U.S. Bureau of Labor Statistics

April 2015

Multiple jobholding over the past two decades

Etienne Lalé
etienne.lale@bristol.ac.uk

Etienne Lalé is an assistant professor in the Department of Economics, University of Bristol, Bristol, England, United Kingdom.

Multiple jobholding declined in the United States during the past two decades. Data from the Current Population Survey show that the trend reflects a lower propensity to moonlight among single jobholders. Multiple jobholders, by contrast, did not become more likely to return to single jobholding.

In 2013, 6.8 million workers in the United States held more than one job. Twenty years before, the figure was 7.5 million, although the total number of workers with a job was lower by 15.9 million. The multiple-jobholding rate—the proportion of multiple jobholders among all employed workers—rose from 6.2 percent in 1994 to a high of 6.8 percent during the summer of 1995. It has declined steadily since then and was at 5.0 percent by the end of 2013. Inspection of data from the Current Population Survey (CPS, see accompanying box) reveals that the downward trend holds across various sociodemographic groups of the working-age population (those 16 to 64 years old).

The Current Population Survey: data and definition

The Current Population Survey (CPS) is a monthly survey of households administered by the U.S. Census Bureau under the auspices of the U.S. Bureau of Labor Statistics. The CPS is the source of U.S. official labor market statistics and is thus a firsthand source of information for the purpose of this article.

Since January 1994, CPS respondents have been asked a series of questions that allows the identification of multiple jobholders. A multiple jobholder is defined as an individual who held more than one job during the reference week of the survey and who usually receives a wage or salary from his or her primary job (i.e., the job at which the individual worked the greatest number of hours during the reference week). Excluded from the definition are individuals who were unpaid family workers on their primary job as well as individuals who were self-employed on their primary job and were either self-employed or unpaid family workers on their second job.

CPS respondents are interviewed for 4 consecutive months, are rotated out of the survey for 8 months, and are then included in the survey again for 4 consecutive months. This approach allows researchers to link (a fraction of) CPS respondents across surveys. In so doing, one can compare the labor market position of an individual in 2 consecutive months and identify transitions into and out of multiple jobholding, as is done later in the text of this article.

This article documents the evolution of multiple jobholding from 1994 to 2013 in the United States, shining a light on workers’ transitions into and out of multiple jobholding. These transitions convey information about the propensity of single jobholders to become multiple jobholders and vice versa, trends that in turn help explain why moonlighting has become less common.

Multiple jobholding is relevant to our understanding of the labor market from a variety of perspectives. At the individual level, moonlighting serves both economic and noneconomic purposes, as earlier studies have shown.1 In May 2004, for instance, most workers who were holding more than one job reported doing so in order to earn extra money (38.1 percent), to meet expenses, or to pay off debt (25.6 percent).2 Meanwhile, a nonnegligible fraction of multiple jobholders (17.6 percent) reported that their primary motivation was the enjoyment they received from their second job. Multiple jobholding is also important from a macroeconomic perspective because moonlighting adds millions of jobs to the economy, as the aforementioned figures show. In 1995, for example, when the multiple-jobholding rate was at its highest level, multiple jobholders worked an average of 13.5 hours per week on their second job, thereby adding about a 100 million hours worked to the economy each week.

Trends in multiple jobholding

Figure 1 shows the evolution of the multiple-jobholding rate in the population of working-age individuals.3 As already mentioned, multiple jobholding has been declining throughout almost the entire period examined. Interestingly, earlier studies based on the May supplements of the CPS reported evidence of a rise in multiple jobholding during the 1980s.4 Therefore, the increase between 1994 and 1995 shown in the figure may well have been the continuation of an earlier trend. During the summer of 1995, however, this trend came to an end, and by 2013 the multiple-jobholding rate was at a decadal low. In figure 1 and in subsequent figures, the recession periods shown are those identified by the National Bureau of Economic Research. Previous research has attempted to correlate moonlighting with the different phases of the business cycle.5 For instance, on the one hand, jobs are seen to be more plentiful during economic expansions, a finding that could result in procyclical multiple-jobholding rates. On the other hand, looser credit constraints during expansions could lead to fewer individuals taking on a second job to meet expenses. Perhaps because of these opposing forces, research on the cyclical behavior of multiple jobholding has not reached definite conclusions. The figures in this article confirm the absence of a clear association between moonlighting and the business cycle.

