No. 16–29 Who Counts as Employed? Informal Work, Employment Status, and Labor Market Slack Anat Bracha and Mary A. Burke Abstract Several recent studies find that as of 2015, a significant share of working-age adults in the United States participates in nonstandard work arrangements. Such arrangements tend to lack long-term employment contracts and are often referred to as “gig economy” jobs. This paper investigates the implications of nonstandard or “informal” work for the measurement of employment status and labor market slack. Using original survey data, we find that as of 2015 roughly 37 percent of nonretired U.S. adults participated in some type of informal work, and roughly 20 percent participated in informal income- generating activities that did not exclusively involve renting out their own property or selling their own goods. The survey also elicits an individual’s employment status according to the definitions of the Bureau of Labor Statistics (BLS). While not the majority, a significant share of those who engage in informal work are classified as not being in the labor force; if all informal workers were counted as employed, the U.S. labor force participation rate (as of 2015) would have been 2 percentage points higher. In addition, individuals who are classified as working “part-time for economic reasons”—those who would like a full-time job but cannot obtain full-time hours—have the highest participation rate in informal work and the highest average hours per month. This latter finding suggests that informal work embodies labor market slack, and we offer several pieces of evidence that support the thesis that workers engage in informal work as a way to compensate for weak labor demand and may therefore drop informal work as formal labor market conditions improve. To estimate the amount of labor market slack embodied in informal work, we convert the total hours of informal work performed by those classified as employed part-time into a number of full-time job equivalents. This exercise yields a figure that ranges from roughly 275,000 to roughly 400,000, depending on the specifics of the calculation. At the same time, we point out that a significant share of informal work hours offer higher wages than what the same individuals earn in their formal jobs. Therefore, formal wages may need to increase by a relatively large margin moving forward in order to attract additional labor into the formal sector. JEL Classifications: J21, J22, J46, E26 Keywords: informal work, gig economy, BLS employment status, labor market slack Anat Bracha and Mary A. Burke are senior economists in the research department at the Federal Reserve Bank of Boston. Their e-mail addresses are [email protected] and [email protected]. This paper presents preliminary analysis and results intended to stimulate discussion and critical comment. The views expressed herein are those of the authors and do not indicate concurrence by the Federal Reserve Bank of Boston, or by the principals of the Board of Governors, or the Federal Reserve System. This paper, which may be revised, is available on the web site of the Federal Reserve Bank of Boston at http://www.bostonfed.org/economic/wp/index.htm. This version: December, 2016 1.Introduction A handful of recent papers and reports use survey responses to measure participation in nonstandard work arrangements in the United States; these jobs are often referred to as informal work, independent work, contract work, contingent work, on-demand work, and/or gig work (Bracha, Burke, and Khachiyan 2015; Katz and Krueger 2016; McKinsey Group International 2016; U.S. Government Accountability Office 2015). While the survey methods differ across studies, all find that as of 2015 among working-age adults in the United States, a significant share participates in nonstandard work arrangements. This paper investigates the implications of informal work for the measurement of employment status and labor market slack. We use original survey data (from the SCE-SIWP 2014–20151) to compare an individual’s employment status based on the official Bureau of Labor Statistics (BLS) definitions to her employment status as revealed by participation in paid informal work activities. We use these findings to consider whether the official BLS estimates may understate U.S. labor force participation and to consider the extent to which informal work embodies labor market slack. Our key findings are as follows: • Among our analysis sample—survey respondents 21 years of age and older who are not retired— 37 percent participate in paid informal work (not including survey work). When we remove those who engage exclusively in informal renting and selling activities, we find that 20 percent of the respondents engage in informal work. • Among individuals classified by the BLS as employed on a part-time basis, the total amount of informal work performed represents an estimated 403,174 full-time job equivalents. Among those classified as employed part-time for economic reasons, informal work represents 320,992 full- time job equivalents (FTEs). • If all informal workers were classified as employed, as of 2015 the employment-to-population ratio would have been higher by 2.5 percentage points and the labor force participation rate would have been just over 2 percentage points higher. • If informal workers with at least 20 hours of informal work per week were counted as employed, both the employment-to-population ratio and the labor force participation rate would increase by between 0.5 and 1 percentage point. • In order to better capture informal