Are Those Who Bring Work Home Really Working Longer Hours? Implications for BLS Productivity Measures* Preliminary: Do not quote.

Transcription

1 Are Those Who Bring Work Home Really Working Longer Hours? Implications for BLS Productivity Measures* Preliminary: Do not quote. Lucy P. Eldridge U.S. Bureau of Labor Statistics 2 Massachusetts Ave., NE Rm Washington, DC Eldridge.Lucy@bls.gov Sabrina Wulff Pabilonia U.S. Bureau of Labor Statistics 2 Massachusetts Ave., NE Rm Washington, DC Pabilonia.Sabrina@bls.gov November 2005 * The authors thank Jay Stewart, Anastasiya Osborne,Younghwan Song, and Leo Sveikauskas for discussions on using the ATUS. All views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of the U.S. Bureau of Labor Statistics.

2 I. Introduction Advancements in information technology have increased workers ability to conduct their jobs in various locations. An ongoing debate surrounding BLS productivity data is that official productivity numbers are perhaps overstated because of an increase in unmeasured hours worked outside the traditional workplace. This paper uses 2003 and 2004 data from the new American Time Use Survey (ATUS) to examine U.S. workers who report doing some work outside their workplace, and particularly those who bring work home. The ATUS provides detailed information on time spent on work, work-related activities, non-work activities, and locations for each activity. The ability to bring work home depends on the nature of work, the worker s degree of autonomy, and technology needed and available to do a job. This is evident in the differences in the tendency to bring work home among ATUS respondents across occupation, industry, and education groups. Contrary to popular perceptions, we show that not all work that is brought home is done by workers in industries with predominately white-collar office workers. However, throughout this paper, we will interchangeably refer to work brought home from the traditional workplace as work brought home from the office, even though there are many different types of workplaces other than office buildings from which workers could bring work home. For example, the traditional workplace for the construction foreman is a construction site rather than an office building and the foreman may do some of his preparation for the following day in his home in the evening. There are many possible reasons why workers may choose to bring their work home. In this study, we analyze time-use behavior to gain insight as to whether workers bring work home

3 to better balance work and family obligations, or to gain an advantage over their competitors. For example, we find some support for a work/family balance explanation because fathers of young children are more likely to bring work home, and workers report watching their children as a secondary activity while working at home as their primary activity. This paper also explores whether those who bring work home work longer hours, or whether they have simply reallocated a portion of their usual hours of work from their traditional workplace to their home. Moreover, based upon the analysis of characteristics of those who bring work home, we infer whether these hours at home are likely to be unpaid, and thus may be missed in official BLS productivity measures. From the ATUS data, we find that approximately 2 percent of production/nonsupervisory employees brought unpaid work home from the office in 2003 and Among these workers, average weekly hours were greater than the average of all production/nonsupervisory workers in 2004, but less than average in A larger portion of nonproduction/supervisory employees brought unpaid work home in 2003 (8.9 percent) than brought unpaid work home in 2004 (5.4 percent). Among the nonproduction/supervisory employees who brought unpaid work home, we find that they worked 10.9 percent greater average weekly hours than all nonproduction/supervisory employees in 2003 and 14.2 percent greater hours in However, only percent of their hours were worked outside the office, and thus potentially unmeasured. Moreover, because BLS uses CPS data to adjust the hours worked for nonproduction/supervisory employees, a portion of these data may be captured in the existing measures. As an upper bound, the hours of only 3 percent of nonfarm business sector workers may be mismeasured due to unpaid hours worked in 2003, and 2.6 percent in 1 We define unpaid work as work that is not done at the office by an employee who is not paid hourly and not paid overtime or commission. 2

4 2004. With successive years off data, we will be able to determine whether this work done at home is unmeasured and increasing. In the next section, we review the literature relevant to this study. In section III, we describe the American Time Use Survey and provide descriptive statistics on workers who bring work home. In section IV, we do a multivariate analysis to determine what characteristics affect the probability of bringing work home. In section V, we discuss labor productivity measurement. In section VI, we examine whether work done outside the traditional workplace is captured by official productivity measures. Section VII concludes this paper. II. Prior Research Previous research on hours worked in other time use surveys and on work at home arrangements are both relevant to this paper; however, only Callister and Dixon (2001) examine workers who work both at the office and at home on the same day. These workers may be more likely to inaccurately state their hours reported in household surveys than hours reported by exclusively homeworkers. This could be the case because the work schedule of these workers is more varied, and they may be less likely to be paid for their work at home. Evidence from household time use diaries shows that respondents to labor force survey questions similar to those in the CPS report higher hours worked compared to the estimates resulting from time diary studies; see for instance Hamermesh (1990), who uses Michigan time use diary data for 1975 and 1981; and Robinson and Bostrom (1994), who use three separate studies for 1965, 1975, and Robinson and Bostrom (1994) also show that this reporting difference is greater for those who worked longer hours, and both papers show that the difference 2 Note that the sample sizes in these studies are smaller than the ATUS sample. 3

