Does temporal and locational flexibility of work increase the supply of working hours? Evidence from the Netherlands

Possenriede, Daniel; Hassink, Wolter H. J.; Plantenga, Janneke Article Does temporal and locational flexibility of work increase the supply of working hours? Evidence from the Netherlands IZA Journal of Labor Policy Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Possenriede, Daniel; Hassink, Wolter H. J.; Plantenga, Janneke (2016) : Does temporal and locational flexibility of work increase the supply of working hours? Evidence from the Netherlands, IZA Journal of Labor Policy, ISSN 2193-9004, Springer, Heidelberg, Vol. 5, Iss. 16, pp. 1-34, https://doi.org/10.1186/s40173-016-0072-y This Version is available at: https://hdl.handle.net/10419/194369 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 DOI 10.1186/s40173-016-0072-y ORIGINAL ARTICLE Open Access Does temporal and locational flexibility of work increase the supply of working hours? Evidence from the Netherlands Daniel Possenriede1,2* , Wolter H.J. Hassink1,2 and Janneke Plantenga1 *Correspondence: [email protected] 1Utrecht University School of Economics, P.O. Box 80125, 3508 TC Utrecht, The Netherlands 2IZA, Bonn, Germany Abstract In recent years, many employees have gained more control over temporal and locational aspects of their work via a variety of flexible work arrangements, such as flexi-time and telehomework. This temporal and locational flexibility of work (TLF) is often seen as a means to facilitate the combination of work and private life. As such it has been recommended as a policy to increase the average number of working hours of part-time workers. To the best of our knowledge, the effectiveness of this policy instrument has not been tested empirically yet. We therefore analyse whether flexi-time and telehomework arrangements increase the number of actual, contracted, and preferred working hours. Based on Dutch household panel data, our results indicate that the link between TLF and working hours is quite weak. Telehomework is associated with moderate increases in actual hours, but not in contracted or preferred hours. Flexi-time generally does not seem to be associated with an increase in hours worked. Despite positive effects on job satisfaction and working time fit, we do not find any convincing evidence of a positive effect of TLF on labour supply. JEL classification: J22, J32, M52, M54 Keywords: Flexi-time, Labour supply, Locational flexibility, Part-time work, Telehomework, Temporal flexibility 1 Introduction In recent years, many employees have gained more control about when, where, and how long they work. This temporal and locational flexibility of work (TLF), which facilitates employees with flexibility in the schedule, location, and duration of their work, is usually implemented via a variety of flexible work arrangements, such as flexi-time, selfscheduling, and telehomework (Fagan 2004; Hill et al. 2008; Plantenga 2003; Rau 2003). This development has been fostered by an increasing relevance of knowledge work and the service industries in general, by new forms of work organisation, but in particular by the proliferation of information and communication technology (ICT), which has facilitated asynchronous and remote exchange of information. In many debates, TLF is primarily viewed as a means to combine work and private life and consequently has become highly topical in the policy debate in a number of countries (see, e.g. CEA 2010; BMFSFJ 2012). As such, TLF arrangements have also become a common policy recommendation to increase labour supply in order to increase economic © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 2 of 34 growth and to prevent labour force shortages in the future (Rürup and Gruescu 2005; Sociaal-Economische Raad 2011; Taskforce DeeltijdPlus 2010). In environments with relatively high labour force participation rates and a lot of (female) part-time workers, the main focus is on increasing the number of working hours with more TLF. It has not, to the best of our knowledge, been tested empirically yet, whether this supposed link between TLF and increasing working hours actually holds, however. The aim of this paper is therefore to analyse whether and to what extent TLF arrangements indeed influence labour supply. In particular, we analyse the impact of TLF arrangements that provide schedule and location flexibility on the number of actual, contracted, and preferred working hours. The analysis is carried out on the basis of a Dutch household panel dataset. The Netherlands are a good test case in this context, because they are a highly developed service society with an excellent ICT infrastructure, which means that the scope for TLF is relatively high. Increasing working hours of parttime employees has been a policy concern for some years now (Sociaal-Economische Raad 2011; Taskforce DeeltijdPlus 2010), and the Dutch labour market is quite flexible already. Employees in the Netherlands have a legal right to both decrease and increase their contracted working hours for example.1Obstacles to adjust working hours are therefore comparatively low, and contracted hours should adapt relatively quickly to new conditions, also within existing employment relations. Our results indicate that the association between TLF and working hours is quite weak. Telehomework is associated with moderate increases in actual hours, but not in contracted or preferred hours, which implies that telehomework is primarily associated with more overtime. Flexi-time generally seems to have an ambiguous effect on working hours. So despite positive effects on job satisfaction and working time fit (see e.g. Possenriede and Plantenga 2014), we do not find any convincing evidence of a positive effect of TLF on labour supply. 2 Theoretical framework Female labour force participation rates have increased tremendously in the Netherlands in the last two decades. In recent years, they have been around 73 % and thus eight to ten percentage points above EU average. Average weekly working hours of females, however, have stagnated at a relatively low 25 hours per week (see Fig. 1). Part-time work is used extensively to combine work and private life in the Netherlands—private life here referring to any other responsibility, activity, or event that is not paid work. The resulting low Fig. 1 Participation rates and average working hours in the EU and the Netherlands, 1992–2011 Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 3 of 34 number of weekly and annual working hours is considered to be problematic in the face of an ageing society, the expected labour force shortages, gender equality, and low economic growth in general. Various policy initiatives to increase female labour supply at the intensive margin have therefore been undertaken, among which calls for more temporal and locational flexibility. The idea is that more TLF and thus more control over working hours will improve work-life fit and induce employees to supply more hours to the labour market. As a result, arrangements such as flexi-time and telehomework can to some extent substitute part-time work as a means to reconcile work and private life. The notion that more TLF may lead to an increase in labour supply has been supported by surveys in which a considerable share of respondents report that they would be willing to supply more hours to the labour market if more flexibility options were available. In a 2009 survey for example, 35–41 % of non-participants and 25–39 % of part-time workers responded that more flexibility would be an important condition to either participate in or supply more hours to the labour market, respectively (Cloïn et al. 2010). The conditions mentioned include better reconciliation of working times and private life, finding a job with the preferred number of hours, working part of the week from home, being able to take a day off if a family member gets sick, and finding a job closer to home. If these responses are sincere, more temporal and locational flexibility and a better fit between work and private life should indeed raise labour supply. In theory, two potential channels can lead from increased flexibility to more working hours. The first one concerns a decrease in commuting time, the second a reduction of schedule constraints and a better match between work and private schedules. Although commuting can be seen as a prerequisite for paid work, commuting time per se is unproductive and inefficient. Commuting time can