LaleFig1

Trends within sociodemographic groups

Figure 2 (a through d) depicts the evolution of multiple jobholding within various sociodemographic groups of the working-age population. To provide the reader with the underlying numbers, table 1 further reports decadal averages and average annual changes in the number of multiple jobholders and of multiple jobholding rates, tabulated both in the aggregate and within those groups.

LaleFig2a

LaleFig2b

LaleFig2c

LaleFig2d

Table 1. Multiple-jobholding levels and rates, aggregate and within different sociodemographic groups, 1994–2013
Table 1. Multiple-jobholding levels and rates, aggregate and within different sociodemographic groups, 1994–2013
Demographic category 1994–2003 (average) 2004–2013 (average)
Number Change Rate Change Number Change Rate Change
 
 
 
 
 
 
 
 
 

Total

7,736 -22 6.18 -0.09 7,223 -46 5.39 -0.05

Gender:

Men

4,095 -40 6.12 -.12 3,606 -34 5.06 -.06

Women

3,639 18 6.26 -.05 3,614 -14 5.76 -.04

Age:

16 to 24

1,121 -18 5.86 -.11 970 -9 5.28 .00

25 to 54

5,911 -40 6.35 -.10 5,180 -69 5.44 -.06

55 to 64

703 36 5.40 .04 1,070 30 5.25 -.04

Education:

Less than high school

483 -6 3.12 -.05 356 -11 2.61 -.01

High school graduate

1,973 -31 4.98 -.08 1,576 -37 4.13 -.06

Some college

2,196 -20 7.22 -.13 2,041 -5 6.21 -.06

College or higher

2,647 31 7.80 -.13 2,819 10 6.59 -.10

Marital status:

Married

4,432 -19 6.01 -.08 3,984 -58 5.21 -.06

Widowed, divorced, or separated

1,228 3 6.99 -.07 1,158 -10 6.16 -.07

Single

2,080 -7 6.14 -.12 2,081 20 5.37 -.02

Note: Levels are reported in thousands; rates are reported in percent. Annual changes are calculated by averaging year-to-year changes over the corresponding period. Levels may not sum to totals because of rounding.

Source: U.S. Bureau of Labor Statistics.

Note: Levels are reported in thousands; rates are reported in percent. Annual changes are calculated by averaging year-to-year changes over the corresponding period. Levels may not sum to totals because of rounding.

Source: U.S. Bureau of Labor Statistics.

Multiple jobholding has become less common for both men and women, although the downward trend has been more pronounced among men. Multiple-jobholding rates among women held almost steady between the 2001 recession and the 2007–2009 recession. Since then, women’s multiple-jobholding rates have been diminishing slightly. On average over the 2004–2013 period, male and female workers had about the same number of multiple jobholders (3.6 million; see table 1).

Twenty years ago, moonlighting was less common among older workers (ages 55 to 64). However, because multiple-jobholding rates have remained fairly stable among these workers and have declined within the other age groups of the population, the proportion of workers who moonlight became uniform across younger (16 to 24), prime-age (25 to 54), and older workers by the end of the period. Further inspection of data on hours worked shows that prime-age workers are more likely to work full time (at least 35 hours per week) at their primary job and part time at their second job, thus having a longer workweek relative to that of other workers who moonlight.

Multiple-jobholding rates are seen to be more disparate across individuals with different education levels. In line with earlier studies on this topic,6 multiple jobholding increases with educational attainment. Meanwhile, despite the difference in levels, moonlighting has been trending downward within all educational groups. On average over the past decade, multiple jobholding has diminished by 0.6 percentage point every year for both high school graduates and workers with some college education. As a result, the decline has not been less statistically significant among workers who have a lower propensity to moonlight.

Multiple jobholding is less frequent among married and single individuals than among those who are widowed, divorced, or separated. The figures mask some slight differences between men and women. For instance, married men are more likely to moonlight than men who are widowed, divorced, or separated. By contrast, multiple-jobholding rates are higher for single than married women.

