Hostname: page-component-cd9895bd7-gxg78 Total loading time: 0 Render date: 2024-12-23T07:39:59.512Z Has data issue: false hasContentIssue false

Flexible working and well-being: evidence from the UK

Published online by Cambridge University Press:  03 December 2024

Margherita Agnoletto*
Affiliation:
Universitá di Torino, Lungo Dora Siena 100, 10153 Torino, Italy Collegio Carlo Alberto, Piazza Vincenzo Arbarello 8, 10122 Torino, Italy

Abstract

Recent technological advancements have facilitated alternative work arrangements. This paper investigates how flextime and working from home (WfH) relate to workers’ well-being using longitudinal data drawn from the Understanding Society study for the UK. It accounts for individual, job, and family characteristics while controlling for individual fixed effects. Additionally, it employs the Oster test to examine the potential influence of unobserved variables. Results show that men experience improved job satisfaction and mental health with flextime arrangements, while women predominantly benefit in terms of job satisfaction. Additionally, women adopting remote work report heightened satisfaction with job and life, and better mental health, whereas men primarily report greater job satisfaction. Interestingly, flextime effects are stronger for men, while WfH is more beneficial for women. Some heterogeneous effects are also found by parental status, age, and income groups.

Type
Research Paper
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re- use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2024. Published by Cambridge University Press in association with Université catholique de Louvain

1. Introduction

Technological advancements have revolutionized work practices, enabling people to work remotely. This shift has facilitated the adoption of flexible job arrangements, allowing employees greater autonomy in determining their work schedules and locations.

The COVID-19 pandemic and associated social distancing measures have further accelerated these changes in work structures, making remote work and flexible scheduling predominant features of modern work environments. During the COVID-19 pandemic, the proportion of the UK workforce predominantly working from home (WfH) has increased from 5.1% in 2019 to 8.5% in 2020 (ONS, 2020). European countries experienced even higher rates, rising from 5.4% in 2019 to 12% in 2020, reaching 13.4% in 2021 (Eurostat, 2011). Since the most profound shift is in how employers and employees feel about and value work, this shift appears to persist long after the conclusion of the pandemic. According to Barrero et al. (Reference Barrero, Bloom and Davis2021), American employers plan to supply 20% of their employees’ workdays from home, four times the pre-COVID level. Moreover, employees are willing to accept a substantial reduction in wages to work from home 2 or 3 days per week. Similar patterns are observed in the UK, as shown by Taneja and Bloom (Reference Taneja and Bloom2021), and in Austria, where Bamieh and Ziegler (Reference Bamieh and Ziegler2022) report a marked increase in employers offering telecommuting options to new hires compared to pre-COVID standards.

At the same time, there has been a parallel shift toward flexible scheduling. In 2019, 39% of employees had the autonomy to determine their workday start and end times (Eurostat, 2019). Due to the COVID-19 pandemic, many employers have become aware of their employees’ childcare needs, leading to the increasing implementation of telecommuting and flexible schedule options.

Despite the growing prevalence of flexible work practices, little is known about their impact on workers’ well-being. This paper aims to address this gap by examining the effects of flextime and WfH on job satisfaction, life satisfaction, leisure satisfaction, and mental health, using data for the UK in the pre-pandemic period from 2010 to 2019. I compare employees working from home and those with flexible office schedules to those without any flexibility.

Results underscore the differential impact of flextime and WfH, with gender playing a significant role. For women, flextime only positively affects job satisfaction, while men experience benefits in terms of both job satisfaction and mental health. On the other hand, WfH leads to increased job satisfaction and better mental health for both women and men, with women also reporting greater life satisfaction.

This paper makes several contributions to the existing literature. First, unlike most previous studies (e.g., Angelici and Profeta, Reference Angelici and Profeta2024), this work takes advantage of a large-scale, nationally representative household sample that is followed over a long period. Thanks to these rich data, I estimate the impact of flexible working on a comprehensible measure of well-being, complementing earlier studies (such as Wheatley, Reference Wheatley2017) which focus on a more limited aspect of well-being.

Second, confounding factors behind the link between flexible working and well-being are addressed in two ways. On the one hand, I exploit the panel nature of the data to account for unobserved time-invariant individual characteristics such as ability and working attitudes. Moreover, using the fixed-effects ordered logit model, I preserve the outcomes’ categorical scale, avoiding the common reduction to a binary scale that results in a significant loss of observations (see section 4). On the other hand, the Oster test further confirms the robustness of findings to a potential omitted variable bias, partially addressing concerns existing in many previous cross-sectional studies (Hill et al., Reference Hill, Grzywacz, Allen, Blanchard, Matz-Costa, Shulkin and Pitt-Catsouphes2008; Weeden, Reference Weeden2005) that tend to overlook the selective nature of flexible workers within firms.

As a third contribution, the paper demonstrates the robustness of findings even when excluding employees who drop out of flexible working. Indeed, the existing literature often fails to consider that the decision to terminate such arrangements is likely to be driven by firm and individual factors. Finally, this paper shows how the effect of flexible working and well-being differs by gender and parental status, building on previous research attributing the remaining gender wage gap to schedule constraints (Goldin, Reference Goldin2014) and emphasizing flexible working as a means to balance family and career (Angelici and Profeta, Reference Angelici and Profeta2024). Findings are also differentiated by age group and income, as the impact of flexible working may vary with technological knowledge and economic opportunities.

In conclusion, the results of this paper may provide helpful suggestions to policymakers in deciding whether to continue incentivizing flexible work arrangements and in targeting them efficiently.

The rest of the paper is organized as follows. Section 2 briefly reviews the existing literature on flexible working and workers’ well-being. Section 3 describes the data. Section 4 outlines the empirical strategy used. Section 5 presents the results and robustness checks, and section 6 discusses the results and concludes.

2. Flexible working outcomes – a review of existing evidence

The role of employment in shaping individuals’ well-being has been extensively acknowledged (e.g., Bassanini and Caroli, Reference Bassanini and Caroli2015). This study assesses the effects of two distinct flexible work arrangements, flextime and WfH, on four indicators of well-being: job satisfaction, life satisfaction, leisure satisfaction, and mental health. Flextime involves employees’ autonomy in selecting their workday's start and end times, while WfH, often known as telecommuting or remote work, empowers workers to determine their work location. For brevity, I will use the expression flexible working when referring to both of these flexible work arrangements.

The concept of flexible working has acquired attention for its potential to alleviate scheduling constraints, consequently enhancing the equilibrium between work and personal life. Moreover, the adoption of remote work reduces commuting time and expenses, potentially leading to enhancements in worker well-being. On the other hand, the introduction of flexible working, particularly the practice of WfH, may introduce new dynamics such as domestic conflicts and increased stress, which could impact overall well-being (Dockery and Bawa, Reference Dockery and Bawa2014). Additionally, flexible workers may face difficulties signaling their effort and commitment to their employer, potentially limiting their career progression (Putnam et al., Reference Putnam, Myers and Gailliard2014).