be reduced with flexible working times, because it is possible to avoid rush hour traffic by commuting at less busy times. Commuting can even be eliminated altogether when one is working at home. This time gain can then be spent at work.2 While the theoretical predictions are to some extent dependent on the assumptions made, a simple model predicts exactly this.3In a graphical representation (Fig. 2), C designates the consumption of goods, Lthe consumption of leisure, and L0maximum Fig. 2 Commuting and labour supply Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 4 of 34 time available. The line L0Adesignates non-labour income, IC is the indifference curve, corresponding to the level of utility obtained by the individual, and BC is the budget line with slope w, the wage rate. If time has to be spent on commuting, the optimal solution is situated at the tangency point E1of the indifference curve with the budget constraint BC1. Here, the individual would supply L1Lchours on labour, spend LcL0=con commuting, and enjoy 0L1hours of leisure.4 If commuting is eliminated, the budget constraint shifts to the right and utility increases. The optimal solution is now at E2and leisure time increases by L1L2.Atthe same time, labour supply increases by c−L1L2and is now L2L0. Part of the time gain due to a reduction in commuting time will thus be spent on additional labour supply. This result holds unambiguously if we assume leisure and consumption to be normal goods (Black et al. 2014).5 The existing empirical evidence, however, suggests that the size of the effect of TLF due to commuting time savings alone is likely to be limited. While there is little direct evidence on the effects of commuting costs on labour supply, indirect evidence (e.g. that commuters seem to attach relatively low value to travel time) suggests that the effect of the length of the commute on labour supply is rather weak (Gibbons and Machin 2006). More recently, Gutiérrez-i-Puigarnau and van Ommeren (2010) even found a small positive effect of commuting on the number of daily and weekly working hours. A second channel via which more temporal and locational flexibility can lead to an increase in labour supply is a reduction of schedule constraints and a better match between schedules of work and private life. Tasks and events of both paid work and private life are not distributed randomly over the day and week. Most of the time, they take place within defined schedules, because in both spheres workers depend on and interact with other individuals. Work is usually carried out in teams within and across firms and many workers deal directly with clients and business partners. Goods and services have to be produced and handled at specific times because they are expected by other workers in the production chain, clients expect them at specific times (e.g. during opening and business hours), or because the goods and services involved are perishable or otherwise time-critical. In addition, working hours and schedules are generally limited due to legal restrictions and social norms. As a result, workers are often constrained in the choice of their working schedule. In the same vein, the timing of leisure tasks and activities often depend on others. The schedules of working parents for example depend on their children’s daycare and school schedules. Informal care often has to be performed at specific times of the day (Hassink and van den Berg 2011). Stores, businesses, and public and health services have limited business and opening hours. Further education classes and recreational activities (sports, clubs, etc.) take place at designated times. Since daycare, school, office, and service hours usually cannot be altered by individual workers, they constitute a binding schedule constraint. Both work and leisure activities thus impose a schedule constraint on workers, meaning that these activities can only be performed at specific times or within a specified time frame. Activities therefore have to be coordinated and their schedules matched. This matching can be achieved more easily the more flexible and controllable schedules are. More flexibility in work schedule and location should therefore improve the fit between work and leisure activities, prevent and eliminate time conflicts, and improve allocative Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 5 of 34 efficiency (Bosch et al. 1994; Golden 2006). TLF has accordingly been shown to increase working-time fit and job satisfaction (Possenriede and Plantenga 2014). The trade-off between working hours and schedules is easily illustrated with an extension of the standard labour supply model (Golden 1996; 2006). According to the standard model, an individual’s well-being is determined by his or her consumption and hours of leisure. This is represented by a utility function (U)with the standard consumption (C) and leisure (L)arguments. But since utility from work is also dependent on work schedule, location, and their flexibility, we add a schedule parameter (S)to the utility function: U=U[C,L,S](1) Srepresents both the schedule and location of work as well as their adaptability. Based on the considerations above, we assume that work schedules and locations that fit in well with leisure activities provide more utility than those that do not. In addition, flexible schedules and locations provide more utility than inflexible ones, because they are more adaptable to changing circumstances for instance. Note that this is not much different from the usual assumptions about consumption and leisure in standard labour economics. Just like we assume that consumption and leisure are normal goods, i.e. that more consumption and more leisure “is better” in the sense that it provides more utility, we assume here that more flexibility in work schedule and location (i.e. more control over timing and location of work) and thus a better schedule and working time fit “is better” as well. It follows that utility is strictly increasing and marginal utility decreasing in all arguments: ∂U ∂C,∂U ∂L,∂U ∂S>0; ∂2U ∂C2,∂2U ∂L2,∂2U ∂S2≤0(2) Under these general assumptions, workers should be willing to trade leisure time or income for more flexibility and vice versa (Golden 2006).6 We need to keep in mind, though, that consumption potentially is another channel via which a trade-off with TLF might take place. Instead of increasing labour supply, workers may be willing to trade part of their wage or future wage increases for increased flexibility and better work schedules (Altonji and Paxson 1988; Baughman et al. 2003; Heywood et al. 2007). This will be addressed in the empirical analysis by controlling for wage. Based on these considerations, we arrive at the following hypothesis: Hypothesis. More temporal and locational flexibility of work is associated with an increase in hours worked. There are a few reasons to believe, however, that the overall impact of TLF on working hours is limited and may even be negative in the aggregate. Employees may for example not be willing to increase labour supply but may just as well enjoy their improved work-life fit from increased TLF. Labour market imperfections may allow them to do so, since TLF seems to be primarily distributed among higher-status jobs with possibly less supply side competition (Felstead et al. 2002; Golden 2008; 2009; Smulders et al. 2011). Norms and societal preferences may reinforce this trend further, in the sense that work norms have eroded due to proliferation of part-time work in the Netherlands (Wielers and Raven 2013) and that it is therefore not “attractive” to increase working hours (see, e.g. Bosch et al. 2010; Booth and van Ours 2013). Furthermore, telehomework may partly be used to just transfer some work home, so working time at the Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 6 of 34 office is substituted for working time at home (Noonan and Glass 2012; Peters and van der Lippe 2007). TLF may also impose externalities on an employee’s partner at home and colleagues at work. From a household perspective, an increase in the number of hours (due to the availability of TLF or otherwise) may at least partially be compensated by a reduction in the hours of the partner. At work, there may be a trade-off between the match between the schedules of work and private life for one employee and the match between work schedules of several colleagues. To improve the latter, fixed schedules may act as a coordination device. So even if TLF is generally available at a workplace, an employee may de facto not be able to make use of it. Similarly, it might not be viable in certain jobs to supply more hours, if those hours can only be supplied late in the evening, on weekends, or at home for example, because those particular jobs require constant interaction with colleagues or clients. TLF may finally not be viewed as a means of more control by the employee, if the utilisation of TLF is imposed by the employer in order to increase operating hours or save on office space for example. So the mere availability of TLF arrangements may not necessarily imply a better fit between work and private life. The question of whether TLF increases the supply of working hours is therefore essentially an empirical matter. One that we are trying to solve in the following sections. 