per worker (%)	0.23	2.46	2.23	‑7.14	2.68
	Log change in value added per worker (main outcome variable)	‑0.06	‑0.04	‑0.02	‑0.08	0.04
	Total employment (level)	98	84	14	‑11	39
	Change in total employment (%)	1.92	1.06	0.86	‑1.54	3.27
	Log change in number of employees (main outcome variable)	0.01	0.00	0.01	‑0.02	0.03
	Average wage (level)	1 969	2 075	‑106	‑260	47
	Change in average wage (%)	0.92	2.61	‑1.68	‑5.75	2.38
	Log change in average wage (main outcome variable)	‑0.03	‑0.01	‑0.03	‑0.07	0.02
	Presence of overtime (dummy)	0.73	0.74	‑0.01	‑0.08	0.05
	Profit situation (5 categories)	2.99	3.06	‑0.07	‑0.24	0.10
	Presence of a work council (dummy)	0.29	0.29	‑0.01	‑0.07	0.06
	Share of white collar workers with a university degree	0.08	0.07	0.01	‑0.01	0.03
	Log change in total investment	‑0.45	‑0.49	0.05	‑0.12	0.22
	Rate of investment in value added (5 categories)	2.53	2.48	0.05	‑0.11	0.21
	Investment in communication / data processing technology	0.49	0.46	0.03	‑0.04	0.11
	Change in the use of overtime (adoption, abandon, stable)	0.03	0.01	0.02	‑0.04	0.09
	Level of the applicable collective agreement (3 categories)	2.22	2.12	0.10	‑0.04	0.24
	Coverage by collective agreement on wage (dummy)	0.44	0.49	‑0.05	‑0.12	0.03
	Firm size (4 categories)	2.06	2.11	‑0.05	‑0.19	0.08
	Industry (7 categories)	5.24	4.95	0.30	‑0.31	0.90
	Share of export in business volume (5 categories)	1.65	1.77	‑0.12	‑0.30	0.06
	Share of full-time employees (5 categories)	4.66	4.65	0.01	‑0.10	0.13
	Share of permanent workers (5 categories)	4.96	4.97	‑0.01	‑0.05	0.02
Korea	Business volume per worker (level)	431	226	206	‑123	535
	Change in business volume per worker (%)	0.06	0.04	0.02	‑0.08	0.11
	Log change in business volume per worker (main outcome variable)	0.02	0.01	0.00	‑0.11	0.12
	Total employment (level)	82	75	8	‑27	43
	Change in total employment (%)	‑0.05	‑0.04	‑0.01	‑0.05	0.03
	Log change in total employment (main outcome variable)	‑0.06	‑0.05	‑0.01	‑0.05	0.04
	Average wage (level)	39	34	5	‑3	12
	Change in average wage (%)	0.06	0.06	0.03	‑0.05	0.12
	Log change in average wage (main outcome variable)	0.04	0.02	0.02	‑0.07	0.11
	Presence of overtime (dummy)	0.93	0.94	‑0.01	‑0.11	0.09
	Profit situation (5 categories)	2.81	2.94	‑0.13	‑0.41	0.15
	Presence of a work council (dummy)	0.13	0.13	0.00	‑0.13	0.13
	Average level of education of the largest occupational group (5 categories)	2.35	2.26	0.09	‑0.20	0.38
	Change in the use of overtime (adoption, abandon, stable)	0.07	0.00	0.07	‑0.06	0.20
	Level of the applicable collective agreement (3 categories)	0.17	0.16	0.01	‑0.17	0.18
	Workplace engaged in wage bargaining last year (dummy)	0.14	0.13	0.01	‑0.12	0.15
	Firm size (3 categories)	1.34	1.32	0.02	‑0.17	0.20
	Industry (7 categories)	1.97	1.97	0.00	‑0.55	0.55
	Change in business volume (5 categories)	2.94	2.97	‑0.03	‑0.42	0.36
	Share of export in business volume (5 categories)	2.22	2.29	‑0.07	‑0.64	0.51
	Share of part-time employees	0.59	2.15	‑1.56	‑3.97	0.86
	Share of employees on fixed-term contracts	6.85	5.07	1.78	‑5.15	8.71
Portugal	Value added per worker (level)¹	20 327	23 804	‑3 477	‑5 434	‑1 520
	Change in value added per worker (%)	‑2.73	‑2.58	‑0.15	‑2.89	2.59
	Log change in value added per worker (main outcome variable)	‑0.10	‑0.08	‑0.02	‑0.06	0.01
	Total employment (level)	4.9	4.9	0.0	‑0.4	0.3
	Change in total employment (%)	2.10	1.39	0.71	‑0.89	2.32
	Log change in total employment (main outcome variable)	0.00	0.00	0.00	‑0.02	0.02
	Average wage (level)¹	13 722	15 585	‑1 863	‑2 506	‑1 220
	Change in average wage (%)	‑0.10	0.88	‑0.98	‑2.78	0.82
	Log change in average wage (main outcome variable)	‑0.03	‑0.01	‑0.02	‑0.03	0.00
	Presence of overtime (dummy)	0.98	0.67	0.30	‑0.48	1.09
	Profit situation (5 categories)	2.97	2.96	0.01	‑0.11	0.13
	Share of middle skill labour	0.17	0.15	0.02	‑0.01	0.05
	Share of high skill labour	0.12	0.13	‑0.01	‑0.04	0.01
	Log change in total investment	‑0.19	‑0.07	‑0.12	‑0.39	0.14
	Rate of investment in business volume (5 categories)	1.72	1.75	‑0.03	‑0.11	0.04
	Investment in intangible assets	0.05	0.06	‑0.01	‑0.03	0.01
	Change in the use of overtime (adoption, abandon, stable)	‑0.01	0.00	0.00	‑0.02	0.01
	Level of the applicable collective agreement (3 categories)	3.45	3.46	‑0.01	‑0.09	0.07
	Firm size (3 categories)	1.09	1.09	0.00	‑0.02	0.02
	Industry (6 categories)	4.93	4.93	0.00	‑0.13	0.13
	Change in business volume (5 categories)	2.74	2.74	0.00	‑0.09	0.08
	Share of export in business volume (5 categories)	1.24	1.24	0.00	‑0.06	0.06
	Share of full-time employees (5 categories)	4.88	4.85	0.03	‑0.01	0.07
	Share of permanent workers (5 categories)	4.47	4.45	0.01	‑0.09	0.11

