Employment Segregation in the Early Twentieth Century

Cliometric Sessions at 1990 ASSA Meetings--December 28, 10:15 AM


1

THE COMPETITIVE DYNAMICS OF RACIAL EXCLUSION: EMPLOYMENT SEGREGATION IN THE SOUTH, 1900-1950*

Robert A. Margo
Vanderbilt University In 1940, the average annual earnings of black men were about 48 percent of the average annual earnings of white men. In 1980, the black- to-white male earnings ratio equalled 61 percent, an increase of 13 percentage points in four decades. Associated with the increase in the earnings ratio, and arguably a better indicator of fundamental change, is the post-World War Two emergence of a "new" black middle class composed of persons employed in a wide variety of white collar and skilled blue-collar occupations. Yet very little change in the earnings ratio appears to have occurred between 1900 and 1940. In 1900 the earnings ratio is estimated to have been 45 percent, just 3 percentage points less than in 1940.

Two frameworks have been advanced to explain the initial stability in the earnings ratio: a supply-side or "human capital" model, and a demand-side, or "institutionalist" model. Proponents of the human capital model argue that the initial stability of the earnings ratio can be explained by large and persistent racial differences in the "quantity" and "quality" of schooling that existed in the first half of the twentieth century. Institutionalists reject the claim that a narrower schooling gap in the first half of the twentieth century would have done much good in fostering black economic progress at that time. Rather, institutionalists believe that, early in the century, the majority of black men were trapped in very low- income jobs, primarily in southern agriculture. To initiate and speed up the process of absorption of black labor into higher paying non-farm jobs, positive "shocks" to the labor market, which permanently increased the non-farm demand for black labor, were required. Large increases in labor demand during the two world wars, for example, resulted in an outflow of blacks from the rural South. But wartime shocks were not sufficient to set in motion a large and sustained rise in the earnings ratio. Additional shocks (the Civil Rights Movement and associated anti-discrimination legislation) were necessary. Despite the emphasis given in the two frameworks to events in the South, previous research has largely been conducted using national aggregate data. This paper uses the public use samples of the 1900, 1910, 1940, and 1950 censuses to examine econometrically the determinants of racial differences in employment outcomes (occupation and industry) in the South during the first half of the twentieth century.

The South began the twentieth century as an agricultural economy -- a majority of male workers, black or white, worked in farming. Agricultural participation rates (the proportion of the male labor force engaged in agriculture) were slightly lower for adult males than for all males in the labor force, but were still substantial. Importantly, racial differences in agricultural participation rates were relatively small at the turn of the century -- 4 percentage points for males ages 10 and over, and 2 percentage points for adult males.

Over the next fifty years the Southern economy "modernizd" -- labor shifted out of agriculture. In 1930 40 percent of the white male labor force (ages 10 and over) was agricultural, a decrease of 20 percentage points from 1900. Black labor, too, shifted out of agriculture, but at a slower pace than white labor -- a decline of 13.4 percentage points between 1900 and 1930. Among adult males, agricultural participation rates declined from 1900 to 1940 for both races, but the decline was greater for whites. During the 1940s, however, black labor shifted out of southern agriculture more quickly than white labor did. Still, the racial gap in agricultural participation rates among adult males was larger in 1950 than in 1900.

An examination of agricultural participation rates by age shows that, prior to World War Two, the shift of labor out of Southern agriculture was a "cohort" phenomenon. Successive generations of younger males had lower agricultural participation rates, while older cohorts remained in agriculture as they aged. Consider the 25-34 age group in 1910: 53 percent of blacks, and 48 percent of whites, were in farming. Among those in the age group still in the South in 1940 (now between the ages of 55 and 64) 59 percent of the blacks and 51 percent of the whites were engaged in agriculture. But agricultural participation rates of 25-34 year olds in 1940 were lower than in 1910 (the same was true of 20-24 year olds). During the 1940s, however, the outflow from agriculture occurred in every age group, blacks to a greater extent than whites.

I next consider the distribution of employment by 1-digit industries and occupations. In 1910 blacks were relatively more numerous than whites in durable-goods manufacturing, transportation-communications- public utilities, and personal services. Black labor was under-represented in the other non-farm industries, especially wholesale and retail trade (by 7 percentage points). In 1940, blacks continued to be over-represented in durable-goods manufacturing and personal services, and were under- represented in mining-construction, non-durable manufacturing, trade, finance and business services, professional services, and government jobs. In 6 of 9 non-farm industries the degree of over- or under-representation of black labor was higher in 1940 than in 1910.