3 Data and variable description The data for the analysis is taken from the Dutch Labour Supply Panel (Arbeidsaanbodpanel, AAP), a biennial panel survey of a representative sample of Dutch households (Sociaal en Cultureel Planbureau (SCP) 2015).7The panel survey is conducted to study developments in labour market behaviour and working conditions in the Netherlands and covers a broad range of work- and life-course-related items. The target population consists of the Dutch labour force aged 16 to 66 years. The AAP has existed since 1985, but questions about (tele)homework were first asked in 2002, so only the waves from 2002 onwards are suitable for an analysis of TLF. This means that we have six waves available for this analysis, for every other year since 2002 to the last publicly available wave from 2012. We restrict the sample to employees (i.e. we exclude self-employed, unemployed, full-time students, etc.), which results in an unbalanced panel of 20,452 observations from 8720 individuals. We use flexi-time and telehomework as indicators for TLF. The flexi-time variable was obtained from the following survey question: “Can you say whether each of the following characteristics does or does not apply to the work you do? [...] Determine start- and end-time myself” The telehomework variable was obtained from the following question: “Do you work at home every now and then in your current job?”8 We only count those respondents as telehomeworkers who state that they work at home at least once a week on average.9On average, 39 % of the respondents in the sample can determine the start- and end-times of their work and 18 % work at home at least once a week. The shares of flexi-timers and telehomeworkers are generally larger for male than for female employees (see Table 1). Over the six waves, 1188 respondents change their Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 7 of 34 Table 1 Share of flexi-time and telehomework by year and gender Male Female Total N %S.E.%S.E.%S.E. Flexi-time 2002 36.41 (1.57) 26.72 (1.41) 31.51 (1.07) 1952 2004 45.54 (1.28) 34.37 (1.37) 40.57 (0.95) 2748 2006 43.46 (1.25) 33.52 (1.26) 38.75 (0.91) 3035 2008 47.92 (1.27) 34.73 (1.24) 41.56 (0.90) 3073 2010 44.01 (1.40) 33.12 (1.34) 38.60 (0.99) 2518 2012 48.67 (1.37) 34.50 (1.31) 41.62 (0.98) 2696 Total 44.81 (0.83) 33.12 (0.81) 39.18 (0.60) 16,022 Telehomework 2002 15.98 (1.19) 13.16 (1.09) 14.55 (0.82) 1952 2004 17.37 (0.97) 14.98 (1.02) 16.30 (0.73) 2748 2006 18.79 (0.98) 16.62 (0.98) 17.76 (0.71) 3035 2008 20.19 (1.01) 18.54 (1.01) 19.39 (0.74) 3073 2010 19.72 (1.12) 17.84 (1.08) 18.78 (0.80) 2518 2012 20.83 (1.10) 18.93 (1.08) 19.88 (0.78) 2696 Total 18.94 (0.63) 16.88 (0.64) 17.95 (0.47) 16,022 S.E. standard error of the mean flexi-time and 845 change their telehomework status (i.e. they were, e.g. working at home in at least one wave and not working at home in another). The availability and use of flexi-time and telehomework varies greatly across sectors (see Table 2). This suggests that job-related factors play an important role here. We use actual, contracted, and preferred hours as outcome variables. Prior research has found a positive relationship between telehomework and hours worked (Eldridge and Wulff Pabilonia 2008; Noonan and Glass 2012; Peters and van der Lippe 2007). This has been mainly attributed to an increase in work demands and overtime, as well as an expansion of the standard working week via an increase in actual hours (Noonan and Glass 2012; Peters and van der Lippe 2007). If only actual and (unpaid) overtime hours increase but contracted and preferred hours do not, workers may not benefit from the increase in working hours through higher income (assuming that the hourly wage would stay the Table 2 Flexi-time and telehomework by sector Sector Flexi-time Telehomework N %S.E.%S.E. Agriculture 32.31 (4.12) 6.92 (2.23) 130 Industry 39.13 (1.12) 10.71 (0.71) 1886 Construction 31.02 (1.74) 10.48 (1.15) 706 Trade, gastronomy, repair 24.25 (0.92) 8.47 (0.60) 2173 Transport 31.59 (1.47) 6.72 (0.79) 997 Business services 55.20 (0.95) 19.83 (0.76) 2743 Care, welfare 30.44 (0.80) 14.16 (0.61) 3285 Other services 44.19 (1.83) 19.73 (1.46) 740 Government 67.21 (1.18) 16.21 (0.92) 1592 Education 29.66 (1.09) 52.37 (1.19) 1770 Total 39.18 (0.39) 17.95 (0.30) 16,022 Note: Share of employees with flexi-time and telehomework by sector S.E. standard error of the mean Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 8 of 34 same after an increase in contracted hours). Furthermore, the work-life fit of employees may not increase but rather decrease if preferred hours stay the same. We therefore not only analyse the impact of TLF on actual working hours but consider contracted and preferred hours as well. The actual (only available from 2004 onwards) and contracted hour variables were obtained from the following survey questions: “How many hours per week do you actually work on average?” and “How many hours do you work according to your contract? Overtime-hours should not be considered.” The preferred hours variable was derived from the following question: “Are you satisfied with the current number of contract hours or would you like to work more or fewer hours? Take into account that your hourly wage does not change and that others in your household will not work more or fewer hours.” The answer categories are as follows: “Yes, satisfied with hours; No, I would like to work XMORE hours per week; No, I would like to work XFEWER hours per week.” Contracted hours were used as the basis for the preferred hours variable, to which Xhours were added or subtracted depending on whether respondents indicated that they wanted to work more or fewer hours. On average, employees in the sample actually work 33.63 h (39.66 and 27.05 h for male and female employees, respectively, see Table 3). Contracted hours are a little lower at 31.12 h (36.58 h for males and 25.24 h for females). Preferred hours are again slightly lower but have been slowly increasing for female employees in the period under consideration (from 23.74 h in 2002 to 25.43 h in 2012; not shown). In order to rule out confounding factors due to differences in individual, household, and job characteristics, we add a number of control variables to our models. These are respondents’ age, marital status, children at home, level of education, work experience, changes in employment (e.g. promotions and demotions within the same job as well as job switches), two or more jobs, contract type, level of occupation, number of supervised employees, sector, firm size, and a time trend. Observations with missing values on any of these variables are dropped from the analysis by listwise deletion, resulting in a net sample of 16,022 observations from 7164 individuals. Table 4 shows the descriptive statistics for both this net sample and the gross sample (N=20, 452) without listwise deletions. 