5 increased over time. However, Jacobs (1998) finds that independent, self-reported measures of working time based on time of departures to and returns from work tend to corroborate labor force survey types of hours questions. Until recently, there have not been any studies comparing hours reports to post-redesign CPS questions with time use diary reports, but presumably the discrepancy is less. 3 Frazis and Stewart (2004) compare CPS reported hours and hours constructed from ATUS time-diaries. Using similar definitions of hours worked, they find that the CPS reported hours of work are very close to those constructed using ATUS time-diaries for the 12 CPS reference weeks in This could be due in part to efforts undertaken in the 1994 redesign of the CPS questionnaire to enhance respondent recall. However, Frazis and Stewart (2004) also find that for weeks outside of the CPS reference weeks in 2003, the ATUS respondents worked fewer hours on average than were reported in the CPS. The ATUS data suggest that for weeks outside the CPS reference weeks, respondents actually work on average approximately 5 percent fewer hours per week. 5 Thus, this preliminary finding suggests that the use of the CPS data to extrapolate monthly, quarterly, or annual hours could be leading to an upward bias in the number of hours worked. A substantial amount of paid work at home occurs globally. For example, the 1999 New Zealand Time-Use Survey reveals that 15.5 percent of non-agricultural weekday workers combined office and home work on their diary day (Callister and Dixon 2001). Murphy and Satherely (2000) find that a greater percentage of these workers live in rural areas rather than urban areas, and there is little difference in work at home by gender. 3 In the 1994 revised CPS, the question on usual hours is asked first, followed by questions about overtime and taking time off for reasons such as illness, slack work, vacation or holiday. Polivka and Rothgeb (1993, p. 16) report that The mean of reported hours measured with the current [pre-1994] wording was 39.0 compared to 37.9 hours measured with the revised [1994- and later] wording. This is a combined survey effect of the employment and hours questions. 4 The CPS reference week is the week of each month that contains the 12 th. 5 Data have been compiled across all months due to the limited number of observations. 4

6 Recent research on work at home arrangements is primarily based upon home-based workers or occasional telecommuters. Recently, Oettinger (2004) used the 1980, 1990, and 2000 Census data to reveal the growth in home-based employment over two decades. He shows that wage penalties for working at home have fallen over time and that the increase in home work is the largest for highly-educated workers. Using the May 1997 CPS Work at Home supplement, Schroeder and Warren (2004) analyzed workers who did any work at home, including homebased workers, occasional telecommuters, and those who bring their work home in the evening. They find that these workers are more likely to be older, better educated, married and white than traditional office workers. They also find that managers and professionals are more likely to report some work from home. Unlike in New Zealand, U.S. workers who do some work at home are more likely to live in an urban area. Using the Canadian Workplace and Employee Survey, Pabilonia (2005) analyzed the decision of employees to do paid work at home during part of their normal working hours (referred to as telecommuters) and the decision of firms to allow these employees to telecommute. In 2001, the 5.9 percent of telecommuters among Canadian workers are more likely to be tech-savvy, experienced white-collar workers than their non-telecommuting counterparts. Unlike the 1997 CPS Supplement used by Schroeder and Warren (2004), data from the 2001 and the recent 2004 CPS Work Schedule and Work at Home Supplements can be used to identify work done at home and in the office on the same day. A future version of this paper will compare the ATUS results with results from these supplements to determine whether there are measurement differences between samples and whether this type of work has grown over the past three years. In addition, these data can help to identify the reasons U.S. workers bring work home. 5

7 III. Data and Descriptive Statistics The American Time Use Survey (ATUS), which began collecting data in 2003, is a survey of how people living in the United States spend their time. The ATUS sample consists of one household member aged fifteen or older from a subset of households completing their final month of interviews for the Current Population Survey (CPS). In 2003, there were 20,720 ATUS interviews. Beginning in December 2003, the sample size was reduced by 35 percent, yielding 13,973 completed diaries in ATUS collects a 24 hour diary of the activities that a respondent was engaged in starting at 4 A.M on the day prior to their interview. These diaries include information on hours worked, such as time at work, time spent on work activities at home, and interruptions of 15 minutes or longer that take place during the work day. 6 The kinds of activities and the time spent on them are available by various demographic characteristics of the respondents, the location where the activity took place, and the people who were with the respondent. In order to analyze hours of work, we aggregated minutes spent on activities coded as work at main job for each ATUS respondent by location from the ATUS activity files, and constructed measures of work time at the workplace, at home, and outside the workplace. We define those who work outside the workplace as respondents who report any minutes of work on their main job at a location other than the workplace. We define those who bring work home as respondents who report any minutes of work on their main job at the workplace and at home on the same day. The restriction to the main job is necessary in order to allow us to focus on those 6 ATUS interviewers are trained to ask for work breaks of 15 minutes or longer any time a respondent reports that he or she worked. Beginning in January 2004, an automated probe was introduced into the survey instrument. If a respondent reports working for more than 4 hours at one time, the interviewer automatically is prompted to ask Did you take any breaks of 15 minutes or longer? If the respondent reports taking a break, the interviewer records the start and stop time and what was done on that break; if no break, the solid work episode is recorded. 6