work activities in official employment statistics, employment surveys might ask about work performed in the previous month in addition to work performed in the previous week, because informal work schedules, for example, may be concentrated within one or two weeks of the month rather than taking place each week. • While not all informal work represents labor market slack, the extent of work that is not captured by the BLS definitions is significant. Among the informal participants who experienced a job loss or other economic loss during or after the Great Recession, 40 percent report engaging in informal work out of economic necessity, and 8.5 percent of all informal workers report that they 1 Survey of Informal Work Participation within the Survey of Consumer Expectations, © 2013 Federal Reserve Bank of New York (FRBNY). The SCE-SIWP data are available without charge at http://www.newyorkfed.org/microeconomics/sce and may be used subject to the license terms posted there. The FRBNY and the Federal Reserve Bank of Boston disclaim any responsibility or legal liability for the analysis and interpretation of SCE-SIWP data. 1 would like to have a formal job. Both of these facts suggest that such individuals would take a formal job if offered one. 2.Background and Related Literature The visibility of nonstandard work arrangements has received a significant boost from the rise of Uber, Lyft, AirBnB, and TaskRabbit, among others. These firms invented mobile platforms that facilitate on- demand work that eschews formal employment contracts. Gig economy work falls within a broad set of alternative work arrangements that go beyond just Internet-mediated work, and encompasses work that departs from the standard employer-employee relationship involving regular hours, a fixed wage or salary, benefits, and a long-term employment contract.2 A growing body of evidence finds that the prevalence of these alternative work arrangements has increased since 2005. For example, Bracha, Burke, and Khachiyan (2015) find that as of early 2015, roughly 55 percent of Americans had engaged in informal work during the previous two years, and this figure represented an increase over the comparable figure of 40 percent from late 2013; Katz and Krueger (2016) find that the share of contingent workers— which includes temporary help agency workers, on-call workers, contract workers, and freelancers— increased from roughly 11 percent to almost 16 percent between 2005 and 2015 (despite being flat between 1995 and 2005). A study by McKinsey Global Institute (2016) finds that 27 percent of working- age adults in the United States are engaged in “independent work,” which this study defines as entailing high worker autonomy, payment by task rather than by hour or salary, and short-term relationships between workers and clients. The U.S. Government Accountability Office (2015) reports that, as of 2010, about 40 percent of employed workers could be characterized as “contingent” according to a very broad definition of that term, and roughly 8 percent of employed workers could be characterized as “core contingent workers” when using a narrower concept of contingent work.3 The growing prevalence of participation in nonstandard forms of work raises numerous questions, such as whether workers are better off under such arrangements than under more standard forms of work (by mid- to late-twentieth century norms at least). This paper discusses the implications of informal work participation for the measurement of labor market slack and the measurement of employment status in the United States. According to the Bureau of Labor Statistics’ Current Population Survey (the CPS or what is also referred to as the “household employment survey”), an individual is considered to be employed if he or she did any work for pay or profit in the previous week.4 There are a few reasons why those 2 This standard represents an obvious simplification, and furthermore this standard only became dominant in the second half of the twentieth century. 3 See Bracha, Burke, and Khachiyan (2015) for a detailed analysis of the changes in informal work in the United States between December 2013 and January 2015. 4 Someone who did not work for pay or profit in the previous week might still get counted as employed, for example if they were on a temporary leave from a job. 2 informal workers who are not also employed in a formal job might be less likely to count as employed based on this criterion: their work may follow an irregular or intermittent schedule, such individuals may not believe that such informal activities “count” as work (see Abraham et al. 2015), or they may not want to report that they are working for reasons related to taxes or immigration status. Given these several possibilities, an increase in nonstandard work arrangements might result in decreased estimates of the employment-to-population ratio and the labor force participation rate, even with no change in the actual number of people engaged in some type of work. Among those informal workers who also have a formal job, their total hours might be understated to the extent that their informal hours are underreported, and thus the aggregate share of workers employed on a full-time basis might be understated. On the other hand, to the extent that informal workers do get counted as employed in the official BLS statistics, increases in employment that are driven by increases in informal work might be valued differently than similar increases in formal employment, just as part-time and temporary payroll employment are valued differently from full-time or long-term payroll employment. Finally, measures of informal work participation may offer additional insights into the extent of labor market slack in the U.S. economy. 