During the 1940s the migration of black labor off the farm found its way into the South's non-durable goods manufacturing plants, narrowing the racial gap in employment shares; and into durable-goods manufacturing, increasing the racial gap there. The proportion of black men employed in trade and professional services also rose over the decade. Black employment in personal services fell (the white share did not), suggesting the relatively high share of black employment in services in 1940 may have been a consequence of the Great Depression. The racial gap in employment increased in mining-construction, financial and business services, and government.

The distribution of employment across occupations in the South was more racially dissimilar than the distribution of employment across industries. At the turn of the century black men were severely under- represented in white-collar jobs. Sixteen percent of white men held white collar jobs, compared with 1.8 percent of black men, or a racial gap of 14.5 percentage points. In the next several decades, black men entered white- collar occupations, increasing the percent so employed to 3.7 percent in 1940 and to 5.5 percent in 1950. But the fraction of white men with white collar jobs rose even faster. By 1940 the racial gap in white collar employment was 22 percentage points. The gap increased to 24 percentage points in 1950. Disaggregation of the data on white-collar employment further reveals that, throughout the period, a majority of black professionals in the South were found in just two occupations, teaching and preaching. Cross-classifications of industry and occupation show that blacks holding managerial positions were mostly self-employed businessmen, in wholesale and retail trade, financial and business services (for example, real estate agencies), or personal services.

If black employment in white-collar work lagged behind white employment, a skilled blue collar job was another means of upward mobility. But blacks were under-represented in skilled blue-collar jobs, and their under-representation increased over time as well. In the 1900 sample 3.8 percent of black men held skilled blue-collar jobs, compared with 9.5 of white men. The black proportion increased to 4.8 percent in 1910, but the increase in the white proportion was larger, so that the racial gap in skilled blue-collar employment rose to 7 percentage points. The fraction of adult black men in 1940 with skilled blue-collar jobs was actually lower than in 1910. Black employment in the skilled trades expanded during the war decade, but growth in white employment was greater, and the racial gap rose to 13 percentage points by 1950.

In the semi-skilled operative category blacks were under- represented slightly in 1900. About 5 percent of black men in the 1900 sample held such jobs compared with 5.7 percent of white men. In the next forty years, the fraction of adult black men in semi-skilled occupations increased, but white employment in semi-skilled jobs rose even faster, to 7.3 percentage points in 1940. But the racial gap closed abruptly during the 1940s, as black men filled newly-created jobs in Southern factories.

If they had problems finding white and skilled blue collar employment, black men had much less trouble getting a low-paying service job or a job as an unskilled laborer. The proportion of black men in service occupations (such as domestic, personal services or protective services) more than doubled over the first half of the twentieth century. In the 1950 sample 10.3 percent of black men held service jobs, compared with only 3.5 percent of white men. The racial gap in domestic employment -- 6.8 percentage points -- was nearly three times as large as in 1900. The proportion of black men working as unskilled non-farm laborers remained roughly constant between 1900 and 1950, at about 23 percent. The proportion of white men in such jobs, however, declined consistently, from 8.1 percent in 1900 to 5.3 percent in 1950. Consequently, the racial gap in unskilled non-farm employment increased, from 15.7 percentage points in 1900 to 17.3 percentage points in 1950.

A summary statistic of racial dissimilarities in employment is a "segregation index". The index I use is:


SI = � |bi - wi|/2 x 100						[1]
      i
where:

bi: share of black labor force in industry or occupation i
wi: share of white labor force in industry or occupation i
The segregation index ranges from 0 to 100. Complete integration (a value of 0) would occur if the black proportion equalled the white proportion in every industry or occupation. Complete segregation (a value of 100) would occur if industries and occupations were either all white or all black -- for every industry or occupation in which wi was positive, bi would be zero -- and vice versa.

The results show that employment segregation in southern industry increased from 1910 to 1940: the value of the index in 1940 (21.7) was 45 percent higher than in 1910. Occupational segregation, too, rose from 1910 to 1940. During the 1940s employment segregation declined in the South. Despite the decline, however, both the industry and occupation indices show that segregation was greater in 1950 than in 1910.