8 who bring work home rather than those who may be doing some part-time work at home in the evenings. 7 We created two samples of workers from the ATUS in order to analyze workers who bring work home. The first sample, which we refer to as the nonfarm worker sample, consists of ATUS nonfarm workers over the age of fifteen, whose diary day was a weekday. The restriction to nonfarm workers is made because the nature of farming would suggest that one could not bring this type of work home. 8 The restriction to over age fifteen was made because official statistics rarely include work done by fifteen-year-olds, and there were only a couple of fifteenyear-olds who brought work home from the workplace. The latter is not unexpected given the types of jobs fifteen-year-olds hold (Pabilonia 2001). We restrict the sample to weekdays because only 40 of the 593 individuals who brought work home for their main job worked on Saturday or Sunday in This is not to say that weekend work at home is not important and may also be missed in official productivity measurement; however, we do not know whether or not the ATUS respondent interviewed on a weekend was working at the workplace during the week and bringing some extra work home to do over the weekend. In future versions of this paper, we hope to be able to infer whether work done at home on the weekend is brought home by examining the occupation and industry codes for the main job and how many hours respondents worked during the week. Table 1 shows the share of workers in this sample, by the location of their work activities. The majority of nonfarm workers (80 percent) work exclusively at the workplace, while a much smaller portion of workers (9 percent) work exclusively away from the workplace. The remaining workers work a combination of hours at the workplace and away from the workplace. 7 Data can be constructed for hours worked for all jobs. This will be done in a future extension of this paper. 8 Of course, some farm work is done exclusively at home on the family farm. 7

9 The largest portion of this group brings work home from the workplace. In 2003, 8.1 percent of nonfarm workers brought some of their work home; in 2004, this share was 7.9 percent. In official productivity statistics, workers in the nonfarm business sector are those who are over fifteen-years-old, work outside of the farm sector, and are classified as employees of private for-profit entities, are self employed, or are unpaid family workers. Our second sample, which we refer to as the nonfarm business employee sample, consists of ATUS employees who are in the nonfarm business sector and whose diary day was a weekday. 9 We will use this sample when we compare hours of work in relation to the major sector productivity statistics. Among those who bring work home in the nonfarm worker sample, we find that more people work at home in the middle of the week than on Mondays and Fridays. Table 2 presents the portion of workers who bring work home by time of day that home work is conducted. We find that much of their work at home was done in the evenings. In 2003, 31 percent did some work at home after 6 PM. A smaller percentage (13 percent) did some work prior to 8 AM. Twelve percent of individuals who worked at home did so while having a child in their care in 2003; however, only two percent did work after their last child had gone to bed. Table 3 presents the portions of workers that bring work home by the number of minutes worked at home. We find that amount of work done at home was economically significant. Of those who brought work home in 2003, slightly more than half worked for more than one hour. Twelve percent of workers who brought work home did less than 15 minutes of work at home. In Table 4, we compare the characteristics of 2003 respondents who bring work home to those who do not bring work home, by sample type. In Table 5, we do the same for 2004 respondents. In 2003, workers in both samples who brought work home were slightly less likely 9 age1>15 & hrswk>0 & teio1cow=4 & teio1icd>180 & tudiaryday>1 & tudiaryday<7 8

10 to be female. Black and Hispanic workers are less likely to work at home. Workers who bring work home were much more likely to be married and have a spouse who also works than those who do not. Workers who bring work home were more likely to hold a bachelor s degree than those who do not. For example, among nonfarm workers, 58 percent of bring work home workers held at least a bachelor s degree while only 29 percent of workers who do not bring work home. Of those that bring work home in the nonfarm worker sample, only 17 percent are considered to be paid hourly in 2003, while 55 percent of nonfarm workers who did not bring work home were paid hourly. In all samples, workers who bring work home are more likely to be managers or professionals than workers who do not bring work home. Among industries, workers who bring work home in either sample are more likely to work in the information industry than workers who do not bring work home. In the 2003 nonfarm worker sample, 28 percent of workers who brought work home worked in the educational and health services industry. This high percentage likely reflects teachers who prepare lessons or do grading in the evening after leaving their school (Callister and Dixon 2001). There are a few other industry differences but they are not consistent between sample years. As mentioned in the introduction, workers who may be working in workplaces other than an office building are also doing some work home. For example, seven percent of workers in construction brought work home. IV. Regression Analysis We estimated the following model on the pooled 2003 and 2004 ATUS data to determine the demographic and job characteristics of workers associated with bringing work home. 9