Some empirical studies shed light on the relationship between flexible working and well-being. In particular, a randomized experiment conducted on an Italian sample (Angelici and Profeta, Reference Angelici and Profeta2024) shows that a fully flexible working arrangement, which gives employees flexibility in their choice of workplace and schedule, improves workers’ well-being and work–life balance. Another randomized experiment involving some Chinese call center employees (Bloom et al., Reference Bloom, Liang, Roberts and Ying2015) finds that WfH increases job and life satisfaction. A more recent randomized control trial of a US-based firm (Bloom et al., Reference Bloom, Han and Liang2022) further supports these findings by revealing increased job satisfaction. However, since these studies primarily concentrate on relatively small samples of workers within the same workplace, the external validity of their findings is limited.

Larger longitudinal surveys have also been employed to investigate the implications of flexible working, particularly focusing on job satisfaction. Overall, they find positive effects of WfH on job satisfaction, although some differences exist based on gender and parental status. Conversely, the empirical evidence concerning the effects of flexible scheduling on job satisfaction remains less conclusive. Arntz et al. (Reference Arntz, Ben Yahmed and Berlingieri2022) find positive impacts of WfH on job satisfaction, albeit statistically significant only for employees without children. This suggests that increased flexibility offers benefits but also introduces challenges, such as greater conflicts between work and personal life. Furthermore, using Italian survey data Esposito et al. (Reference Esposito, Mendolia, Scicchitano and Tealdi2024) have recently found that WfH increases job satisfaction only for women with specific personality traits.

Drawing upon data from the British Household Panel Survey and Understanding Society, Wheatley (Reference Wheatley2017) investigates the effects of flexible working arrangements. Through analysis of covariance testing and a change-score analysis, the study suggests a positive effect of homeworking on job satisfaction, especially for women. Conversely, the influence of homeworking on leisure satisfaction seems to benefit men exclusively. Furthermore, flexible scheduling emerges as a dual-edged sword: positively enhancing job satisfaction among men while yielding adverse effects for women. Differently, my empirical model addresses time-invariant unobserved heterogeneity among workers, examines potential omitted variable bias performing the Oster test, and extends the analysis to include mental health. Furthermore, I describe how the relationship between flexible working and well-being differ across demographic groups, such as parental status, age group, and income, offering a more comprehensive understanding. My findings confirm that remote work positively impacts job satisfaction, particularly for women, and that flexible schedules increase men's job satisfaction. However, I do not find a significant increase in leisure satisfaction for men WfH. Additionally, my results show that women with flexible schedules also report higher job satisfaction.

Flexible working also influences time allocation decisions, with teleworkers dedicating more time to non-market work and leisure during workdays compared to their commuting counterparts. This allocation strategy leads to greater job satisfaction among men but not women (Giménez-Nadal et al., Reference Giménez-Nadal, Molina and Velilla2020). Possenriede and Plantenga (Reference Possenriede and Plantenga2014) find that WfH and flexible schedules positively correlate with job satisfaction. The latter also increases work–life balance (see also Bender et al., Reference Bender, Donohue and Heywood2005), unlike the former.

The influence of flexible working extends beyond job satisfaction to mental health. A study of working adults in the USA shows a correlation between higher levels of time-flexible work policies and fewer reported stress symptoms (Halpern, Reference Halpern2005). However, previous research also recognizes potential adverse consequences of flexible working. For instance, WfH may lead to feelings of isolation and lack of support (Tavares, Reference Tavares2017), and unclear boundaries between work and non-work spheres, which may impact work–life balance ambiguously (Bellmann and Hübler, Reference Bellmann and Hübler2021). Additionally, home-based work may increase work–family conflict and job stress (Eddleston and Mulki, Reference Eddleston and Mulki2017). In particular, studies conducted during the COVID-19 pandemic suggest a negative impact of full-time WfH on workers’ mental health (Gueguen and Senik, Reference Gueguen and Senik2023; Bertoni et al., Reference Bertoni, Cavapozzi, Pasini and Pavese2021), although it is crucial to consider the exceptional circumstances of the pandemic-induced remote work. Developing a novel distinction, Yang et al. (Reference Yang, Kelly, Kubzansky and Berkman2023) find that WfH during regular hours is associated with better well-being and higher job satisfaction but heightened work–family conflicts. In contrast, WfH outside regular hours diminishes well-being and exacerbates work–family conflicts.

3. Data

This study relies on high-quality data from the Understanding Society study, also known as the United Kingdom Household Longitudinal Study, conducted by the Institute for Social and Economic Research, at the University of Essex. It is a nationally representative household panel studyFootnote 1 started in 2009 with 40,000 households, for a total of about 100,000 individual interviews.

The survey, conducted annually, provides information on various aspects of individuals’ lives. Since the COVID-19 pandemic has affected many aspects of life and work differently, only observations collected before 2020 are considered. For this study, waves 2, 4, 6, 8, and 10, covering 2010–2019, are selected. These waves specifically include data regarding time and location flexibility in work arrangements, making the Understanding Society dataset unique as one of the few reporting this information.

Sample selection is conducted meticulously to ensure the robustness of the analysis. The study includes individuals who are not self-employed, not in education or training, and fall within the age range of 20–66, as this most represents active participants in the labor market. Additionally, employees WfH without the option to arrange their working time are excluded since it is a relatively infrequent practice.Footnote 2 These restrictions yield an unbalanced panel of about 30,000 observations from over 6,000 individuals.

The analysis examines flextime and WfH's effects on individual well-being. Individuals are categorized into three distinct groups: those lacking any flexibility, those working in an office setting with the ability to manage their working hours, and those engaged in regular home-based work with the flexibility to manage their working hours.

Dependent variables are job satisfaction, life satisfaction, leisure satisfaction, and mental health, each treated as distinct ordinal dependent variables.Footnote 3 Job, life, and leisure satisfaction are evaluated through employee-rated scores ranging from 1 (completely dissatisfied) to 7 (completely satisfied). Mental health is assessed using the General Health Questionnaire (GHQ), which contains 12 questions covering feelings of strain, depression, inability to cope, anxiety-based insomnia, and lack of confidence (see Goldberg and Williams, Reference Goldberg and Williams1988). Responses are graded on a four-point scale indicating the frequency or intensity over the last few weeks. The GHQ Likert score, running from 0 (the least distressed) to 36 (the most distressed), is derived from the summed and recalibrated scores. For ease of interpretation, the GHQ scale is reversed, meaning that a higher mental health index corresponds to lower levels of distress in the individual.

This analysis exploits the extensive information available in the Understanding Society survey, and the choice of regressors follows previous literature. All specifications include controls for individual characteristics such as age, education, health status, and marital status, job-related characteristics such as working hours, labor income, and occupation class, and family attributes such as having children by age group, region of residence, and living in a city. Table A1 lists all the explanatory variables.

As the determinants of subjective well-being may differ by gender (Fugl-Meyer et al., Reference Fugl-Meyer, Melin and Fugl-Meyer2002) and, women and men often value job and workplace characteristics differently (Sloane and Williams, Reference Sloane and Williams2000; Bender et al., Reference Bender, Donohue and Heywood2005), the analysis is performed separately for females and males.

Table 1 presents summary statistics by gender and flexible working status. The female sub-sample comprises approximately 56% who enjoy a flexible schedule, while only 5% of them regularly work from home. In contrast, the male sub-sample consists of about 48% employees with flexible schedules and 8% employees working from home.

Table 1. Summary statistics by flexible working status and gender

Notes: Mean and standard deviation (in parentheses).

