New England job changes during the recession: the role of self-employment

Katharine L. Bradbury

One of the puzzles regarding the recent recession in New England was a divergence between the two major series on employment, one based on establishments and one based on households. Both show the recession to have been much more severe in the region than in the nation. But while the establishment series shows New England to have lost one in 10 jobs as compared with U.S. recession losses of less than one in 50, the household series shows the region's losses amounting to less than one in 13. The inclusion of unincorporated self-employed individuals in the household employment count but not the establishment series explains part of the divergence. The number of unincorporated self-employed individuals--defined as those who work in their own unincorporated business, profession, or trade--grew between 1988 and 1992, while all the other major classes of workers--farm and nonfarm private, government, incorporated self-employed, and those working without pay--shrank in the region. Some of the additional self-employed were undoubtedly former payroll employees who began working on their own because they lost establishment jobs.

The shift into self-employment represents one part of a set of changes in the mix of workers and jobs that reflects the nature of the region's downturn and the economic adjustments it entailed. This article examines patterns of job and income change for different classes of workers in New England from the pre-recession peak year of 1988 to the recession-low year of 1992, with an emphasis on the role of the self-employed.

Income data suggest that the self-employed fared better than the unemployed during the recession, but their earnings declined more, on average, than the earnings of individuals still working for other employers in 1992. Thus, self-employment apparently represented a successful stopgap measure, for some, to keep earning after the loss of a wage and salary job, but typically at a lower level. A key question is the degree to which these adjustments will be reversed as the New England economy recovers. Although data are not yet available to track the amount of self-employment in New England after 1992, a marked reversal since mid-1992 of the recession's divergence between the household and establishment series job counts suggests that self-employment may be shrinking in the region. Certainly those newly self-employed individuals who have low earnings and few fringe benefits would be expected to seek wage and salary employment as the number of establishment jobs expands. Nonetheless, anecdotal evidence suggests that a subset of the newly self-employed may be unlikely to resume payroll employment.

I. Overall Patterns of Employment Change in the Downturn

The count of payroll jobs at nonagricultural establishments in New England peaked in February 1989 and declined fairly steadily to a low in December 1991; it then moved very little until after June 1992 when it began growing. Losses from peak to trough amounted to 650,000 payroll jobs. When employment is measured based on a survey of households, rather than establishments, the number of employed persons in New England declined by about 500,000 from its peak in February 1989 to its trough in June 1992.(1) (See Appendix A for a more detailed examination of the differences between the two series' job tallies.)

While total employment declined markedly (according to either measure), one type of employment was growing--the unincorporated self-employed. Table 1 reports data collected during the Current Population Surveys of March 1989 and March 1993, which refer to the preceding calendar years.(2) The number of unincorporated self-employed workers expanded noticeably in New England while all other classes of workers shrank between 1988 and 1992.(3) Even the count of incorporated self-employed declined during the recession, presumably because the red tape involved in incorporating, while not onerous, makes incorporating a less fluid adjustment mechanism. The remainder of this article uses the term "self-employed" to refer to the unincorporated self-employed,(4) and characterizes the three remaining employment classes who receive pay as "employees," since the incorporated self-employed are considered (for tax purposes as well as survey purposes) to be employees of their corporations.(5) The relatively small class of unpaid volunteers and family workers is dropped from the analysis.

Table 1
New England Adults by Labor Force Status and Class of Worker
Thousands

                                                                 1988-92
Change Labor Force Status, Class of Worker     1988       1992       Number
  Percent

Employed Persons                        7,616.2    7,380.7    -235.5      -3.1

Private Employees                       5,904.8    5,698.8    -206.0      -3.5

Government Employees                      971.1      928.7     -42.4      -4.4

Self-Employed
 Incorporated                             243.4      197.7     -45.7     -18.8
 Not Incorporated                         478.4      547.4      69.1      14.4

Employed without Pay                       18.5        8.1     -10.4     -56.3

Persons with Zero Weeks Employment      2,683.6    3,030.1     346.5      12.9

Not in Labor Force (NILF) All Year      2,635.2    2,869.6     234.4       8.9
Unemployed All Year                        11.1       82.2      71.1     641.0
Mixed NILF and Unemployed                  37.3       78.2      41.0     109.9

Total Persons Age 15 and Older         10,299.8   10,410.7     110.9       1.1

Memo: "Employees"(a)                    7,119.3    6,825.2    -294.2      -4.1

Note: Employed persons are individuals employed at least part of the year.
Class of worker (private, government, etc.) refers to longest job in year.

a Private and government employees plus incorporated self-employed.

Source: Author's calculations based on data from U.S. Bureau of the Census,
Current Population Survey, March 1989 and March 1993, machine-readable data
files.