Note: Bold means significant at the 5% level.

1: In Portugal, the sample is not balanced with regards to levels of value added per worker and levels of average wage (although it is balanced when considering for percentage and log changes). This means that firms that reduce their hours have a significantly higher level of value added per worker and pay a higher average wage at t‑1. This does not affect the identification strategy used in section 5.2.3, since the analysis is based on growth rather than levels. Yet, to correct for this imbalance, controls for pre‑change levels of value added per worker and average wage are added in the baseline analysis for Portugal and therefore reflected in the results presented in Figure 5.8 above.

Variable	Control group mean	Treated group mean	Difference of the means	Means difference 95% CI lower bound	Means difference 95% CI upper bound
Log change in value added per worker (main outcome variable)	‑0.06	‑0.06	0.00	‑0.06	0.06
Log change in number of employees (main outcome variable)	0.02	0.02	0.00	‑0.02	0.03
Log change in average wage (main outcome variable)	‑0.01	0.01	‑0.02	‑0.06	0.02
Value added per worker (level)	65 982	61 824	4 158	‑8 503	16 820
Number of employees (level)	102	118	‑16	‑47	15
Average wage (level)	2 278	2 375	‑97	‑275	80
% change in value added per worker	0.15	‑1.30	1.45	‑3.86	6.76
% change in number of employees	2.79	2.48	0.30	‑2.37	2.98
% change in average wage	1.74	2.42	‑0.69	‑4.61	3.24
Firm size (4 categories)	2.39	2.39	0.00	‑0.13	0.13
Industry (7 categories)	5.05	5.04	0.02	‑0.71	0.74
Share of full-time employees (5 categories)	4.71	4.62	0.09	‑0.04	0.21
Share of permanent workers (5 categories)	4.95	4.98	‑0.04	‑0.08	0.01
Share of white collar workers with a university degree	0.10	0.11	‑0.01	‑0.04	0.01
Profit situation (5 categories)	2.75	2.73	0.02	‑0.16	0.21
Share of export in business volume (5 categories)	2.10	2.25	‑0.15	‑0.41	0.10
Log change in total investment	‑0.41	‑0.33	‑0.07	‑0.26	0.11
Rate of investment in value added (5 categories)	2.56	2.46	0.10	‑0.06	0.27
Investment in communication / data processing technology (dummy)	0.53	0.56	‑0.03	‑0.12	0.05
Presence of overtime (dummy)	0.86	0.85	0.01	‑0.05	0.07
Change in the use of overtime (adoption, abandon, stable)	0.01	‑0.03	0.04	‑0.02	0.09
Coverage by a collective agreement on wage (dummy)	2.26	2.24	0.02	‑0.14	0.17
Presence of a work council (dummy)	0.37	0.41	‑0.03	‑0.12	0.05
Level of the applicable collective agreement (3 categories)	0.42	0.43	0.00	‑0.09	0.08

Notes

← 1. Authors would like to acknowledge the contribution of Bayram Cakir to this chapter, in particular for his research assistantship in dealing with the Portuguese data in Section 5.2.3.