The human capital and institutionalist models offer very different explanations of the evolution of racial differences in employment in the South. The human capital explanation has several parts. The odds of entering the non-farm economy in the South were a positive function of schooling. As each succesive birth cohort came of age and entered the labor force, better-educated members of the cohort, black or white, were more likely to find a non-farm job. But, because racial differences in the quantity and quality of schooling were persistently large -- and, in the case of racial differences in school quality, increasing early in the century --the black shift into the non-economy lagged behind the white shift, particularly in the expansion of blue and white-collar employment. In addition, the shift of black labor out of southern agriculture may have been slowed initially by "spatial mismatch": the initial concentration of blacks in rural counties, where non-farm jobs were few and far between. Spatial mismatch diminished in importance, however, as industrialization spread throughout the South, leading to a more uniform geographic distribution of people and jobs.

The institutionalist view is well expressed by Gavin Wright in his recent book, Old South, New South. According to Wright, a dualistic labor market emerged in the South before 1950 in which white and black workers were "non-competing groups" in the non-farm labor market. Wright rejects the argument that this dualism can be attributed to racial differences in schooling because "schooling had little to do with job requirements" in most of the South's expanding non-farm industries, such as textiles. Wright also rejects the claim that racial differences in employment can be explained by differences in the geographic distribution of white and black labor within the South because "[s]egregation followed industry lines rather than geography" and employment segregation in industry was a consequence of historical accident and fixed costs. Within industries occupational segregation was a matter of racial prejudice and privelege. There were "black" jobs, primarily menial, and "white" jobs. The "old" black middle class, composed of black merchants and professionals (including clergy and teachers), serviced a segregated clientele, but the number and average size of black-owned establishments was too small to provide a significant alternative source of non-farm employment for blacks.

In normal times, most individual southern firms, owned or managed by whites, had few or no incentives to deviate from these social norms; and once the norms were in place, individual blacks could overcome them only by enormous effort and, not infrequently, at great personal risk. To dislodge the competitive dynamics of racial exclusion the South had to be "shocked" out of regional isolation and segregationist ideology. World War One was an initial shock; while it did usher in the beginnings of an exodus of black labor from the South, for a variety of reasons it did not fundamentally alter racial hiring patterns in southern non-farm industries. World War Two had a much bigger impact. In the early 1940s labor markets were extremely tight; the demand for non-farm labor skyrocketed. Although the South was slow to respond initially, a black breakthrough in non-farm employment, concentrated in semi-skilled operative jobs, had occurred there as well by 1944. The effects of the war were, first, to reduce employment segregation in the South; and second, to permanently raise wage levels in Southern agriculture, which provided the impetus for agricultural mechanization and further displacement of farm labor in the late 1940s and throughout the 1950s. But, by itself, World War Two was not enough; the Southern economy was still highly segregated in 1950. Further progress awaited a further shock, the Civil Rights Movement and its associated anti-discrimination legislation.

I use the public use samples of the 1900, 1910, 1940, and 1950 censuses to distinguish between the human capital and institutionalist interpretations of the history of employment segregation in the South. The analysis is based on least squares regressions of the form:

p = X� + e 							[2]
where p is the probability an individual would be employed in a particular industry or occupation, the X's are personal characteristics (for example, age and years of schooling), the �'s are coefficients to be estimated, and e is a random error term. The independent variables are taken from the census samples: age; literacy (1900, 1910); years of schooling (1940 and 1950); census region; degree of urbanization; marital status; and an indicator of geographic mobility, whether the person's state of residence differed from his state of birth. The hypothesis is that, if spatial mismatch was important, black interstate migrants should have been employed more frequently in non-farm occupations and industries.

Two sets of estimations were performed. In the first set the white and black samples were pooled, and a dummy variable indicating race was included among the independent variables. The second set of estimations is race-specific. Later in the paper I use the regression coefficients to calculate segregation indices under various assumptions about racial differences in the independent variables.

The first principal finding is that race per se (that is, holding other factors constant) was an economically significant determinant of the distribution of occupations in the South. The importance of "pure" racial over- or under-representation, however, varied across occupations. Much of the over-representation of blacks in the farm laborer category can be explained by factors other than race. It is also noteworthy that, in 1940 and 1950, black under-representation in white-collar employment -- and, to a much lesser extent, in the semi-skilled category in 1940 -- was considerably smaller once factors other than race are controlled for. However, blacks were still over-represented among unskilled non-farm laborers and in domestic and personal service. Factors other than race cannot explain this over-representation.