In sum, table 1 and figure 2 show that multiple jobholding has become less common within most sociodemographic groups over the past two decades. The downward trend in the aggregate thus does not reflect a compositional change in the working-age population of those groups of workers who were already less likely to moonlight 20 years ago—at least not for the sociodemographic characteristics considered in figure 2.7 Finally, within all subgroups, there is no apparent relationship between multiple jobholding and the business cycle.

Occupation and industry of employment

Delving further into the findings just described, figure 3 (a through d) shows multiple-jobholding rates separately for workers with a different occupation or industry of employment in their primary job. The occupation or industry of the primary job is relevant for a number of reasons. To begin with, some occupations or industries may entail a work schedule that does not lend itself to holding a second job. Also, different occupations and industries pay different wages, a fact that may affect the need to work at a second job. Finally, to the extent that skills may be specific to the occupation or industry of the primary job, workers may have different opportunities to use these skills or to acquire new skills at a second job.

LaleFig3a

LaleFig3b

LaleFig3c

LaleFig3d

In line with the preceding discussion, figure 3 shows that there are occupations and industries with either low or high multiple-jobholding rates. Manual workers—for instance, those working in the mining, construction, or manufacturing industry—are less likely to work at a second job. By contrast, a considerable proportion of workers in professional and service occupations hold more than one job. One such example is teachers in elementary, middle, or secondary schools, whose multiple-jobholding rates are no less than 13 percent, as found in earlier studies.8 A relatively more convenient work schedule is probably an explanation for these workers’ high multiple-jobholding rates.

As regards the evolution of multiple jobholding, figure 3 unambiguously shows that changes to the occupation or industry structure of the economy do not explain the downward trend in the aggregate—at least, not for the broad occupation and industry categories presented in the figure. Indeed, multiple jobholding has declined steadily among workers, irrespective of the occupation or industry of their primary job.

Workers’ transitions into and out of multiple jobholding

From an accounting point of view, the pool of multiple jobholders is a labor market stock: a quantity that can be measured by using a snapshot of the labor market in any given month. A better understanding of its evolution can be gained by looking at labor market flows, which requires following the same workers in 2 consecutive months in order to identify their transitions. In so doing, one can determine whether the downward trend in moonlighting is a consequence of having fewer single jobholders who take on a second job, more multiple jobholders who give up their second job, or a combination of both.

In principle, transitions into and out of multiple jobholding can be measured with respect to many labor market positions of origin and destination: single jobholding, unpaid work, unemployment, inactivity, and more. A closer look at the data reveals that (i) most transitions into and out of multiple jobholding involve single jobholding rather than some other reason, (ii) among single jobholders, part-time and full-time workers have markedly different propensities to take on or to give up a second job, and (iii) transitions between multiple jobholding and either unemployment, inactivity, or unpaid work are quantitatively negligible. Altogether, these observations suggest that transitions into and out of multiple jobholding (M) ought to be measured with respect to full-time single jobholding (F), part-time single jobholding (P), and nonemployment (N), where the latter category effectively lumps together all working-age individuals who are either unpaid family workers, unemployed, or out of the labor force.

From single to multiple jobholding

Figure 4 (a through l) shows monthly transition rates into multiple jobholding within the various sociodemographic groups of the working-age population. Figures 4a, 4d, 4g, and 4j display the proportion of full-time single jobholders who become multiple jobholders the next month (F à M); figures 4b, 4e, 4h, and 4k show the proportion of part-time single jobholders who become multiple jobholders the next month (P à M); and figures 4e, 4f, 4i, and 4l depict the proportion of nonemployed individuals who become multiple jobholders the next month (N à M). The picture that emerges is that individuals of working age have become less likely to move into multiple jobholding during the past decades, irrespective of their sociodemographic characteristics. That is, all transition rates into multiple jobholding exhibit a downward trend.

