VoxEU Column Labour Markets Global crisis Gender

A tale of hours worked for pay from home before and after the Great Recession: Learning from high-frequency diaries

The COVID-19 pandemic has meant that more people are working from home and women are disproportionately losing paid work. This column uses daily activity diaries from the American Time Use Survey to look back at the impact of the unemployment benefit extensions that were triggered by the Great Recession on hours worked from home. The overall picture is one of increased gender inequality in the labour market, with women but not men increasing work hours and effort in response to the Great Recession and the consequent changes in the duration of unemployment benefits. 

The current COVID pandemic has dramatically affected the labour market, with the home becoming the predominant place of work for many and women losing more jobs than men (Alon et al. 2020, Baldwin 2020, Barrero et al. 2020, Taneja et al. 2021). The literature to date on work performed from home concludes that working from home may well affect workers’ wellbeing and productivity (Schieman et al. 2009, Bloom et al. 2015, Morikawa 2021). In a new paper (Kapteyn and Stancanelli 2021), we expand on these studies by examining how the Great Recession and the state unemployment benefit extensions it triggered affected work done from home, paying particular attention to gender differences in responses. 

The Great Recession, which officially started in December 2007 and ended in June 2009, was the most severe recession since the Great Depression of the 1930s and its adverse impact on the American labour market extended well beyond 2009 (Yagan 2019). In response to this, unemployment benefit durations were extended up to 99 weeks, from the usual 26 weeks, with large variation across US states in Extended Benefit (EB) generosity and timing. Expansions and cuts in unemployment benefit duration have been shown to affect unemployment and employment to different degrees (e.g. Lalive 2008, Farber et al. 2015). In particular, using Current Population Survey (CPS) data over the period 2008-2014 on the CPS longitudinal sample of unemployed to estimate a competing risk model and distinguishing the impact of benefit expansions from that of benefit cuts, Farber et al. (2015) conclude that both benefit expansions and cuts reduced labour force exits, by increasing labour market attachment, but neither expansions nor cuts in unemployment benefits had any impact on the job finding rate. As they also argued, because benefit expansions are implemented when the labour market is slack and benefit cuts when it is tight, labour demand conditions may mediate individual labour supply responses to benefit cuts and expansions.

Increases in the duration of unemployment benefits are likely to lower the costs of job loss but also make jobs more valuable as eligibility for unemployment benefits is tied to prior employment. Reductions in the duration of unemployment benefits may work in the opposite direction. Changes in the duration of unemployment benefits may impact reservation wages and induce job seekers to accept work from different locations than they would otherwise. Specifically, if working from home is valued differently than working from the employer’s workplace, changes in the duration of unemployment benefits may impact the individual propensity to take up remote work. Changes in the duration of unemployment benefit may also impact individuals beyond those eligible for unemployment insurance, and especially so at times of high unemployment rates. 

In particular, during recessions workers that fear losing their job may increase work effort, by, for example, bringing more work home. Indeed, Lazear et al. (2013) find a significant increase in output per worker during the Great Recession, which appears to be driven by an increase in work effort, based on daily panel data for a large US firm. The authors hypothesize that during recessions workers are willing to work harder for the same wage, as they fear losing their job and the value of alternatives decreases. 

We examine over 150,000 continuous daily activity diaries1 from the American Time Use Survey (ATUS) 2003-2019, linked to Current Population Surveys (CPS), which are merged with state monthly unemployment rates, from local area unemployment (LAU) statistics of the Bureau of Labor Statistics (BLS), and exhaustive start/end dates of state Extended Benefits collected by the Department of Labor’s Employment and Training Administration. 