A sizable fraction of the 69,100 net additions to self-employment were probably individuals who had held a private or government job in 1988. Employment in the other classes of work (private, government, incorporated self-employed, and unpaid volunteer and family workers) fell by more than 300,000 between 1988 and 1992. Many appear to have left the labor force, while others became unemployed for at least part of the year.

While the bulk of New England's adjustment to employee cutbacks thus took the form of changes in labor force status, the increase in self-employment raises some interesting questions. Who were the self-employed in the region and how did their characteristics change during the slowdown? In which industries or occupations might those losing private and government jobs have gone into business for themselves, as evidenced by a decrease in the number of employees accompanied by gains in self-employment? What happened to the incomes of employees as compared with those of individuals in the growing pools of self-employed and unemployed? Did these additions to self-employment represent "good" jobs?

An interesting study of Vermont workers laid off by General Electric from 1988 through 1990 (Kessel and Maher 1991) found a high level of self-employment (38 percent) among exempt workers when follow-up contact was made in early 1991. Their findings provide a preview, in some dimensions, of the apparent effects of New England's general economic slowdown, to be described in this article. The authors speculate that these former G.E. workers' motivations for self-employment ranged from a desire to be one's own boss to a lack of other opportunities, especially for older workers. They found the self-employed former G.E. workers in a wide range of industries and occupations, including innkeepers, artisans, real estate agents, construction, car repair, and retailing.

Nine out of 10 of the self-employed respondents felt that "their current job was 'better' than their G.E. job" and they wanted to remain in it. However, neither earnings, work hours, nor fringe benefits made self-employment "better." Three-fourths were earning less than at G.E., most were working longer hours, and many fewer had fringe benefits.

II. Who Were the Self-Employed in New England and How Did Their Mix of Characteristics Change during the Recession?

Table 2 reports the industry composition of New England employment changes over the 1988-92 period, based on household survey data. In the late 1980s, the self-employed in New England, as in the nation, were concentrated in farming, construction, and services; substantial numbers also worked in retail trade, although the self-employed do not comprise a large fraction of retail workers. During New England's 1988-92 downturn, all the industry categories shown in Table 2 except for retail, some categories of services, and public administration saw cutbacks in the number of employees, while self-employment increased in a broad range of industries--all but natural resources, trade, and business services.(6)

The construction industry accounted for the largest numerical self-employment gains (the four years saw a net addition of over 30,000 self-employed construction workers). Nevertheless, total construction employment in the region shrank 20 percent (by more than 110,000 workers) over the four years. Presumably, a subset of the employees let go by construction firms set up their own small-scale construction TABULAR DATA OMITTED businesses. By contrast, the large professional services industry (which includes health, engineering, management, legal, and educational services) gained private and government employees as well as (indeed faster than) self-employed.(7)

Table 3 reports the composition of employees and the self-employed by broad occupational group. As in most recessions, the hardest-hit occupations were blue-collar jobs associated with the highly cyclical construction and manufacturing industries: precision production, craft and repair; machine operators, assemblers, and inspectors. The other occupational groups losing jobs over the four years were administrative support (including clerical) and sales occupations. The growth in self-employment was concentrated in executive, administrative, and managerial occupations, along with "other" (not private household, not protective) service occupations, and precision production, craft and repair (the mirror image of the construction industry gains in self-employment shown in Table 2). Apparently, many people who started up and ran their own (unincorporated) business classified themselves as executives or managers, unless they held credentials to offer a specific service or construction skill.(8) Technicians also made gains; they tend to be employees of private and government enterprises, not self-employed.

Table 4 indicates that virtually all of the self-employed in New England work in very small firms (defined here as those with fewer than 25 employees). Furthermore, all the growth in self-employment occurred in this size category.(9)

TABULAR DATA OMITTED

The self-employed are also somewhat more likely to work part-time or part-year than employees. Being one's own boss may allow some of the self-employed greater flexibility in setting a part-time work schedule; others of the self-employed may face inadequate demand to support full-time work.