4 Empirical analysis Our analysis starts out with a simple cross-tabulation of the TLF and working hour variables (Table 5). A comparison of working hours of employees with and without TLF seems Table 3 Average working hours by gender Working hours Male Female Total N Mean S.E. Mean S.E. Mean S.E. Actual hours 39.66 (0.10) 27.05 (0.13) 33.63 (0.10) 14,046 Contracted hours 36.58 (0.07) 25.24 (0.11) 31.12 (0.08) 16,022 Preferred hours 35.85 (0.08) 25.29 (0.11) 30.76 (0.08) 16,022 Note: Employees’ average working hours by gender S.E. standard error of the mean Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 15 of 34 Table 9 TLF arrangements on working hours—robustness checks (Continuation) Telehomework 0.944** 0.723* 1.098** 0.271 −0.024 0.495* 0.302 0.121 0.403 (0.216) (0.293) (0.311) (0.146) (0.174) (0.238) (0.186) (0.258) (0.261) Observations 14046 7325 6721 16022 8299 7723 16022 8299 7723 Individuals 6409 3333 3076 7164 3681 3483 7164 3681 3483 2004–2012 only Flexi-time 0.223 0.555* −0.189 −0.008 0.192 −0.256 0.041 0.284 −0.312 (0.208) (0.278) (0.307) (0.168) (0.202) (0.270) (0.200) (0.273) (0.284) Telehomework 0.920** 0.681* 1.103** 0.199 −0.089 0.438 0.215 −0.141 0.573* (0.218) (0.298) (0.312) (0.155) (0.192) (0.248) (0.202) (0.274) (0.292) Observations 14046 7325 6721 14046 7325 6721 14046 7325 6721 Individuals 6409 3333 3076 6409 3333 3076 6409 3333 3076 Note: Parameter estimates of TLF arrangements on working hours. Standard errors in parentheses (clustered at employee level) *p<0.05, **p<0.01 Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 16 of 34 More generally, our estimates do not allow for drawing strong conclusions about the causal effect of flexi-time and telehomework on labour supply, however. This is due to the following issues. Self-selection and other sources of endogeneity, like common shocks that may at the same time influence the availability and usage of TLF on the one hand and the number of working hours on the other, may bias our estimates. Many employees may for example choose working hours and working conditions simultaneously as parts of a whole employment package at the start of a contract.17 Time-varying family commitments like elderly care or the birth of a (second) child may lead to both decreases in hours worked and higher demand for TLF. Employers may also award TLF to employees with the largest productivity, status, or authority (cf. Golden 2009; Winder 2009; Noonan and Glass 2012). Since we control for changes in employment as well as for time-fixed-effects, both of these confounding factors are only relevant, however, if they are time-varying and independent of promotions and job switches for example. Underlying the specification as depicted by Eq. 4 is furthermore that the idiosyncratic error term it is strictly exogenous, i.e. that time-varying unobserved variables from any time period may not be correlated with any of the explanatory variables from all time periods. This assumption would be violated for example, if there were reverse causality or a feedback mechanism going on. This might be the case for TLF, because instead of being a policy for improved working conditions and work-life fit, employees might get more TLF, when (actual) working hours and workloads increase in order to get the job done (Noonan and Glass 2012). This means that causality would run from increased hours to TLF and not the other way around. Prior research has indeed shown that employees at the high end of the hour spectrum have considerably more access to flexi-time for example than those with a standard 40-h working week. Part-time employees, however, enjoy greater availability than full-time employees as well, so access to TLF seems to actually be U- shaped in working hours (Golden 2008; 2009). So there might be a feedback mechanism that may work in both directions, which means that the overall impact on our estimates is unclear. We have considered the following procedures to remedy these potential issues. First, we took alternative estimation techniques like differences in differences or regression discontinuity into account. Unfortunately, these cannot be applied, because there is no exogenous variation in the availability of schedule and location flexibility during the period under consideration and there is no exogenous variation in the dependent variables either. The closest is the Working Hours Adjustment Act, which was implemented in 2000. Important variables for our analysis, among them the number of actual work hours and availability of telehomework, are not available in the AAP before 2004 and 2002, respectively, however. Next, we tried to find suitable external instrumental variables for TLF. We tested various autonomy measures as IVs for telehomework and flexi-time such as “I can determine how I do my job”, “I can determine my work speed”, or “I can determine in which order I do my work”, for instance. All of these variables turned out to be weak instruments in the fixedeffects specification, however, and would therefore lead to inconsistent estimates. Hence, we were not able to find any suitable external instrumental variables in our data. We were able to use flexi-time and telehomework from the previous wave as internal instruments. This again leads to a significant drop in sample size, because only those Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 17 of 34 respondents with observations from at least two subsequent waves can be used. All coefficients indicate no significant association between TLF and working hours (see Table 9, panel 2). Next we experimented with the inclusion of lags of flexi-time and telehomework to take possible adjustment lags into account. Again, the results indicate a significant drop in sample size and no significant association between TLF and working hours (panel 3). We also winsorised the working hour variables to the 1st and 99th percentile, limiting the range to 3 to 60 h per week, to rule out that outliers may be driving our results. In addition, we excluded the 2002 wave from our sample, since the answer categories for the frequency of telehomework, on which our telehomework variable ultimately relies, differ in this wave, and to rule out that the differences in effects with respect to actual, contracted, and preferred is merely caused by differences in data availability. Most coefficients in these two specifications are statistically insignificant, and where they are significant, the results are comparable to the baseline estimates (panels 4 and 5).18 As a next step, we also estimated Eq. 4 with various interactions (see Table 10). First, we estimated the model with an interaction between flexi-time and telework. This interaction is statistically insignificant in all of the specifications tested, which means that schedule and location flexibility are independent types of flexibility (panel 1). Second, we respectively interacted flexi-time and telehomework with whether or not there are children at home or whether the individual is below or above the median age of 42 years. One may assume that these groups differ in their preferences for TLF and their propensity to adjust working hours (panels 2 and 3). Third, we created indicator variables for white-collar workers (i.e. employees with a “higher” or “scientific” level of occupation) and service sector jobs (i.e. those in business services, other services, government and education sectors, as opposed to agriculture, industry, construction, trade, gastronomy, repair, transport, care, and welfare). For both these groups, we may assume that they have better access to TLF and can more easily adjust working hours. We interact both variables with flexi-time and telehomework, respectively (panels 4 and 5). Finally, we added an interaction between flexi-time and telehomework and whether or not the wave is post 2008 to account for possible effects due to the Great Recession. While the occasional coefficient of these interactions may be significantly different from zero, the bigger picture is that the above-mentioned interactions are generally not significant and that the main effects are quite similar to the baseline specification. This means that according to our estimates there seem to be little to no differences in the associations between TLF and the supply of working hours for different groups of employees. 