← 2. The term teleworking can be defined in many different ways (OECD, 2021[1]). Throughout the chapter, teleworking is taking to mean the possibility for employees to work remotely, from home or a location other than the employers’ premises, usually or occasionally. In practice, this often corresponds to the newer term of “hybrid working”, rather than to workers working exclusively from home.

← 3. Defining long hours is not an easy task and different thresholds are considered across studies, surveys and measurement frameworks. One possibility is to consider regulation, e.g. for instance in Europe the 2003 European Union Working Time Directive that establishes that maximum weekly hours cannot exceed 48 hours including overtime. Another way to define long hours is to refer to the distribution of hours worked in the country/population studied: for instance the OECD job quality framework considers that individuals working very long hours are those working more than 60 hours per week when defining job strain in emerging economies. However, many papers simply define long hours as overtime work, which is imprecise but available in many surveys.

← 4. Solving the “healthy worker” identification problem is difficult when only cross-sectional data are available, but easier using panel data which allows controlling for past health levels, or estimating fixed-effect models with an exogenous variation that can be used as an instrument. Other challenges in empirical studies of the effect of working hours on health include the fact that working hours are not randomly assigned, which introduces biases since potential omitted unobserved factors might influence both hours and health, or the fact that estimates of the impact of hours are usually confounded by the influence of hours on income, which has an important independent effect on health. The data collected might also already be biased: for instance, the negative effect of some arrangements on workers’ well-being might be under-estimated if workers quite because of them (and therefore are not recorded in the data).

← 5. These results are driven by the positive associations between working between 30 and 44 hours per week (which is the modal group) and the three satisfaction outcomes (e.g. life, job and free time satisfaction. Indeed, additional analyses (not shown here and available on request) show that in France, the probability to be “life‑satisfied” is highest for those in the 40‑44 hours group, to be “job-satisfied” is highest for those in the 30‑34 hours group, and satisfaction with free time is most likely for those in the 35‑39 hours group.

← 6. Even if effects differ by working time arrangements, by well-being measures and between different workers’ groups.

← 7. According to recent surveys of workers and business across OECD countries (Barrero, Bloom and Davis, 2021[150]; Criscuolo et al., 2021[155]).

← 8. Of the four studies, only Nikolova and Graham (2014[70]) differentiate between voluntary or involuntary adoption of part-time work. Benson et al. (2017[67]), Cho (2018[68]) and Beham et al. (2019[69]) consider only general part-time status, irrespective of workers’ reason for take‑up.

← 9. Although see the discussion on theoretical models allowing for the possibility of wage restraint by unions negotiating wages when hours are reduced in Kapteyn et al. (2004[79]).

← 10. Theoretical papers in the literature are therefore inconclusive: while imposing an upper limit on working hours where it did not exist is expected to have a positive effect on employment, a “too strong” (i.e. “too low”) working-time constraint would reduce the demand for workers (Contensou and Vranceanu, 2000[139]; Marimon and Zilibotti, 2000[137]). Reducing working hours might increase aggregate employment in a context of high pre‑existing unemployment, but it might worsen the labour market situation of low-unemployment countries (Rocheteau, 2002[138]). Calmfors and Hoel (1988[81]) expect a reduction in statutory hours to reduce employment, as firms substitute overtime for workers (if the overtime premium is not raised); however, hours reduction might create employment, they argue, in the context of a fixed output level, when complemented by an increase in the overtime premium.

← 11. This somewhat simple idea is behind the “work-sharing” theory, which used to be popular in policy making circles, although it was never grounded in economic theory. In that logic, working time reduction could foster employment creation through work-sharing, i.e. through sharing the same quantity of hours of work between more workers doing less hours each. This idea does not easily stand up to economic reasonning taking into account, notably, the fixed costs associated with hiring a worker, or the frictions created by the imperfect substituability between different workers. This idea is also incompatible with the idea that reducing working time could improve firms’ productivity through organisational innovations, and/or through workers’ reduced level of fatigue and increased levels of engagement. In either case, firms would not need to hire new staff to maintain the same output level (e.g. there would be no work sharing).