Employees who work from home or have flexible schedules are typically characterized by higher educational attainment, increased earnings, and longer commuting times compared to on-site employees without flexible schedules. Moreover, flexible working is more common in large firms and highly skilled occupations. Regarding family characteristics, flexible workers are more likely to be married and have children.

The Pearson correlation matrix (Table A2) provides valuable information on the relationships among outcomes. Overall, correlations between outcomes are modest, with most falling below 50%. However, life and leisure satisfaction display stronger correlations, reaching 55% for females and 53% for males. Additionally, Table A2 examines the correlation between control variables and outcomes, which helps understand how each control variable is related to the outcomes of interest. Finally, Table A2 also reports that none of the included control variables have high correlations with other control variables, reducing concerns about multicollinearity affecting my results.

4. Empirical strategy

To study the effect of flexible working on workers’ well-being, I estimate an ordered logit model, in which I introduce individual fixed effects to control for unobservable time-invariant differences between individuals.

To introduce fixed effects, other studies often reduce a categorical satisfaction scale to a binary scale following Chamberlain (Reference Chamberlain1980)'s method. However, this estimation strategy comes at the expense of losing many observations since only individuals moving across the cut-off point can be used in the analysis. This large loss may also increase measurement errors. A different common approach is to dichotomize every individual separately by choosing individual-specific cut-off points (Ferrer-i Carbonell and Frijters, Reference Ferrer-i Carbonell and Frijters2004). However, Baetschmann et al. (Reference Baetschmann, Staub and Winkelmann2015) show that individual-specific dichotomized estimators are biased in finite samples. Another common practice in the literature is to use the ordinary least squares (OLS) estimator, although it assumes the cardinality of dependent variables. A major advantage of this assumption is the ability to easily apply a fixed-effects estimator. Nevertheless, cardinality assumptions are quite strong. For example, for a job satisfaction question measured on a scale of 1–7, the cardinality assumption implies that the relative difference between responses at 1 and 2 is the same as the relative difference between responses at 4 and 5. In contrast, ordinality assumptions supporting the ordered logit estimator accept that the relative difference between the answers at different points of the scale is unknown.

For these reasons, I follow the empirical strategy of Baetschmann et al. (Reference Baetschmann, Staub and Winkelmann2015), who implement the so-called blow-up and cluster (BUC) estimator. This estimator is a consistent (though not the most efficient) estimator, in contrast to the estimator proposed by Ferrer-i Carbonell and Frijters (Reference Ferrer-i Carbonell and Frijters2004). Further, the BUC estimator performs well in small samples and can be a useful alternative for applied work with many regressors (Baetschmann et al., Reference Baetschmann, Staub and Winkelmann2015). The approach has been used in, e.g., Geishecker et al. (Reference Geishecker, Riedl and Frijters2012) and Possenriede and Plantenga (Reference Possenriede and Plantenga2014). Every individual's observations are replaced by K − 1 copies of itself, and then, each copy is dichotomized at a different cutoff point. Finally, the entire inflated sample is used to estimate the coefficient by applying the conditional maximum likelihood estimator. Because an individual's copies of observation are not independent of each other, standard errors are clustered at the individual level.

The empirical model relates the score of the dependent variables (job satisfaction, life satisfaction, leisure satisfaction, and mental health) to flextime and working from home through an ordered logit specification:

(1)$$S_{it}^{\ast } = \beta_{1} {Flextime}_{it} + \beta_{2} {Wf\,H}_{it} + {X^{'}_{it}} \beta_{3} + \alpha_{i} + \omega_{t} + \epsilon_{it},\; \quad i = 1,\; \ldots ,\; N$$

where $S_{it}^\ast$ is the latent level of well-being for individual i at time t, Flextime it is equal to 1 if the individual is an on-site worker with flexible schedules and 0 otherwise, WfH it is equal to 1 if the individual works from home with flexible schedules and 0 otherwise, the vector X it includes demographic, job, and family characteristics, α i represents individual fixed effects, ω t refers to year fixed effects, and $\epsilon _{it}$ is an error term.

The non-linear function τ ties the latent variable $S_{it}^\ast$ to the observed ordered variable S it through the thresholds τ ik:

(2)$$S_{it} = k \text{ if } \tau_{ik}< S_{it}^\ast \leq\tau_{ik + 1} \quad k = 1,\; \ldots,\; K$$

To interpret the magnitude of the results, I estimate odds, which refers to the ratio between the probability of a certain event and the complementary probability (Baetschmann et al., Reference Baetschmann, Ballantyne, Staub and Winkelmann2020). In the ordered logit, the odds of individual i in period t having S it above category k relative to below or equal to k is

(3)$${Odds \ ratio}( k,\; \Delta {\bf x}_{it}) \equiv {Pr( S_{it}> k \vert {\bf x}_{itl}) \over Pr( S_{it}\leq k \vert {\bf x}_{it}) } = \exp( {\bf x}'_{it}\beta - \tau_{it}) $$

The change in the odds, if the variable of interest l is modified (i.e., if the individual benefits from flexibility in time or place), depends on β and the variable's shift as follows:

(4)$$Odds \ ratio ( k,\; \Delta {\bf x}_{itl}) \equiv {Odds( k,\; {\bf x}_{it} + \Delta {\bf x}_{itl}) \over Odds( k,\; {\bf x}_{it}) } = \exp( \Delta {\bf x}_{itl}'\beta) $$

Therefore, having flexibility at work changes the odds by exactly {exp (β l) − 1} × 100%.

In the analysis, the estimated coefficient on flexible working reflects the experience of individuals who change the working category over the years of the panel survey. Table A3 shows the transition matrix estimating the probability of changing flexible work status, relating the status at wave t to that at wave t + 1. The most common outcome is immobility from one wave to the next, as shown in the diagonal cells. However, there is a degree of mobility evident in the remaining cells and this variability is exploited in the fixed-effects specification.

The individual fixed effects control for both observable and unobservable time-invariant heterogeneity among individuals which may influence well-being outcomes and the likelihood of being a flexible worker, such as personality type, family values, gender, ethnicity, ability, and individual background. The year fixed effects account for time-varying but individual-constant unobserved characteristics such as social and cultural trends. However, endogeneity concerns arise from changes in time-varying but not individual-constant factors that affect, at the same time, the availability and use of flexible work arrangements and individual well-being. To attenuate this source of bias, several time-varying factors are controlled for, including education, marital status, parental status, health condition, area of residence, and job changes. Indeed, education not only impacts income but also shapes expectations regarding job and overall life satisfaction; hence, the net effect of education on happiness is shown by the difference between the increase in expectations and the one in income (Clark, Reference Clark, Adler and Fleurbaey2016). Marital status exhibits mixed effects on well-being, with research indicating both negative and positive associations with job satisfaction (Gazioglu and Tansel, Reference Gazioglu and Tansel2003; Clark et al., Reference Clark, Oswald and Warr1996). On the other hand, life satisfaction and mental health have been shown to be positively correlated with being married (Deklyen et al., Reference Deklyen, Brooks-Gunn, Mclanahan and Knab2006; Brown, Reference Brown2000). Moreover, parental responsibilities may influence the propensity for seeking flexible work arrangements and simultaneously impact well-being (Booth and Van Ours, Reference Booth and Van Ours2009). Similarly, changes in health conditions may influence well-being and the likelihood of engaging in flexible working. For instance, flexible jobs may accommodate individuals facing temporary health challenges, easing their work-related burdens. In addition, moving to the countryside may be associated with an increase in the likelihood of flexible working because of a greater commuting time given that the labor market is mainly centered in the urban area. At the same time, moving from rural to urban areas, or vice versa, may influence well-being differently across individuals depending on which life aspects they value more. Furthermore, past literature shows that public employees differ from private ones suggesting distinct satisfaction premiums associated with non-pecuniary job aspects (Ghinetti, Reference Ghinetti2007). Not surprisingly, changing the job may affect job satisfaction and the probability of flexible working. Chadi and Hetschko (Reference Chadi and Hetschko2021) find initial increases in life satisfaction followed by diminishing returns over time. Table A4 shows how the transition in independent variables correlates with changes in relevant observable characteristics. This analysis helps to identify the main factors driving the transition process, emphasizing the importance of accounting for these observable characteristics in subsequent analyses. For instance, transitioning to flexible work arrangements shows a positive correlation with longer working hours and employment in the private sector, and a negative correlation with union presence or tenure with the same employer, irrespective of gender.