11 y it * = α + βx it + ε it (1) y 1it = 1 if y it *>0 y 1it = 0 if y it * 0 where y it * is the unobserved propensity to bring work home, y 1it is an observed dichotomous variable indicating whether or not a worker brings home work on their diary day, X it is a vector of observed characteristics of i at time t; α and β are parameters to be estimated; and ε it is a stochastic disturbance term assumed to follow a normal distribution. X it includes human capital factors (educational degree attainment indicators), demographic characteristics (gender, age and age squared, race, Hispanic ethnicity, indicators for married or divorced, an indicator for spouse works, indicators for having a child who was an infant, preschooler, elementary school student, or adolescent, and indicators for the interaction of these latter child variables with gender), job characteristics (indicators for multiple job holder, paid hourly, six occupations, and thirteen industries), four region indicators and a year indicator. Marginal effects and standard errors from a reduced-form probit regression are reported in Table All standard errors are weighted to account for sample design. Holding all else equal, results as a whole indicate that highly-educated, salaried workers are more likely to bring work home than less-educated, hourly workers. Black workers are less likely to bring work home than whites and Hispanic workers are less likely to bring work home than non-hispanic workers. In addition, results suggest that fathers of young children (not school aged) are more likely to bring work home than men without children and that fathers with infants have the largest effect on the probability of bringing work home. Therefore, some fathers may bring work home to better balance work and family responsibilities when the children are young. Results for 10 A test for whether all the coefficients in this model were identical for females and males could not be rejected. 10

12 mothers of young children are not significant, and the variables representing the interaction of female with either an indicator variable for infant, preschooler, or elementary student consistently indicate a negative effect of having a young child on the probability of bringing work home. It is possible that this result is biased due to sample selection issues arising from mothers of young children choosing not to participate in the labor force. Alternatively, mothers of young children as opposed to fathers may choose not to do any work at home because they traditionally spend and may be expected to spend more time on childcare and household production than their male spouses. Contrary to the descriptive statistics, married workers are not more likely to bring work home than single workers. The previous result may be due to differences in the types of jobs married versus single workers hold. V. Use of Hours Data in U.S. Productivity Measurement Productivity trends are widely watched by businesses, policymakers, and many others interested in understanding business cycles and assessing U.S. competitiveness. BLS produces quarterly measures of labor productivity for major sectors of the economy, including business, nonfarm business, nonfinancial corporations, manufacturing, and durable and nondurable manufacturing. 11 Labor productivity measures the difference between output growth and labor hours growth. Labor productivity reflects many sources, including increases in the quantities of nonlabor inputs (i.e., capital services, fuels, other intermediate materials, and purchased services); changes in technology; economies of scale; improvements in management techniques; and improvements in the skills of the labor force. 11 BLS also produces measures of multifactor productivity for major sectors and labor productivity for selected detailed industries. 11

13 To calculate labor productivity for the U.S. nonfarm business sector, the BLS combines indexes of real output from the National Income and Product Accounts (NIPA) produced by the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce. The BLS Office of Productivity and Technology constructs quarterly measures of hours worked for all persons as the relevant labor input for measuring major sector productivity. There exist two surveys in the U.S. that collect monthly measures of hours and employment, the BLS Current Employment Statistics program (CES) and the Current Population Survey (CPS). 12 The CES is the primary source of data used to construct hours for productivity measures. 13 The CES is a monthly payroll survey that collects data on employment for all employees from a sample of 400,000 nonfarm establishments. The CES also collects data on the average weekly hours-paid for production workers in goods-producing industries and nonsupervisory workers in serviceproducing industries. 14 The CES average weekly hours paid data are adjusted to an hours-atwork concept using a ratio of hours-worked to hours-paid because changes in vacation, holiday, and sick pay are best viewed as changes in labor costs. 15 Because they are not collected in the CES, BLS estimates the average weekly hours of nonproduction/supervisory workers using data from the CPS. 16 Data from the CPS are used to construct a ratio of the average weekly hours 12 The CPS, the U.S. labor force survey of approximately 60,000 households, collects information on employment, hours worked, and demographics for all persons, including employees, proprietors, and unpaid family workers. 13 The CES sample is over six times larger than the CPS sample. In addition, the CES is benchmarked annually to levels based on administrative records of employees covered by state unemployment insurance tax records. There is no direct benchmark for the CPS employment data. Adjustments to the CPS underlying population base are made annually using intercensal estimates and every ten years using the decennial census. Also, establishment hours data are more consistent with the measures of output used to produce productivity measures; output data are based on data collected from establishments. In addition, establishment data provide reliable reporting and coding on industries and thus are well-suited for producing industry-level measures. Measures for industries based on household reports tend to produce industry estimates with considerable variance, even in a survey as large as the CPS. Thus, BLS s official measures by industry come from establishment surveys wherever possible. 14 The BLS is planning to collect CES data on earnings and hours for all employees and publish estimates in The hours worked to hours paid ratio is constructed using information from the National Compensation Survey program; prior to 2000, the annual Hours at Work Survey was used. 16 In August 2004, BLS introduced this new method of constructing estimates of hours for nonproduction and supervisory workers; see Eldridge, Manser, and Otto (2004). 12