3. Defining Informal Work and Measuring Informal Work Participation Anat Bracha, Mary Burke, and Arman Khachiyan designed the Survey of Informal Work Participation (SCE-SIWP), administered within the Federal Reserve Bank of New York’s Survey of Consumer Expectations (SCE). The survey was fielded in December 2013 (SCE-SIWP 1), January 2015 (SCE-SIWP 2), and December 2015 (SCE-SIWP 3) on three separate and nationally representative samples. In an earlier paper based on SCE-SIWP 1 and 2, Bracha, Burke, and Khachiyan (2015) reported that the informal work participation rate increased from an estimated 40 percent of nonretired workers 21 years of age and older in SCE-SIWP 1 to roughly 55 percent of the corresponding population in SCE- SIWP 2. In that earlier study, an informal participant was defined as someone who (1) reported that they engaged in some type of informal work (other than paid survey completion5) during the previous two years,6 and (2) reported nonzero “typical hours per month” of paid informal work during the previous two-year period. In addition to asking whether individuals had engaged in paid informal work during the preceding two years, our three surveys also asked whether individuals were currently engaged in paid informal 5 We exclude paid survey work because all of our survey respondents would qualify as informal participants simply by participating in our survey. 6 Individuals were presented with a list of examples of informal work, such as babysitting, dog-walking, and driving for a ride- sharing service, as well as a write-in option to indicate some other type of informal work. An individual had to check at least one item on the list, including the “other” option, to potentially be considered an informal worker. For the complete text of the questions, see the Appendix. 3 work, again based on a checklist of specific activities. In SCE-SIWP 2 and 3 only, a respondent who indicated that he or she was currently engaged in informal work was asked—separately for each type of informal work selected from the list of current activities—to quantify his/her typical hours and earnings per month in the given activity.7 The responses to the questions about current informal work participation naturally offer a more accurate measure of the extent of informal work participation at a given point in time than do responses to the questions about participation in the preceding two years; the same responses have the added benefit of containing information on earnings and hours by type of informal work. However, a snapshot of the informal participation rate within a specific month may reflect seasonal labor market influences to a greater extent than would an informal participation rate based on activities performed at any time in the preceding two-year period. Consistent with monthly seasonal patterns in payroll employment, we find that informal participation rates based on current engagement are lower in SCE-SIWP 2, fielded in the month of January, than for either SCE-SIWP 1 or SCE-SIWP 3, both of which were fielded in the month of December. For our present purposes, we believe that the benefits of defining informal participation based on current engagement outweigh the drawbacks. Hereafter in this paper, a survey respondent is defined as an informal participant if the individual (1) reports being “currently engaged” in some type of informal work (not including the exceptions described below), and (2) across the informal work activities in which the individual reports being currently engaged, the total number of paid informal hours is strictly positive (not including hours spent on excluded activities, described below). To minimize the seasonality problem, we combine the (weighted) responses from SCE-SIWP 2 and 3 and report the average participation rates between the surveys. As a result, our estimates represent an average of the “current” participation rate (or current hours, earnings, and so on) between January 2015 and December 2015.8 To focus on alternative paid work activities that require significant labor input and that do not rely on owning valuable assets, we do not count someone as an informal participant if their informal work consisted exclusively of renting out their own property and/or selling their own used goods.9 Consistent with this decision, average informal work hours are lower among individuals who engage exclusively in renting and selling than among those who engage in other activities. We continue to exclude those who engaged only in paid survey work, and again we restrict the sample to nonretired individuals 21 years of 7 In choosing to use these questions from SCE-SIWP 2 and 3 to help define informal work participation and to measure these hours and earnings, we drop SCE-SIWP 1 from most of the current analysis because comparable questions are not available in that survey. 8 Due to the potential seasonality issue in the estimates of the current informal work participation rates, we do not make pronouncements about changes in participation between SCE-SIWP 2 and SCE-SIWP 3. 