Over time, a significant change in the relationship between male andfemale unemployment rates has occurred. Between 1970 and 1981, thefemale unemployment rate averaged 1.
5 percentage points higher than themale rate. However, in 1982, the male unemployment rate (9.9 percent)exceeded the female rate (9.4 percent) for the first time since suchdata were recorded beginning in 1947. This reversal in unemploymentrates is the apparent culmination of a narrowing of the differentialthat began in 1978. (Shee chart 1.) Although male unemployment rates generally increase more thanfemale rates during recessions (see the shaded areas in chart 1), therelative worsening experienced by men during the 1981-82 recession wasgreater than in previous downturns.
(And, as noted, the female-maleunemployment rate differential began to narrow prior to the recession,which is inconsistent with historical patterns.) Are we witnessing along-term improvement in the unemployment situation of women relative tomen? To what extent are the observed changes due to trends ininterindustry growth rates in employment which may favor one sex overthe other? This article addresses these questions using a modified version of shift-share analysis (see appendix A) to estimate the effectthat change in employment patterns among industries has had on thefemale-male unemployment rate differential since 1964, and to projectlikely future effects through 1995. Shirt-share analysis is commonlyused to disagregate regional employment change in an industry in orderto identify the components of that change. The applciation ofshift-share analysis in this article, however, is to disaggregate annualchanges in the male-female unemployment differential into threecomponents. Many researchers have observed the procyclical nature of thefemale-male unemployment rate differential. Because men tend to beconcentrated in those industries which are most sensitive to thebusiness cycle (particularly manufacturing, construction, and mining),it is not surprising that male unemployment rates rise relative tofemale rates during recessions and fall during recoveries.
Butindustries also change their employment requirements in response toforces other than the business cycle. For example, in recent years,automobile and steel manufacturing employment has experienced a secular decline because of increased foreign competition and laborsaving technological changes. Such longer term trends have an impact onunemployment differentials between men and women.
The effect that the growth (or decline) of a given industry has onthe female-male unemployment rate differential depends on severalfactors, inclunding: * The rate of growth (or decline) of the industry. * the percentage of total employment in the industry which isfemale (or male); * the interindustry mobility of men and women in response tochanges in employment opportunities in the industry; and * the labor force mobility of men and women in response to changesin employment opportunities in the industry. Information on the first two factors is presented in table 1. Itshows the average annual rate of growth of employment in nine broadlydefined industries during 1964-82. Clearly, employment grew mostrapidly in those industries which employ the highest proportion ofwomen, particularly services, and finance, insurance, and real estate.
This trend in industry growth rates has contributed to the narrowing ofthe female-male unemployment rate differential. However, it is importantto note that the mobility of men and women between industries and intoand out of the labor force must be “less than perfect” forchanges in the industrial composition of employment to have an effect onthe unemployment differential. Otherwise, an increase in unemployment inan industry would quickly be offset by the movement of unemployedworkers to other industries (interindustry mobility) or by an exit ofunemployed workers from the labor force (labor force mobility).Industry growth differentials would then have no direct effect on maleand female unemployment rates. With perfect mobility, men who losetheir manufacturing jobs would quickly join the growing serviceindustries or drop out of the labor force.
Research has shown, however,that unemployed men and women do not exhibit perfect interindustry andlabor force mobility. in sum, it appears that the four factors previously cited wouldtend to decrease the female-male unemployment rate differential. First,female-dominated service-producing employment is growing faster thanmale-dominated goods-producing employment. Second, becauseinterindustry and labor force mobility is less than perfect, variationsin employment demand will influence unemployment rates. The trendtowards slower goods-producing growth rates relative to servicesimplies, then, that the recent reversal in the female-male unemploymentrate differential could be the result of secular growth differentialsamong industries as well as the severe recession.
The following section presents a shift-share technique which isused to measure the effects of relative changes in industry employmenton the female-male unemployment rate differential from 1964 to 1982. Ina subsequent section, this technique is applied to BLS employmentprojections to predict how expected future trends in industry employmentgrowth would affect female-male unemployment rate differentials. Theappendices develop the methodology in greater detail. Components ofchange in differentials Shift-share analysis has frequently been used to analyze the sourceof regional employment growth, but seldom to disaggregate the componentsof change in unemployment differentials. (See table 2.
) The purpose ofthe shift-share analysis is to dissect the year-to-year change in thefemale-male differential into three components: national share effect,industry mix effect, and employment shift effect. The sum of theseeffects equals the total change in the unemployment differential. Theanalysis starts with very restrictive assumption regarding labor forceand interindustry employments trends and proceeds to relax theseassumptions one at a time. National share effect. This effect is computer by assuming thatmale and female employment in each industry changes at the same rate astotal national employment. The male and female labor forces are eachassumed to grow at the same rate as the total labor force. The nationalshare effect shows how the female-male unemployment rate differentialwould have changed from year to year if: (1) the proportion of men andwomen in each industry remained unchanged, (2) the proportion of men andwomen in the labor force remained unchanged, (3) the share of eachindustry’s employment in total employment was constant, and (4)total employment and the labor force grew at their actual rates. Under these assumptions, male and female employment and laborforces change at the same rate.