Out of the 3.5 hours worked on average per person and per day (considering the population aged under 70 and excluding students, the military and agricultural sectors, and the self-employed), less than three hours are worked from people’s workplace and over 40 minutes from locations other than the workplace (home, elsewhere or commuting).2

Figure 1 State monthly unemployment rates


Note: The chart plots the state monthly unemployment rates for the 50 states and the district of Columbia. The data are drawn from the Local Unemployment Area statistics of the Bureau of Labor Statistics and merged with the American Time Use Surveys data. The state unemployment rate in the month before answering the survey is linked to ATUS, based on the date of survey participation and the state of residence of ATUS respondents. The two red vertical lines are drawn, respectively, to denote the technical start of the great recession, in December 2007, and its end, in June 2009.  

We exploit the large upswings in monthly unemployment rates across US states (see Figure 1) to capture the impact of labour market slack on hours worked from home and other locations (the workplace, the car, or elsewhere). The variation in the timing of unemployment insurance extensions by state helps us identifying their impact on hours worked from home and other places. While state EB programmes are triggered by the state unemployment rate, the begin and end dates of these extensions are unlikely to be perfectly anticipated by individuals. Often states decided upon starting or ending unemployment benefit extensions with little advance notice, if any at all. Therefore, we implement a regression discontinuity design, in which the running variable is the days elapsed between the ATUS interview and the begin/end date of the EB. Benefit expansions and cuts are allowed to have a differential impact on work done from home and other locations, by estimating separate RDD models for EB starts and ends. EB expansions (reductions) make unemployment benefits available for longer (shorter) durations and thus capture lower (higher) costs of job loss but possibly also additional (lesser) value from employment due the additional job protection. Therefore, state EB expansions and cuts may impact individuals beyond the those eligible for unemployment insurance and especially so at times of high unemployment rates when workers may fear losing their job. Ours is the first study to investigate the effects of the Great Recession and the unemployment benefit extensions it triggered on hours worked from home and other locations. 

Figure 2 The impact of benefit extension start/end on hours of work


Notes: The figures plot the residuals of hours of work (measured in minutes per day) against the days elapsed since, respectively, the beginning or the end of the Benefit Extension program. The hours of work residuals are estimated by regressing hours of work on individual demographics (a quadratic in age, sex, education, race, marital status, dummies for number and age of kids), state of residence, metropolitan area of residence, home type, day of the week, month, year and industry fixed effects, and state monthly unemployment rate in the month before the day of the interview. This approach enables us to focus on the impact of benefit extensions on hours of work, once other explanatory variables have been accounted for. The dots represent the data averages, while the thick line are the estimates from a triangular Kernel and the thin lines are the 95% confidence intervals around those estimates. 

We find that the Great Recession, as tracked by the huge variation in unemployment rates across states, reduced overall employment but increased the probability of working for pay from home for women. We also conclude that EB cuts significantly increased overall employment (by over four percentage points, which corresponds to a 9% increase in employment) and hours3 (see Figure 2), with work done from home increasing by 24% and work from the workplace by 7%. These findings conceal substantial heterogeneity as these effects are large and strongly significant for women but are not statistically significant for men, a stark contrast with the fact that employment and hours dropped more from men than for women with the Great Recession. 

In particular, we find that EB cuts increased women’s employment by 8.5 percentage points and women’s hours worked by half an hour per day, on average. For women, EB cuts increased the probability of working from the workplace by seven percentage points, from home by five percentage points and commuting by six percentage points. The probability of doing some work both from the workplace and from home, and at unsocial hours (night-time or on a Sunday) almost doubled with EB cuts for women, suggesting that their work effort increased considerably. Moreover, EB expansions significantly increased women’s employment (only significant at the 10% level) and women’s commuting.

These estimates are robust to several checks, including narrowing the RDD bandwidth, as well as dropping covariates, or dropping treated states one by one, or eliminating observations for days close to the cut-off from the estimation sample. In addition, these findings hold when controlling for occupation fixed effects (in addition to several hundred 4-digit industry fixed effects included in the RDD model) or for the duration of state regular and total unemployment benefits. 