14 worked by nonproduction/supervisory workers relative to the average weekly hours worked by production/ nonsupervisory workers. 17 This ratio is combined with CES hours and employment information to arrive at a measure of total hours worked for nonproduction/supervisory workers. Hours worked data on employees of farms, proprietors, and unpaid family workers are taken directly from the CPS; remaining data are obtained from various sources. 18 To construct hours worked for the U.S. business sector, BLS removes the hours worked by employees of nonprofit institutions. VI. Implications for Productivity Measures: Are Hours at Home Being Counted? As U.S. productivity showed unexpected growth around the 2001 recession, debates began about the possible overestimation of BLS official productivity numbers by understating hours worked. Some critics suggested that IT innovations have allowed workers the flexibility to work outside the traditional workplace and that these hours worked are not being properly measured. 19 This criticism is mainly presented in terms of the most widely watched measure of productivity, the quarterly labor productivity measures for the nonfarm business sector; therefore, this discussion will focus on the nonfarm business sector. It is important to note, 17 In goods-producing industries, workers are divided into production and nonproduction workers. Nonproduction workers include professional specialty and technical workers; executive, administrative, and managerial workers; sales workers, and administrative support workers, including clerical. In service-producing industries, workers are divided into supervisory and nonsupervisory workers. Supervisory workers include all executives and administrative and managerial workers. 18 Employment counts for employees in agricultural services, forestry and fishing come from the BLS s 202 program, based on administrative records from the unemployment insurance system. These counts are moved forward in the current period using limited information on employment in agricultural services (part of SIC 07). The number of employees of government enterprises comes from the BEA. These are moved forward using information from the CES. Average weekly hours for employees of government enterprises are from the CPS. 19 Steven Roach of Morgan Stanley Dean Witter argued that the government may be undercounting the number of hours people are working. He wrote that many white collar workers are in fact working longer workdays than the official U.S. data show, as a result of the new portable technologies of the information age laptops, cellular telephones, home fax machines, and beepers; see Roach (1998, p.6). 13

15 however, that an underestimation of hours worked would affect measures of productivity growth only if underestimated hours grow differently than measured hours and affect a significant portion of the working population. Eldridge (2004) found that a hypothetical hours series constructed by combining CPS average weekly hours data and CES employment data produces slightly higher hours levels, but showed a similar decline from The measures of hours worked used to measure productivity are constructed separately for nonemployees, production/nonsupervisory employees, and nonproduction /supervisory employees. Therefore, we analyze hours worked for each group of workers separately. Figure 1 shows the share of nonfarm business sector hours worked and employment for each group. Production/nonsupervisory employees account for the majority of all nonfarm business sector hours worked, while nonemployees account for the smallest share of hours worked. Figure 1. Share of Nonfarm Business Sector Hours and Employment, by Type of Worker: % 12% 19% Production and Nonsupervisory Nonproduction and Supervisory 17% 69% Nonemployees 71% Hours Worked Employment 14

16 Nonemployees First, the BLS measure can be divided into employees and nonemployees. Hours worked for nonemployees account for approximately 12 percent of nonfarm business sector hours and employment, and are obtained directly from the CPS. There may be reason to be skeptical of these data because of the small sample size and the occurrence of proxy reporting; however, the CPS is the sole source for these data. Since previous research on household surveys and selfreporting suggests that there is a tendency toward over-reporting hours worked, it is unlikely that nonemployee hours are biased downward. 20 Production and nonsupervisory employees As mentioned above, BLS uses different procedures to construct hours for production/nonsupervisory employees and for nonproduction/supervisory employees. Hours worked for production/nonsupervisory employees account for approximately 69 percent of nonfarm business sector hours worked and 71 percent of employment. Hours worked by these employees are constructed by multiplying average weekly hours paid from the CES, employment from the CES, the hours-worked to hours-paid ratio, and 52 weeks in a year. The hours-worked to hours-paid ratio removes paid vacation, holidays and sick leave. If hours for production and nonsupervisory employees are being understated it is only to the extent that workers are working without pay. Using the 2003 and 2004 ATUS data we find that 12.7 percent of production/ nonsupervisory employees did some work outside the workplace on their diary day in 2003, and 20 Future research will look at the ATUS to gain insights on reporting by nonemployees. 15