9 Among those respondents who engaged in informal labor, such as babysitting, in addition to renting property and/or selling their own used goods, hours and earnings for the renting/selling tasks are excluded from our calculations of the individual’s total informal hours and total informal earnings. Such exclusions are not possible in Survey 1, which did not ask respondents to report hours and earnings separately by task. 4 age and older.10 Based on the differences between the definition of informal participation employed in the present study and the definition used in Bracha, Burke, and Khachiyan (2015), the estimates of informal participation rates are not comparable between these two studies. Nonetheless, all the previous results remain accurate based on the definitions employed at the time. In addition to asking questions about participation in informal work activities, the survey collected basic demographic information about the respondents and their households. A subset of questions taken from the CPS enable us to determine an individual’s detailed employment status according to the official definitions used by the BLS.11 Based on the answers to these questions we calculate the labor force participation rate, the unemployment rate, and the employment-to-population ratio within our unrestricted survey sample (N=2,233), taking averages between SCE-SIWP 2 and 3. As shown in Figure 1, the employment statistics for our survey population are comparable to the BLS’s published employment statistics for the United States for the same time period (taking an average of the employment statistics from January 2015 and December 2015).12 The 2015 labor force participation rate in our sample is 2.8 percentage points higher than the BLS’s official U.S. participation rate, but the 95 percent confidence interval for our estimate includes (at the lower bound) a value that almost matches the official U.S. rate. Similarly, the employment-to-population ratio for our sample is somewhat higher than the U.S. rate but most likely does not differ significantly from it, and our unemployment rate is slightly lower than the U.S. rate.13 These results indicate that our sample is reasonably representative of the U.S. population with regard to employment status. For the purposes of analyzing informal work participation, we restrict the samples from SCE- SIWP 2 and SCE-SIWP 3 to respondents 21 years of age and older who are not retired, as we are primarily interested in the activities of potential candidates for participation in the formal U.S. labor market. We also exclude a small number of respondents who reported individual incomes (from a formal job) of $600,000 per year or more, as well as those with missing values for the responses of interest. The resulting analysis sample consists of a total of 1,377 respondents—690 from SCE-SIWP 2 and 687 from SCE-SIWP 3. Table 1 describes some selected characteristics of the analysis sample, combining the results from SCE-SIWP 2 and SCE-SIWP 3.14 Unlike the unrestricted sample, the analysis sample is not 10 Additional sample restrictions are described below. 11 We did not include all the questions related to employment status that are included in the CPS household survey. See the Appendix for the set of questions used to determine BLS employment status. 12 Our survey data are not seasonally adjusted. Therefore we use nonseasonally adjusted BLS data to calculate the comparable employment statistics for the United States. 13 We do not have confidence intervals for the U.S. employment statistics and therefore we cannot test formally for whether our estimates differ significantly from the U.S. labor market statistics generated by the BLS. 14 The statistics from the combined surveys are calculated by taking the weighted mean (or other weighted statistic) within each survey and then taking the simple average of the statistics from the two surveys. In this way, each survey is given equal weight in the calculation of the combined statistics. The weights for a given survey are designed so that a selected set of the weighted sample characteristics match those of the Current Population Survey for the same time period. 5 representative of the overall U.S. population in terms of average employment status because we select only nonretired individuals who are 21 years of age and older. In addition, the analysis sample population has a relatively high level of educational attainment compared with the U.S. population, and a slightly higher average income compared with total population estimates.15 Within the analysis sample, roughly 37 percent of the respondents are “currently engaged” in some type of paid informal work, not including survey work.16 Excluding those who engage only in informal renting and selling activities, roughly 20 percent of the analysis sample (241 individuals) are then classified as informal work participants.17 (From this point forward, the discussion of “informal work participants” will refer only to this latter group.) As shown in Figure 2, Panel A, the informal labor force participation rate is significant within all the BLS employment categories, including those classified as being not in the labor force (NILF), and yet the informal participation rate is the highest for those who are classified as being employed part-time for economic reasons (PTER). To give a conservative estimate—based on the lower bound of the 95 percent confidence interval—at least 30 percent of those classified as PTER participate in informal work based on our definition. As shown in Figure 2, Panel B, monthly income from informal work—calculated among the subset of informal participants only—is statistically and economically significant across all employment classes. Across all groups, the average monthly informal earnings amount to $405. The average informal earnings per month appears somewhat greater among employed individuals compared with those who are either unemployed or not in the labor force. Among employed individuals, those employed part-time have higher average earnings than those employed full-time, and those classified as PTER may have higher earnings than other part-time employees. However, given that some of these estimates are imprecise, we cannot be confident that the earnings from informal paid work are significantly different across the different employment classes. 