As the economy deteriorated between 1988 and 1992, the fraction of the self-employed who worked less than the full year (fewer than 50 weeks of employment) rose noticeably.(10) When examining descriptive data such as these, it is important to keep in mind that two factors are at work in the changing characteristics of the self-employed: (1) changes in the characteristics of individuals who were self-employed throughout the period and (2) differences between the characteristics of those joining the ranks of the self-employed during the period and the initial group. As a case in point, Table 5 indicates that the rise in the fraction of the self-employed working part-year or part-time was not attributable to declines in the number of self-employed individuals working full-time or full-year, but rather was TABULAR DATA OMITTED entirely due to much faster growth in part-time/ full-year and full-time/part-year self-employment than in the number of self-employed with other work schedules. One interpretation of the large number of part-year self-employed in 1992 is that many of them were previous employees who experienced a spell of unemployment and then became self-employed.(11)

Indeed, as shown in the first panel of Table 6, a TABULAR DATA OMITTED sizable fraction of the net additional self-employed over the 1988-92 period were individuals who experienced a spell of unemployment more than two weeks long in 1992. In addition, for some fraction of the self-employed who were not unemployed in 1992, such a transition through unemployment may have occurred in 1989, 1990, or 1991.(12)

Table 6

New England Labor Force Status and Earnings

Adults with Earnings, by Class of Worker for Longest Job Held

                                                               Percent
                                              1988     1992    Change

Number (Thousands):
Adults Working 50+ Weeks:
Employees(a)                                  5,068    4,857    -4.1
Self-Employed                                   340      369     8.5
Adults Unemployed [is greater than]2 Weeks
Employees(a)                                    533      895    68.1
Self-Employed                                    23       59   152.1
Others in the Work Force(b)
Employees(a)                                  1,519    1,073   -29.4
Self-Employed                                   115      120     4.0
Average Annual Earnings (1993 Dollars):
Adults Working 50+ Weeks:
Employees(a)                                 32,050   31,050    -3.2
Self-Employed                                30,650   28,700    -6.4
Adults Unemployed [is greater than]2 Weeks   14,350   11,650   -18.7
Others in the Work Force(b)                  10,300    9,600    -7.1

Note: This table excludes adults with zero earnings or with no job all year.
Earnings include all earnings in calendar year; employment class refers to
longest job in year. Earnings inflated to 1993 dollars using the personal
consumption expenditure (PCE) deflator. Annual earnings rounded to nearest
$50.

a Private and government employees and incorporated self-employed.

b Individuals in the labor force for at least one week who were employed for
less than 50 weeks, unemployed for two or fewer weeks, and out of the labor
force for the remainder of the year.

Source: See Table 1.

III. Income Changes during the Regional Slowdown

The region's deep recession affected not only the employment status of New England residents, but also their incomes. Average annual earnings declined in real terms over the 1988-92 period in New England. Not surprisingly, real earnings declined the most for adults who were unemployed for part of the year (as shown in the lower panel of Table 6). But among those working all year (50-plus weeks), annual earnings declined twice as fast in real terms for the self-employed as for employees.

Real earnings declined overall in part because job losses meant that fewer adults had any earnings in 1992 than in 1988, in part because the fractions of earning adults who worked part-time and/or part-year rose, and in part because of a decline in real average hourly earnings among those with earnings. Average annual earnings of the self-employed declined almost 14 percent in real terms between 1988 and 1992, about four times the pace of real earnings losses for employees.

Average weeks employed declined more for the self-employed than for other workers; self-employed workers averaged more weeks of work in 1988 than employees, but by 1992 they averaged slightly less. Average weekly earnings also declined more steeply for the self-employed--over 10 percent as compared with less than 3 percent for other employees. Just as for annual earnings and weeks of work, the self-employed began with higher weekly earnings but fell below other workers by 1992. Furthermore, the usual workweek declined more for the self-employed than for other workers. Hourly earnings also declined more for the self-employed, but they still remained higher than the average hourly earnings of employees.

All the relative losses experienced by the self-employed that are shown in Tables 6 and 7 suggest that the circumstances of those who became self-employed during this period were not the same as the circumstances of either the initial pool of self-employed or those who retained their private, government, or incorporated self-employed jobs. Indeed, the workweek and earnings of the self-employed became more like those of the unemployed during this period, perhaps not surprisingly since some of the newly self-employed became so after a stint of unemployment. Even the self-employed working 50 or more weeks resembled the unemployed more in 1992 than they had in 1988.