Finally, I perform the Oster test, which extends the method proposed by Altonji et al. (Reference Altonji, Elder and Taber2005), to evaluate the robustness of the results to a potential omitted variable bias. This method calculates the level of selection on unobserved variables, proportionally to the level of selection on observed variables, required to drive the treatment effect to zero. The assumption behind the estimation of the level of selection, denoted by the parameter δ, is a value for the R-max defined as the R-squared from a hypothetical regression of the outcome on the treatment and both observed and unobserved controls. Following Oster (Reference Oster2019), I set the R-max equal to 1.3 times the R-squared from a regression of the outcome on the treatment and observed control variables.Footnote 4 Since the Oster test is appropriate only for a linear model, I perform it by exploiting a linear regression model with fixed effects. Although the linear model can produce biased estimates, which is why I have previously chosen an ordered logit model with fixed effects, it is suitable for the Oster test. Indeed, the OLS and ordered logit models exhibit similar statistical significance for the included covariates (Riedl and Geishecker, Reference Riedl and Geishecker2014; Ferrer-i and Carbonell Frijters, Reference Ferrer-i Carbonell and Frijters2004).

5. Results

5.1. Flextime, WfH, and well-being

In the current study, I examine four measures of individual well-being: job, life, and leisure satisfaction, and mental health. Table A5 shows the linear fixed effects (FE) and BUC estimates of flexible working schedules and WfH on these outcomes. Results are qualitatively the same for both the linear FE and BUC estimators. Given that the magnitude of the linear FE estimates may be biased, as previously detailed in section 4, I focus the discussion of the results solely on the BUC estimates. Figure 1 displays the odds ratios of flextime at the office and WfH, along with the corresponding confidence intervals.Footnote 5

Figure 1. Effect of flextime at the office and WfH on well-being by gender. Notes: Odds ratios based on equation (4) by gender, with females indicated by red squares and males by blue circles. All equations include year fixed effects, demographic controls, job controls, and family controls.

Flextime reveals a positive association between flextime and job satisfaction, indicating that the ability to determine the start- and end-times of work increases the probability of being satisfied with job overall. More specifically, having flexible schedules increases the odds ratio by about 42%Footnote 6 for women and 44% for men. By contrast, the effect is null on life and leisure satisfaction except for a marginal effect on male life satisfaction (16% at the 10% significant level). In terms of mental health, flextime leads to a 10% increase in the odds ratio for women, although not statistically significant at conventional levels, and a 28% increase for men.

Men appear to derive greater benefits from flextime compared to women, potentially attributed to men using the flexibility in their work hours to relax or pursue hobbies, while women may prioritize balancing work and domestic responsibilities, as suggested by Dentinger and Clarkberg (Reference Dentinger and Clarkberg2002).

Conversely, the relationship between WfH and workers’ well-being is stronger for women than for men. WfH for women increases the odds ratio by 95% for job satisfaction, 50% for life satisfaction, 26% for leisure satisfaction (albeit significant at 10% level), and 34% for mental health. While still positive, the association between home-based work and male workers’ well-being is weaker. WfH leads to a 50% increase in job satisfaction and a 30% increase in mental health for men (significant at 10% level), but no significant effects on life and leisure satisfaction. These findings corroborate previous literature, emphasizing the positive impact of WfH, particularly for women (Arntz et al., Reference Arntz, Ben Yahmed and Berlingieri2022).

5.2. Oster test

Results from the Oster test are shown in Table A5, which displays the δ values associated with fixed-effect OLS models illustrating the relationship between flextime at the office, WfH, and various well-being aspects such as job satisfaction, life satisfaction, leisure satisfaction, and mental health. These models are evaluated separately for the female and male subsamples. The δ statistic measures the extent to which the remaining characteristics of individuals that remain unobservable (i.e., the unobservable determinants of employees’ well-being) should be correlated with flexible working status to reduce the β coefficient to zero.

For instance, in column 1 of Table A5, the δ for β 1 in the female sample amounts to approximately 34. This means that to conclude that there is no difference in well-being between flextime and non-flexible workers, the unobservable characteristics would have to be about 34 times more correlated with flextime status than the observed characteristics. Such a high correlation is arguably implausible. In general, all δ estimates are very high and always greater than 1 (in absolute value).Footnote 7 This indicates, according to Oster (Reference Oster2019), that the findings remain robust to a substantial degree of selection on unobservables, and, thus, the Oster test confirms the credibility of my estimates. A negative δ indicates that if the observables are positively correlated with the treatment, the unobservables must be negatively correlated with the treatment to nullify the effect (β = 0).

5.3. Additional analyses

In this section, I examine the robustness of my main findings regarding flexible schedules and WfH by including additional covariates and focusing on particular groups of individuals (Table 2).

Table 2. Robustness checks for flextime and WfH: BUC estimates

Notes: Further control variables included are as in Fig. 1. Standard errors clustered at the individual level in parentheses.

*** p < 0.01, ** p < 0.05, *p < 0.1.

First, the inclusion of region-by-year fixed effects takes into account unobserved regional trends that might reflect regional policies or labor market conditions. Since the results do not change (Table 2, panel A), these trends do not drive my estimates.

The transition to a new employer or a different job role within the firm may lead to simultaneous changes in well-being and flexible working status. Despite already controlling for changes in employers in the main specification, estimates may be driven by such career transitions. To address this concern, I replicate the analysis focusing only on employees who remain in the same position and the same employer (Table 2, panel B). The positive effect of WfH decreases for women, especially on mental health, and remains significant only on job satisfaction. When replicating the analysis for employees remaining with the same employer but potentially altering job positions (with results available upon request), the estimates remain significant and consistent with the main findings. The implication is that the simultaneous change of job position and WfH status explains a significant portion of the increased well-being for female employees, especially in terms of mental health. On the contrary, outcomes for male employees exhibit greater stability throughout these analyses.