4. Does Informal Work Represent Labor Market Slack? Participation in informal work raises the question: what is labor market slack? Defining slack as the available labor hours that are not utilized in the formal labor market, the number of informal work hours may be indicative of the amount of slack that exists in the formal labor market. To determine whether there is indeed substitution between formal and informal hours—in which case informal hours would represent slack—we examine the relationship between changes in informal work (participation and informal hours) and formal labor market conditions for different demographic groups and for two broad 15 The population estimates refer to 2015 Current Population Survey, Table PINC-04. 16 Consistent with the definition given above, we also require that respondents indicated nonzero hours per month in at least one informal activity other than survey work. 17 Again, these statistics represent simple averages of the weighted statistics from SCE-SIWP 2 and 3, respectively. 6 groups of states that differ in their formal labor market conditions. We also examine informal work patterns among part-time workers, especially those who are working part-time for economic reasons, because the informal work performed by these individuals may be more likely to represent labor market slack than the informal work performed by members of some other employment groups, such as those who are already employed full-time or those who are not in the labor force and are not interested in finding work. We then attempt to quantify how much slack may be embodied in the informal work performed by part-time workers. This estimate is done by calculating the full-time job equivalent of the informal work hours reported by part-time workers. On balance, our results suggest that informal work does embody a significant amount of labor market slack. In particular, we find that formal labor market conditions are negatively correlated with informal work hours, and those who are employed part-time for economic reasons are especially likely to turn to informal work, as might be expected if informal hours represent a substitute for formal work. We find that most informal workers (regardless of their formal work status) engage in this work for the money; informal workers who lost a job or experienced other negative economic fallout from the Great Recession state directly that their decision to engage in informal work represented a response to those economic shocks. The total number of informal work hours performed by all part-time workers represents 403,174 full-time job equivalents (FTEs), and the total informal hours performed by those who are working part-time for economic reasons amounts to 320,992 FTEs. Considering the recent payroll employment growth estimates from the BLS, these numbers represent between two and three months’ worth of nonfarm payroll employment gains. However, since the official job growth figures do not differentiate between full-time and part-time jobs, the average payroll job represents less than one FTE. In the next section we show that the average earnings from informal work are substantial for all the BLS employment groups. 4.1. Formal and Informal Labor Substitution If informal work serves as a substitute for formal or traditional employment, informal work participation and informal work hours should vary negatively with labor demand in the formal sector. Our survey data do not allow us to observe directly whether a given individual substitutes away from informal work as formal labor demand improves over time (or vice-versa). Instead, we test for substitution between formal and informal work within aggregated groups of individuals, as the labor market conditions specific to those groups change over time. Separately, we analyze patterns in informal work participation among individuals who work part-time, especially those who are working part-time for economic reasons, including the relationship between the formal and informal wages of such workers. Finally, the survey 7 contains suggestive evidence on the individual level related to the substitutability of informal and formal work, which we describe below. 4.1.1. Group-Level Analysis State Groups To link formal labor market conditions to informal work behavior, we first define two separate measures of a state’s formal labor market conditions as of 2015 (averaging between January 2015 and December 2015). The first measure is the state’s unemployment rate gap in 2015, defined as the difference between its 2015 unemployment rate and its average unemployment rate over the 2005–2007 period. (The latter measure is a proxy for the state’s natural rate of unemployment.) The second measure will be referred to as the state’s “PTER gap” in 2015, which is defined as the difference between the state’s share of employed workers classified as part-time for economic reasons as of 2015 and the average value of this share in the state for the 2005–2007 period. Given these measures, we first split states into a “high” unemployment gap group and a “low” unemployment gap group, where the high-gap states have zero or positive unemployment