Among blacks, schooling had a large, negative effect on the probability of employment as a farm laborer; and, as the century progressed, a negative effect on the probability of employment as a farm operator or as an unskilled non-farm laborer. Schooling improved the chances a black man would be employed in service jobs (primarily personal service), skilled blue-collar and white-collar occupations and --in 1940 but not 1950 -- as a semi-skilled operative. Education reduced the probability a white man would be employed in agriculture or as an unskilled non-farm laborer but (except in 1940) had little effect on employment chances in services. Early in the century better-educated whites were more likely to be employed in the skilled blue-collar trades, but as the century progressed, increasingly opted for white-collar employment. It is important to note that the positive effects of schooling on white-collar employment (and skilled blue collar employment in 1900 and 1910) were higher for whites than for blacks.

The distribution of occupations in the South was not neutral with respect to migrant status. Interstate migrants, black or white, were far more likely to be employed in the non-farm sector. In terms of upward mobility in the non-farm economy, however, interstate migration had a bigger impact on whites. Among blacks, interstate migrants were significantly more likely to be employed as unskilled non-farm laborers or in service occupations, but any positive effects of migration on blue-collar or white collar employment were small and generally statistically insignificant. White interstate migrants, by contrast, were more likely than blacks to find employment in skilled blue-collar or white-collar occupations. The impact of migration, however, was much smaller in 1950 than earlier in the century, suggesting that any spatial mismatch between jobs and people diminished over time as the South industrialized.

The results of the industry regressions broadly confirm those from the occupation regressions. Race per se influenced the distribution of employment across industries. Controlling for factors other than race, blacks were over-represented to a significant extent in agriculture, durable goods manufacturing (except in 1940), and personal services. Blacks were under-represented in mining-construction, wholesale and retail trade, non- durable goods manufacturing (which includes textiles), and government. Educated men of both races were more likely to work outside of agriculture, and schooling had its biggest positive impact on employment in services, not in manufacturing. Interstate migrants were more likely to be employed in the non-farm sector, particularly mining-construction and durable goods manufacturing. Consistent with the occupation results, the impact of interstate migration declined over time.

I use the regression coefficients to calculate counterfactual segregation indices under various assumptions about racial differences in the independent variables. The first set of indices reveal levels and trends in employment segregation, adjusting for all factors (in the regressions) other than race. Controlling for factors other than race lowers occupational segregation by 10 to 14 percent in the early twentieth century; the reductions are larger for 1940 and 1950, but the 1940 and 1950 regressions use a much better measure of educational attainment (years of schooling instead of literacy). Controlling for factors other than race lowers industrial segregation by about a third in 1910 and 1940. However, after adjusting for other factors, employment segregation in the South was higher in 1950 than earlier in the century. It is noteworthy that pure racial segregation continued to worsen during the 1940s, despite the large shift of black labor out of agriculture.

The next set of indices are calculated under various assumptions about racial differences in schooling. The calculations are based on employment distributions predicted from the occupation and industry regressions. Racial differences in educational attainment (literacy and years of schooling) contributed to employment segregation, but the impact was modest. A small fraction of occupational segregation around the turn of the century can be attributed to racial differences in literacy. The percent of occupational segregation explained by racial differences in years of schooling was 21 percent in 1940 and 17 percent in 1950. Had black and white literacy rates been the same in 1910, the industry segregation index would have been 13.6 instead of 15.0, a decline of 9.3 percent. If mean years of schooling in 1940 had been the same for both races, the industrial segregation index would have equalled 17.2 instead of 21.7, or 20.7 percent (= 1-17.2/21.7) lower. Controlling for racial differences in educational attainment does not alter the fundamental finding that employment segregation in the South was worse in 1950 than in 1900 or 1910.

The adjustments for schooling can be criticized, however, because they do not take into account racial differences in the quality of schooling. I adjust for racial differences in school quality by making an assumption that, for racial differences in the quantity and quality of schooling to be truly equal, mean black years of schooling had to equal the white mean plus three additional years. Thus, for example, a black man completing nine years of schooling is assumed to have been as well educated as a white completing 6 years of school. The basis for such an adjustment is that black scores on standardized tests were lower than white test scores.

Racial differences in the quality of schooling certainly were a factor in employment segregation. The indices of occupational segregation in 1940 and 1950 would have been 29 percent smaller had both school quantity and quality been equalized. Nevertheless, racial differences in the quantity and quality of schooling do not explain much of employment segregation in the South. Race, not schooling, was the principal factor limiting the participation of black labor in certain industries and occupations.