Moreover, the decision to initially embrace flexible working is more likely to be driven by factors exogenous to the firm and the individual, whereas the decision to terminate such arrangements is more likely to be driven by firm and individual factors. These factors could refer to aspects such as the quality of the work arrangement or changes in childcare obligations, some of which may not always be observable. Hence, I exclude from the analysis workers who drop out from flexible working, and findings (Table 2, panel C) remain stable to the baseline model.

In addition, the impact of flexible working arrangements may be influenced by an individual's partner's employment status and flexible work arrangements. To address this potential influence, I include three variables in the estimation model to capture this information: partner flextime, partner WfH, and partner employment status. The main estimates are barely affected when controlling for these partner's characteristics (Table 2, panel D).

Although all analyses are weighted using longitudinal weights to correct for non-response and potential attrition throughout the waves, I conduct a robustness check with a balanced panel (Table 2, panel E). The relationship between flexible working and well-being remains largely unaffected.

While year fixed effects in the main specifications control for trends over time, seasonal trends may introduce bias into results (Banks Xu, Reference Banks and Xu2020). Therefore, I conducted an additional robustness check by including months of interview fixed effects (Table 2, panel F) to address the possibility that transitions to flexible working may be more common during certain periods, such as holidays, to facilitate childcare arrangements. Yet, the estimates for flextime and WfH remain robust.

5.4. Heterogeneous effects

Table 3 explores the heterogeneity of the main results across some groups. First, I examine the heterogeneity between workers with and without young children in the household (see Fig. A1 for women and Fig. A2 for men). Female flexible workers with young children report lower well-being than their counterparts with older children or childless. However, this difference is statistically significant only regarding life satisfaction. This heterogeneous analysis may confirm that childcare responsibilities translate into a professional disadvantage for workers with children. Although the adoption of flexible working arrangements may help to reconcile work and childcare responsibilities, providing parents with advantages similar to those enjoyed by non-parents who enjoy flexibility in managing their professional and personal lives, the advantage associated with the increased flexibility may be counteracted by potential drawbacks, such as irregular working hours (e.g., evening shifts) for parents. These irregular hours not only reduce leisure time but also raise conflicts between the job and the private sphere. In contrast, male flexible workers with young children tend to benefit more from flexible working than their counterparts, although the difference is significant only in terms of job satisfaction.

Table 3. Flextime at the office, WfH, and well-being across individuals with or without young children, by age groups and income levels

Notes: The table shows the BUC estimates based on equation (1). Control variables are equivalent to Fig. 1. Standard errors clustered at the individual level in parentheses.

*** p < 0.01, **p < 0.05, *p < 0.1.

Furthermore, I perform a heterogeneous analysis based on age, dividing individuals into three groups using tertiles (Figs A3 for females and A4 for males): the youngest cohort (20–36 years old), the middle-aged cohort (37–48 years old), and the oldest cohort (49–66 years old). The positive effect of WfH on women's mental health is mainly driven by the youngest cohort. On the other hand, among males, the oldest cohort of remote workers reports lower job satisfaction and leisure satisfaction.

Finally, the heterogeneous analysis by income groups is shown in Figs A5 and A6 for females and males, respectively. The low-income group is represented by those employees with a household net income below the sample median (i.e., £3,300). There are no heterogeneous effects by income groups in the female sample. By contrast, men with flexible schedules report high levels of job satisfaction and mental well-being, particularly in households with lower-income levels. Individuals with lower household incomes may place a higher value on flexible schedules for several factors. For example, individuals with limited financial resources may find greater utility in flexible schedules as they facilitate the management of responsibilities outside of work, such as childcare and household. In contrast, individuals from higher-income households may have the option to delegate such tasks to others, thereby reducing their reliance on flexible scheduling for balancing work and personal obligations.

6. Discussion and conclusion

In light of the growing importance of well-being in the context of employment and the evolving landscape of flexible work arrangements, this paper aims to contribute to the existing literature by investigating the relationship between flexible working and worker well-being, with a specific focus on job satisfaction, life satisfaction, leisure satisfaction, and mental health indicators.

Flexible work arrangements are important in current policy debates about working conditions experienced during the COVID-19 pandemic and the combination of work and private life. Despite their importance, there is little research on how these arrangements shape workers’ well-being and importantly, how these effects vary with workers’ characteristics.

This paper comprehensively explains how two flexible work arrangements, flextime and WfH, can impact worker well-being and contribute to informed decision-making for individuals and firms alike. Results suggest that male workers would benefit from flexible schedules in terms of job satisfaction and mental health, while female workers only in terms of job satisfaction. On the other hand, females WfH report a great increase in all the well-being measures, especially in terms of job satisfaction, life satisfaction, and mental health. However, for male workers, WfH leads to greater job satisfaction.

Flexible working may enhance job satisfaction through many mechanisms. First, it improves the fit between paid work and other activities (Possenriede and Plantenga, Reference Possenriede and Plantenga2014). Second, it allows employees to work during times more suited to their personal needs and more productive for themselves, potentially decreasing work- and commuting-related stress (Moen et al., Reference Moen, Kelly, Fan, Lee, Almeida, Kossek and Buxton2016). Third, flexible working may signal employees that their employers trust them and care about their well-being (Casper and Harris, Reference Casper and Harris2008).

Furthermore, results shed light on the gendered aspects of flexible working. With women often bearing a disproportionate burden of household and childcare responsibilities (Del Boca et al., Reference Del Boca, Oggero, Profeta and Rossi2022), remote work may offer significant benefits to women by facilitating better work–life balance and enabling increased time spent with children, which can positively impact their life satisfaction (Nomaguchi et al., Reference Nomaguchi, Milkie and Bianchi2005).

Additionally, flexible working arrangements afford workers greater control over their time commitments, which may lower the risk of developing depression and anxiety (Griffin et al., Reference Griffin, Fuhrer, Stansfeld and Marmot2002). Moreover, reduced commuting time associated with home-based work may improve workers’ psychological health. Consistently with my findings that show greater benefit from remote work for women than for men, a previous study shows that women's psychological health is adversely affected by commuting time while men's is not (Roberts et al., Reference Roberts, Hodgson and Dolan2011).

Considering the established correlation between heightened workers’ satisfaction and favorable workplace outcomes – including reduced turnover rates, decreased absenteeism, and heightened worker productivity (Bubonya et al., Reference Bubonya, Cobb-Clark and Wooden2017; Oswald et al., Reference Oswald, Proto and Sgroi2015; Clark et al., Reference Clark, Georgellis and Sanfey2012; Clark, Reference Clark2001; Akerlof, Reference Akerlof1986; Freeman, Reference Freeman1978) – the elevated levels of satisfaction resulting from flexible working carry advantageous implications for both employers and employees alike. Moreover, promoting flexible work arrangements could yield substantial cost savings in healthcare through the positive impact on mental health. It is particularly relevant in the UK, where mental healthcare expenditure represents a significant portion of GDP (10.2% in 2019; World Health Organization, 2019).

The main limitation of this analysis is the difficulty in providing strong causal evidence on the relationship between flexible working and workers’ well-being in the absence of an exogenous variation in flexible working. Although the fixed-effects model accounts for the common concern of unobservable time-invariant heterogeneity, it may not fully control for some events affecting workers’ well-being and their likelihood of flexible working, driving the results. Therefore, I control for various covariates capturing random events, such as changing jobs, marital status, and parenting status. Further, the results are robust to additional analyses that take into account unobserved regional trends, career changes, employees dropping out of flexible working, partner characteristics, and seasonal trends. Finally, the Oster test provides even more credibility to the paper's findings.