17 15.3 percent in See Tables 7 and 8 for these results. The average weekly hours of those who worked outside the workplace were only 0.4 percent greater than the average of all production/ nonsupervisory employees in 2003, and almost 3 percent greater in However, the average daily hours reported to the ATUS were less than all production/ nonsupervisory employees average daily hours in both years. This result is not surprising because this group may include part time workers. It is reasonable to expect that employees that are paid a salary are more likely to have unpaid hours outside the workplace. We find that 6.1 and 6.5 percent of production/nonsupervisory workers who worked outside the workplace were paid a salary and not paid overtime in 2003 and 2004 respectively. Among these employees, we find that average weekly hours were 2.3 percent greater weekly hours than all production/ nonsupervisory employees in 2003, and 6.6 percent greater in These findings suggest that there may exist unmeasured hours for production/nonsupervisory employees that work outside the workplace, and that these hours may be increasing. However, we find that this group of employees, production/ nonsupervisory employees who worked outside the workplace are paid a salary and not paid overtime, accounted for 4.3 percent of nonfarm workers in 2003 and 4.5 percent of nonfarm workers in Among production/nonsupervisory employees, approximately 3.9 percent brought work home from the workplace on their diary day in 2003 and 6.1 percent in In 2003, these employees had average weekly hours that were 4.1 percent lower than average for all production/ nonsupervisory employees, but 7.8 percent higher average weekly hours in Among those 21 We further restrict the sample to those who have the same occupation and usual duties as they reported to the CPS. This reduces the number of respondents that bring work home from 246 to 140 in We further restrict the sample to those who have the same occupation and usual duties as they did when they reported to the CPS. This reduces the number of respondents who bring work home from 246 to 140 in

18 employees who brought work home, we find that their average daily hours are greater than the average for all production/nonsupervisory employees, but their average hours in the workplace are lower than for all production/nonsupervisory employees. Thus, we do find that those who bring work home are reallocating some of their work across locations. From the 2003 ATUS data, we find that over 73 percent of production/nonsupervisory employees that brought work home were not paid hourly; in 2004, 55 percent were not paid hourly. For these employees, average weekly hours in 2003 were 0.3 percent higher than the average for all production/nonsupervisory employees; however, they were 5.1 percent higher in These production/nonsupervisory employees who brought work home and were paid a salary and not paid overtime accounted for approximately two percent of all production/nonsupervisory workers in both years. They accounted for 1.4 percent of nonfarm worker employment in 2003 and 1.7 percent in Their average weekly hours were seven percent lower in 2003 and ten percent higher in Nonproduction and supervisory employees Hours worked by nonproduction/supervisory employees account for approximately 19 percent of nonfarm business sector hours worked and 17 percent of employment. Hours worked for these employees are constructed by multiplying employment from the CES, the ratio of nonproduction/supervisory average weekly hours to production/nonsupervisory average weekly hours, average weekly hours paid from the CES, the hours worked to hours paid ratio, and 52 weeks in a year. We find that 26.4 percent of nonproduction/supervisory employees did some work outside the workplace on their diary day in 2003; this fell to 23.9 percent in See results in 17

19 Tables 9 and 10. Among these employees, the majority did not receive overtime pay or commissions and were not paid hourly. In 2003, 19.5 percent of nonproduction/supervisory employees did some work outside the workplace and were not paid overtime and were not paid hourly; this decreased to 18.8 percent in The average weekly hours of these employees was approximately 5.7 percent greater than more average weekly hours in 2003 and 5.8 percent greater in However, their average daily hours reported to the ATUS were 11 percent less than average in 2003 and 15 percent less in We find more striking results among nonproduction/ supervisory employees who brought work home from the workplace on their dairy day. In 2003, 10.9 percent of nonproduction/ supervisory employees brought work home; this decreased to 7.1 percent in The average weekly hours of those who brought work home were approximately 11 percent greater than the average of all nonproduction/supervisory employees in both years. Moreover, their average daily hours reported to the ATUS were 21 percent greater in 2003 and 14 percent greater in From the ATUS data, we find that the majority of nonproduction/supervisory employees who brought work home were not paid hourly nor any overtime/commissions/tips. In 2003, 8.9 percent of nonproduction/supervisory employees brought work home and were not paid overtime and were paid salary; this decreased to 5.4 percent in Among these employees, average weekly hours were 10.9 percent greater in 2003 and 14.2 percent greater in However, only 15.8 percent of their hours were worked outside the workplace. It is clear that those nonproduction/supervisory employees that bring work home and are not paid overtime and are paid salary are not reallocating work across locations but are truly working longer hours. Because BLS uses CPS data to adjust the hours worked for nonproduction/supervisory employees, a portion of these data may be captured in the existing measures. Nonproduction/supervisory employees who bring work home, are not paid 18

20 overtime, and are paid a salary account for 1.6 percent of nonfarm workers in 2003 and 0.9 percent of workers in Although the percent of workers in this group is decreasing, their hours outside the traditional workplace are increasing. VII. Conclusion Using the ATUS, we examined the 8.7 percent of U.S. workers in 2003 and 9.9 percent of U.S. workers in 2004 that performed their main job both in the workplace and in a location outside their workplace on the same diary weekday. The majority of these workers bring their work home, and over half work more than one hour at home. Forty-three percent of workers who bring work home did some of their work at home outside the traditional core business hours of 8 AM to 6 PM. Results from a probit regression suggest that fathers of young children are more likely to work at home than men who are not fathers. In addition, over ten percent of bring-work-home workers report a child in their care while working at home. These results provide evidence that workers bring work home at least in part to better balance work and family responsibilities. Results from the probit model also indicate that highly-educated, salaried workers are much more likely to bring work home than less-educated, hourly workers. These findings suggest that the nature of the work and the desire to stay ahead of the competition may also drive workers to work longer hours and that they choose to do that extra work at home, perhaps due to the before-mentioned family responsibilities. We compared average weekly and daily hours worked for those who worked exclusively in the workplace to those who also worked outside the workplace. The share of nonproduction/ supervisory employees that do some work outside the workplace is almost twice that of 19

