in the economy, albeit with increased skepticism about its effectiveness as the period unfolded.
The mean unemployment rate of 4.7 percent in the first (nonintervention) period was far lower than the 6.09 percent rate observed from 1960 to 1989, or the 9.3 percent average rate in the transitional era of 1930–59. While unemployment variation was somewhat lower in the 1960–89 period than in the early decades (standard deviation of 1.60 versus 2.29), Romer’s insight suggests that this almost entirely reflects faulty data rather than a real phenomenon. Using this tripartite division, the evidence generally points to the conclusion that the unemployment situation was better in the relatively laissez-faire era before the Great Depression than in periods since.
Spatial Variations in Unemployment
In addition to substantial intertemporal variations in unemployment, interstate differentials in joblessness have been great throughout the period. Before 1960, good state unemployment data were not collected on an annual basis, although information was gathered as part of several of the decennial population censuses.7 In table 1.2, indicators of interstate variation in unemployment are presented for a number of years.
TABLE 1.2 INTERSTATE VARIATIONS IN UNEMPLOYMENT, VARIOUS YEARS, 1930–88a
Note that throughout the period for which data are available, there are areas of the United States where the incidence of unemployment is three to four times that of other areas. Typically in recent years, there are states in which the unemployment rate is 10 percent or more, and other states where the rate is about 3 percent.
Do unemployment differentials between states tend to persist over time? The evidence suggests that in the very long run, there is no correlation between unemployment rates for different time periods. States that in one time period had high unemployment rates did not have a clear tendency to have relatively high rates in the later period. For example, the correlation between the unemployment rates for the forty-eight contiguous states and the District of Columbia in 1930 and 1988 is actually negative (-.13). Yet the “very long run” is actually quite long—perhaps forty or fifty years. There are relatively long periods in twentieth-century American history where unemployment variations have seemed to persist in something of a pattern. For example, the correlation between unemployment in 1961 and in 1988 for the forty-eight contiguous states and the District of Columbia is a fairly high. 47. Between 1930 and 1950 it was even greater, .62.
Some states seem to have persistently high or low rates of unemployment. For example, for twelve of the thirteen years examined from 1930 through 1988, the unemployment rate in West Virginia exceeded or equaled the national average. By contrast, in every single year the unemployment rate in Nebraska and South Dakota was less than the national average. Thus it is clearly possible that there are some regional variations in the normal or natural rate of unemployment.
Demographic Variations in Unemployment
It is a well-established fact that in the late twentieth century unemployment rates tended to be dramatically higher for nonwhite than for white workers, and that younger workers have a much higher incidence of unemployment than is true among older persons in the labor force. On the other hand, gender differences in the incidence of unemployment are relatively small. Have these patterns persisted throughout the twentieth century?
RACIAL DIFFERENCES
The answer to that question, as table 1.3 suggests, is “no.” Turning first to racial distinctions, note that the white-nonwhite unemployment rate differential widened dramatically in percentage terms as the twentieth century proceeded. The 1900 data are not comparable with other years, since they reflect unemployment flows over a twelve-month period, rather than the stock of unemployed as of a specific date. Data on the average duration of unemployment by race suggest that blacks and other nonwhites tended to be unemployed for shorter periods than whites in 1900. Some 55.4 percent of nonwhites were unemployed for three months or less, compared with 47.7 percent of whites.8 The proportion of whites unemployed for seven or more months was nearly twice as high as for nonwhites.9
The effect of the differential duration is to bias the the reported race differential in the direction of overstating it. To illustrate, suppose that over a given year there were ten black workers, four of whom were each unemployed for three months, with each of their unemployment not overlapping with the others. Further assume there were ten whites, one of whom was unemployed for the entire year. The incidence of unemployment, twelve worker months, is the same for both groups. Yet the reported unemployment rate using the 1900 census procedure would have been 40 percent for nonwhites (four were unemployed out of ten), and 10 percent for whites. Using current measurement procedures, the reported rate for both racial groups would have been 10 percent.
Making a number of assumptions, it is possible to estimate a point-in-time unemployment rate for 1900 which is highly consistent with the official (Lebergott) annual data.10 Doing so, we obtain a white unemployment rate of 6.47 percent and a nonwhite rate of 7.57 percent. The racial differential is about 17 percent, with the absolute differential being slightly more than one percentage point. The 1930 data suggest actually a slightly lower unemployment rate among nonwhites—5.17 percent—than among whites—6.19 percent. Assuming the 1900 and 1930 observations are reasonably representative of intervening years, it would seem reasonable to conclude that the white/nonwhite unemployment differential was of a negligible magnitude in the period from 1900 to 1930.
TABLE 1.3 AGE, RACE AND GENDER DIFFERENTIALS IN UNEMPLOYMENT RATES, 1900–88
Further examination of the data suggests the large racial differential in unemployment rates had its genesis in the 1930s and 1940s, with an additional aggravation of that difference in the 1970s. While Americans often pride themselves on reducing racial economic distinctions over time, the evidence on unemployment is highly inconsistent with that sanguine interpretation. Other data on employment confirm the deterioration in black job opportunities.11 It would appear that over time, racial variations in the natural rate of unemployment have grown very substantially.
GENDER DIFFERENCES
To a dramatically smaller extent, the same phenomenon exists with respect to male-female differentials. The 1900 data show a moderately higher female unemployment rate, but the 1930 numbers actually show women with a significantly lower incidence of unemployment, while in 1940 there was virtually no gender differential. Summarizing the data to 1940, it probably would be safe to conclude that gender differentials before 1940 were not systematic and minor in magnitude. By contrast, since 1950, female unemployment rates have tended to be consistently higher than male ones, especially in the 1960s and 1970s, although the differentials have narrowed significantly in recent years.
TEENAGE UNEMPLOYMENT
Teenage unemployment has consistently been higher than that for older Americans, although the differential reported in the 1930 census was very modest. Between 1930 and 1940, the teenage unemployment differential soared, then narrowed somewhat in the 1950s. On balance, the teenage differential has widened since 1950, as is demonstrated by comparing 1950 and 1989, years in which the aggregate unemployment rate was identical. In 1989, the teenage