Future research might seek to identify a flexible working policy that would address this limitation, possibly providing an exogenous variation in flexible working conditions. For instance, a national policy that makes some employees work from home or have flexible schedules without affecting their well-being in other ways, as, unfortunately, the WfH policy adopted during the COVID-19 pandemic has done.

Acknowledgments

I thank the associate editor, Prof. Libertad González, and two anonymous referees for excellent comments and a very professional review process. I am also grateful to my supervisors, Silvia Mendolia and Francesco Devicienti, and my host supervisor, Sarah C. Dahmann, for their valuable suggestions and comments.

Competing interests

None.

Appendix A:

Table A1. Definition of variables

Table A2. Pearson correlations for outcomes and some relevant control variables

Table A3. Transition matrix of flexible working by gender

Table A4. Correlations of transition in flexible working and changes in some relevant observable characteristics

Table A5. Flextime at the office, WfH, and employees’ well-being

Table A6. Step-by-step inclusion of controls in the regression

Figure A1. Women: odds ratio of flextime at the office, WfH on well-being by parental status. Notes: The effects on parents are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure A2. Men: odds ratio of flextime at the office, working from home on well-being by parental status. Notes: The effects on parents are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure A3. Women: odds ratio of flextime at the office, WfH on well-being by age groups. Notes: Individuals are divided into three age groups according to tertiles: the youngest cohort (20–36 years old), the middle-aged cohort (37–48 years old), and the oldest cohort (49–66 years old). The effects on the middle-aged cohort and the oldest are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure A4. Men: odds ratio of flextime at the office, working from home on well-being by age groups. Notes: Individuals are divided into three age groups according to tertiles: the youngest cohort (20–36 years old), the middle-aged cohort (37–48 years old), and the oldest cohort (49–66 years old). The effects on the middle-aged cohort and the oldest are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure A5. Women: odds ratio of flextime at the office, WfH on well-being by income groups. Notes: The low-income group reports a household net income under the sample median, £3,500. The effect on the low-income group is calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure A6. Men: odds ratio of flextime at the office, working from home on well-being by income groups. Notes: The low-income group reports a household net income under the sample median, £3,500. The effect on the low-income group is calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Footnotes

1 The analyses are weighted using a longitudinal weight suggested and provided by Understanding Society that corrects for non-response and attrition over the years.

2 Employees WfH without the ability to arrange their working time are 90 female observations and 59 male observations. My results remain robust if those individuals are included in the WfH group.

3 Recently, Bond and Lang (Reference Bond and Lang2019) have raised concerns about the use of ordinal measures. They argue that, when assessing subjective well-being (SWB) on an ordinal scale, the average ranking of observed SWB among different groups can be unexpectedly inverted through specific monotonic increasing transformations applied to the SWB data. However, this issue becomes less likely as the number of response categories increases. This is one reason why I opt to retain the 7-point satisfaction rating and the 36-point mental health scale. Additionally, my results are also robust to using a linear model rather than an ordered logit model as recommended by Bond and Lang (Reference Bond and Lang2019) (see Table A5).

4 Following Hener et al. (Reference Hener, Rainer and Siedler2016), my results are found to be robust to an R-max equal to 2.2 times the R-squared from a regression of the outcome on the treatment and observed control variables. Results will be provided on request.

5 Table A6 sequentially shows the inclusion of individual fixed effects, year fixed effect, demographic controls, job controls, and family controls.

6 I.e., ${( \exp ( \beta _1) -1) } \times 100\% = {( 1.416 - 1) } \times 100\% \approx 42\%$.

7 Oster (Reference Oster2019) interprets a limit of 1 for the sensitivity parameter δ as equal selection on observed and unobserved variables.

Notes: A star refers to 5% significant level or better.

Notes: A star refers to 5% significant level or better.

Notes: Linear FE estimates and BUC estimates based on equation (1) refer to separate estimations for male and female employees. All equations include year fixed effects, demographic controls (age, highest qualification obtained, general health, physical health limiting work, marital status), job characteristics (working hours, private sector, firm size, presence of a union, commuting time, labor income, changing employer dummy, occupational class), and family characteristics (having children by four age groups based on the school system, living in a city, region of residence). Standard errors clustered at the individual level in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

Standard errors clustered at the individual level in parentheses.