gaps and the low-gap states have strictly negative unemployment gaps. Separately, we partition states into a “high” PTER gap group and a “low” PTER gap group, where the PTER gaps are greater than 1 in the high-PTER-gap states and less than or equal to 1 in the low-PTER- gap states.18 For a given partition—for example, between those states with high unemployment gaps and those states with low unemployment gaps—we place each observation in our survey sample into either the high-gap group or the low-gap group based on each respondent’s state of residence. For both groups we then calculate the change in the average informal work participation rate and the change in average informal hours between January 2015 and December 2015. We proceed analogously in the case of the partition based on the PTER gaps. We find, consistent with the assertion of substitutability between formal and informal work, that the “low” unemployment gap states (meaning those with better formal labor market conditions) experienced smaller changes in both informal work participation and informal hours compared with the “high” unemployment gap states. As shown in Table 2, the informal participation rate actually declined in the low unemployment gap states and increased in the high unemployment gap states. The average informal hours increased in both groups of states, but such hours increased by only a very slight margin in the low-gap states and increased by a much larger amount in the high-gap states. When we compare states with high PTER gaps to those with low PTER gaps, again we find that between January and December 2015, the low-gap states experienced smaller changes in informal work 18 A list of the states that belong to each group—high unemployment gap, low unemployment gap, high PTER gap, low PTER gap—is provided in the Appendix. 8 participation and informal hours compared with the high-gap states. As shown in Table 2, the informal participation rate increased in both groups of states, but increased by a larger margin in the group of states with high PTER gaps, while informal hours decreased by a small amount in the low PTER gap group and increased significantly in the high PTER gap group. Demographic Groups In this subsection we use the variability in formal labor market conditions across demographic groups within our analysis sample to examine how informal work responds to business cycle fluctuations. First we divide the analysis sample into 16 demographic groups based on combinations of gender, education level (split between those with at least some college and those with no college), and age group (25–39 years, 40–54 years, 55–64 years, and 65 years and older).19 Based on its small sample size, we drop the group of females 65 years of age and older with at least some college. Then, using regression analysis we relate group-level changes in informal work outcomes between January 2015 and December 2015 to contemporaneous changes as well as one-year lagged changes in formal labor market conditions within the same groups.20, 21 The informal work outcomes examined are the informal participation rate by group and (separately) the average informal work hours by group. Several measures of formal job market conditions are included as explanatory variables: (1) the percent change between January 2014 and December 2014 in the share of workers (by demographic group) that are classified as working part-time for economic reasons, (2) the percent change in the unemployment rate within each demographic group between January and December 2014, and (3) the percent change by demographic group in average “formal” hours (across all individuals regardless of employment status), again between January 2014 and December 2014. These three variables are constructed using the monthly micro-data from the BLS’s Current Population Survey (Flood et al. 2015) and are not seasonally adjusted. Each pertains to a one-year lag relative to the time period over which the dependent variables are measured, since informal work outcomes are likely to respond to changes in formal labor market conditions with a lag, and because contemporaneous changes in labor market 19 We omit those who are 21–24 years old from the regression analysis because individuals in that age range tend to have different formal labor market behavior than those who are 25–39 years old, and our sample size of those who are 21–24 years of age (especially when subdivided by sex and education) is too small to form a separate group. 20 The reader may be concerned about any seasonal changes in informal participation between January 2015 and December 2015. Assuming that seasonality affected all of our demographic groups similarly in terms of percent changes in informal work habits between those dates, the common seasonal component of the dependent variables should be captured in the constant term of the regression. Even if seasonal effects differ across our demographic groups, the fact that we use nonseasonally adjusted data to construct our explanatory variables helps to control for that possibility. 21 The three waves of our survey give rise to two changes in (average) informal work outcomes for each age-education-sex demographic group—from December 2013 to January 2015, and from January 2015 to December 2015. Based on the aforementioned changes in the survey design over time, some of the relevant explanatory variables cannot be constructed for the period between the first two survey waves. Therefore we can only make use of the group-level changes in informal work for the period between January 2015 and December 2015. 9