21 production/nonsupervisory employees, while the share of nonproduction/supervisory employees that bring work home from the workplace is almost three times that of production/nonsupervisory employees. We find that production/nonsupervisory employees who worked outside the workplace and who are paid a salary and not paid overtime reported that their average weekly hours were 2.3 percent greater than weekly hours for all production/nonsupervisory employees in 2003, and 6.6 percent greater in This finding suggests that there may exist unmeasured hours for production/nonsupervisory employees that work outside the workplace. This group of employees account for 4.3 percent of nonfarm worker employment in 2003 and 4.5 percent in The findings among production/nonsupervisory employees who brought work home and were not paid hourly and not paid overtime, approximately 1.4 percent of nonfarm worker employment in 2003 and 1.7 percent in 2004, were inconclusive. In 2003, 8.9 percent of nonproduction/supervisory employees brought work home and were not paid overtime and were paid salary; this decreased to 5.4 percent in The average weekly hours of those employees were 10.9 percent greater than the average for all nonproduction/supervisory employees in 2003 and 14.2 percent greater in However, only 15.8 percent of their hours were worked outside the workplace. It is clear that those nonproduction/ supervisory employees that bring work home and are not paid overtime and are paid a salary are not reallocating work across locations but are truly working longer hours. Because BLS uses CPS data to adjust the hours worked for nonproduction/supervisory employees, a portion of these data may be captured in the existing measures. Nonproduction/supervisory employees that bring work home and are not paid overtime and are paid a salary account for 1.6 percent of nonfarm workers in 2003 and 0.9 percent of workers in 20

22 2004. As an upper bound, the hours of only 3 percent of nonfarm business sector workers may be mismeasured due to unpaid hours worked in 2003, and 2.6 percent in In addition to the future research mentioned previously in this paper, we intend to examine the time diaries for evidence of more and/or longer breaks in the work day for those who bring some work home. The 2004 data may show more breaks overall, since the interviewers were automatically prompted to ask for breaks of more than 15 minutes during work hours. This examination will help us to determine whether workers are doing work at home to make up for missed work time at the workplace. 21

23 REFERENCES Callister, Paul and Sylvia Dixon. New Zealanders Work Time and Home Work Patterns: Evidence from the Time Use Survey New Zealand Department of Labour. Occasional Paper 2001/5. Eldridge, Lucy P., Marilyn E. Manser and Phyllis F. Otto, Alternative Measures of Supervisory Employee Hours and Productivity Growth, Monthly Labor Review, April 2004, pp Eldridge, Lucy P. Hours Measures for Productivity Measurement and National presented to Paris Group on Measuring Hours of Work, September Accounting, Frazis, Harley and Jay Stewart. What Can Time-Use Data Tell Us About Hours of Work? Monthly Labor Review, December 2004, pp Hamermesh, Daniel S. Shirking or Productive Schmoozing: Wages and the Allocation of Time at Work. Industrial and Labor Relations Review, 1990, vol. 43, no. 3, pp. 121S- 133S. Jacobs, Jerry A. Measuring Time at Work: Are Self-reports Accurate? Monthly Labor Review, December 1998, pp Murphy, Bridget and Paul Satherley. New Zealand Time Use Survey 1999 Key Statistics 2000, p Oettinger, Gerald. The Growth in Home-Based Wage and Salary Employment in the States, : How Much and Why? Presentation at the Society of Labor Economist Meetings United Pabilonia, Sabrina Wulff. Evidence on Youth Employment, Earnings, and Parental Transfers in the National Longitudinal Survey of Youth 1997, Journal of Human Resources, 36(4) Pabilonia, Sabrina Wulff. Working at Home: An Analysis of Telecommuting in Canada, Working Paper Polivka, Anne E. and Jennifer M. Rothgeb. Redesigning the CPS Questionnaire. Monthly Labor Review, September 1993, pp Roach, Stephen S. The Boom for Whom: Revisiting America s Technology Paradox. Morgan Stanley Dean Witter, Special Economic Study, January 9, Robinson, John, and Ann Bostrom. The Overestimated Workweek? What Time Diary Measures Suggest. Monthly Labor Review, 1994, vol. 117, no. 8, pp