← 12. Models exploring the latter situation tend to assume that hourly productivity increases following a reduction in normal hours, but not enough to fully compensate the rise in unit labour costs.

← 13. For instance, the identification strategy in (Hunt, 1999[85])is based on the strong assumption that hours change at the industry level are exogenous because they are agreed to in advance. However, while negotiated hours cannot indeed be amended as a reaction to unforeseen economic changes, they cannot be considered to be exogenous with regard to anticipated changes, which means that industry-specific trends should be included. When they are, results are insignificant, which suggests that hours change at the industry level are in fact not exogenous to industrial trends in employment and productivity. Without that crucial hypothesis holding, the identification strategy is equivalent to an industry-level fixed effect regression, which is suggestive of association not causation. Hence this study is classified as “correlational”.

← 14. Brown and Hamermesh look at the effect of the introduction of an overtime premium in the United States (which is equivalent to introducing an upper bound on working hours) on unemployment, and find that the latter had no effect on unemployment in the long run.

← 15. E.g. For instance, nominal wages are kept constant, but not fully adjusted for inflation, thus limiting the increase in hourly wage,or that unions engage in wage restraint, or that firms boost productivity enough – through e.g. work re‑organisation, increased investments, or workforce re‑structuration centred around more productive workers – to compensate for the rise in hourly pay.

← 16. For instance, Kramarz et al. (2008[87]) argue that a working time reduction law can help employment to reach its maximum level if it forces a monopsony to set individual working times at the competitive level, thereby improving the welfare of workers – see also e.g. Marimon and Zilibotti (2000[137]) or Contensou and Vranceanu (2000[139]). Moreover, as shown in Chapter 3, local labour markets are heterogeneous in their degree of monopsony. As competitive conditions are different across local markets in the same country, national estimates reported in the literature are likely to result in the aggregation of positive and negative effects occurring in different markets. This means that normal hours reduction implemented in a context of monopsony, or even to counteract a situation of monopsony in a given market, might be more appropriately designed at a local level, where the degree of monopsony can be more accurately estimated.

← 17. While the two channels discussed in this paragraph might increase hourly worker productivity in a linear way, this is probably not the case for total productivity per worker : brought to the extreme, while hourly productivity might still increase if working hours were reduced to one hour per week, total productivity per worker would very likely decrease.

← 18. Hart and Krall (2007[145]) find that shorter shifts were associated with a greater hourly productivity of physicians in emergency departments in the United States. Olds and Clarke (2010[151]) find that medication errors and needle stick injuries are statistically related to nurses working more than 40 hours per week in the United States, and Rogers et al. (2004[152]) find similar results for nurses working shifts longer than 12 hours. Shepard and Clifton (2000[146]) find that overtime reduced output per worker in the United States manufacturing firms, while Schank (2005[147]) find no difference in productivity between German plants using overtime and those that did not. Similar evidence of a fatigue effect causing productivity to decrease with long hours have been found in various sectors, with results aligning across contexts as different as paramedics in Mississippi (Brachet, David and Drechsler, 2012[148]), British munition workers during World War I (Pencavel, 2015[149]) and American factory workers in the 1920s (Dolton, Howorth and Abouaziza, 2016[108]). This detrimental fatigue effect is found to be persistent over time in the absence of adequate recovery time (Pencavel, 2016[107]). Yet, one apparent exception is the paper from Lu and Lu (2017[144]). Authors observe a performance decrease following the introduction of laws prohibiting mandatory overtime in some nursing homes in the United States. However, they conclude that this is in fact an indirect effect of the increased use of contract workers following the change.

← 19. The relatively timid exploration of the potential productivity effect of reducing normal hours is all the more surprising that it is theoretically a more promising means of generating employment than the work-sharing theory that has been much more explored in the literature (see note 12): if reducing hours can enhance productivity through one or both or the channels discussed above, it could potentially generate employment in the mid-to longer term as a second order effect of this productivity increase.

← 20. The effect is identified by comparing changes in the outcomes of interest around a reform reducing working hours between “high exposure” sectors (i.e. sectors with long average working hours before the reform, which will be affected by it) and “low exposure” sectors (i.e. sectors with relatively short average working hours before the reform, which will not be affected by it).