6Conclusions In this paper, we analyse the effect of temporal and locational flexibility of work (TLF) on the number of working hours using Dutch household panel data spanning the period from 2002 to 2012. We test the claim that more TLF is associated with an increase in labour supply due to a better fit between work and private life. An increase of TLF has been a common policy recommendation to boost labour supply in order to enhance economic growth and to prevent labour force shortages in the future. According to our estimates, the general association between TLF and the number of working hours is small at best. According to our baseline results, telehomework is Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 18 of 34 Table 10 TLF arrangements on working hours—robustness checks: interaction Actual hours Contracted hours Preferred hours Total Male Female Total Male Female Total Male Female Interaction between flexi-time and telehomework Flexi-time 0.189 0.547 −0.262 −0.053 0.205 −0.374 −0.017 0.348 −0.493 (0.222) (0.298) (0.329) (0.166) (0.198) (0.271) (0.197) (0.272) (0.282) Telehomework 0.817* 0.653 0.922 0.408 0.193 0.552 0.475 0.450 0.453 (0.340) (0.480) (0.476) (0.222) (0.287) (0.329) (0.291) (0.439) (0.380) Flexi-time ×telehomework 0.169 0.042 0.345 −0.201 −0.281 −0.104 −0.188 −0.411 0.043 (0.399) (0.541) (0.574) (0.253) (0.313) (0.400) (0.336) (0.493) (0.461) Interaction between flexi-time, telehomework, and median age Flexi-time 0.213 0.796 −0.371 −0.192 0.239 −0.668 −0.152 0.231 −0.604 (0.337) (0.435) (0.504) (0.247) (0.279) (0.406) (0.285) (0.376) (0.415) Telehomework 1.198** 0.584 1.569** 0.347 −0.055 0.491 0.422 0.300 0.283 (0.368) (0.512) (0.510) (0.245) (0.297) (0.392) (0.310) (0.411) (0.445) >Median age 0.644* 0.225 1.092** 1.043** 0.546 1.496** 0.928** 0.335 1.541** (0.288) (0.403) (0.390) (0.226) (0.278) (0.342) (0.265) (0.355) (0.376) Flexi-time ×> median age 0.009 −0.376 0.313 0.171 −0.143 0.498 0.168 0.058 0.223 (0.363) (0.484) (0.532) (0.266) (0.300) (0.446) (0.320) (0.420) (0.473) Telehomework ×> med age −0.447 0.138 −0.829 −0.111 0.100 −0.096 −0.109 −0.175 0.218 (0.420) (0.579) (0.580) (0.271) (0.319) (0.443) (0.356) (0.455) (0.546) Interaction between flexi-time, telehomework, and child(ren) at home Flexi-time 0.141 0.668 −0.299 0.025 0.484 −0.263 −0.185 0.321 −0.661 (0.330) (0.448) (0.478) (0.247) (0.314) (0.386) (0.308) (0.446) (0.416) Telehomework 0.500 0.326 0.505 −0.061 −0.272 −0.063 0.094 0.107 −0.078 (0.384) (0.539) (0.540) (0.257) (0.347) (0.376) (0.300) (0.429) (0.419) Child(ren) −1.292** 0.036 −2.519** −0.927** 0.563 −2.300** −0.880** 0.301 −1.988** (0.308) (0.398) (0.464) (0.263) (0.319) (0.394) (0.295) (0.421) (0.398) Flexi-time ×child(ren) 0.142 −0.176 0.171 −0.190 −0.541 −0.222 0.217 −0.084 0.282 (0.365) (0.492) (0.532) (0.279) (0.358) (0.423) (0.333) (0.478) (0.458) Telehomework ×child(ren) 0.666 0.536 0.992 0.554 0.417 0.927* 0.430 0.114 0.904 (0.415) (0.589) (0.583) (0.292) (0.383) (0.443) (0.359) (0.509) (0.504) Interaction between flexi-time, telehomework, and white collar worker Flexi-time 0.247 0.387 0.043 −0.019 0.150 −0.225 0.166 0.341 −0.079 (0.248) (0.347) (0.343) (0.190) (0.241) (0.292) (0.226) (0.331) (0.302) Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 19 of 34 Table 10 TLF arrangements on working hours—robustness checks: interaction (Continuation) Telehomework 0.794* 0.812 0.577 0.246 0.059 0.314 0.365 0.636 −0.143 (0.318) (0.477) (0.414) (0.233) (0.312) (0.358) (0.300) (0.437) (0.402) White collar 0.492 0.249 0.500 0.475* 0.368 0.471 0.539 0.483 0.474 (0.279) (0.379) (0.405) (0.223) (0.282) (0.335) (0.276) (0.386) (0.388) Flexi-time ×white collar −0.003 0.487 −0.622 −0.125 0.102 −0.442 −0.515 −0.127 −1.097* (0.320) (0.418) (0.493) (0.246) (0.293) (0.409) (0.291) (0.402) (0.428) Telehomework ×white collar 0.218 −0.222 0.949 0.078 −0.088 0.325 −0.004 −0.736 1.020* (0.373) (0.524) (0.526) (0.268) (0.335) (0.433) (0.353) (0.491) (0.499) Interaction between flexi-time, telehomework, and service-sector jobs Flexi-time 0.205 0.639 −0.399 −0.194 0.266 −0.739* −0.071 0.403 −0.663* (0.264) (0.350) (0.389) (0.195) (0.241) (0.309) (0.237) (0.347) (0.312) Telehomework 1.094** 0.974* 1.150** 0.487* 0.077 0.862** 0.520* 0.331 0.668 (0.307) (0.441) (0.400) (0.213) (0.272) (0.324) (0.264) (0.375) (0.366) Services 0.566 0.284 0.874 0.259 0.351 0.155 0.270 0.319 0.197 (0.344) (0.417) (0.552) (0.291) (0.341) (0.481) (0.333) (0.435) (0.499) Flexi-time ×services 0.088 −0.190 0.563 0.218 −0.276 0.809 0.046 −0.297 0.466 (0.374) (0.481) (0.569) (0.279) (0.339) (0.448) (0.341) (0.456) (0.500) Telehomework ×services −0.336 −0.572 −0.066 −0.355 −0.179 −0.571 −0.275 −0.289 −0.282 (0.406) (0.561) (0.570) (0.279) (0.333) (0.463) (0.366) (0.469) (0.577) Interaction between flexi-time, telehomework, and post 2008 Flexi-time 0.346 0.656* −0.135 −0.108 0.127 −0.444 −0.080 0.300 −0.656* (0.223) (0.288) (0.335) (0.163) (0.197) (0.264) (0.192) (0.256) (0.283) Telehomework 0.828** −0.583 0.974** 0.236 −0.026 0.404 0.304 0.154 0.341 (0.247) (0.345) (0.346) (0.169) (0.212) (0.267) (0.219) (0.305) (0.306) Post 2008 −0.056 −0.340 0.031 −0.142 −0.167 −0.264 −0.439* −0.644* −0.353 (0.200) (0.279) (0.284) (0.162) (0.214) (0.240) (0.210) (0.302) (0.294) Flexi-time ×post 2008 −0.338 −0.253 −0.184 0.042 0.104 0.097 0.093 −0.074 0.468 (0.224) (0.293) (0.329) (0.183) (0.220) (0.288) (0.237) (0.318) (0.349) Telehomework ×post 2008 0.267 0.264 0.366 0.150 0.092 0.262 0.109 −0.003 0.284 (0.292) (0.402) (0.413) (0.212) (0.233) (0.351) (0.306) (0.411) (0.441) Observations 14046 8299 6721 16022 8299 7723 16022 8299 7723 Individuals 6409 3681 3076 7164 3681 3483 7164 3681 3483 Note: Parameter estimates of TLF arrangements on working hours. Standard errors in parentheses (clustered at employee level) *p<0.05, **p<0.01 Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 20 of 34 positively associated with actual hours and our results indicate an increase of an hour per week. Contracted and preferred working hours are generally not affected significantly by telehomework, except for a small positive association between telehomework and contracted hours for females. Telehomework therefore does not seem to be associated with a structural increase in contracted nor preferred working hours but seems to be primarily associated with an increase in actual working hours. At least part of the positive effect of telehomework on actual hours seems to be driven by an increase in unpaid overtime hours. Preliminary estimates indicate that unpaid overtime hours increase by 1.5 h per week for male employees and 0.75 h per week for female employees who work at home at least once a week.19 This suggests that TLF may also be used for work intensification and an increase in overtime hours, a result that has been discussed before (Noonan and Glass 2012; Peters and van der Lippe 2007). An alternative interpretation is that employees may reciprocate TLF availability by exerting extra effort (Akerlof 1982; Kelliher and Anderson 2010). Previous findings that TLF increases job satisfaction (Possenriede and Plantenga 2014) and job performance (Baltes et al. 1999; Bloom et al. 2015; Eaton 2003; Gajendran and Harrison 2007; Hill et al. 1998) support this interpretation. Both explanations, i.e. telehomework as a means for employers to intensify work on the one hand and more work effort from employees in exchange for more flexibility on the other, may nevertheless apply, particularly at different ends of the job spectrum. With respect to flexi-time, the results are even more limited. Almost all coefficients are statistically insignificant at the 5 %-level except for a small positive association between flexi-time and actual hours for males. Flexi-time therefore does not seem to be associated with an increase in hours worked. Overall, we cannot convincingly reject the null hypothesis of no effect of TLF on working hours. Even though we do find some small, positive associations between TLF and working hours here and there, the bigger picture is that after considering and controlling for mitigating factors, such as preferences, job, and household characteristics, the association between TLF and working hours vanishes (almost) completely. In addition, our sensitivity analyses do not reveal systematic differences in the associations between TLF and working hours for different groups. So there is no convincing evidence to support the claim that more schedule and location flexibility leads to a larger supply of working hours in general. The merits of this study are that we consider two TLF arrangements, namely flexi-time and telehomework, at the same time and that we utilised data spanning 12 years from different sectors. The results further indicate the importance of controlling for unobserved heterogeneity in jobs and individuals in these types of analyses. A limitation of our study is that we cannot completely rule out endogeneity and reverse causality and therefore do not identify a true causal effect. Since the association between TLF and working hours after controlling for several individual and job-related factors turns out to be almost absent, this does not seem to be a major issue, however. Future research could extend this analysis with other TLF arrangements, like self-scheduling or working time accounts. Overall, the hypothesis that more temporal and locational flexibility of work leads to an increase in hours worked appears to be rejected. The findings suggest that TLF does not have significant effects on labour supply at the intensive margin with the Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 21 of 34 exception of telehomework and actual hours. This implies that the arguments regarding increases in labour supply in the debate about policy support for TLF are not empirically supported. This does not take away that there may be other good reasons to support policies for more TLF, e.g. higher productivity (Bloom et al. 2015), increased job satisfaction and working-time fit (Possenriede and Plantenga 2014), and less absenteeism (Possenriede et al. 2014). Endnotes 1Every employee who has worked for a company with ten or more employees for at least one year can request a working hours adjustment. This right can be exercised once a year. The employer may only dismiss a request if it is a severe impediment to business interest. The Working Hours Adjustment Act (Wet Aanpassing Arbeidsduur) has been effective since mid 2000. Equal treatment of part-time and full-time employees with respect to employment conditions is furthermore stipulated in the Equal Treatment Working Hours Act (Wet verbod op onderscheid naar arbeidsduur), effective since 1996. 