Because the unemployment rate isdefined as: 1 – number employed/, number in labor force this results inproportionate changes in male and female unemployment rates. Thenational share effect on the female-male differential is thusprocyclical but trivial in magnitude. Industry mix effect. To calculate the industry mix effect, theassumption that each industry grows at the national rate is dropped.Employment in each industry is postulated to grow at its actual rate,but it is assumed that the proportion of men and women employed in eachindustry remains the same as in the previous period. If employment infemale-male-dominated industries, as appears indicated in table 1, theindustry mix effect will reduce the unemployment rate of women relativeto that of men.
When employment increases in an industry, the additional workerswill be drawn into employment from the ranks of the unemployed and fromoutside the labor force. Therefore, an assumption is needed about howthis effect changes the labor force. It is assumed that men and womenwho “enter” employment as a result of the industry mix effectcome from the unemployment pool and from outside the labor force in thesame proportions as they actually did during the previous year.Similarly, when the industry mix effect causes a decrease in employment,it is assumed that men and women who exit employment leave the laborforce or become unemployed in the same proportions as they actually didduring the previous year. (This procedure is discussed in detail inappendix A.) It is also assumed, in the computation of theinterindustry effect, that there is no net interindustry mobility oflabor. The industry mix effect shows how differing industry growth ratesaffect the female-male unemployment rate differential when there aredifferent percentages of men and women in each industry.
(See table 2.)When the effect is negative, female-dominated industries are growingfaster (or declining less) than male-dominated industries, reducing thefemale-male unemployment rate differential. When the effect ispositive, male-dominated industries are growing faster (or decliningless) than female-dominated industries, thereby increasing thedifferential.
The industry mix effect appears to have both a cyclical componentand a secular trend. The cyclical component is suggested by theindustry mix effect always being negative during recession (for example,1970-71, 1974-75, and 1981-82) and positive only during expansions.This is because employment is more cyclically variable in male-dominatedindustries than in those which are female-dominated. For example, thethree industries that are most sensitive to the business cycle (mining,construction, and manufacturing) are very much male-dominated. (Seetable 1.) The industry mix effect shows smaller positive changes in eachsuccessive expansion and generally larger negative changes in eachsuccessive recession, which suggests that there may be a long-term trendwhich lowers female unemployment rates relative to male rates. (Seetable 2.
) To determine whether there is a significant trend in theindustry mix effect which is independent of the business cycle, aregression equation was estimated for the 1964-82 period which predictsthe impact of the business cycle (as measured by the help-wantedadvertising index) and trend variables on changes in the industry mixeffect over time. The regression results presented in appendix B, showthat the trend and the business cycle were both highly significantpredictors of change. After controlling for cyclical effects, thefemale-male differential declined on average by about 0.
2 percentagepoints per year. These results indicate that the differentialemployment growth rates of industries have tended to favorfemale-dominated industries and that this has caused a narrowing in thefemale-male unemployment rate differential, even after accounting forthe short-term effects of the business cycle. Employment shift effect. This effect is the change in themale-female unemployment differential that remains after accounting forthe national share and industry mix effects. Two factors determine thesign and the magnitude of this effect. The first is the difference inthe rates of growth in the male and female labor force. The fact thatthe female labor force has been expanding more rapidly than the malelabor force tends to cause unemployment rates of women to be greaterthan those of men.
The second factor which determines the employmentshift effect is the change in the male-female employment compositionwithin industries. If an industry increases the proportion of women itemploys, the unemployment rate of women will decrease relative to thatof men. The following tabulation presents the proportion of womenemployed in each industry during 1964 and 1982 and the average annualpercentage change in that proportion. These data show significantdifferences among industries in the rates at which the proportions offemale employment have increased. The employment shift effect can be thought of as representing theability of industries to respond to changes in labor force participationrates of men and women by altering the distribution of their employmentbetween sexes. Perfect accommodation to changes in labor forceparticipation would result in an employment shift effect which equalszero.
However, if the share of female employment within industries doesnot rise by enough to accommodate the increase in female labor forceparticipation, the employment shift effect would be positive. Thiswould tend to increase female unemployment rates relative to the malerate. And finally, where the share of female employment in the industryadvances by more than enough to accommodate the increase in female laborforce participation, the employment shift effect would be negative.This would tend to decrease the female unemployment rate relative to themale rate.
We note that the employment shift does not exhibit the same kind ofcyclical behavior as the industry mix effect. For example, during the1970-71 and 1974-75 recessions, the employment shift effect favored men,but during the 1980 and 1981-82 recessionary period it favored women.(See table 2.) This is a potentially important development because itmay represent a change in the ability and willingness of individualindustries to absorb women into employment. Regression results show,however, that on average, during the 1964-82 period, the employmentshift effect shows no significant trend or cyclical response. (Seeappendix B.) In recent years (1979-82), all three effects–the national shareeffect, industry mix effect, and employment shift effect-contributed toreducing the female-male unemployment rate diffeential. The industrymix effect indicates that, as in previous recessions, the 1980 and1981-82 downturns affected male-dominated industries more severely thanfemale-dominated ones.
But there is also a trend in the industry mixeffect independent of the business cycle. This means that long-termindustry-specific employment trends have favored women’s employmentbecause of their greater concentration in those industries with thehighest long-term growth rates. Finally, an examination of theemployment shift effect shows that since 1979 many industries morereadily employed women entering the work force, but that there has beenno such long-term trend. Employment projections Will employment trends continue to improve the unemploymentsituation of women relative to men? The preceding analysis suggeststhat this will depend to a large extent on the future growth rates offemale- versus male-dominated industries. The Bureau of LaborStatistics projections of employment by industry make it possible toanalyze the probably impact of the industry mix effect on the future ofthe female-male unemployment diffeential.
Table 3 presents the averageannual rates of change in projected employment between 1982 and 1990 andbetween 1990 and 1995. The BLS made three sets of projections for eachtime frame: the first assumes low rates of economic growth; the second,moderate growth rates; and the third, high growth rates. Valerie A.
Personick describes the moderate growth scenario as follows: Table 3 shows significant differences in projected employmentgrowth rates among industries under each of the three growth scenarios.It also indicates that, expect for the construction industry, women arecurrently overrepresented in the high-growth-rate industries (forexample, services and finance, insurance, and real estate). Womenrepresent only 25.5 percent of total employment in the five industrieswhich are projected in the moderate scenario to grow by 13.
2 percentbetween 1982 and 1990. However, women constitute 51.6 percent ofemployment in the four service-oriented industries projected to increaseby 18.9 percent.
It appears that future trends in employment willcontinue to favor a reduction of the unemployment rates of womenrelative to men’s. What are the implications of these trends for the female-maleunemployment rate differential? The following tabulation presents theresults of a partial shift-share analysis of changes in female-maleunemployment rate differentials which would occur between 1982 and 1990and between 1982 and 1995 under each of the three economic growthscenarios: Because BLS does not project male and female employment byindustry, it is possible to calculate only the industry mix effect. Itscomputation assumes that employment in each industry grows at itsprojected rate and that the proportions of men and women in eachindustry remain at the 1982 levels. Also, male and female labor forceentry and exit patterns are assumed identical to those of 1982.
Underthese assumptions, the female unemployment rate would decrease by about2 percentage points relative to the male rate between 1982 and 1990 andwould decrease by approximately 2.4 percentage points between 1982 and1995. The industry mix effect would continue its 1964-82 trend,exerting downward pressure on the female-male unemployment ratedifferential by about 0.
2 percentage points per year. It should be noted that the impact of the changing industry mix onthe differential is likely to be modified by seveal factors which arenot measured in the partial shift-share analysis. First, the BLSprojections of employment growth between 1982 and 1995 do not allow forcyclical variation, apart from the current recovery. The results for1964-82 imply that the industry mix effect is strongly affected by thebusiness cycle, and thus the results reported in the tabulationrepresent only the trend component of this effect. There willundoubtedly be substantial year-to-year cyclical variation in thefemale-male unemployment differential during 1982-95.
Second, maleinterindustry mobility may increase over past rates as the relativessecular decline in goods-producing industries continues. Men mayincrease their employment share in the rapidly growing industries,decreasing their projected unemployment rate. Third, female labor forceparticipation rates will continue to rise during the next decade, andwomen’s attachment to the labor force has also been increasing.These factors would tend to boost female unemployment rates over theirindustry mix levels. Both of these trends–the possible rise in themale share of rapidly growing industries, and the continuing increase inthe female participation rate–would be reflected in a positiveemployment shift effect over the 1982-95 period. Still, the projected relative secular decline in goods producingindustries will tend to increase the male unemployment rate relative tothe female rate at least in the near term. There is no recent evidencethat the employment shift effect will offset this negative industry mixeffect.
On the contrary, in 4 of the 5 years since 1978, the employmentshift has been negative. The most plausible scenario for thefemale-male unemployment rate differential is for the male rate to dropbelow the female rate during the current cyclical recovery, and for thefemale rate to again be lower than the male rate in the next recession.Beyond that, it seems likely that the female rate will remain below themale rate well into the 1990’s.