Table 7

Real Earnings of New England Workers

Earnings in 1993 Dollars

                                                             Percent
                                    1988         1992        Change

Number (000s):
All Adults                         10,300        10,411         1.1
Adults with Earnings                7,598         7,373        -3.0
Employees(a)                        7,119         6,825        -4.1
Self-Employed                         478           547        14.4
Average Annual Earnings:
All Adults                        $19,200       $17,650        -8.1
Adults with Earnings               26,050        24,950        -4.2
Employees(a)                       25,950        25,050        -3.5
Self-Employed                      27,350        23,550       -13.9
Average Weeks Worked in Year:
Adults with Earnings                   45.1          44.7       -.9
Employees(a)                           45.1          44.7       -.9
Self-Employed                          46.0          44.6      -3.0
Average Weekly Earnings:
Adults with Earnings                $ 547.3       $ 528.8      -3.4
Employees(a)                          545.3         529.9      -2.8
Self-Employed                         576.2         515.5     -10.5
Usual Weekly Hours:
Adults with Earnings                   38.9          37.7       -.8
Employees(a)                           37.9          37.6       -.8
Self-Employed                          40.4          39.1      -3.2
Average Hourly Earnings:
Adults with Earnings                 $ 14.17       $ 14.02     -1.1
Employees(a)                           14.07         13.93     -1.0
Self-Employed                          15.64         15.25     -2.5
Percentage Poor:(b)
All Adults                              6.8           8.1       1.3(c)
Adults with Earnings                    3.1           4.6       1.5(c)
Employees(a)                            3.0           4.3       1.3(c)
Self-Employed                           4.9           7.8       2.9(c)

Note: Earnings include all earnings in calendar year: employment class refers
to longest job in year. Earnings inflated to 1993 dollars using the personal
consumption expenditure (PCE) deflator. Annual earnings rounded to nearest
$50. Average weekly earnings calculated as annual earnings divided by weeks
employed. Average hourly earnings calculated as average weekly earnings
divided by usual weekly hours.

a Private and government employees and the incorporated self-employed.

b "Poor" is defined as in a family with income below the poverty line.

c Percentage point change.

Source: See Table 1.

The relative declines in hours and weeks of work imply that the self-employed were not as successful at finding as much work as the average employee (with a job), and by inference, as much as they would have liked. These findings are thus consistent with the stories of self-employment as a stopgap measure, aimed at maintaining some earnings after the loss of a payroll job, albeit a stopgap that was less successful for some of the self-employed than for others. For instance, the data suggest that the average hourly earnings of the newly self-employed in 1992 were lower than those of employees, although they were probably higher than those of all earners with a spell of unemployment in 1992.(13) Thus, the good news is that some individuals who might otherwise have been entirely unemployed were garnering earnings through self-employment. The bad news, however, is that making ends meet was a struggle for a noticeable fraction of the self-employed.

The self-employed were considerably more likely than other employed individuals to be poor (as shown at the bottom of Table 7). Poverty rates reflect the combination of an individual's family situation (the presence of other workers and the size of their earnings and unearned income) and the individual's own earnings and unearned income. Consistent with the changes in real earnings, poverty rates also rose more for the self-employed between 1988 and 1992 than for other adults.

These changes in annual earnings are summarized in Figure 2, which displays the percentage of employees and self-employed workers falling into three annual earnings classes. All 1988 workers were ranked from lowest to highest annual earnings and then cutoffs between the classes were defined so that each class contained one-third of all 1988 workers. As the chart makes clear, in 1988 the self-employed were much more likely to fall into the bottom and top earnings classes than were employees, indicating greater earnings inequality among the self-employed.(14)

The same constant-dollar cutoffs between the classes, applied to 1992, indicate that the fraction of all workers in the bottom earnings class increased noticeably (but not by a large amount, rising from 33.2 percent in 1988 to 35.2 percent in 1992), with offsetting declines in the middle and top classes, as the weakening economy caused the entire earnings distribution to shift down in real terms. This decline in the number of workers with high earnings was much more pronounced for the self-employed: 37 percent of the self-employed were in the top class in 1988, but only 25 percent were in 1992, with some of the loss taken up by the middle earnings class and some by the bottom class.(15)

Although real earnings declined as the economy weakened between 1988 and 1992, job-related fringe benefit coverage changed very little. In 1988, the self-employed enjoyed much less health coverage than other workers, and they were much less likely than other employees to be included in their employer's or union's pension or other retirement plan. The large number of self-employed with health coverage not offered by their own employer or union include those purchasing health insurance individually and those with coverage through other employed family members. Since the self-employed are their own employers, and most are in very small firms, these facts may be unsurprising.

While the fraction of adults covered by private health insurance actually rose slightly in New England over this period, the fraction with health insurance offered by their own employer or union declined noticeably overall and especially for the self-employed. Access to pension or other retirement plans, by contrast, expanded over the period for both the self-employed and employees. Even after these gains, however, the self-employed were much less likely than employees to be included in a retirement plan.

Such changes in available benefits as well as income and employment status of self-employed individuals in New England obviously reflect recession-induced shifts in the demand side of the labor market, as well as supply. Many firms cut back employment as demand for their products declined, and some former employees responded to these cutbacks by employing themselves. A second-round effect was that some firms explicitly opted for shifting some production to non-employees. Observers have noted increased "outsourcing" of a variety of economic adjustments (during both slack and boom periods) in the United States, and establishments' increased reliance on the self-employed, from executive consultants to cleaning staff, may be a case in point, at least in New England. A related phenomenon is many industries' increased use of temporary workers.(16) (See the Box for a description of temporary workers in New England.)