2Flexible working times may also induce employees to travel to work earlier and leave from work later to avoid traffic congestion, increasing work duration as a result (Arnott et al. 1993; Gutiérrez-i-Puigarnau and van Ommeren 2010). 3Predictions differ depending on whether one distinguishes between monetary and time costs of commuting, whether workdays, daily and total hours are allowed to vary, and whether one considers a static or dynamic approach (See e.g. Manning 2003; Gutiérrez-i-Puigarnau and van Ommeren 2010; Black et al. 2014). 4The amount of commuting time is exaggerated in the figure for better visibility. 5We assume here that individuals are able to choose their preferred levels of consumption and leisure without any other constraints of course. In addition, a decrease in commuting costs and thus a shift in the budget constraint to the right reduces the size of the kink in the budget line. This reduction of the fixed costs of work not only increases labour supply at the intensive margin, but also induces non-working individuals to participate and thus raises labour force participation as well (e.g. Oi 1976; Cogan 1981; Black et al. 2014). Since Dutch labour participation rates are relatively high already, though, we focus on the effects on hours worked in this study. 6Note that this model also captures workers who do not have binding private schedule constraints in the above sense, but just a preference for work at certain intervals, e.g. due to certain life-style choices. Nevertheless the degree to which private schedule constraints are binding certainly differs between workers and depends inter alia on whether they have care responsibilities or not. 7The panel formerly known as the OSA Labour Supply Panel is now conducted on behalf of the Social Cultureel Planbureau (http://www.scp.nl/english/). The data and its documentation are in Dutch. 8While the question refers to work at home and not explicitly to telework, it is the firstquestioninthesectiontitled“telework”inthesurvey.Inaddition,only2.1% of the respondents who work at home do not use ICT. Hence we label this variable telehomework. 9Telehomeworkers were asked how often they were working at home on average. From 2004 onwards the answer categories were less than once per month, less than twice per Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 22 of 34 month, once per week or twice or more often per week. We only count the latter two categories as telehomeworkers. In 2002 the answer categories were once per month, twice per month, three times per month, more than three times per month. We include only the latter as telehomeworkers. 10A superscript to indicate the different outcome variables is omitted. 11We estimated the model with a random-effects specification as well. The crucial assumption of a random-effects specification, however, is that the individual-specific error term αiis not correlated with the right-hand side variables Zit, otherwise the estimated coefficients will be biased. Since the availability and usage of TLF and the number of working hours quite likely depend on various job and individual characteristics this assumption seems rather strong. The random-effects specification was thus firmly rejected by a Hausman specification test in favour of the fixed-effects specification for all models and (sub-)samples considered. 12Note that since the flexi-time and telehomework indicators are binary, we effectively estimate linear probability models for these TLF variables. We do not control for the other TLF arrangement in these models (i.e. flexi-time is not controlled for in the telehomework regression and vice versa). The linear probability models for flexi-time behave well, as no observations are predicted outside the unit-interval. For telehomework, only 4.4 %, 12.8 %, and 9.3 % of the observations are predicted outside the unit-interval for the total, male, and female sample respectively. 13The random-effects specification was again firmly rejected by a Hausman specification test for all samples. 14The division bias refers to the fact that hourly wages are calculated by dividing net wages per month by hours per month. This causes measurement error in hours to enter both sides of Eq. 4 and results in a spurious negative correlation between wages and hours. 15We use the same control variables as in the baseline specification in this and all following specifications in this paragraph. 16In their study on labour supply and commuting, Gutiérrez-i-Puigarnau and van Ommeren (2010) also find that the inclusion of an instrumented wage variable does not affect their results. 17Note however that more than half of the employees for which TLF status changes have no change in their employment status. This indicates that a considerable share of employees in our sample does not seem to make these choices simultaneously. 18It might seem desirable to estimate this model on the sub-sample of part-time working (female) employees as well. Part-time employment may be an alternative strategy to combine work and private life and one thus might expect the largest effects of schedule and location flexibility here. Empirically, this is incorrect, however, because one would selectthesampleonthedependentvariableandthusgetbiasedestimates.Furthermore weareinterestedintheneteffectofTLF,notjusttheeffectonpart-timers.Giventhedistribution of working hours across gender in the Netherlands, i.e. male employees mostly working full-time and female employees mostly working part-time, one could interpret gender as a proxy for part-time/full-time employment, however. 19We estimated a model like Eq. 4 on paid and unpaid overtime hours. These results are only indicative, however, due to the large number of employees with zero overtime hours. Estimates are therefore not shown. Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 23 of 34 Appendix Table 11 Control variables on working hours Actual hours Contracted hours Preferred hours Total Male Female Total Male Female Total Male Female Age -0.019 -0.133* 0.127* 0.004 -0.080 0.108** 0.134** 0.114* 0.167** (0.040) (0.057) (0.056) (0.029) (0.042) (0.041) (0.037) (0.056) (0.049) Marital status Ref: Married Cohabiting 0.909 -0.356 2.244** 0.707 -0.491 1.912** 0.488 -0.530 1.487* (0.491) (0.559) (0.770) (0.374) (0.362) (0.608) (0.459) (0.573) (0.697) Single 1.114 -1.676* 3.100** 1.090* -0.268 2.207** 1.793** 0.594 2.807** (0.699) (0.814) (1.001) (0.548) (0.449) (0.852) (0.597) (0.585) (0.911) Child(ren) -1.063** 0.111 -2.238** -0.886** 0.390 -2.185** -0.689** 0.302 -1.703** (0.257) (0.306) (0.403) (0.207) (0.224) (0.334) (0.249) (0.326) (0.358) Education Ref: Primary School Lower secondary 1.543 0.850 2.029 0.961 0.295 1.841 1.118 0.720 1.434 (1.006) (1.224) (1.722) (0.836) (1.038) (1.320) (0.943) (1.245) (1.263) Higher secondary 2.126* 0.754 3.462 1.438 0.438 2.775* 1.950* 1.147 2.823* (1.077) (1.312) (1.820) (0.878) (1.083) (1.393) (0.988) (1.303) (1.338) Vocational college 4.418** 2.623 6.191** 3.118** 1.610 4.957** 3.237** 1.957 4.572** (1.180) (1.447) (1.965) (0.963) (1.182) (1.535) (1.064) (1.387) (1.490) Academic 4.864** 3.086 6.512** 3.742** 1.904 6.001** 4.233** 2.608 6.088** (1.356) (1.661) (2.274) (1.098) (1.312) (1.810) (1.190) (1.508) (1.772) Work experience 0.008 -0.006 0.009 -0.007 -0.011 -0.011 -0.017 -0.021 -0.021 (0.028) (0.042) (0.039) (0.019) (0.030) (0.024) (0.025) (0.039) (0.033) Permanent contract 0.873** 0.604 1.005* 0.599* 0.664 0.407 -0.020 -0.118 -0.049 (0.329) (0.465) (0.447) (0.276) (0.379) (0.381) (0.296) (0.438) (0.392) Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 24 of 34 Table 11 Control variables on working hours (Continuation) Empl. status change 0.222 -0.256 0.645** 0.092 -0.170 0.314 0.093 -0.107 0.228 (0.158) (0.223) (0.217) (0.118) (0.157) (0.171) (0.146) (0.213) (0.196) 2nd job -0.913* -1.045 -1.015 -1.419** -1.808** -1.285* -1.202** -1.541* -1.116 (0.448) (0.716) (0.563) (0.403) (0.565) (0.551) (0.423) (0.604) (0.579) Supervised employees Ref: None 1-9 employees 1.192** 0.929** 1.434** 0.542** 0.331 0.715** 0.385* 0.151 0.651* (0.213) (0.253) (0.348) (0.152) (0.170) (0.263) (0.180) (0.231) (0.278) 10-49 employees 2.006** 1.917** 2.000** 0.672** 0.465 0.876 0.722* 0.773* 0.443 (0.376) (0.423) (0.697) (0.248) (0.264) (0.490) (0.302) (0.352) (0.556) 50 or more employees 3.045** 3.113** 2.506 1.397 1.472* 0.935 1.253 2.165** -1.722 (0.805) (0.866) (1.683) (0.717) (0.600) (2.005) (0.875) (0.771) (2.327) Occupational level Ref: Medium Elementary -1.711** -2.418** -1.180 -1.197* -1.759** -0.627 -1.138* -1.750** -0.588 (0.606) (0.810) (0.943) (0.488) (0.594) (0.813) (0.520) (0.591) (0.904) Lower -0.687** -0.613* -0.572 -0.556** -0.631** -0.330 -0.195 -0.189 -0.086 (0.205) (0.275) (0.297) (0.169) (0.219) (0.249) (0.206) (0.309) (0.273) Higher 0.439* 0.365 0.368 0.332* 0.289 0.300 0.230 0.221 0.153 (0.195) (0.255) (0.304) (0.157) (0.192) (0.255) (0.191) (0.246) (0.296) Scientific 0.446 0.731 -0.160 0.430* 0.561* 0.152 0.270 0.297 0.220 (0.331) (0.415) (0.548) (0.212) (0.264) (0.351) (0.284) (0.348) (0.486) Sector Ref: Agriculture Industry 0.585 -1.488 4.033 1.481 -0.158 4.651 0.543 -0.868 3.300 (1.414) (1.277) (2.298) (1.066) (0.631) (2.407) (1.064) (0.874) (2.223) Construction 0.672 -1.236 3.563 2.426* 0.869 5.346* 1.143 -0.109 3.273 (1.510) (1.390) (2.531) (1.104) (0.703) (2.708) (1.092) (0.901) (2.552) Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 31 of 34 Table 13 TLF arrangements on working hours (Continuation) Sector Ref: Agriculture Industry 0.658 -1.360 3.985 1.477 -0.136 4.599 0.547 -0.819 3.238 (1.420) (1.290) (2.312) (1.065) (0.630) (2.409) (1.062) (0.871) (2.207) Construction 0.732 -1.127 3.539 2.418* 0.891 5.315* 1.142 -0.064 3.240 (1.514) (1.399) (2.541) (1.103) (0.701) (2.708) (1.090) (0.900) (2.542) Trade, gastronomy, repair -0.274 -2.123 3.220 0.672 -0.834 3.942 -0.046 -1.309 2.679 (1.448) (1.327) (2.363) (1.090) (0.662) (2.441) (1.085) (0.927) (2.204) Transport 1.280 -1.096 6.021* 1.854 -0.460 7.127** 0.554 -1.156 4.242 (1.554) (1.475) (2.686) (1.215) (0.886) (2.740) (1.216) (1.100) (2.504) Business services 1.003 -1.482 5.377* 1.851 0.118 5.439* 0.776 -0.604 3.624 (1.436) (1.282) (2.405) (1.070) (0.637) (2.442) (1.074) (0.875) (2.247) Care, Welfare 1.177 -2.264 5.682* 1.729 -1.017 5.719* 1.095 -1.341 4.499 (1.484) (1.415) (2.451) (1.142) (0.875) (2.517) (1.155) (1.173) (2.298) Other services 0.577 -2.157 5.181* 0.941 -0.965 4.581 0.244 -1.430 3.333 (1.484) (1.434) (2.424) (1.129) (0.795) (2.516) (1.168) (1.116) (2.320) Government 0.867 -1.396 4.881* 1.109 -0.661 4.615 0.694 -0.914 3.675 (1.440) (1.326) (2.449) (1.088) (0.723) (2.478) (1.097) (0.956) (2.270) Education 3.442* 0.363 8.314** 3.102* 0.422 7.515** 2.168 -0.903 6.495** (1.620) (1.740) (2.597) (1.230) (0.999) (2.613) (1.264) (1.234) (2.436) No. of employees (/1000) 0.024 0.055 -0.030 0.011 0.031 -0.017 0.022 0.036 0.009 (0.026) (0.029) (0.051) (0.020) (0.022) (0.045) (0.028) (0.031) (0.064) Constant 29.352** 44.701** 10.959** 27.001** 38.909** 11.996** 21.982** 30.338** 11.419** (2.239) (2.595) (3.592) (1.732) (1.813) (3.250) (1.961) (2.507) (3.197) Observations 14046 7325 6721 16022 8299 7723 16022 8299 7723 Individuals 6409 3333 3076 7164 3681 3483 7164 3681 3483 Note: Parameter estimates of TLF arrangements on working hours. Year (wave) dummies included. Standard errors in parentheses (clustered at employee level) *p<0.05, ** p<0.01 Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 32 of 34 Acknowledgements The authors would like to thank two anonymous referees as well as participants at ILERA European conference 2013 and IZA/RIETI workshop “Changing Demographics and the Labor Market” for valuable comments and suggestions. Data provision by Sociaal en Cultureel Planbureau and financial support by Instituut Gak are gratefully acknowledged. Responsible editor: Juan Jimeno Competing interests The IZA Journal of Labor Policy is committed to the IZA Guiding Principles of Research Integrity. The authors declares that they have observed these principles. Received: 27 July 2015 Accepted: 24 June 2016 References Akerlof GA (1982) Labor contracts as partial gift exchange. Q J Econ 97(4):543–569. http://www.jstor.org/stable/1885099 Altonji JG, Paxson CH (1988) Labor supply preferences, hours constraints, and hours-wage trade-offs. J Labor Econ 6(2):254–276. doi:10.2307/2535044 http://www.jstor.org/stable/2535044 Anderson DJ, Binder M, Krause K (2003) The motherhood wage penalty revisited: experience, heterogeneity, work effort, and work-schedule flexibility. Ind Labor Relat Rev 56(2):273–294. doi:10.2307/3590938. http://www.jstor.org/stable/ 3590938 Arnott R, Palma Ad, Lindsey R (1993) A structural model of peak-period congestion: a traffic bottleneck with elastic demand. Am Econ Rev 83(1):161–179. doi:10.2307/2117502. http://www.jstor.org/stable/2117502 Baltes BB, Briggs TE, Huff JW, Wright JA, Neuman GA (1999) Flexible and compressed workweek schedules: a meta-analysis of their effects on work-related criteria. J Appl Psychol 84(4):496–513. doi:10.1037/0021-9010.84.4.496. http://dx.doi.org/10.1037/0021-9010.84.4.496 Baughman R, DiNardi D, Holtz-Eakin D (2003) Productivity and wage effects of “family-friendly” fringe benefits. Int J Manpow 24(3):247–259. doi:10.1108/01437720310479723. http://dx.doi.org/10.1108/01437720310479723 Black DA, Kolesnikova N, Taylor LJ (2014) Why do so few women work in New York (and so many in Minneapolis)? Labor supply of married women across US cities. J Urban Econ 79:59–71. doi:10.1016/j.jue.2013.03.003. http://dx.doi.org/10. 1016/j.jue.2013.03.003 Blau FD, Kahn LM (2007) Changes in the labor supply behavior of married women 1980 –2000. J Labor Econ 25(3):393–438. doi:10.1086/513416. http://www.jstor.org/stable/10.1086/513416 Bloom N, Liang J, Roberts J, Ying ZJ (2015) Does working from home work? Evidence from a Chinese experiment. Q J Econ 130(1):165–218. doi:10.1093/qje/qju032. http://qje.oxfordjournals.org/content/130/1/165 BMFSFJ (2012) Zeit für Familie. Familienzeitpolitik als Chance einer nachhaltigen Familienpolitik. Achter Familienbericht. Bericht der Sachverstä, Bundesministerium für Familie, Senioren. Frauen und Jugend, Berlin. http://www.bmfsfj.de/ RedaktionBMFSFJ/Abteilung2/Pdf-Anlagen/Achter-familienbericht.pdf Bonke J, Datta Gupta N, Smith N (2004) The timing and flexibility of housework and men and women’s wages. In: Hamermesh DS, Pfann GA (eds). The economics of time use, Contributions to Economic Analysis. Elsevier, Amsterdam Vol. 271. pp 43–77. http://dx.doi.org/10.1016/S0573-8555(04)71003-2 Booth AL, van Ours JC (2013) Part-time jobs: what women want? J Popul Econ 26(1):263–283. doi:10.1007/s00148-012-0417-9. http://dx.doi.org/10.1007/s00148-012-0417-9 Borjas GJ (1980) The relationship between wages and weekly hours of work: the role of division bias. J Human Res 15(3):409–423. doi:10.2307/145291. http://www.jstor.org/stable/145291 Bosch G, Dawkings P, Michon F (1994) Times are changing. Working time in 14 industrialised countries. International Institute for Labour Studies - International Labour Organization (ILO), Geneva Bosch N, Deelen A, Euwals R (2010) Is part-time employment here to stay? Working hours of Dutch women over successive generations. LABOUR 24(1):35–54. doi:10.1111/j.1467-9914.2010.00470.x. http://dx.doi.org/10.1111/j.1467-9914.2010. 