*** p < 0.01, ** p < 0.05, *p < 0.1.

References

Akerlof, G. A. (1986). Efficiency wage models of the labor market. Cambridge University Press.CrossRefGoogle Scholar
Altonji, J., Elder, T., and Taber, C. (2005). Selection on observed and unobserved variables: Assessing the effectiveness of catholic schools. Journal of Political Economy, 113(1), 151184. https://doi.org/10.1086/426036CrossRefGoogle Scholar
Angelici, M., and Profeta, P. (2024). Smart working: Work flexibility without constraints. Management Science 70(3), 16801705.CrossRefGoogle Scholar
Arntz, M., Ben Yahmed, S., and Berlingieri, F. (2022). Working from home, hours worked and wages: Heterogeneity by gender and parenthood. Labour Economics, 76, 102169. https://doi.org/10.1016/j.labeco.2022.102169CrossRefGoogle Scholar
Baetschmann, G., Staub, K. E., and Winkelmann, R. (2015). Consistent estimation of the fixed effects ordered logit model. Journal of the Royal Statistical Society. Series A (Statistics in Society), 178(3), 685703. https://doi.org/10.1111/rssa.12090CrossRefGoogle Scholar
Baetschmann, G., Ballantyne, A., Staub, K. E., and Winkelmann, R. (2020). feologit: A new command for fitting fixed-effects ordered logit models. The Stata Journal, 20(2), 253275. https://doi.org/10.1177/1536867X20930984CrossRefGoogle Scholar
Bamieh, O., and Ziegler, L. (2022). Are remote work options the new standard? Evidence from vacancy postings during the COVID-19 crisis. Labour Economics, 76, 102179. https://doi.org/10.1016/j.labeco.2022.102179CrossRefGoogle ScholarPubMed
Banks, J., and Xu, X. (2020). The mental health effects of the first two months of lockdown during the COVID-19 pandemic in the UK. Fiscal Studies, 41(3), 685708. https://doi.org/10.1111/fisc.v41.3CrossRefGoogle Scholar
Barrero, J. M., Bloom, N., and Davis, S. J. (2021). Why working from home will stick. Working Paper 28731, National Bureau of Economic Research.CrossRefGoogle Scholar
Bassanini, A., and Caroli, E. (2015). Is work bad for health? The role of constraint versus choice. Annals of Economics and Statistics, 119/120, 1337. https://doi.org/10.15609/annaeconstat2009.119-120.13CrossRefGoogle Scholar
Bellmann, L., and Hübler, O. (2021). Working from home, job satisfaction and work–life balance–robust or heterogeneous links?. International Journal of Manpower 42(3), 424441.CrossRefGoogle Scholar
Bender, K. A., Donohue, S. M., and Heywood, J. S. (2005). Job satisfaction and gender segregation. Oxford Economic Papers, 57(3), 479496. https://doi.org/10.1093/oep/gpi015CrossRefGoogle Scholar
Bertoni, M., Cavapozzi, D., Pasini, G., and Pavese, C. (2021). Remote working and mental health during the first wave of COVID-19 pandemic. IZA Discussion Papers 14773, Institute of Labor Economics (IZA).Google Scholar
Bloom, N., Liang, J., Roberts, J., and Ying, Z. J. (2015). Does working from home work? Evidence from a Chinese experiment. The Quarterly Journal of Economics, 130(1), 165218. https://doi.org/10.1093/qje/qju032CrossRefGoogle Scholar
Bloom, N., Han, R., and Liang, J. (2022). How hybrid working from home works out. Working Paper 30292, National Bureau of Economic Research.CrossRefGoogle Scholar
Bond, T. N., and Lang, K. (2019). The sad truth about happiness scales. Journal of Political Economy, 127(4), 16291640. https://doi.org/10.1086/701679CrossRefGoogle Scholar
Booth, A. L., and Van Ours, J. C. (2009). Hours of work and gender identity: Does part-time work make the family happier?. Economica, 76(301), 176196. https://doi.org/10.1111/ecca.2009.76.issue-301CrossRefGoogle Scholar
Brown, S. L. (2000). The effect of union type on psychological well-being: Depression among cohabitors versus marrieds. Journal of Health and Social Behavior, 41(3), 241255. https://doi.org/10.2307/2676319CrossRefGoogle ScholarPubMed
Bubonya, M., Cobb-Clark, D. A., and Wooden, M. (2017). Mental health and productivity at work: Does what you do matter?. Labour Economics, 46, 150165. https://doi.org/10.1016/j.labeco.2017.05.001CrossRefGoogle Scholar
Casper, W. J., and Harris, C. M. (2008). Work–life benefits and organizational attachment: Self-interest utility and signaling theory models. Journal of Vocational Behavior, 72(1), 95109. https://doi.org/10.1016/j.jvb.2007.10.015CrossRefGoogle Scholar
Chadi, A., and Hetschko, C. (2021). How job changes affect people's lives – Evidence from subjective well-being data. British Journal of Industrial Relations, 59(2), 279306. https://doi.org/10.1111/bjir.v59.2CrossRefGoogle Scholar
Chamberlain, G. (1980). Analysis of covariance with qualitative data. The Review of Economic Studies, 47(1), 225238. https://doi.org/10.2307/2297110CrossRefGoogle Scholar
Clark, A. E. (2001). What really matters in a job? Hedonic measurement using quit data. Labour Economics, 8(2), 223242. https://doi.org/10.1016/S0927-5371(01)00031-8CrossRefGoogle Scholar
Clark, A. (2016) SWB as a measure of individual well-being. in Adler, M. A. and Fleurbaey, M. (eds.), The Oxford Handbook of Well-Being and Public Policy, Oxford Handbooks. Oxford University Press, https://doi.org/10.1093/oxfordhb/9780199325818.013.17Google Scholar
Clark, A., Oswald, A., and Warr, P. (1996). Is job satisfaction U-shaped in age?. Journal of Occupational and Organizational Psychology, 69(1), 5781. https://doi.org/10.1111/joop.1996.69.issue-1CrossRefGoogle Scholar
Clark, A. E., Georgellis, Y., and Sanfey, P. (2012). Job satisfaction, wage changes, and quits: Evidence from Germany. Research in Labor Economics, 35(1), 499525. https://doi.org/10.1108/rlecCrossRefGoogle Scholar
Deklyen, M., Brooks-Gunn, J., Mclanahan, S., and Knab, J. (2006). The mental health of married, cohabiting, and non-coresident parents with infants. American Journal of Public Health, 96, 1836–41. https://doi.org/10.2105/AJPH.2004.049296CrossRefGoogle ScholarPubMed
Del Boca, D., Oggero, N., Profeta, P., and Rossi, M. C. (2022). The impact of COVID-19 on the gender division of housework and childcare: Evidence from two waves of the pandemic in Italy. IZA Journal of Labor Economics, 11(1).Google Scholar
Dentinger, E., and Clarkberg, M. (2002). Informal caregiving and retirement timing among men and women: Gender and caregiving relationships in late midlife. Journal of Family Issues, 23(7), 857879. https://doi.org/10.1177/019251302236598CrossRefGoogle Scholar
Dockery, A., and Bawa, S. (2014). Is working from home good work or bad work? Evidence from Australian employees. Australian Journal of Labour Economics (AJLE), 17(2), 163190.Google Scholar
Eddleston, K. A., and Mulki, J. (2017). Toward understanding remote workers’ management of work–family boundaries: The complexity of workplace embeddedness. Group & Organization Management, 42(3), 346387. https://doi.org/10.1177/1059601115619548CrossRefGoogle Scholar
Esposito, P., Mendolia, S., Scicchitano, S., and Tealdi, C. (2024). Working from home and job satisfaction: The role of gender and personality traits. IZA Discussion Paper 16751, Institute of Labor Economics (IZA).CrossRefGoogle Scholar
Eurostat(2021). Eurostat data: Employed persons working from home as a percentage of the total employment. https://ec.europa.eu/eurostat/databrowser/view/LFSA_EHOMP_custom_4671617/default/table?lang=en.Google Scholar
Ferrer-i Carbonell, A., and Frijters, P. (2004). How important is methodology for the estimates of the determinants of happiness?. The Economic Journal, 114(497), 641659. https://doi.org/10.1111/j.1468-0297.2004.00235.xCrossRefGoogle Scholar
Freeman, R. B. (1978). Job satisfaction as an economic variable. The American Economic Review, 68(2), 135141.Google Scholar
Fugl-Meyer, A., Melin, R., and Fugl-Meyer, K. (2002). Life satisfaction in 18- to 64-year-old swedes: In relation to gender, age, partner and immigrant status. Journal of Rehabilitation Medicine, 34, 239–46. https://doi.org/10.1080/165019702760279242CrossRefGoogle ScholarPubMed
Gazioglu, S., and Tansel, A. (2003). Job satisfaction in Britain: Individual and job related factors. ERC working papers, ERC – Economic Research Center, Middle East Technical University.Google Scholar
Geishecker, I., Riedl, M., and Frijters, P. (2012). Offshoring and job loss fears: An econometric analysis of individual perceptions. Labour Economics, 19(5), 738747. https://doi.org/10.1016/j.labeco.2012.05.018CrossRefGoogle Scholar
Ghinetti, P. (2007). The public–private job satisfaction differential in Italy. Labour (Committee on Canadian Labour History), 21(2), 361388.Google Scholar
Giménez-Nadal, J., Molina, J., and Velilla, J. (2020). Work time and well-being for workers at home: Evidence from the American time use survey. International Journal of Manpower 41(2), 184206.CrossRefGoogle Scholar
Goldberg, D., and Williams, P. D. (1988). A user's guide to the General Health Questionnaire.Google Scholar
Goldin, C. (2014). A grand gender convergence: Its last chapter. American Economic Review, 104(4), 10911119. https://doi.org/10.1257/aer.104.4.1091CrossRefGoogle Scholar
Griffin, J. M., Fuhrer, R., Stansfeld, S. A., and Marmot, M. (2002). The importance of low control at work and home on depression and anxiety: Do these effects vary by gender and social class? Social Science Medicine, 54(5), 783798. https://doi.org/10.1016/S0277-9536(01)00109-5CrossRefGoogle ScholarPubMed
Gueguen, G., and Senik, C. (2023). Adopting telework: The causal impact of working from home on subjective well-being. British Journal of Industrial Relations 61(4), 832868.CrossRefGoogle Scholar
Halpern, D. F. (2005). How time-flexible work policies can reduce stress, improve health, and save money. Stress and Health, 21(3), 157168. https://doi.org/10.1002/(ISSN)1532-2998CrossRefGoogle Scholar
Hener, T., Rainer, H., and Siedler, T. (2016). Political socialization in flux?: Linking family non-intactness during childhood to adult civic engagement. Journal of the Royal Statistical Society. Series A (Statistics in Society), 179(3), 633656. https://doi.org/10.1111/rssa.12131CrossRefGoogle Scholar
Hill, E. J., Grzywacz, J. G., Allen, S. M., Blanchard, V. L., Matz-Costa, C., Shulkin, S., and Pitt-Catsouphes, M. (2008). Defining and conceptualizing workplace flexibility. Community, Work & Family, 11, 149163. https://doi.org/10.1080/13668800802024678CrossRefGoogle Scholar
Moen, P., Kelly, E. L., Fan, W., Lee, S.-R., Almeida, D., Kossek, E. E., and Buxton, O. M. (2016). Does a flexibility/support organizational initiative improve high-tech employees’ well-being? Evidence from the work, family, and health network. American Sociological Review, 81(1), 134164. https://doi.org/10.1177/0003122415622391CrossRefGoogle Scholar
Nomaguchi, K. M., Milkie, M. A., and Bianchi, S. M. (2005). Time strains and psychological well-being: Do dual-earner mothers and fathers differ?. Journal of Family Issues, 26(6), 756792. https://doi.org/10.1177/0192513X05277524CrossRefGoogle Scholar
Oster, E. (2019). Unobservable selection and coefficient stability: Theory and evidence. Journal of Business & Economic Statistics, 37(2), 187204. https://doi.org/10.1080/07350015.2016.1227711CrossRefGoogle Scholar
Oswald, A. J., Proto, E., and Sgroi, D. (2015). Happiness and productivity. Journal of Labor Economics, 33(4), 789822. https://doi.org/10.1086/681096CrossRefGoogle Scholar
Possenriede, D., and Plantenga, J. (2014). Temporal and locational flexibility of work, working-time fit, and job satisfaction. IZA Discussion Papers 8436, Institute of Labor Economics (IZA).CrossRefGoogle Scholar
Putnam, L. L., Myers, K. K., and Gailliard, B. M. (2014). Examining the tensions in workplace flexibility and exploring options for new directions. Human Relations, 67(4), 413440. https://doi.org/10.1177/0018726713495704CrossRefGoogle Scholar
Riedl, M., and Geishecker, I. (2014). Keep it simple: Estimation strategies for ordered response models with fixed effects. Journal of Applied Statistics, 41(11), 23582374. https://doi.org/10.1080/02664763.2014.909969CrossRefGoogle Scholar
Roberts, J., Hodgson, R., and Dolan, P. (2011). “It's driving her mad”: Gender differences in the effects of commuting on psychological health. Journal of Health Economics, 30(5), 10641076. https://doi.org/10.1016/j.jhealeco.2011.07.006CrossRefGoogle ScholarPubMed
Sloane, P. J., and Williams, H. (2000). Job satisfaction, comparison earnings, and gender. Labour (Committee on Canadian Labour History), 14(3), 473502.Google Scholar
Taneja, S. M. P., and Bloom, N. (2021). Working from home is revolutionising the UK labour market, VoxEU.ORG.Google Scholar
Tavares, A. I. (2017). Telework and health effects review. International Journal of Healthcare, 3(2), 30. https://doi.org/10.5430/ijh.v3n2p30CrossRefGoogle Scholar
Weeden, K. A. (2005). Is there a flexiglass ceiling? Flexible work arrangements and wages in the United States. Social Science Research, 34(2), 454482. https://doi.org/10.1016/j.ssresearch.2004.04.006CrossRefGoogle Scholar
Wheatley, D. (2017). Employee satisfaction and use of flexible working arrangements. Work, Employment and Society, 31(4), 567585. https://doi.org/10.1177/0950017016631447CrossRefGoogle Scholar
World Health Organization (2019). Global health expenditure database. apps.who.int/nha/database. Accessed 14 September 2023.Google Scholar
Yang, D., Kelly, E. L., Kubzansky, L. D., and Berkman, L. (2023). Working from home and worker well-being: New evidence from Germany. ILR Review, 76(3), 504531. https://doi.org/10.1177/00197939221148716CrossRefGoogle Scholar
Figure 0