Table 8

Fringe Benefits of New England Workers

Percent or Percentage Points

                                                            Percentage
                                                              Point
                                    1988        1992         Change

Percentage with Government
Health Coverage:(a)
All Adults                          21.5        22.9         1.4
Adults with Earnings                 7.9         7.4         -.5
Employees(b)                         7.5         7.5          .0
Self-Employed                       13.7         6.0        -7.7
Percentage with Private
Health Insurance:
All Adults                          75.3        75.9          .6
Adults with Earnings                82.0        83.1         1.1
Employees(b)                        82.9        84.1         1.2
Self-Employed                       70.1        71.1         1.0
Percentage with Health
Insurance Offered by Own
Employer or Union:
All Adults                          46.0        42.7        -3.3
Adults with Earnings                57.5        54.4        -3.1
Employees(b)                        59.9        57.2        -2.7
Self-Employed                       22.6        19.1        -3.5
Percentage Whose Employer or
Union Had a Pension Plan:
All Adults                          37.8        38.1          .3
Adults with Earnings                51.1        53.9         2.8
Employees(b)                        54.0        57.3         3.3
Self-Employed                        8.5        11.3         2.8

a Government coverage includes Medicare, Medicaid, CHAMPUS,
VA, or military health care coverage.

b Private and government employees and incorporated self-employed.

Source: See Table 1.

The supply response of some workers who lost establishment jobs--dropping out of the labor force or offering self-employed labor--thus met an altered mix of employer demand for employees, altered both by recession shrinkages and by structural changes such as the move toward outsourcing. For players on both sides of the market, these changes have advantages and disadvantages. Some of the self-employed might like to go back to being employees, while others may value more highly the independence of self-employment. Similarly, hiring consultants or temporary workers looks to be a long-term money-saver for some employers but others plan to sign on additional permanent employees as soon as product demand picks up enough to justify it.

IV. Will Self-Employment Shrink as New England Recovers?

Given these pluses and minuses for both employers and workers, a key question, looking forward, is the degree to which the recession's growth in self-employment will be reversed as the New England economy expands. Current Population Survey data comparable to those used here are not yet available for 1993 or 1994. Thus, it is impossible to directly estimate changes in the number of self-employed individuals in New England in recent years. Nonetheless, changes in the monthly payroll and household employment totals in the last two years may provide some clues regarding possible changes.

Since the recession low point in mid-1992, measured job growth has been much faster according to the establishment series than according to the household series. The former shows a net gain of 215,000 jobs from June 1992 to June 1994 and the latter less than 90,000. Just as a gain in self-employment was (correctly) inferred from the divergence of the two measures during the recession (Bradbury 1993), it seems reasonable to conclude from their recent convergence that some shrinkage in self-employment has occurred over the two years of recovery in the region.(19)

Many of the newly self-employed whose self-employment earnings (and fringe benefits) fell short of their previous wages or salaries (and fringes) can be expected to seek establishment jobs as they become available. Nonetheless, two groups seem less likely to resume payroll employment: Some self-employed individuals may have been "scarred," becoming less attractive to employers because of their stints of unemployment or self-employment; others, like many of the self-employed former Vermont G.E. workers (Kessel and Maher 1991), may have realized that they strongly prefer being on their own. In addition, some establishments will continue to find it advantageous to "outsource" selected operations or functions to self-employed individuals.

V. Conclusions

The economic adjustments engendered by the severe 1989-92 downturn in New England fell heavily on those who lost their jobs. All classes of employees--private, government, the incorporated self-employed--shrank in size from 1988 to 1992; the only net growth occurred in unincorporated self-employment. These shifts imply that some individuals began working for themselves as a way to maintain at least some earnings after losing their establishment jobs.

Anecdotal evidence suggests that the recession's effects on New England labor markets will not unravel totally even when the region's severe employment losses have been reversed.(20) Changes in the way establishments view their operations, including relationships with outside suppliers, wariness about adding to "permanent" payrolls after so recently facing the pain of large-scale separations, and changes in individual attitudes towards self-employment are likely to constrain the amount of establishment hiring that is associated with increases in output. Thus, the regional economy may well continue to have a somewhat larger class of independent workers than would have occurred in the absence of the recession and the adjustments it engendered.

1 These household data are adjusted (additively) for a January 1990 discontinuity that is due to rebenchmarking by the U.S. Bureau of Labor Statistics to the 1990 Census baseline. Without this adjustment, the 1989-90 job losses according to the household series would be even smaller, and the measured discrepancy between the household and establishment peak to trough losses would be even bigger than the text indicates.

2 The overall employment losses according to the annual household data shown in Table 1 are noticeably smaller than those based on the monthly data cited in the preceding text paragraph. This is partly because the individual month in which employment peaked obviously showed a higher job count than the average for the year in which employment peaked and, by the same token, the monthly trough was lower than the annual trough. In addition, the annual data, as tallied here, count as employed any adult working at least one week during the year, while the monthly data refer to an individual's employment status in the current month.

3 Comparable data for the United States do not show unincorporated self-employment as the only growing class of worker. The numbers of private

employees, government employees, and unincorporated self-employed expanded in the United States from 1988 to 1992 (by 1.0, 5.3, and 4.2 percent, respectively), while the numbers of incorporated self-employed and employed without pay shrank (by 0.2 and 30.7 percent, respectively).

4 The only exceptions to this statement are some explicit references to the incorporated self-employed which are clearly labeled as such.

5 Ample precedent exists for this grouping. In summary Census classifications, private and government workers are considered "wage and salary" workers, and self-employed incorporated are included with private workers. In Census income data, the earnings of the incorporated self-employed are reported as wages and salaries, not self-employment earnings. Furthermore, the incorporated self-employed are tallied as wage and salary employees in the establishment series.

6 The Current Population Survey's March 1989 and March 1993 samples include just under 10,000 adults in New England to represent just over 10 million adult residents. While this is an ample sample for examining broad trends, when cross-tabulations are undertaken, some cells are quite small and therefore some percentages are unreliable (they have large sampling errors). For example, only six of the New England observations for 1992 were mining employees. For this reason, mining was combined with agriculture, forestry, and fisheries in Table 2. Small sample size probably also accounts for the extremely large measured growth in the self-employed handlers, etc. occupation group shown in Table 3.

7 The "employee" gains in professional services are not attributable to the inclusion of incorporated self-employed with private and government employees. Indeed, the number of incorporated self-employed individuals in professional services shrank between 1988 and 1992, but gains in private and government employees more than offset these losses.

8 The number of executives, administrators, and managers also expanded for the incorporated self-employed, but these gains were much smaller than those for private and government employees in these occupations which, as the table makes clear, were in turn much smaller than those of the unincorporated self-employed. The number of incorporated self-employed in professional specialty occupations shrank, just as (indeed, much faster than) did the number of unincorporated self-employed with professional specialties.

9 Note, however, that the category of very small firms can grow for two reasons: (1) new firms start up and are small and (2) larger firms shrink and are then counted in the small category. These explanations, especially the latter, actually apply to all size classes except the largest. Furthermore, a sizable part of this difference in average size of firm between the self-employed and employees must be attributable to the fact that, by definition, it is not possible to have more than one or several (if partners) self-employed persons working in any one firm, with all the remaining workers being counted as employees.

10 In 1988, the average person for whom self-employment was the longest job spent 46 weeks working, 1 week unemployed, and 5 weeks not in the labor force; by 1992, the average self-employed New Englander's weeks at work had fallen to 44.6 and weeks unemployed had risen to 2.6. About 6 percent of the self-employed in New England experienced one or more weeks of unemployment in 1988, a figure that had risen to 11 percent by 1992 (slightly below the corresponding percentages for employees).

11 Note that "weeks worked" includes all jobs held during the calendar year, while "self-employed" and "employee" status refer to the longest job during the year.

12 Recall that the Current Population Survey does not follow specific individuals over the entire 1988-92 time span. Rather, the population and its characteristics are estimated on the basis of a sample that would have turned over completely between 1988 and 1992.

13 The rough estimate behind this statement assumes that 478,000 of New England's self-employed (the number in 1988) saw their hourly earnings decline at the 1988-92 pace of employees' hourly earnings, and calculates the hourly earnings implied for the 69,000 net new self-employed, given the average hourly earnings shown in Table 7 for all the 1992 self-employed. Since the changes experienced by both employees and the self-employed actually ranged across a distribution, this "average" calculation is only illustrative.

14 These data confirm the hints of a more unequal distribution of earnings provided by the fact that the self-employed had higher average earnings but also a higher poverty rate than private and government employees in 1988.

15 In fact, the actual number of self-employed individuals with high earnings declined by one-fifth between 1988 and 1992 even as total self-employment increased.

16 Other documented examples of such outsourcing include auto manufacturers' requirements that their suppliers absorb price cuts and the costs of "just-in-time" inventory practices.

17 Data on industry (and hence temporary worker status) are not available for individuals who were unemployed all year or not in the labor force all year. Such individuals report no "longest job" for which industry can be ascertained.

18 New England workers for whom personnel supply services was the industry of longest job averaged 33 weeks of work in 1992, 12 weeks of unemployment, and 7 weeks out of the labor force. Over 40 percent of temporary workers experienced one or more weeks of unemployment in 1992, as compared with fewer than 14 percent of workers in other industries.

19 Since the region's unemployment rate has fallen to about the national average but not significantly below, it is unlikely that a marked increase in dual job-holding explains the convergence.

20 This will not happen soon, in any case. According to New England Economic Project forecasts, total employment in New England will not exceed its 1989 peak until almost 10 years later, in the final quarter of 1998 (Kodrzycki 1994).

21 This is not the case for the nation; both the establishment and household counts show U.S. job losses of 1.8 million from peak to trough.

22 This discrepancy is quite large, but at the time it appeared to be even larger. Each spring the U.S. Bureau of Labor Statistics revises the establishment employment estimates for the preceding two calendar years on the basis of more complete data that are available only with a lag. In March of 1994, these revisions to the 1992 and 1993 data were very sizable and positive. Thus, before those revisions, the total recession job losses were even bigger, according to the establishment series, than those reported in the text. Note also that the loss figures in Table A-1 do not match the figures cited in the text at the beginning of the article, because of the difference between monthly and annual data discussed in footnote 2.

23 The Current Population Survey (household series) for the month of March includes data for the longest job each individual held in the calendar year prior to the March survey date. These data, by industry, are compared with the corresponding annual averages of establishment data in Table A-1.

24 Service jobs declined only briefly during the recession even in hard-hit New England, peaking in March 1990 and bottoming out in February 1991.

25 A second important difference between the two sources is that dual job holders are counted as two jobs in the establishment survey but one employed individual in the household survey. A further source of difference is that both series are based on samples, one of households and one of establishments. The establishment survey includes nearly all the large establishments in the region; of necessity, a smaller fraction of small firms is used to represent that universe.

26 Other researchers (Devine 1993, 1994; Aronson 1991) have found a long-term shift in the composition of U.S. self-employment toward women and some minorities. That research, however, does not specifically track the recession period. As the text notes, the male percentage of the self-employed in the United States rose between 1988 and 1992 (as in New England, but counter to the long-term U.S. trend), while the white percentage fell slightly (unlike what occurred in New England, but in line with the long-term U.S. trend).

References

Aronson, Robert L. 1991. Self-Employment: A Labor Market Perspective. Ithaca, New York: ILR Press.

Bradbury, Katharine L. 1993. "Shifting Patterns of Regional Employment and Unemployment: A Note." New England Economic Review, September/October, pp. 3-12.

Devine, Theresa J. 1993. "The Recent Rise in U.S. Self-Employment." Pennsylvania State University, Department of Economics, Working Paper no. 2-92-3 (August).

-----. 1994. "Characteristics of Self-Employed Women in the United States." Monthly Labor Review, vol. 117, no. 3 (March), pp. 20-34.

Kessel, Herbert, and Frederick Maher. 1991. "Worker Readjustment to G.E. Layoffs." Prepared for Vermont Department of Employment and Training by Saint Michael's College, Center for Social Science Research. Fall.

Kodrzycki, Yolanda K. 1994. "New England Economic Outlook." New England Economic Project Economic Outlook 1994-1998. May, pp. 9-14.

U.S. Bureau of the Census. 1993, 1989. "Current Population Survey," machine-readable data files. March.

U.S. Bureau of Labor Statistics. Various dates. "Nonfarm Payroll Employment," machine-readable data files. Monthly.

TABULAR DATA OMITTED

Appendix A

Discrepancies between Household Series and Establishment Series Job Counts, 1988 to 1992

During the severe 1989-92 downturn in New England, the two basic measures of employment, based on surveys of establishments and households, gave very different readings on the magnitude of the region's employment losses.(21) From the pre-recession peak year 1988 to the recession-trough year of 1992, the establishment series reported the loss of 581,400 nonfarm jobs while the household series data, after excluding jobs in agriculture, forestry, and fisheries, indicated that "only" 208,700 jobs were lost.(22)

Just as for total employment, recession losses by industry were generally greater according to the establishment series than according to the household series.(23) The biggest differences occurred in two of the three biggest industries--trade and services. According to the establishment tally, almost 180,000 payroll jobs were lost in wholesale and retail trade between 1988 and 1992, but the household survey found only 8,000 fewer people reported employed in this industry in 1992 than in 1988. The number of jobs in services actually expanded over this period(24)--by 177,000 according to the household series but only 77,000 according to the establishment count. Manufacturing, the second largest industry group, showed the biggest job losses according to either source; construction, a relatively small industry, also showed large losses according to both series. Employment in finance, insurance, and real estate declined almost 9 percent according to the establishment series but grew about 4 percent according to the household-based estimates.