00470.x CEA (2010) Work-life balance and the economics of workplace flexibility. Executive Office of the President – Council of Economic Advisers, Washington, D.C. http://www.wh.gov/files/documents/100331-cea-economics-workplace- flexibility.pdf Cloïn M, Schols M, van den Broek A (2010) Tijd op orde? Een analyse van de tijdsorde vanuit het perspectief van de burger. SCP-publicatie, Sociaal en Cultureel Planbureau, Den Haag. http://www.scp.nl/Publicaties/Alle_publicaties/ Publicaties_2010/Tijd_op_orde Cogan JF (1981) Fixed costs and labor supply. Econometrica 49(4):945–963. doi:10.2307/1912512. http://www.jstor.org/ stable/1912512 Eaton SC (2003) If you can use them: flexibility policies, organizational commitment, and perceived performance. Ind Relat J Econ Soc 42(2):145–167. doi:10.1111/1468-232X.00285. http://dx.doi.org/10.1111/1468-232X.00285 Eldridge LP, Wulff Pabilonia S (2008) Are those who bring work home really working longer hours? Implications for BLS productivity measures. In: OECD (ed). Productivity measurement and analysis. OECD Publishing, Paris. pp 179–209. http://dx.doi.org/10.1787/9789264044616-10-en Fagan C (2004) Gender and working time in industrialized countries. In: Messenger JC (ed). Working time and workers’ preferences in industrialized countries. Finding the balance, Institute for Labour Studies of the International Labour Organisation series, Routledge. pp 108–146 Felstead A, Jewson N, Phizacklea A, Walters S (2002) The option to work at home: another privilege for the favoured few? New Technol Work Employment 17(3):204–223. doi:10.1111/1468-005X.00105. http://onlinelibrary.wiley.com/doi/10. 1111/1468-005X.00105/abstract Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 33 of 34 Gajendran RS, Harrison DA (2007) The good, the bad, and the unknown about telecommuting: meta-analysis of psychological mediators and individual consequences. J Appl Psychol 92(6):1524–1541. doi:10.1037/0021-9010.92.6.1524. http://www.apa.org/pubs/journals/releases/apl-9261524.pdf Gariety BS, Shaffer S (2001) Wage differentials associated with flextime. Mon Labor Rev 124(3):68–75. http://www.bls.gov/ opub/mlr/2001/03/art4full.pdf Gariety BS, Shaffer S (2007) Wage differentials associated with working at home. Mon Labor Rev 130(3):61–67. http:// www.bls.gov/opub/mlr/2007/03/art5full.pdf Gibbons S, Machin S (2006) Transport and labour market linkages: empirical evidence, implications for policy and scope for further research. Background Paper for the Eddington Report to the Department of Transport, HM Treasury, London. http://collections.europarchive.org/tna/20070402093752/http://www.hm-treasury.gov.uk/media/38C/A0/ eddington_researchannex1.2_011206.pdf Golden L (1996) The economics of worktime length, adjustment, and flexibility. Rev Soc Econ 54(1):1–45. doi:10.1080/00346769600000001. http://dx.doi.org/10.1080/00346769600000001 Golden L (2006) Hours of labor supply. A more flexible approach. In: Altman M (ed). Handbook of contemporary behavioral economics. Foundations and developments. M.E. Sharpe, Armonk. pp 479–496 Golden L (2008) Limited access: Disparities in flexible work schedules and work-at-home. J Fam Econ Iss 29(1):86–109. doi:10.1007/s10834-007-9090-7. http://dx.doi.org/10.1007/s10834-007-9090-7 Golden L (2009) Flexible daily work schedules in U.S. jobs: formal introductions needed? Ind Relat J Econ Soc 48(1):27–54. doi:10.1111/j.1468-232X.2008.00544.x. http://dx.doi.org/10.1111/j.1468-232X.2008.00544.x Gutiérrez-i-Puigarnau E, van Ommeren JN (2010) Labour supply and commuting. J Urban Econ 68(1):82–89. doi:10.1016/j.jue.2010.03.003. http://dx.doi.org/10.1016/j.jue.2010.03.003 Hassink WHJ, van den Berg B (2011) Time-bound opportunity costs of informal care: consequences for access to professional care, caregiver support, and labour supply estimates. Soc Sci Med 73(10):1508–1516. doi:10.1016/j.socscimed.2011.08.027. http://www.sciencedirect.com/science/article/pii/S027795361100534X Heywood JS, Siebert WS, Wei X (2007) The implicit wage costs of family friendly work practices. Oxf Econ Pap 59(2):275–300. doi:10.1093/oep/gpm006. http://dx.doi.org/10.1093/oep/gpm006 Hill EJ, Miller BC, Weiner SP, Colihan J (1998) Influences of the virtual office on aspects of work and work/life balance. Pers Psychol 51(3):667–683. doi:10.1111/j.1744-6570.1998.tb00256.x. http://dx.doi.org/10.1111/j.1744-6570.1998.tb00256.x Hill EJ, Grzywacz JG, Allen S, Blanchard VL, Matz-Costa C, Shulkin S, Pitt-Catsouphes M (2008) Defining and conceptualizing workplace flexibility. Commun Work Fam 11(2):149–163. doi:10.1080/13668800802024678. http://dx. doi.org/10.1080/13668800802024678 Johnson NB, Provan KG (1995) The relationship between work/family benefits and earnings: a test of competing predictions. J Socio-Econ 24(4):571–584. doi:10.1016/1053-5357(95)90003-9. http://dx.doi.org/10.1016/1053- 5357(95)90003-9 Kelliher C, Anderson D (2010) Doing more with less? Flexible working practices and the intensification of work. Human Relations 63(1):83–106. doi:10.1177/0018726709349199. http://dx.doi.org/10.1177/0018726709349199 Lowen A, Sicilian P (2009) “Family-friendly” fringe benefits and the gender wage gap. J Labor Res 30(2):101–119. doi:10.1007/s12122-008-9046-1. http://dx.doi.org/10.1007/s12122-008-9046-1 Manning A (2003) The real thin theory: monopsony in modern labour markets. Labour Econ 10(2):105–131. doi:10.1016/S0927-5371(03)00018-6. http://www.sciencedirect.com/science/article/pii/S0927537103000186 Noonan MC, Glass JL (2012) The hard truth about telecommuting. Mon Labor Rev 135(6):38–45. http://www.bls.gov/ opub/mlr/2012/06/art3full.pdf Oi W (1976) Residential location and labor supply. J Polit Econ 84(4):S221–S238. doi:10.2307/1831110. http://www.jstor. org/stable/1831110 Peters P, van der Lippe T (2007) The time-pressure reducing potential of telehomeworking: the Dutch case. Int J Human Resource Manag 18(3):430–447. doi:10.1080/09585190601167730. http://dx.doi.org/10.1080/09585190601167730 Peters P, den Dulk L, van der Lippe T (2009) The effects of time-spatial flexibility and new working conditions on employees’ work-life balance: the Dutch case. CWF 12(3):279–297. doi:10.1080/13668800902968907. http://dx.doi. org/10.1080/13668800902968907 Plantenga J (2003) Changing work and life patterns: examples of new working-time arrangements in the European member states(Giele JZ, Holst E, eds.). Elsevier, Oxford. http://dx.doi.org/10.1016/S1040-2608(03)08006-7 Possenriede D, Plantenga J (2014) Temporal and locational flexibility of work, working-time fit, and job satisfaction. IZA Discussion Paper 8436, Institute for the Study of Labor (IZA), Bonn. http://ftp.iza.org/dp8436.pdf Possenriede D, Hassink WHJ, Plantenga J (2014) Does temporal and locational flexibility of work reduce absenteeism. TKI Discussion Paper 14-09, Utrecht University School of Economics, Utrecht. http://hdl.handle.net/1874/298963, universiteit Utrecht Rau B (2003) Flexible work arrangements(Kossek EE, Pitt-Catsouphes M, eds.). Sloan Work and Family Research Network, Chestnut Hill, MA. http://repo.library.upenn.edu/storage/content/2/h94f69es38oo6k6s/1/ Flexible_Work_Arrangements.pdf Rürup B, Gruescu S (2005) Familienorientierte Arbeitszeitmuster - Neue Wege zu Wachstum und Beschäftigung. Bundesministerium für Familien, Senioren, Frauen und Jugend, Berlin. http://www.bmfsfj.de/RedaktionBMFSFJ/ Abteilung2/Pdf-Anlagen/familienorientierte-arbeitszeitmuster.pdf Smulders P, Kraan K, Pot F (2011) Telewerkers vormen een elite. Economisch Statistische Berichten 96(4616):493–495. http://www.economie.nl/artikel/telewerkers-vormen-een-elite Sociaal-Economische Raad (2011) Tijden van de samenleving: Slimmer organiseren van tijd en plaats van arbeid en dienstverlening. Advice 2011/06, Sociaal-Economische Raad, Den Haag. http://www.ser.nl/~/media/DB_Adviezen/ 2010_2019/2011/b29686.ashx Sociaal en Cultureel Planbureau (SCP) (2015) Arbeidsaanbodpanel 1985 t/m 2012. DANS doi:10.17026/dans-x8q-46p7. http://dx.doi.org/10.17026/dans-x8q-46p7 Possenriede et al. IZA Journal of Labor Policy (2016) 5:16 Page 34 of 34 Taskforce DeeltijdPlus (2010) De discussie voorbij. Eindrapport Taskforce DeeltijdPlus. Ministerie van Sociale Zaken en Werkgelegenheid, Den Haag. http://www.rijksoverheid.nl/bestanden/documenten-en-publicaties/kamerstukken/ 2010/04/06/de-discussie-voorbij-eindrapport-taskforce-deeltijdplus/129-2010-3-14091.pdf Wielers R, Raven D (2013) Part-time work and work norms in the Netherlands. Eur Sociol Rev 29(1):105–113. doi:10.1093/esr/jcr043. http://dx.doi.org/10.1093/esr/jcr043 Winder KL (2009) Flexible scheduling and the gender wage gap. The BE J Econ Anal Policy 9(1). doi:10.2202/1935-1682.2197. http://dx.doi.org/10.2202/1935-1682.2197 Submit your manuscript to a journal and benefi t from: 7 Convenient online submission 7 Rigorous peer review 7 Immediate publication on acceptance 7 Open access: articles freely available online 7 High visibility within the fi eld 7 Retaining the copyright to your article Submit your next manuscript at 7 springeropen.com