← 21. Information comes from multiple sources: the CBR Labour Regulation Index (Adams, Bishop and Deakin, 2010[156]) complemented and cross-checked with information available in the ILO Travail Database and the European Union Commission LABour market REForm (LABREF) database (European Commission, 2021).

← 22. Results for Panel B go in the same direction, but coefficients have to be interpreted as the relative effect of going from 0 to 100% of exposed workers: in sectors where all workers are affected by the reduction in hours, hours drop by 6% relative to those sectors where all workers were already working less than the reform threshold, and the share above the threshold decreases by 33 percentage points.

← 23. See details in Section 5 of Batut et al. (2022[109]), “The Employment effects of Working Time Reductions in Europe”.

← 24. This explanation can however not be confirmed by this analysis which uses gross labour compensation, e.g. a variable that includes social security contributions and would also have captured social security reductions; however, the coefficient estimated for the average effect of the reforms on compensation in “high exposure” versus “low exposure” sectors being positive, this goes against the idea that labour costs were fully compensated by social security reductions.

← 25. Although firm-level data is used to identify episodes during which firms reduced their contractual hours, the decisionary basis behind these episodes of reduction cannot be derived from the data: firm-level reductions of contractual hours could be the outcome of a unilateral decision on the part of employers at the firm level, or of a negotiation process between workers and management at the firm-level, but it could also reflect a decision bargained at the sectoral or national level, or even a legislative reform at the national level.

← 26. Results estimated at t+1 capture the evolution between t and t+1, while those at t+2 capture the evolution between t and t+2, i.e. the cumulated effect of treatment over both post-change years observed.

← 27. The difference with the number of observations reported under the figures below is due to the fact that all four years in each treated spells are in the treated group, so there are more treated observations as there are treated spells.

← 28. Which concentrate respectively 15.7, 17.6 and 18.1% of cases versus 4.8% in each year on average.

← 29. To dig further into the negative effect on employment growth, Equation 5.3 is estimated with growth of separations and new hires as outcome variables. The growth of separations is negatively and significantly related to treatment at t+1 (‑29.6%), which suggests that treated firms retain more workers than control firms. Further, the growth of new hires is also negatively and significantly related to treatment (‑33.7%), which suggests that firms that reduce contractual hours also hire less than firms that do not. These combined results suggest that the negative effect observed for employment growth might be due to the differential between the movement in the growth rates of separations and hires, with both diminishing but the growth of entries more so than that of exits, rather than to an increase of separations. In other words, it is possible that workers’ retention potential increased in firms reducing normal hours (although note that the data does not allow knowing what happens within separations, i.e. how dismissals and quits evolve separately).

← 30. Respectively 19.3% in 2012 and 17.7% of cases in 2013.

← 31. According to Hijzen and Thewissen (2020[110]), the reform contained a number of measures to ease potential negative impacts on employers including an exemption from the obligation to pay the overtime premium for firms with less than five employees, as well as temporary reductions of the overtime premium for the first four hours of overtime.

← 32. Analyses differentiating the employment impact of teleworking in urban vs. rural regions would appear relevant, but are still missing today. This is a fruitful area for future research.

← 33. General conclusions on the productivity effect of teleworking cannot be derived from studies using data collected during the COVID‑19 pandemic; indeed, as explained already in Box 5.1 above, too many factors characterising that period might confound productivity at the time. Nonetheless, the overall message arising from pandemic time studies is that self-reported productivity was maintained or increased among employees who started teleworking during the pandemic (Ker, Montagnier and Spiezia, 2021[154]). Among the minority of workers who reported decrease productivity, the main identified cause included the lack of interaction with colleagues, conflicting family care duties, difficulties in accessing work-related information, additional hurdles in getting to work done, inadequate physical work spaces and lacking internet speed.

← 34. This section is focusing on studies looking at how formal, paid teleworking hours affect productivity. For an analysis on how unpaid work at home affect productivity in the United States, see Elridge and Pabilonia (2010[153]).

← 35. To dig further into this, the model is run with the share of new hires and the share of separations. There is a significant negative association between treatment and separations at t+2, and an insignificant (positive) association with new hires at t+2 (the regression cannot be run at t+1 as there is not enough treated observations at t+1 in this specification). This suggests indeed that the positive effect on employment might have more to do with improved worker retention, rather than a significant increase in hirings.

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Countries

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Publications

Publications

Featured publications

Data

Data

Featured data

News & Events

News & Events

Featured Events

About OECD

About

Engage with us

Work with us

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment