Flows to and from working part time for economic reasons and the labor market aggregates during and after the 2007-09 recession.
Canon, Maria E. ; Kudlyak, Marianna ; Luo, Guannan 等
While the unemployment rate is one of the most cited economic
indicators, economists and policymakers also examine a wide array of
other indicators to gauge the health of the U.S. labor market. One such
indicator is the U-6 index, an extended measure of the unemployment rate
published by the Bureau of Labor Statistics (BLS). In addition to
unemployed workers, the U-6 index includes individuals who are working
part time for economic reasons and individuals who are out of the labor
force but are marginally attached to the labor market. Individuals are
classified as working part time for economic reasons (henceforth, PTER)
if they work fewer than 35 hours per week, want to work full time, and
cite "slack business conditions" (1) or an inability to find a
full-time job as a reason for not working full time. On average, from
1994-2014, 2.4 percent of the civilian noninstitutionalized population
16 years and older are classified as PTER. In 2009, this share reached
3.8 percent. (2)
Part-time employment for economic reasons has become a concern
since the 2007-09 recession because, even though the numbers of
unemployed and marginally attached individuals have been decreasing
since 2009, the number of individuals who are working part time for
economic reasons has remained elevated. (3) During the 2014 Economic
Symposium in Jackson Hole, Wyo., Fed Chair Janet Yellen noted that the
elevated number of workers who are employed part time but desire
full-time work might imply that the degree of resource underutilization
in the labor market is greater than what is captured by the standard
unemployment rate (Yellen 2014).
In this article, we first use cross-sectional data to evaluate
whether part-time employment for economic reasons differs from full-time
employment or part-time employment for noneconomic reasons such as
childcare or other family reasons (henceforth, PTNER) along dimensions
other than hours (i.e., observable characteristics of workers and
wages). We then examine whether the changes in the labor market flows in
and out of PTER during and in the aftermath of the 2007-09 recession can
account for any of the changes in unemployment.
We find that PTER workers are typically less educated than fulltime
or other part-time workers and are typically employed in middleor
low-skill occupations. On average, PTER workers earn 19 percent less
than full-time workers and 9 percent less (per hour) than PTNER workers,
even after controlling for sociodemographic and occupational
characteristics. The differences persist if we compare wages of PTER to
wages of other workers within broad occupational categories. More
research, however, is needed to understand whether PTER workers are
workers who cannot find full-time jobs because of bad luck or because of
structural reasons.
We now turn to the question of PTER and unemployment. Note that the
number of PTER workers at any point in time (i.e., stock) is affected by
the number of workers who worked PTER in the previous period and
continue to do so, as well as the number of workers who transition
(i.e., flow) into PTER from full-time employment, other part-time
employment, unemployment, and out-of-the-labor-force (OLF) and the
number of workers who transition from PTER into these other labor market
statuses. Similarly, the flows into and out of PTER impact other labor
market aggregates--full- and part-time employment, unemployment, and
OLF. In this article, we decompose the changes in stocks of full-time
employed, PTNER, unemployed, and OLF due to the changes in the flows of
workers to and from PTER in the aftermath of the 2007-09 recession. Of
course, the flows are in turn determined by fundamental factors
affecting households' and firms' behavior. Nevertheless, it
can be instructive to look at such decomposition. To this end, we
perform a counterfactual exercise by fixing the transition probabilities
between PTER and other labor force statuses at their respective sample
means, and constructing the counterfactual time series of the labor
market aggregates. The exercise is similar in spirit to the exercise
presented by Shimer (2012) for the contribution of different labor
market flows to changes in the unemployment rate.
The accounting exercise shows that changes in the transition
probabilities to and from PTER after 2009 were associated with changes
in stocks of full-time employed, PTER, and PTNER, but had almost no
impact on the changes in stocks of unemployed or OLF. In the
counterfactual exercise, the main drivers of the changes in the stocks
of full-time employed, PTER, and PTNER were transition probabilities
between PTER and full-time work and between PTER and PTNER. If the
transition probabilities to PTER from either full-time or PTNER had
remained at their sample means throughout 1994-2014, the population
share of PTER in 2014 would have been 0.47 percentage points (pp) lower
at the expense of full-time work and PTNER. If the transition
probabilities from PTER to full-time work and to PTNER had remained at
their sample means throughout 1994-2014, the population share of PTER in
2014 would have been 0.43 pp lower at the expense of full-time work and,
to a lesser extent, of PTNER. In contrast, this same exercise yields
counterfactual unemployment that is essentially identical to the one
actually observed.
Thus, our results show that changes in the transition probabilities
to and from PTER in the aftermath of the 2007-09 recession mainly impact
the composition of employment (full versus part time, and the reasons
for working part time) instead of the distribution of individuals
between employment and non-employment. Consequently, policymakers'
attention to PTER potentially implies a broader definition of resource
underutilization in the labor market than the one captured by the
standard unemployment rate. In particular, in addition to working
fewer-than-desired hours, underutilization in the labor market can take
the form of workers being overqualified for their jobs. For example,
Abel, Deitz, and Su (2014) provide evidence of an upward trend in
underemployment of recent college graduates whereby the graduates are
employed in jobs that do not require a college degree. Importantly, the
challenge for policymakers lies in determining how much of such changes
in the quality of employment represent structural changes in the
economy.
Finally, regarding the future of PTER, an examination of the series
of PTER over time reveals that the ratio of the number of PTER workers
to the number of unemployed workers typically increases during economic
recoveries. The increase is fueled by PTER workers who cite an inability
to find full-time work as a reason for part-time employment (the number
of PTER workers who cite "slack work" declines during economic
recoveries). PTER workers' share is highest in nonroutine manual
(typically low-wage) occupations. Given the recent work on job
polarization (Autor [2010], among others), which shows that medium-wage
jobs are disappearing but jobs on the high- and low-end of the wage
distribution are growing, it thus becomes a challenging task to
disentangle cyclical versus structural factors behind an increased
number of PTER workers after the 2007-09 recession. Thus, the following
questions might represent fertile ground for future research: (1) To
what extent is PTER an important mechanism of labor market adjustment
during recoveries from recessions? (2) What is the impact of
trend-related developments like job polarization on such an adjustment,
especially after deep recessions? (3) To what extent does the burden of
adjustment fall more on certain demographic and socioeconomic groups
than on the others?
The rest of the article is structured as follows. Section 1
describes the construction of the PTER series in the CPS data. Section 2
presents basic facts about PTER. Section 3 presents the main results.
Finally, section 4 concludes.
1. MEASUREMENT OF PTER IN THE CPS
The data in the analysis are from the Current Population Survey
(CPS) monthly microdata files from January 1994 to August 2014. The
survey features a rotating panel structure in which households are
surveyed for four months, taken out of the sample for eight months, and
then surveyed for another four months to complete their participation.
The CPS allows us to classify each individual into one of five labor
force statuses: employed full time, employed part time for economic
reasons, employed part time for noneconomic reasons, unemployed, and
OLF. (4)
The survey asks respondents about their hours worked during the
reference week, their desire and availability for full-time work if they
work part time, and their reason for working part time. The individuals
who work fewer than 35 hours per week are considered part-time workers.
(5) For the part-time work to be classified as "for economic
reasons," the worker must desire full-time work and cite an
economic reason as the primary reason for not working full time. Such
economic reasons are "slack work or business conditions,"
"could only find part-time work," and seasonal work.
Noneconomic reasons are child care problems, other family/personal
obligations, health/medical limitations, school/training, retired/Social
Security limit on earnings, full-time workweek is less than 35 hours,
weather affected job, military/civic duty, labor dispute, holiday, own
illness, vacations, and other (unspecified) reasons.
The 1994 CPS redesign affected the PTER series. Prior to 1994, the
CPS did not specifically ask whether part-time workers wanted to or were
available to work full time. (6) Additionally, the survey did not
distinguish between respondents who usually worked full time and those
who usually worked part time; it only asked about actual hours worked.
The effect of the CPS redesign on the PTER series after 1994 is
therefore twofold: (1) it decreased the number of part-time workers
classified as PTER because it excludes those who do not want to work
full time; and (2) it may have increased the total number of part-time
workers because it includes those who usually, but not actually in the
reference week, work fewer than 35 hours per week. (7) Consequently,
caution needs to be exercised while constructing a longer series of PTER
that begins prior to 1994. Another change the redesign introduced was
"seasonal work" as an economic reason for working part time.
(8) Prior to 1994, only slack work, not being able to find a full-time
job, and a job starting or ending during the reference week were
considered economic reasons for working part time. Therefore, our
analysis focuses on the 20-year period following the 1994 CPS redesign
so that we can use the BLS U-6 definition of PTER.
2. BASIC FACTS ABOUT WORKING PART TIME FOR ECONOMIC REASONS
Wages, Hours, and Occupations
Table 1 shows average weekly hours and real hourly wages over the
1994-2014 period for three different groups of the employed: full-time,
PTER, and PTNER workers. (9) During 1994-2014, a full-time worker's
average real hourly wage is $17.02 (in 2013 U.S. dollars), while it is
$13.66 for a PTNER worker and $11.81 for a PTER worker.
PTER workers report working 23 hours per week on average as
compared to 45 hours reported by those working full time. (10) They also
work on average 1.5 hours more per week than PTNER workers. As can be
seen from Figure 1, these gaps persist throughout 1994-2014.
[FIGURE 1 OMITTED]
Table 2 presents demographic characteristics of the three groups of
employed workers, shedding some light on the difference in hourly wages
between PTER workers and other employed persons. For example, fulltime
workers are more likely to have finished high school or college than
part-time workers; among part-time workers, PTNER workers tend to be
more highly educated than PTER workers (41.4 percent of full-time
workers, 33.0 percent of PTNER workers, and 22.0 percent of PTER workers
have a college degree or higher). PTER workers tend to be younger, with
a comparatively large share of 20-24 year olds.
To further understand the differences between wages of PTER workers
and the rest of the employed population, we tabulate the shares of
different types of workers across different occupations. Following
Jaimovich and Siu (2012) (see also Autor, Levy, and Murnane [2003] and
Acemoglu and Autor [2011]), we classify the occupations into four
different groups: non-routine cognitive, routine cognitive, routine
manual, and non-routine manual occupations. (11) Routine occupations are
typically middle-skill occupations. (12) As discussed in Autor (2010)
and Jaimovich and Siu (2012), the U.S. labor market is experiencing a
job polarization phenomenon where employment in routine occupations is
shrinking while employment in non-routine cognitive and non-routine
manual occupations is growing.
Table 3 shows the distribution of full-time, PTER, and PTNER work
across four broad occupational groups with cognitive-manual and
routine/non-routine classifications. Part-time workers represent a
significantly higher fraction of low-skill and medium-skill occupations
than of high-skill occupations. Interestingly, among the highest skill
occupations, classified as non-routine cognitive, the share of PTER
workers is only 0.03 while the share of PTNER workers is 0.16. The share
of PTER workers is highest among non-routine manual occupations (0.11),
which are typically low-skill occupations.
To understand whether the differences in wages between full-time,
PTNER, and PTER workers can be explained by the differences in their
sociodemographic characteristics and/or their occupations, we estimate a
linear regression of the logarithm of the real hourly wage on
educational level, occupation, race, gender, year, and employment type
dummy variables. The omitted category for employment type is PTER. The
coefficients for the type of employment show the difference in the (log
of the) real hourly wage between PTER and working full time or PTNER,
after controlling for sociodemographic and occupational characteristics.
The results of this regression are presented in Table 4. (13) On
average, full-time workers earn 19 percent more and PTNER workers earn 9
percent more (per hour) than PTER workers, taking into account
education, age, and broadly defined occupational categories.
To further understand the wage differences, instead of occupational
and employment dummy variables, we include a full set of interactions
between seven occupational categories and the three types of employment
(full time, PTER, and PTNER). If, for example, better workers (either
employed full or part time) are employed in higher-paying occupations,
then one should compare the wages of full- and part-time workers in
these occupations in order to estimate the differences in earnings
between full- and part-time workers. Table 5 contains the results of the
regression with the interaction terms. (14) We then perform a series of
pairwise t-tests comparing the coefficient for the interaction term of
full-time work (and similarly PTNER) to the coefficient for the
interaction term of PTER with each of the seven occupational categories.
In each of the seven occupational categories, we find that PTER workers
receive lower wages than full-time or PTNER workers. For example, on
average, PTER workers in service occupations are paid 18 percent less
than full-time workers and 7 percent less than PTNER workers in service
occupations. (15)
The regression results in Tables 4 and 5 also show that the year
dummies are positive during the 2008-11 recession years and turn
negative in the post-recession years, 2012-14, which points to a
somewhat lagged response of wages during the 2007-09 recession. (16)
Working Part Time for Economic Reasons Over the Years
Figure 2 shows the population shares of full-time, PTNER, and PTER
workers. As can be seen, there is a notable drop in the share of
fulltime workers and an increase in PTER workers during the 2007-09
recession. Figure 3 shows a close-up of the PTER series. The PTER
population share was higher in the 2007-09 recession than in the 2001
recession. In 2009, the series reached 3.8 percent. In 2014, the PTER
population share stands at 3.0 percent. (17)
Figure 4 examines PTER by reason: slack work, could only find
part-time work, and "other," which includes a job
starting/ending during the reference week (such that hours add up to
less than 35) and seasonal work. The first two reasons account for the
majority of the PTER workers. Notably, during the 2007-09 recession the
share of workers who reported slack work/business conditions increased
to a much higher level than during the previous recession. While the
share of workers reporting "slack work" has declined
substantially since 2009, the share of workers who are working part time
because they could only find part-time work has remained elevated since
2009. A similar cyclical pattern is observed during previous downturns.
(18)
[FIGURE 2 OMITTED]
Figure 5 shows the ratio of PTER workers to the number of
unemployed workers in the economy. Interestingly, the ratio was about 10
percentage points higher at the trough of the 2007-09 recession than at
the trough of the 2001 recession. The ratio appears procyclical,
indicating that during recessions PTER grows at a slower rate than
unemployment. The most recent growth started in 2010, increasing from
0.60 in 2010 to 0.74 in 2014.
[FIGURE 3 OMITTED]
3. THE TRANSITION PROBABILITIES OF PTER FLOWS DURING 2007-09 AND
EFFECTS ON EMPLOYMENT, UNEMPLOYMENT, AND OLF
In this section, we focus on the transition probabilities to and
from the stock of PTER and other states of the labor market. We
decompose changes in the stocks of the labor market aggregates--full-
and part-time employment, unemployment, and out-of-the-labor-force
(OLF)--into the changes in these transition probabilities during the
2007-09 recession. The counterfactual exercises show that these changes
were not associated with the changes in the stocks of unemployment or
OLF, but they were associated with the decrease of the stocks of
full-time and PTNER employment.
Transition Probabilities to and from PTER
As mentioned above, each individual in the population can be
classified into one of the following five labor force statuses: employed
full time (FT), PTER, PTNER, unemployed (U), and OLF. The labor market
is characterized by the flows of individuals among these statuses. The
stocks and the transition probabilities among them are linked via the
following equation
S (t) = P (t)S(t - 1), (1)
where S(t) is the vector of stocks (expressed in population
shares), and P(t) is the matrix of discrete transition probabilities.
(19)
[FIGURE 4 OMITTED]
The change in the stock of PTER can be decomposed into components
representing changes in the probabilities of entering and exiting PTER
as well as components representing changes in the transition
probabilities between the remaining labor force statuses. Likewise,
changes in entry and exit to/from PTER are associated with the changes
in the stocks of FT, PTNER, U, and OLF.
[FIGURE 5 OMITTED]
To construct the transition probabilities matrix we match
individuals between consecutive months in the CPS following the matching
procedure described in Shimer (2012). Because the unit of observation is
the physical address, we use sex, age, and race in addition to the
household identification number to produce matches. The transition
probability from state i in month t-1 to state j in month t is the flow
of individuals moving from state i to state j divided by the total
number of individuals in state i in month t-1 (out of those that can be
matched). We call this the "exit probability from" state i to
state j, or the "entry probability to" state j from state i.
Table 6 shows the mean of annual average transition probabilities
among the five labor market statuses during 1994-2014. A PTER worker has
probability 0.31 of transitioning to full-time employment next month.
This probability is 0.30 for a PTNER worker and 0.13 for an average
unemployed worker. Thus, in their propensity to join full-time work,
PTER workers are closer to PTNER than to unemployed workers. The data
reveal the substantial flows between PTER and PTNER. An unemployed
worker and a PTNER worker have similar probabilities of transitioning
into PTER, 0.048 and 0.044, respectively.
Panels A and B of Figure 6 show the transition probabilities from
and to PTER, respectively, by labor force status. The observations from
the figure can be summarized as follows. First, the transition
probability from PTER to FT declined during 2007-09 and has remained low
since then. Second, the transition probability from PTER to PTNER
declined during 2007-09 and has only slightly increased since then.
Third, the transition probability from FT to PTER increased during
2007-09 and has decreased since then. Fourth, the transition probability
from PTNER to PTER increased during 2007-09 and has remained elevated
since 2009.
Counterfactual Exercises with the Transition Probabilities to and
from PTER
To separately examine the effects of exit and entry, we perform a
series of counterfactual exercises using equation (1). The exercises are
as follows:
1. fix all transition probabilities from PTER (to FT, to PTNER, to
U, and to OLF);
2. fix transition probabilities from PTER to FT;
3. fix transition probabilities from PTER to PTNER;
4. fix transition probabilities from PTER to U;
5. fix transition probabilities from PTER to OLF;
6. fix all transition probabilities to PTER (from FT, from PTNER,
from U, and from OLF);
7. fix transition probability from FT to PTER;
8. fix transition probability from PTNER to PTER;
9. fix transition probability from U to PTER; and
10. fix transition probability from OLF to PTER.
[FIGURE 6 OMITTED]
[FIGURE 7 OMITTED]
To perform these exercises, we fix the respective probabilities at
their 1994-2014 sample means and construct the monthly counterfactual
time series of the fixed labor force status stocks using equation (1)
recursively, setting [t.sub.0] = 1994. (20)
We start in 1994 because of the changes in the series after the
1994 CPS redesign (mainly PTNER and FT). In the exercises, the diagonal
elements of the transition matrix (i.e., the probability of remaining in
the same status) are adjusted accordingly so that the column elements
add to 1. Figures 7 and 8 show the resulting counterfactuals using
annual averages of monthly series. (21) Figure 7 shows the
counterfactuals with fixed exit rates from PTER. All stocks are
expressed as shares of the population. The effect of the counterfactual
transition probabilities of exiting PTER on the aggregate stocks is as
follows:
1. PTER (Figure 7, Panel C): If all exits from PTER are fixed at
their sample means, the counterfactual share of PTER in 2014 is 0.43 pp
lower than the actual share. PTER is primarily affected by exits from
PTER to FT and from PTER to PTNER.
2. FT (Figure 7, Panel A): If the exit from PTER to FT is fixed at
its 1994-2014 sample mean, the population share of FT in 2014 increases
by 0.69 pp (as compared to its 44.4 pp level in 2014). Other exits from
PTER do not have a substantial impact on the share of FT workers.
3. PTNER, U, and OLF (Figure 7, Panels B, D, and E): The relative
magnitudes of the effect of the fixed exits on PTNER, U, and OLF are
much smaller than the effect of the counterfactual exits on the share of
FT workers.
Figure 8 shows the counterfactuals with fixed transition
probabilities to PTER. The effect of the counterfactual transition
probabilities of entering PTER on the aggregate stocks is as follows:
1. PTER (Figure 8, Panel C): We observe that if all transition
probabilities to PTER are fixed at their sample means, PTER in 2014 is
0.47 pp lower than the actual population share observed. As with the
case of fixed exit rates, PTER is primarily affected by transition
probabilities from FT and from PTNER.
2. FT (Figure 8, Panel A): If the transition probability from FT to
PTER remains at its 1994-2014 sample mean, the population share of FT in
2014 increases by 0.39 pp (as compared to its 44.4 pp level in 2014).
Other entries to PTER do not have a substantial impact on the share of
FT workers.
3. PTNER (Figure 8, Panel B): If the transition probability from
PTNER to PTER remains at its 1994-2014 sample mean, the population share
of PTNER in 2014 is 0.27 pp higher than the actual share observed. Other
entries to PTER do not have a substantial impact on the share of PTNER
workers.
4. U and OLF (Figure 8, Panels D, and E): The fixed transition
probabilities into PTER have essentially no effect on U or OLF.
[FIGURE 8 OMITTED]
As can be seen from Figures 7 and 8, the transition probabilities
to PTER contribute substantially to the cyclical behavior of the share
of PTER workers, while the exit rates do not drive much of the cyclical
fluctuations.
4. CONCLUSIONS
The elevated number of PTER workers in the aftermath of the 200709
recession has raised a concern of whether the extent of resource
underutilization in the labor market is greater than that captured by
the standard unemployment rate.
In this article, we find that the changes in the transition
probabilities to and from PTER in the aftermath of the 2007-09 recession
have been mainly associated with the composition of employment (full
versus part time, and part time for economic versus for noneconomic
reasons) instead of with the distribution of individuals between
employment and non-employment.
We also find that, in general, part-time workers represent a
significantly higher fraction of low-skill and medium-skill occupations
than of high-skill occupations. Among the highest skill occupations,
classified as non-routine cognitive, the share of PTER workers is only
0.03 while the share of PTNER workers is 0.16. The share of PTER workers
is highest among non-routine manual occupations, which are typically
low-skill occupations. The educational achievement of PTER workers is
typically lower than of those working full time or part time for
noneconomic reasons. PTER workers typically earn less per hour than
full-time or PTNER workers, even after controlling for age, education,
and broadly defined occupational groups. Given the recent work on job
polarization (Autor 2010), it thus becomes a challenging exercise to
disentangle the effect of cyclical versus structural factors on driving
up the number of PTER following the deep recession of 2007-09.
Canon is an economist at the Federal Reserve Bank of St. Louis. Luo
is an assistant professor at the City University of Hong Kong. Kudlyak
is an economist and Reed is a research associate at the Federal Reserve
Bank of Richmond. The authors are very grateful to Alex Wolman, Felipe
Schwartzman, Christian Matthes, and Peter Debbaut for useful comments
and suggestions. The views expressed here are those of the authors and
do not reflect those of the Federal Reserve Bank of Richmond, the
Federal Reserve Bank of St. Louis, the Federal Reserve System, or any
other institution with which the authors are affiliated. Email:
Marianna.Kudlyak@rich.frb.org.
REFERENCES
Abel, Jaison R., Richard Deitz, and Yaqin Su. 2014. "Are
Recent College Graduates Finding Good Jobs?" Federal Reserve Bank
of New York Current Issues in Economics and Finance 20 (1): 1-8.
Acemoglu, Daron, and David Autor. 2011. "Skills, Tasks and
Technologies: Implications for Employment and Earnings." In
Handbook of Labor Economics, Volume 4B, edited by David Card and Orley
Ashenfelter. Elsevier: 1,043-171.
Autor, David. 2010. "The Polarization of Job Opportunities in
the U.S. Labor Market: Implications for Employment and Earnings."
Center for American Progress and The Hamilton Project (April).
Autor, David, Frank Levy, and Richard J. Murnane. 2003. "The
Skill Content of Recent Technological Change: An Empirical
Exploration." Quarterly Journal of Economics 118 (November): 1,
279-333.
Cajner, Tomaz, Dennis Mawhirter, Christopher Nekarda, and David
Ratner. 2014. "Why is Involuntary Part-Time Work Elevated?"
FEDS Notes, April 14. Available at
www.federalreserve.gov/econresdata/notes/feds-notes/2014/why
is-involuntary-part-time-work-elevated-20140414.html.
Elsby, Michael W. L., Donggyun Shin, and Gary Solon. 2014.
"Wage Adjustment in the Great Recession and Other Downturns:
Evidence from the United States and Great Britain." Mimeo,
University of Edinburg. Available at
https://sites.google.com/site/mikeelsby/papers.
Frazis, Harley J., Edwin L. Robison, Thomas D. Evans, and Martha A.
Duff. 2005. "Estimating Gross Flows Consistent with Stocks in the
CPS." Monthly Labor Review 128 (September): 3-9.
Hornstein, Andreas, Marianna Kudlyak, and Fabian Lange. 2014.
"Measuring Resource Utilization in the Labor Market." Federal
Reserve Bank of Richmond Economic Quarterly 100 (First Quarter): 1-21.
Jaimovich, Nir, and Henry E. Siu. 2012. "The Trend is the
Cycle: Job Polarization and Jobless Recoveries." Cambridge, Mass.:
National Bureau of Economic Research Working Paper No. 18334.
Kearns, Jeff, and Jeanna Smialek. 2014. "Part-Time Workers a
Full-Time Headache on Yellen Radar: Economy." Bloomberg, August 18.
Available at www.bloomberg.com/news/2014-08-18/yellen-dashboard-warning
light-glows-as-millions-work-part-time.html.
Polivka, Anne E., and Jennifer M. Rothgeb. 1993. "Overhauling
the Current Population Survey: Redesigning the CPS Questionnaire."
Monthly Labor Review 116 (September): 10-28.
Shimer, Robert. 2012. "Reassessing the Ins and Outs of
Unemployment." Review of Economic Dynamics 15 (April): 127-48.
Valetta, Rob, and Leila Bengali. 2013. "What's Behind the
Increase in Part-Time Work?" Federal Reserve Bank of San Francisco
Economic Letter 2013-24, August 26.
Yellen, Janet. 2014. "Labor Market Dynamics and Monetary
Policy." Remarks at the Federal Reserve Bank of Kansas City
Economic Symposium, Jackson Hole, Wyo., August 22.
(1) This is the term ("slack work/business conditions")
used in the CPS questionnaire as opposed to the term "slack"
used in recent policy discussions that typically describes a degree of
labor utilization below a level of full resource utilization.
(2) The data in this paragraph are from HAVER.
(3) See, for example, Kearns and Smialek (2014) for a summary of
policy discussions about individuals working PTER. For research on
working PTER, see Valetta and Bengali (2013) and Cajner et al. (2014).
For alternative measures of resource utilization in the labor market
that incorporate PTER, see Hornstein, Kudlyak, and Lange (2014).
(4) We restrict the analysis to the civilian noninstitutionalized
population 16 years and older (henceforth, population).
(5) We use actual hours worked in the reference week to
differentiate full-time and part-time workers. We count those workers
who are absent from work (and thus whose actual hours are not available
in the survey) as full-time workers if they report that they usually
work full-time hours. Workers who are absent from work and report that
they usually work part-time hours are excluded from our analysis (for
example, 0.62 percent of the population in 2013) because they are not
asked to provide a reason for why they work part time.
(6) That is, after the 1994 redesign, if the respondents do not
desire full-time work, they are asked to choose from only noneconomic
reasons. If the respondents desire fulltime work, they are asked for the
primary reason for working part time, with the option to provide an
economic or noneconomic reason. Therefore, in order to be considered
working part time for economic reasons after 1994, workers must desire
full-time work in addition to citing economic reasons. Prior to 1994,
the survey does not separate those who do and those who do not want
full-time jobs.
(7) See Polivka and Rothgeb (1993) for a thorough treatment of the
effect of the redesign on part-time work calculations and for an
explanation of how to adjust the series to be consistent over time.
(8) Seasonal work, however, does not constitute a large portion of
PTER.
(9) To construct hourly wages, we use hourly earnings (if they are
reported) or construct the wage by dividing weekly earnings by usual
weekly hours. The reported wage statistics are based on non-imputed
data. We also calculated the statistics incorporating imputed data and
the results do not differ significantly. In the calculations we use
outgoing rotation group weights.
(10) We take the mean of each year's average actual hours
worked at all jobs in the reference week from 1994 to 2014. Usual hours,
which are used to construct hourly earnings for non-hourly workers, are
lower for full-time workers and higher for both voluntary and
involuntary part-time workers.
(11) Non-routine cognitive occupations include management,
business, and financial occupations and professional related. Routine
cognitive occupations include sales and office occupations and office
and administrative support occupations. Routine manual occupations
include construction and extraction; installation, maintenance, and
repair; production; and transportation and moving material occupations.
Non-routine manual occupations include service occupations.
(12) Following Autor (2010), high-skill occupations include
managers, professionals, and technicians. Middle-skill occupations
include sales; office and administration; production, craft, and repair;
and operators, fabricators, and laborers. Finally, low-skill occupations
include protective services; food preparation; building and grounds
cleaning; and personal care and personal services.
(13) In Table 4, the dependent variable is the log of real hourly
wage as described in the note to Table 1. The explanatory variables are
type of employment dummies, occupation, education, age, race, gender,
and a set of annual time dummies. The omitted categories are working
part time for economic reasons, less than high school education, service
occupations, male, and white. *** denotes statistical significance at
the 1 percent level for a two-sided test. All data are from 1994 to
August 2014 and include employed working age persons in months four and
eight of the CPS sample except for those in the armed forces or farming,
fishing, and forestry occupations. The regression is estimated by OLS
with heteroscedasticity robust standard errors, with the CPS outgoing
rotation group sampling weights. See footnote 14 for the details about
the occupational classification.
(14) In Table 5, the dependent variable is the log of real hourly
wage as described in the note to Table 1. The explanatory variables are
type of employment interacted with occupation, education, age, race,
gender, and a set of annual time dummies. The omitted categories are
working part time for economic reasons interacted with service
occupations, less than high school education, male, and white. See
footnote 13. The occupational classification used in the regression is
as follows (accounting for the change in coding in 2002): (1) Healthcare
support occupations; protective service; food preparation and serving
related; building and grounds cleaning and maintenance; personal care
and service occupations (post-2002) and service occupations (pre-2002);
(2) Construction trades, extraction workers; transportation and material
moving occupations (post-2002); and operators, fabricators, and laborers
(pre-2002); (3) Installation, maintenance, and repair workers and
production occupations (post-2002) and production occupations
(pre-2002); (4) Sales and related occupations (post-2002) and sales
occupations (pre-2002); (5) Office and administrative support
occupations (post-2002) and administrative support occupations,
including clerical (pre-2002); (6) Computer and mathematical;
architecture and engineering; life, physical, and social science;
community and social services; legal occupations; education, training,
and library; arts, design, entertainment, sports, and media; healthcare
practitioners and technical occupations (post-2002); and professional
specialty and technicians and related support occupations (pre-2002);
(7) Management; business and financial operations occupations
(post-2002); and executive, administrative, and managerial occupations
(pre-2002). Occupation 1 is non-routine manual; occupations 2-3 are
routine manual; occupations 4-5 are routine cognitive; and occupations
6-7 are non-routine cognitive.
(15) However, more analysis is needed to examine how much of the
wage difference can be attributed to worker fixed effect. Such analysis
is beyond the scope of the article.
(16) See Elsby, Shin, and Solon (2014) for a detailed exploration
of wage adjustment in the 2007-09 recession.
(17) This figure is calculated using January 1994-August 2014 data.
(18) However, due to the changes to the CPS described in section 1,
most of these observed downturns are not strictly comparable.
(19) In the analysis, we also include inflows and outflows into the
population.
(20) Due to the rotating panel structure of the CPS, at most 75
percent of the observations may be matched to the following month when
we exclude individuals in months four and eight in the survey, and thus
the labor force stocks tabulation from unmatched monthly CPS data may
differ from the labor force stocks tabulation from the matched
month-to-month files (see, for example, Frazis et al. [2005]). We
therefore employ a procedure that ensures that in every period the
recursion delivers the distribution of the labor force stocks consistent
with the one observed in the unmatched CPS monthly files.
(21) Additionally, we impute missing data in unmatchable months,
i.e., we take the average of each stock and probability of the adjacent
months for June-September 1995 and May 2004. We employ the same
procedure for September 1998 and September 2009 for FT and PTNER to
remove the effect of full-time workers being classified as part time for
noneconomic reasons due to Labor Day--this is the only national holiday
occurring in any reference week after 1994 and would constitute a
significant spike in the series if not adjusted.
Table 1 Average Weekly Hours and Real Hourly Wages,
Full- and Part-Time Employment, 1994-2014
Full Time PTER PTNER
Weekly hours 44.49 23.31 21.80
Hourly wage, $2013 17.02 11.81 13.66
Notes: The table shows mean of annual averages, 1994-2014. For
2014, the average is taken over the first eight months for which
the data are available at the time of publication. To calculate
hourly wage, we use hourly wages for hourly workers and compute
hourly wages for salaried workers by dividing usual weekly
earnings by usual weekly hours. Zero wages are dropped. All
calculations employ the CPS outgoing rotation group sampling
weights. Hourly wages are in 2013 dollars. Calculations are based
on the CPS microdata basic files.
Table 2 Education and Demographic Characteristics of Full-
and Part-Time Employment, 1994-2014
Group's Share In
Group Full Time PTER PTNER
Female 40.24 49.84 63.21
High school degree 91.71 79.88 85.14
Associate's, 41.44 21.95 33.01
Bachelor's or
higher degree
Master's, 10.91 4.06 8.11
professional, or
Doctorate degree
Average age 40.82 36.62 39.12
Under 20 y.o. 1.54 8.05 13.82
20-24 y.o. 8.18 17.52 13.26
25-34 y.o. 24.19 23.32 16.90
35-44 y.o. 26.80 21.15 18.54
45-54 y.o. 24.26 17.69 16.39
Over 55 y.o. 15.03 12.28 21.08
Notes: The table shows mean of annual averages, 1994-2014. For
2014, the average is taken over the first eight months for which
the data are available at the time of publication. Authors'
calculations using the CPS microdata basic files.
Table 3 Full-Time and Part-Time Employment Shares, by
Occupation 1994-2014
Non-routine cognitive Routine cognitive
Year Full time PTNER PTER Full time PTNER PTER
1994 0.781 0.196 0.024 0.713 0.251 0.036
1995 0.787 0.190 0.023 0.717 0.248 0.035
1996 0.786 0.192 0.022 0.715 0.251 0.033
1997 0.794 0.187 0.020 0.723 0.247 0.030
1998 0.775 0.207 0.018 0.706 0.267 0.027
1999 0.796 0.188 0.016 0.728 0.248 0.024
2000 0.811 0.175 0.014 0.742 0.236 0.022
2001 0.797 0.187 0.016 0.733 0.241 0.025
2002 0.802 0.180 0.018 0.735 0.236 0.029
2003 0.807 0.176 0.017 0.724 0.243 0.033
2004 0.805 0.179 0.017 0.719 0.249 0.032
2005 0.808 0.176 0.016 0.723 0.246 0.031
2006 0.809 0.176 0.015 0.727 0.245 0.028
2007 0.815 0.170 0.015 0.730 0.241 0.030
2008 0.815 0.167 0.019 0.726 0.235 0.040
2009 0.782 0.187 0.031 0.689 0.247 0.064
2010 0.799 0.169 0.031 0.701 0.229 0.070
2011 0.803 0.168 0.030 0.703 0.229 0.068
2012 0.809 0.162 0.029 0.707 0.228 0.065
2013 0.817 0.155 0.028 0.715 0.221 0.064
2014 0.814 0.160 0.025 0.710 0.230 0.061
Routine manual Non-routine manual
Year Full time PTNER PTER Full time PTNER PTER
1994 0.811 0.142 0.047 0.606 0.321 0.073
1995 0.812 0.142 0.046 0.616 0.316 0.068
1996 0.814 0.145 0.042 0.624 0.311 0.065
1997 0.824 0.137 0.039 0.629 0.309 0.062
1998 0.806 0.159 0.036 0.628 0.318 0.054
1999 0.828 0.139 0.033 0.643 0.309 0.048
2000 0.840 0.129 0.032 0.654 0.300 0.045
2001 0.825 0.137 0.038 0.644 0.305 0.051
2002 0.827 0.130 0.042 0.641 0.301 0.057
2003 0.830 0.122 0.047 0.624 0.307 0.068
2004 0.834 0.121 0.045 0.625 0.308 0.067
2005 0.838 0.121 0.041 0.632 0.305 0.063
2006 0.839 0.120 0.040 0.642 0.301 0.057
2007 0.840 0.118 0.042 0.638 0.303 0.058
2008 0.820 0.119 0.061 0.631 0.294 0.075
2009 0.765 0.136 0.099 0.593 0.293 0.114
2010 0.793 0.119 0.088 0.600 0.280 0.119
2011 0.798 0.121 0.081 0.600 0.281 0.118
2012 0.814 0.117 0.069 0.606 0.280 0.114
2013 0.822 0.115 0.064 0.609 0.279 0.112
2014 0.824 0.118 0.058 0.607 0.288 0.106
Notes: The table shows shares of FT, PTER, and PTNER in each of
the four occupational groups, annual averages of monthly series.
Non-routine cognitive occupations include management, business,
and financial occupations and professional related. Routine
cognitive occupations include sales and office occupations and
office and administrative support occupations. Routine manual
occupations include construction and extraction; installation,
maintenance, and repair; production; and transportation and
moving material occupations. Non-routine manual occupations
include service occupations. Authors' calculations using the CPS
microdata basic files.
Table 4 Hourly Wage, Demographic and Socioeconomic
Characteristics, 1994-2014
Variable Coefficient Variable Coefficient
Fulltime .1915106 *** Dummy 1996 -.0123845 ***
(.000) (.000)
Part time (PTNER) .0914258 *** Dummy 1997 -.0019732
(.000) (.348)
Construction/ .2215752 *** Dummy 1998 .0278192 ***
transportation (.000) (.000)
Production and repair .3131558 *** Dummy 1999 .0382212 ***
(.000) (.000)
Sales and related .0678285 *** Dummy 2000 .0424516 ***
(.000) (.000)
Office/administrative .2269783 *** Dummy 2001 .0544279 ***
support (.000) (.000)
Professional specialty .4906422 *** Dummy 2002 .0624557 ***
(.000) (.000)
Management/executive .3944512 *** Dummy 2003 .0585759 ***
(.000) (.000)
High school .1300564 *** Dummy 2004 .0471874 ***
(.000) (.000)
Some college .1753108 *** Dummy 2005 .0319903 ***
(.000) (.000)
College .3062498 *** Dummy 2006 .0284528 ***
(.000) (.000)
Graduate degree .2087941 *** Dummy 2007 .0293244 ***
(.000) (.000)
Age .0413028 *** Dummy 2008 .0228746 ***
(.000) (.000)
[Age.sup.2] -.0004105 *** Dummy 2009 .0470142 ***
(.000) (.000)
Black -.057996 *** Dummy 2010 .0310914 ***
(.000) (.000)
American Indian/ -.0468621 *** Dummy 2011 .0099154 ***
Alaskan Native (.000) (.000)
Asian -.0150465 *** Dummy 2012 -.0020422
(.000) (.372)
Other race -.0095699 *** Dummy 2013 -.0093538 ***
(.003) (.000)
Female -.1564769 *** Dummy 2014 -.0136535 ***
(.000) (.000)
Dummy 1995 -.0093544 *** Constant 1.260337 ***
(.000) (.000)
Mean(log real 2.630377
wages)
N 1,483,262
[R.sup.2] .4109
Notes: See footnote 13.
Table 5 Hourly Wage, Demographic and Socioeconomic
Characteristics with Occupation-Employment Type Interactions,
1994-2014
Variable Coefficient Variable Coefficient
Full time X .1823325 *** Black -.0578995 ***
Service (.000) (.003)
Full time X .398573 *** American -.046481 ***
Constmction/ (.000) Indian/Alaskan (.000)
transportation Native
Full time X .4920935 *** Asian -.01469 ***
Production (.000) (.000)
and repair
Full time X .2403244 *** Other race -.0098147 ***
Sales and (.000) (.000)
related
Full time X .410917 *** Female -.1569275 ***
Office/ (.000) (.000)
administrative
support
Full time X .6591924 *** Dummy 1995 -.0092787 ***
Professional (.000) (.000)
specialty
Full time X .5672558 *** Dummy 1996 -.0123659 ***
Management/ (.000) (.000)
executive
Part time .0716929 *** Dummy 1997 -.0019273
non-economic (.000) (.359)
(PTNER) X
Service
PTNERX .2472299 *** Dummy 1998 .0279902 ***
Constmction/ (.000) (.000)
transportation
PTNER X .4085932 *** Dummy 1999 .0381384 ***
Production (.000) (.000)
and repair
PTNER X Sales .1543721 *** Dummy 2000 .0425298 ***
and related (.000) (.000)
PTNERX .3011755 *** Dummy 2001 .0545344 ***
Office/ (.000) (.000)
administrative
support
PTNER X .5966031 *** Dummy 2002 .0623247 ***
Professional (.000) (.000)
specialty
PTNERX .5108225 *** Dummy 2003 .0584042 ***
Management/ (.000) (.000)
executive
PTERX .2226177 *** Dummy 2004 .0470143 ***
Constmction/ (.000) (.000)
transportation
PTER X .2740908 *** Dummy 2005 .0318157 ***
Production (.000) (.000)
and repair
PTER X Sales .060803 *** Dummy 2006 028199 ***
and related (.000) (.000)
PTER X Office/ .1890605 *** Dummy 2007 .0291946 ***
administrative (.000) (.000)
support
PTER X .4600883 *** Dummy 2008 .0227143 ***
Professional (.000) (.000)
specialty
PTERX .3030132 *** Dummy 2009 .0469685 ***
Management/ (.000) (.000)
executive
High school .1296502 *** Dummy 2010 .0310504 ***
(.000) (.000)
Some college .1751928 *** Dummy 2011 009801 ***
(.000) (.000)
College .3063701 *** Dummy 2012 -.0020417
(.000) (.371)
Graduate .2079177 *** Dummy 2013 -.0093253 ***
degree (.000) (.000)
Age .0409877 *** Dummy 2014 -.0135823 ***
(.000) (.000)
[Age.sup.2] -.0004068 *** Constant 1.278611 ***
(.000) (.000)
Mean(log 2.630377
real wages)
N 1,483,262
[R.sup.2] .4115
Notes: See footnote 14.
Table 6 Average Transition Probabilities, 1994-2014
To PTER To PTNER To FT To U To OLF
From PTER .3702 .2146 .3092 .0614 .0447
From PTNER .0437 .5744 .2995 .0179 .0645
From FT .0156 .0797 .8795 .0094 .0158
From U .0482 .0705 .1263 .5185 .2364
From OLF .0033 .0218 .0175 .0260 .9314
Notes: We take the mean of each yearly average, 1994-2014.