LaleFig4a

LaleFig4b

LaleFig4c

LaleFig4d

LaleFig4e

LaleFig4f
LaleFig4g

LaleFig4h

LaleFig4i

LaleFig4j

LaleFig4k

LaleFig4l

A large fraction of multiple jobholders work part time at both their primary and second job. Not surprisingly, then, single jobholders who work part time have the highest probability of taking on a second job in the next month. About 2.5 percent do so in each month of the two decades examined, almost twice the percentage for full-time workers who take on a second job (1.3 percent) and more than tenfold the percentage for nonemployed workers (0.2 percent).

The fact that transitions from both full-time and part-time single jobholding have become less frequent is of consequence to the analysis of the downward trends in multiple jobholding. Indeed, a further examination of hours worked in part-time jobs indicates that those hours have increased over the past two decades. A less flexible work schedule for part-time workers could account for the lower propensity to moonlight documented in the rest of this article. However, the fact that full-time workers, too, have become less likely to moonlight suggests that the main explanatory factor is to be sought elsewhere.

Figure 4 offers several insights into some other features of the multiple-jobholding rates depicted in Figure 2. First, the divergent trends in multiple-jobholding rates between men and women are not explained by differences in their propensity to move into multiple jobholding; indeed, their transition rates have actually become more similar over the past two decades. Second, part of the convergence of the multiple-jobholding rates of younger workers and prime-age workers (ages 25 to 54) toward those of older workers can be attributed to the convergence of their own transition rates into multiple jobholding. Third, over the period examined, the relatively stable differences in multiple-jobholding rates among workers with different education levels and workers with different marital statuses were accompanied by relatively stable differences (or the absence of any differences) in transitions into multiple jobholding among these sociodemographic groups. In addition, figure 2 reveals that, although the proportion of workers who moonlight changes with marital status, transitions from single jobholding into multiple jobholding are similar for workers who differ in their marital status.

From multiple to single jobholding

Figure 5 (a through l) draws attention to transitions out of multiple jobholding: workers who give up their second job. Most of these transitions are toward single jobholding: only about 1.6 percent of workers move directly into nonemployment. Summing the probabilities of moving toward single jobholding, one notices that multiple jobholding is a highly temporary position: more than 30 percent of multiple jobholders return to single jobholding the following month. This high degree of turnover can be uncovered only by looking at data on labor market transitions.

LaleFig5a

LaleFig5b

LaleFig5c

LaleFig5d

LaleFig5e

LaleFig5f

LaleFig5g

LaleFig5h

LaleFig5i

LaleFig5j

LaleFig5k

LaleFig5l

Unlike figure 4, figure 5 does not deliver an unambiguous message as to trends in workers’ transition patterns. That is, transitions toward part-time single jobs have remained fairly stable over the 1994–2013 period, whereas transitions toward full-time single jobs trended differently in different subperiods of the period examined. Between 1996 and 2000, transitions toward full-time single jobs became more likely, thus effectively explaining part of the decline in multiple jobholding during the late 1990s. But from 2000 until the 2007–2009 recession, the proportion of multiple jobholders who returned to a full-time single job declined, generally at a steady pace. Furthermore, the recession seems to have exacerbated this trend. Since the recession, however, transitions toward full-time single jobs have remained stable.

Clearly, the downward trend in multiple jobholding is not explained by a higher propensity of multiple jobholders to return to a single job with a part-time schedule or by more frequent transitions towards nonemployment. Still, during the years from 1996 to 2000, part of the decline in the multiple-jobholding rate can be attributed to a higher propensity to take on a full-time single job. This trend reversed in the subsequent period and would have resulted in an increase in multiple jobholding, had transitions from single jobholding into multiple jobholding not diminished in greater proportion.9

Assessing the significance of the time trends

It is possible to demonstrate formally that the time trends (or, as the case may be, the absence of a trend) identified in figures 4 and 5 are statistically significant. To this end, the time series of each transition rate is regressed against a linear time trend. The results of these calculations are displayed in table 2. In addition, the table shows corresponding results for the (not reported) transition rates computed within each occupation and industry of the primary job.10

Table 2. Trends in workers' transitions into and out of multiple jobholding
Table 2. Trends in workers' transitions into and out of multiple jobholding
Demographic, occupational, or industry category Into multiple jobholding(1) Out of multiple jobholding(2)
F → M P → M N → M M → F M → P M→ N
 
 
 
 
 
 
 