By replicating the analysis also for other subgroups of the population, we also conclude that employment and hours increased significantly and largely in response to EB changes for racial and ethnic minorities. For this subgroup, employment increased by over eight percentage points and hours worked by about an hour per day, with either EB expansions or reductions. Commute also increased significantly for this subgroup of the population but not work performed from home. We conclude that the increase in hours worked from home concerned mainly white women with less than college education. White women and women from racial and ethnic minority groups increased work done from the workplace and commute, as well as work performed at unsocial hours (night-time or on a Sunday) in response to EB changes. 

These findings are unexpected as employment and hours of men dropped more than employment and hours of women with the Great Recession and therefore, one might have expected men to react more than women to changes in the duration of state unemployment benefits. Furthermore, women were not overrepresented compared to men among home workers or those working from the car before these changes in the generosity of state unemployment insurance: over 9% of women and about 11% of men were doing some work from home and about 35% of women and 45% of men were working from the car, on average. The overall picture is one of increased gender inequality in the labour market with women but not men increasing work hours and effort in response to the Great Recession and the consequent changes in the duration of unemployment benefits.    


Alon, T, M Doepke, J Olmstead Rumsey, and M Tertilt (2020), “The Impact of Covid-19 on Gender Equality”, NBER Working Paper 26047. 

Baldwin, R (2020), “COVID, hysteresis, and the future of work”,, 29 May.

Barrero, J, M, N Bloom and S J Davis (2020), “COVID-19 and labour reallocation: Evidence from the US”,, 14 July. 

Bloom, N, J Liang, J Roberts, and Z J Ying (2015), “Does Working from Home Work? Evidence from a Chinese Experiment”, Quarterly Journal of Economics 130(1): 165-218.

Farber, H, S, J Rothstein and R G Valletta (2015), “The Effect of Extended Unemployment Insurance Benefits: Evidence from the 2012-2013 Phase-Out”, American Economic Review 105(5): 171-76. 

Kapteyn, A and E Stancanelli (2021), “A tale of hours worked for pay from home before and after the Great Recession:  learning from high-frequency diaries”, mimeo.

Lalive, R (2008), “How do Extended Benefits affect Unemployment Duration? A Regression Discontinuity Approach”, Journal of Econometrics 142(2): 785-806.  

Lazear, E, P, K L Shaw and C Stanton (2016), “Making Do with Less: Working Harder During Recessions”, Journal of Labor Economics 34(S1): 333-360.

Morikawa M (2021),“The productivity of working from home: Evidence from Japan”,, 12 March. 

Schieman, S, M A. Milkie and P Glavin (2009), “When Work Interferes with Life: Work-Nonwork Interference and the Influence of Work-Related Demands and Resources”,  American Sociological Review 74: 966-988. 

Taneja, S, P Mizen, and N Bloom (2021), “Working from home is revolutionising the UK labour market”,, 15 March. 

Yagan, D (2019), “Employment Hysteresis From the Great Recession”, Journal of Political Economy 127(5): 2505-2508.  


1 ATUS diary are collected every day from 1 January to 31 December of each year, including on weekends and vacation days. The day of the interview is not chosen by the survey participants, who are randomly allocated to interview days by the Bureau of the Census. The ATUS activity diary collects information on the activities carried out over a 24-hour period, beginning in the middle of the previous night.

2 The following possible alternative locations are reported in the diary for each activity: the workplace; the respondent’s home; someone else’s home; restaurant or bar; place of worship; grocery store; other store or mall; school; outdoors away from home; library; bank; gym, health club; post-office; other or unspecified place; car, track or motorcycle (driver); car, track or motorcycle (passenger); bus; subway, train; boat, ferry; taxi, limousine service; airplane; other mode of transportation; unspecified mode of transportation.  

3 Work is defined as including all hours of work, including time spent on work/related meals and activities and other income-generating activities, as well as time spent commuting to work, as is standard in this literature.

2,835 Reads