The final calculation gives the industry segregation indices under the assumption that the black interstate migration rate equalled the white interstate migration rate. Industrial segregation would have been little changed had blacks been as mobile across state lines as whites were. Similar results were obtained for occupational segregation. Spatial mismatch limited the participation of black labor in the non-farm economy but it was not a major factor behind employment segregation in the South.

Prior to World War Two, the shift of labor out of southern agriculture was a cohort phenomenon. Schooling and migration -- "human capital" -- were integral to this shift. Better-educated, geographically mobile blacks (and whites) left farming; the illiterate and immobile stayed behind. The quantitative significance of illiteracy and immobility can be revealed by using the agricultural industry regressions to calculate the probability an uneducated, immobile (that is, non-interstate migrant) young black male (ages 20 to 24) would be employed in agriculture. This probability exceeded 70 percent in 1910 and 1940. But the probability fell to below 50 percent in 1950. The best explanation of the decline is the one offered by Wright -- an increase in the non-farm demand for black labor, coupled with rising agricultural wages leading to displacement of farm workers. I have already shown that many blacks who left agriculture in the 1940s found employment as semi-skilled operatives. Before World War Two schooling and black semi-skiled employment were positively related, but the influx of rural, less educated blacks reversed the sign of the relationship during the 1940s.

Data from the 1940 and 1950 public use samples reveal that the black-to-white ratio of average weekly earnings of adult males in the South rose substantially between 1940 and 1950. Because agricultural wages were lower than non-farm wages (including wages in semi-skilled occupations) the greater relative (black - white) shift of black labor out of agriculture may have raised the earnings ratio. But it is also true that racial differences in educational attainment were smaller in 1950 than in 1940, as better-educated blacks began to enter the southern labor force. This decline in racial differences in years of schooling might also have increased the earnings ratio.

To distinguish between the two hypotheses, I estimated race-specific earnings regressions for southern males ages 25 to 64, using samples from the 1940 and 1950 public use tapes. The dependent variable is the log of weekly earnings, and the independent variables are dummy variables for age group, years of schooling, location in the South (region and an SMSA dummy), marital status, and dummy variables for economic sector (agriculture and services; the left-out sector was manufacturing). The results show that between 26 and 36 percent of the increase in the earnings ratio can be attributed to the greater relative shift of black labor out of agriculture. Declining racial differences in years of schooling were less important, accounting for 5 to 11 percent of the increase in the earnings ratio.

Recent studies have argued that the Civil Rights Movement and its associated anti-discrimination legislation played a minor role in raising the national earnings ratio in the 1960s and 1970s. The earnings ratio increased during the 1940s (also in the 1950s) before social change had occurred and civil rights legislation fully enacted. According to Smith and Welch, the pre-Civil Rights increase in the earnings ratio "suggests that ... slowly moving historical forces [eg. education] ... were the primary determinants of the long-term black economic improvement". The increase in the earnings ratio in the South during the 1940s was not a consequence of "slowly moving historical forces" but of abrupt changes in labor demand in the context of large sectoral differences in wages. The experience of the 1940s supports the institutionalist argument that, historically, black economic progress and labor demand were closely linked.

In conclusion, analysis of the census samples reveals much about the determinants of employment in the South during the first half of the twentieth century. Racial differences in the quantity and quality of schooling limited the participation of blacks in the non-farm southern economy. In the words of Roger Ransom and Richard Sutch illiteracy "helped to trap the black farmer in southern agriculture". Consistent with the human capital model, a narrower racial gap in the quantity and quality of schooling would have improved the employment prospects of southern blacks, leading to a higher earnings ratio before World War Two. But the quantitative impact of racial differences in schooling was modest, and the impact was concentrated in certain occupations and industries. The "old" black middle class would have been bigger but its composition would still have been very different from that of the "new" black middle class.

Race, not schooling or spatial mismatch, was the principal factor behind employment segregation in the South. Although employment segregation declined during the 1940s World War Two did not fundamentally alter the social norms. Controlling for factors other than race, employment segregation in the South was higher in 1950 than in 1940.

In the 1950s and 1960s the dualism of southern labor markets finally came into conflict with the long-term increase in black schooling. Blacks entering the southern labor force in the 1950s and 1960s were better-educated than previous generations. For them (and their parents) the wait to end segregation had been long enough. Eventually the new generations had the federal government as an ally. It was no accident that federal enforcement of anti-discrimination legislation was initially targeted at the South for that is where enforcement was needed the most.

*This paper is an abridged version of Chapter 6 of R. Margo, Race and Schooling in the South, 1880-1950: An Economic History (University of Chicago Press, 1990)