Total

-0.037 -0.051 -0.007 -0.133 0.024 0.004

Gender:

Men

-.045 -.058 -.009 -.155 .054 .014

Women

-.027 -.047 -.006 -.054 -.046 -.009

Age:

16 to 24

-.041 -.051 -.012 -.242 .078 .028

25 to 54

-.039 -.058 -.006 -.095 .034 .006

55 to 64 years

-.015 -.007 .000 -.169 -.056 -.008

Education:

Less than high school

-.022 -.041 -.004 -.136 -.056 .005

High school graduate

-.036 -.050 -.004 -.197 .047 .014

Some college

-.044 -.073 -.014 -.161 .061 .006

College or higher education

-.048 -.066 -.013 -.067 .050 .020

Marital status:

Married

-.037 -.046 -.004 -.124 .007 .006

Widowed, divorced, or separated

-.033 -.062 -.006 -.100 .045 .003

Single

-.041 -.054 -.011 -.115 .007 -.001

Occupation: (3)

Managerial and professional

-.042 -.052 -.081 .051

Technical, sales, and administrative

-.038 -.052 -.082 .011

Service

-.053 -.051 -.172 -.054

Farming, forestry, and fishing

-.049 -.063 -.227 .035

Precision production, craft, and repair

-.038 -.049 -.143 .082

Operatives and laborers

-.029 -.057 -.208 .004

Industry: (3)

Agriculture, forestry, and fishing

-.051 -.053 -.214 -.053

Mining

-.022 -.042 .754 .050

Construction

-.035 -.039 -.239 .108

Manufacturing

-.028 -.036 -.077 .021

Transportation, communication, and other utilities

-.041 -.093 .162 -.003

Wholesale and retail trade

-.037 -.046 -.179 .021

Finance, insurance, and real estate

-.042 -.061 -.013 -.001

Business and repair services

-.045 -.064 -.224 -.022

Personal and recreation services

-.038 -.056 -.128 -.038

Professional and related services, and public administration

-.051 -.051 -.076 -.003
Footnotes:

(1) F → M = transition from full-time single job to multiple jobholding; P → M = transition from part-time single job to multiple jobholding; N → M = transition from nonemployment to multiple jobholding.

(2) M → F = transition from multiple jobholding to full-time single job; M → P = transition from multiple jobholding to part-time single job; M → N = transition from multiple jobholding to nonemployment.

(3) By construction, transitions from nonemployment into multiple jobholding are not available by occupation and industry because there is no primary job in the current month. For symmetry, transitions from multiple jobholding into nonemployment computed within each occupation and industry category also are omitted from the table.

Note: An entry in the table is the time trend (multiplied by 12 to obtain a yearly figure; see note 11 in the text,) estimated by regressing the corresponding time series on a linear trend. indicates that the estimate is not significant at p < .01; all other estimates are significant at p < .01.

Source: U.S. Bureau of Labor Statistics.

Footnotes:

(1) F → M = transition from full-time single job to multiple jobholding; P → M = transition from part-time single job to multiple jobholding; N → M = transition from nonemployment to multiple jobholding.

(2) M → F = transition from multiple jobholding to full-time single job; M → P = transition from multiple jobholding to part-time single job; M → N = transition from multiple jobholding to nonemployment.

(3) By construction, transitions from nonemployment into multiple jobholding are not available by occupation and industry because there is no primary job in the current month. For symmetry, transitions from multiple jobholding into nonemployment computed within each occupation and industry category also are omitted from the table.

Note: An entry in the table is the time trend (multiplied by 12 to obtain a yearly figure; see note 11 in the text,) estimated by regressing the corresponding time series on a linear trend. indicates that the estimate is not significant at p < .01; all other estimates are significant at p < .01.

Source: U.S. Bureau of Labor Statistics.

The first three columns of table 2 confirm the presence of a downward trend in single jobholders’ probability of taking on a second job; the trends are all statistically significant at the 1-percent level. In the aggregate, each year of the period examined reduced the probability of moving into multiple jobholding by 0.04 percentage point for full-time workers and by 0.05 percentage point for part-time workers.11 Similar patterns are present in the time series computed with each occupation and industry in which the primary job is held. Finally, most trends in the probabilities of moving from nonemployment to multiple jobholding are also significant.