24 Schroeder, Christine and Ronald S. Warren, The Effect of Home-Based Work on Earnings Unpublished manuscript

25 Table 1: Percent of Workers Working by Location of Work Activities Nonfarm Worker Sample Nonfarm Business Employee Sample Work exclusively in the workplace Work exclusively away from workplace Bring work home Number of respondents 4,014 2,652 3,444 2,263 Table 2. Proportion of Nonfarm Workers Who Bring Work Home Working During Different Times of the Day at Home, Weekday Only Time of Day AM-6AM AM-8AM AM-10AM AM-12PM PM-2PM PM-4PM PM-6PM PM-8PM PM-10PM PM-12AM Number of observations Proportions are weighted to account for sampling design. Numbers are rounded and do not sum to 100 because a worker could be working in more than one time period. Table 3. Proportion of Nonfarm Workers by Minutes Worked at Home, Weekday Only Minutes per day < Number of observations Proportions are weighted to account for sampling design. Numbers are rounded. 24

26 Table 4. Means and Proportions of 2003 ATUS Respondents, by Bring-Work-Home Status and Sample Type Nonfarm Worker Sample bring work home Nonfarm Worker Sample do not bring work home Nonfarm Business Employee Sample bring work home Nonfarm Business Employee Sample do not bring work home Female Age (0.70) (0.24) (0.90) Black Other race Hispanic Married Divorced Spouse works Multiple job holder High school dropout High school degree Some college Bachelor s degree Advanced degree Paid hourly Any child Infant Preschooler Elementary student Adolescent (0.28) Management and professional occupation Service occupation Sales and office occupations Farming, fishing, and forestry occupations Construction and maintenance occupations Production, transportation, and material moving occupations Number of Observations 498 4, ,230 Note: Sampling weights are used to account for survey design. Standard errors in parentheses. 25

27 Table 4 Continued. Means and Proportions of 2003 ATUS Respondents, by Bring- Work-Home Status and Sample Type Nonfarm Worker Sample bring work home Nonfarm Worker Sample do not bring work home Nonfarm Business Employee Sample bring work home Nonfarm Business Employee Sample do not bring work home Agriculture, forestry, fishing and hunting Mining Construction Manufacturing Wholesale and retail trade Transportation and utilities Information Financial activities Professional and business services Educational and health services Leisure and hospitality Other services Public administration Number of Observations 498 4, ,230 Note: Sampling weights are used to account for survey design. Standard errors in parentheses. 26

28 Table 5. Means and Proportions of 2004 ATUS Respondents, by Bring-Work-Home Status and Sample Type Nonfarm Worker Sample bring work home Nonfarm Worker Sample do not bring work home Nonfarm Business Employee Sample bring work home Nonfarm Business Employee Sample do not bring work home Female Age (0.83) (0.28) (1.04) Black Other race Hispanic Married Divorced Spouse works Multiple job holder High school dropout High school degree Some college Bachelor s degree Advanced degree Paid hourly Any child Infant Preschooler Elementary student Adolescent (0.34) Management and professional occupation Service occupation Sales and office occupations Farming, fishing, and forestry occupations Construction and maintenance occupations Production, transportation, and material moving occupations Number of Observations 333 3, ,096 Note: Sampling weights are used to account for survey design. Standard errors in parentheses. 27

29 Table 5 Continued. Means and Proportions of 2004 ATUS Respondents, by Bring- Work-Home Status and Sample Type Nonfarm Worker Sample bring work home Nonfarm Worker Sample do not bring work home Nonfarm Business Employee Sample bring work home Nonfarm Business Employee Sample do not bring work home Agriculture, forestry, fishing and hunting Mining Construction Manufacturing Wholesale and retail trade Transportation and utilities Information Financial activities Professional and business services Educational and health services Leisure and hospitality Other services Public administration Number of Observations 333 3, ,096 Note: Sampling weights are used to account for survey design. Standard errors in parentheses. 28

30 Table 6. Effect of Selected Covariates on the Probability of Bringing Work Home from the Office, Weekday Only Nonfarm Business Nonfarm Worker Sample Employee Sample Female (0.009) (0.011) Age (0.002) (0.002) Age squared/ (0.017) (0.021) Black *** (0.009) *** (0.010) Other race (0.012) (0.016) Hispanic *** (0.009) *** (0.009) Married (0.009) (0.012) Divorced (0.011) (0.013) Spouse works (0.007) (0.009) High school degree 0.038** (0.018) (0.018) Some college 0.054*** (0.020) 0.038* (0.021) Bachelor's degree 0.083*** (0.025) 0.068*** (0.026) Advanced degree 0.128*** (0.033) 0.110*** (0.037) Part time (0.009) (0.011) Paid hourly *** (0.007) *** (0.008) Infant 0.032* (0.020) 0.045* (0.024) Infant * Female *** (0.011) ** (0.015) Preschooler 0.026* (0.014) 0.029* (0.017) Preschooler * Female (0.014) (0.017) Elementary student (0.014) (0.017) Elementary * Female * (0.012) (0.019) Adolescent (0.012) (0.014) Adolescent * Female (0.016) (0.024) F-statistic Number of observations 8,812 5,748 Notes: All regressions include region, occupation, industry, and year indicators as well as a constant. Standard errors are in parentheses. Significance levels: * = p<.10; ** = p<.05; *** = p<