One important difference between the two job counts is that the household data are more inclusive: The establishment data include only individuals on the payrolls of nonfarm establishments, while the household measure includes, in addition, the unincorporated self-employed, unpaid volunteers and family workers, and agricultural workers.(25) Table A-2 reports 1988-92 changes in these TABULAR DATA OMITTED components of the household-based measure of employment, comparing several subsets with the establishment series totals. With no adjustments, the household measure of all employed individuals was 16 percent larger than the and fisheries) from the household measure reduces the discrepancies to 14 and 21 percent. The gaps are reduced even further, to 7 and 13 percent, by dropping the unincorporated self-employed from the household total.

While the differences in employment counts in the two years are thus about half explained by these known differences in definition, the measured employment losses remain much smaller according to the household series (even the most comparable measure) than the establishment series. Thus, the different treatment of self-employed individuals, farm workers, and those working without pay contributes to the discrepancy between the two employment measures, but these factors do not explain it. The bulk of the difference between the two measures of loss must be attributable to inconsistencies between the two series, including sampling techniques and misreporting by employers or individuals.

TABULAR DATA OMITTED

Appendix B

The Demographic Characteristics of the Self-Employed

Table B-1 takes another cut at describing the self-nonfarm establishment measure in 1988 and 23 percent larger in 1992. Dropping individuals working without pay and workers associated with fanning (agriculture, forestry, employed, summarizing their personal characteristics rather than the characteristics of their jobs (as was done in text Tables 2 through 5). In general, the self-employed in New England were more likely to be male and white than were employees; and the self-employed were also typically somewhat older and more educated than employees. The self-employed in the United States also show a higher fraction male, white, age 25 and older, and with at least some college than U.S. employees.

In New England, these four compositional differences between the self-employed and other employees became more pronounced between 1988 and 1992. U.S. self-employment data also show a 1988-92 shift toward men and toward greater age and education, but away from whites.(26)

Temporary Workers

Another phenomenon often seen as closely related to growing self-employment is the increasing use of temporary workers. Nationally, a sizable fraction of recent employment growth has been attributed to expanded employment in the "personnel supply services" category of the business services industry; firms in this industry employ workers in a variety of occupations, whom they then supply on an hourly, weekly, or longer-term basis to firms in a variety of industries. In New England as of 1992, about 50,000 employees of personnel supply services comprised about two-thirds of 1 percent of all workers; the industry was comparable in size to the apparel, aircraft, or security brokerage and investment industries in the region.

Temporary workers in New England were considerably more likely than workers in other industries to have experienced a mixture of employment and unemployment in 1992 or a mixture of employment and "not in labor force" status;(17) employees in other industries were much more likely to have been employed for 50 or more weeks in the year. About 40 percent of temporary workers looked for work in one or more stretches, while this was true of only 13 percent of other industry workers.(18) By the same token, one-third of temporary workers had more than one employer (at separate times) during the year, while only one-eighth of other industry workers moved among employers. Temporary workers, unlike the self-employed, were more likely than workers in other industries to be female and nonwhite; like the self-employed, an above-average fraction of employees in the personnel service industry worked part-time, defined as usually working fewer than 35 work hours per week.

Workers supplied by personnel agencies represented a wide range of occupations but were considerably more concentrated than workers in other industries in administrative support occupations, which include the clerical workers many people associate most strongly with temporary help. Temporary help workers in New England were also over-represented in two of the blue-collar groups: machine operators, assemblers, and inspectors; and handlers, equipment cleaners, and helpers. Interestingly, a slightly above-average fraction of those in the personnel service industry were in executive, administrative, and managerial occupations.

Employees in the personnel supply service industry earned less per year, on average, and were less likely to have fringe benefits associated with their jobs than workers in other industries. Total annual earnings averaged $18,500 for temporary workers and $25,000 for workers in other industries in New England in 1992 (earnings in 1993 dollars). Most of this difference in annual earnings was attributable to shorter hours and fewer weeks of work, however; average weekly and hourly earnings of temporary workers were much closer to those of workers in other industries. Only one-fifth of temporary help workers in New England had health insurance offered through (and partly or fully paid for by) their employer or union, as compared with over half the workers in other industries. And only 16 percent of personnel supply service workers were included in a company or union pension plan, as compared with 41 percent of workers in other industries.

Katharine L. Bradbury Assistant Vice President and Economist, Federal Reserve Bank of Boston. The author thanks Lynn Browne, Yolanda Kodrzycki, and Jane Little for helpful comments and Catherine Jew for able research assistance.

COPYRIGHT 1994 Federal Reserve Bank of Boston
COPYRIGHT 2004 Gale Group