By contrast, a large number of probabilities of moving out of multiple jobholding exhibit no significant time trend. This finding is true especially for transitions toward a single job operated on a part-time basis or toward nonemployment. As regards transitions from multiple jobholding toward a full-time single job (the M à F column), table 2 shows that, in most instances, the dominant trend has been a negative one. That is, the increase between the years 1996 and 2000 did not offset the subsequent decrease that occurred from 2001 onward. This finding confirms that the decline in multiple jobholding over the past two decades is not explained by a higher probability of multiple jobholders returning to single jobholding.

Multiple jobholding has become less common in the United States over the past two decades. The downward trend cannot be attributed either to changes in the sociodemographic composition of the working-age population or to shifts in the occupation or industry structure of the economy. Instead, this article shows that the trend originates from a lower propensity of single jobholders to take on a second job. Multiple jobholders, in contrast, did not become more likely to give up their second job. One explanation for these findings is that workers may have become increasingly reliant on alternative sources of income to meet expenses or to pay off debt. Another, noneconomic explanation is that looking for enjoyment through a second, different job may have become more unusual. Future research could delve into these explanations or even examine alternative explanations.

Suggested citation

Etienne Lalé, "Multiple jobholding over the past two decades," Monthly Labor Review, U.S. Bureau of Labor Statistics, April 2015, https://doi.org/10.21916/mlr.2015.7.

Notes


1 For evidence based on the Current Population Survey, see Jennifer L. Martel, “Reasons for working multiple jobs,” Monthly Labor Review, October 2000, pp. 42–43, https://www.bls.gov/opub/mlr/2000/10/atissue.pdf. For evidence from the Panel Study of Income Dynamics, see Christina H. Paxson and Nachum Sicherman, “The dynamics of dual job holding and job mobility,” Journal of Labor Economics, vol. 14, no. 3, 1996, pp. 357–393.

2 See Steven F. Hipple, “Multiple jobholding during the 2000s,” Monthly Labor Review, July 2010, pp. 21–32, especially table 7, p. 30, https://www.bls.gov/opub/mlr/2010/07/art3full.pdf.

3 All time-series displayed in this article are adjusted as follows: first, seasonality is removed with the use of the U.S. Census Bureau’s X13-ARIMA-SEATS program; then, high-frequency variations are smoothed out by calculating the 3-month centered moving average of the time series.

4 See, for example, John F. Stinson, Jr., “Multiple jobholding up sharply in the 1980’s,” Monthly Labor Review, July 1990, pp. 3–10, https://www.bls.gov/opub/mlr/1990/07/art1full.pdf; and “New data on multiple jobholding available from the CPS,” Monthly Labor Review, March 1997, pp. 3–8, https://www.bls.gov/opub/mlr/1997/03/art1full.pdf.

5 Catalina Amuedo-Dorantes and Jean Kimmel, “Moonlighting over the business cycle,” Economic Inquiry, October 2009, pp. 754–765, provides a discussion of the underlying mechanisms. The authors explain why sample selection into employment (i.e., the fact that individuals in employment at various phases of the business cycle represent a nonrandom sample of the population) is likely to affect the measured cyclicality of multiple-jobholding rates.

6 See, for instance, Thomas Amirault, “Characteristics of multiple jobholders, 1995,” Monthly Labor Review, March 1997, pp. 9–15, https://www.bls.gov/opub/mlr/1997/03/art2full.pdf.

7 These findings also hold for multiple-jobholding rates in different states and separately for metropolitan and nonmetropolitan areas.

8 See, for instance, Amirault, “Characteristics of multiple jobholders,” table 4.

9 Transitions from nonemployment to multiple jobholding diminished, too, but their role is quantitatively negligible.

10 By construction, transitions from nonemployment into multiple jobholding are not available by occupation and industry because there is no primary job in the current month. For symmetry, transitions from multiple jobholding into nonemployment computed within each occupation and industry category are omitted from table 2.

11 The frequency of observations is monthly. Therefore, the coefficients have been multiplied by 12 to measure the effects of an additional year on the probability of transition.