Table 1. Summary statistics by flexible working status and gender

Figure 1

Figure 1. Effect of flextime at the office and WfH on well-being by gender. Notes: Odds ratios based on equation (4) by gender, with females indicated by red squares and males by blue circles. All equations include year fixed effects, demographic controls, job controls, and family controls.

Figure 2

Table 2. Robustness checks for flextime and WfH: BUC estimates

Figure 3

Table 3. Flextime at the office, WfH, and well-being across individuals with or without young children, by age groups and income levels

Figure 4

Table A1. Definition of variables

Figure 5

Table A2. Pearson correlations for outcomes and some relevant control variables

Figure 6

Table A3. Transition matrix of flexible working by gender

Figure 7

Table A4. Correlations of transition in flexible working and changes in some relevant observable characteristics

Figure 8

Table A5. Flextime at the office, WfH, and employees’ well-being

Figure 9

Table A6. Step-by-step inclusion of controls in the regression

Figure 10

Figure A1. Women: odds ratio of flextime at the office, WfH on well-being by parental status. Notes: The effects on parents are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure 11

Figure A2. Men: odds ratio of flextime at the office, working from home on well-being by parental status. Notes: The effects on parents are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure 12

Figure A3. Women: odds ratio of flextime at the office, WfH on well-being by age groups. Notes: Individuals are divided into three age groups according to tertiles: the youngest cohort (20–36 years old), the middle-aged cohort (37–48 years old), and the oldest cohort (49–66 years old). The effects on the middle-aged cohort and the oldest are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure 13

Figure A4. Men: odds ratio of flextime at the office, working from home on well-being by age groups. Notes: Individuals are divided into three age groups according to tertiles: the youngest cohort (20–36 years old), the middle-aged cohort (37–48 years old), and the oldest cohort (49–66 years old). The effects on the middle-aged cohort and the oldest are calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure 14

Figure A5. Women: odds ratio of flextime at the office, WfH on well-being by income groups. Notes: The low-income group reports a household net income under the sample median, £3,500. The effect on the low-income group is calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.

Figure 15

Figure A6. Men: odds ratio of flextime at the office, working from home on well-being by income groups. Notes: The low-income group reports a household net income under the sample median, £3,500. The effect on the low-income group is calculated through the lincom command in Stata, which computes point estimates and confidence intervals for linear combinations of coefficients displayed as odds ratio.