Aaron Dymarskiy has a challenging piece reproduced at Aporia magazine.com, which complete with scholarly references, puts the case that not only do US Blacks face little economic discrimination at present, but he argues that this applies as well to much of the past. The argument is long and involved. But here is the conclusion: "By reviewing the published literature, this article has shown that in various occupations at various times and places, there was little or no economic discrimination against blacks—either because there was no wage differential in the first place, or because adjusting for relevant factors reduced it to insignificance. In some cases, the data was more ambiguous, but there was never any clear evidence for substantial discrimination.
Claims of substantial economic discrimination against blacks are therefore unconvincing, especially when considering that there has not been any major improvement in black income, relative to white income, over the last 80 years—during which time prejudice has largely disappeared.
Many of the authors cited in this article maintained that discrimination was the main cause of racial income differentials in the 19th and 20th centuries. Some notable exceptions are Robert Higgs, Gavin Wright and Stephen DeCanio. Their views were more complicated, but mostly leaned toward discrimination playing a smaller role than market forces.
How is it possible for me to conclude that racial discrimination was a relatively small factor, despite relying primarily on the work of those who strongly disagree with that view? The answer surely lies in historians' liberal bias, which was particularly strong in the 1970s, when left-wing radicals came to dominate the field (see Graber, 2019). Some authors even analyzed their own data incorrectly: Mill and Stein (2016) took at face value an estimate that was statistically insignificant; Foote and colleagues (2003) did not consider that the foundry premium, if applied equally, would only raise black wages by 1.54%; and Margo (1984b, 1990) did not adjust his black-white teacher wage ratios by the number of days they taught per year.
The fact that such errors seemed to have always favoured the discrimination hypothesis, and were made by researchers whose analyses were otherwise intricate, suggests that liberal bias played a large role in their conclusions."
With demands being made for billions in reparations, analyses such as thisknocks the stuffing out of the case. Perhaps he needs to do a similar analysis for Australia, regarding the Aboriginal population?
https://www.aporiamagazine.com/p/how-bad-was-anti-black-discrimination
"Research has consistently demonstrated that black people in the US face little economic discrimination (Last, 2022; Wright et al., 2024). The reason blacks have lower incomes than whites is simply that they are less productive workers—whether this is measured subjectively (e.g., supervisor ratings) or objectively through tests (Roth et al., 2003). And the primary reason they are less productive is that they have lower average cognitive ability (Herrnstein & Murray, 1996; Murray, 2021). The view that blacks do not face substantial economic discrimination is relatively mainstream, held even by public figures like Ben Shapiro. A less common view is that there has never been substantial economic discrimination against blacks. This is the view argued here.
Background
Trends in income, employment and occupational status. If employers discriminated against blacks, one would expect blacks to have lower employment rates and higher unemployment rates. While unemployment rates for blacks and whites differ significantly today, this was not the case until the 1940s and '50s (Murray, 1984, ch. 5). Indeed, according to Robert Higgs (1977a), the 1890 Census showed that 58% of blacks and 47% of whites were "gainfully occupied", with most of the difference being due to a higher employment rate among black women (and see Fairlie & Sundstrom, 1997; Vedder & Gallaway, 1992).
The employment data, then, are not consistent with a high level of discrimination, given that the gap was close to zero in the late 19th century and early 20th century when prejudice was highest, but increased in the '40s and '50s when prejudice was rapidly declining.
The data show also that, between 1910 and 1950, there was almost no change in the relative occupational statuses of blacks and whites within the same region (Becker, 1957):
More data is available regarding wages. For 1880, Ng and Virts (1993) report that the income ratio between Southern blacks and whites was 0.53, meaning that the average black Southerner had an income 47% lower than the average white Southerner.
Since 1940, income data has been available for all regions, and a time series has been constructed by Bayer and Charles (2017). In 1940, Southern blacks earned 58% less than Southern whites, meaning that the gap actually widened in the intervening sixty years. (The table below uses natural logs, meaning that setting e to the power of the coefficient tells you what percentage of the median white income was earned by the median black in that year and region.) A comparison of the first and last rows in the table shows that, in all regions but the South, blacks got poorer relative to whites between 1940 and 2014—despite the decline in prejudice.
By itself, however, this data does not show that economic discrimination against blacks was rare. It is only indicative. What we need is data on the wages for blacks and whites in the same occupation.
Definitions. Before proceeding, it is necessary to define 'discrimination' and explain how its effects can be measured. In The Economics of Discrimination, Gary Becker (1957) states that discrimination is occurring when an economic actor loses or forgoes something of economic value in order to act on his prejudice or racial preference. Black wages will be depressed by economic discrimination because blacks must offer to work for less in order to be hired by employers with a prejudice against them. This is the sense in which I use the term.
Note that economic segregation and economic discrimination are often confused, even though they are distinct: it is possible for an entire business or industry to exclude a race without depressing its income (Becker, 1957; Whatley & Wright, 1994).
A Note on qualitative evidence. This article's conclusions are based on the weight of quantitative evidence. This is in contrast to many economic historians, especially those writing about racial discrimination, who have relied largely on qualitative evidence. I believe that this is a mistake as qualitative evidence is more open to interpretation. In the words of Lord Kelvin, "When you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind".
The most extensive review of the qualitative evidence, based on written statements from the relevant time period, can be found in Stephen DeCanio's Agriculture in the Postbellum South (1974), where it is said that "a strong case can be built on either side"—for profuse discrimination or for a competitive market where there was little racial preference. DeCanio demonstrates this fact by first writing at length in support of the discrimination hypothesis, and then doing the same on behalf of the opposing view. It is worth quoting his concluding words on the topic:
Traditional historical methods simply cannot resolve the postbellum Southern agricultural system sufficiently to distinguish which of the major alternative hypotheses were true in the aggregate … Clearly, if further progress is to be made toward understanding these historical problems, more powerful techniques [than analysis of impressionistic evidence] must be developed and applied.
Attitudes versus action. The economist Gary Becker (1957) famously made the point that prejudice does not always lead to discrimination. If discrimination is thought of as a preference, individual acts of discrimination by employers are trade-offs between economic value and non-economic value. Suppose applicant A, who is black, can produce $100 of economic value per hour, while applicant B, who is white, can produce $90. In that case, the cost of favoring applicant B would be huge; if he worked for forty hours a week, hiring him would cost the employer $20,000. As Milton Friedman (1962) notes:
This same phenomenon is of much wider scope. It is often taken for granted that the person who discriminates against others because of their race, religion, color, or whatever, incurs no costs by doing so but simply imposes costs on others. This view is on par with the very similar fallacy that a country does not hurt itself by imposing tariffs on the products of other countries. Both are equally wrong. The man who objects to buying from or working alongside a Negro, for example, thereby limits his range of choice. He will generally have to pay a higher price for what he buys or receive a lower return for his work. Or, put the other way, those of us who regard color of skin or religion as irrelevant can buy some things more cheaply as a result.
The fact that economic interests can dissuade people from acting on their prejudices was shown in an early study by Richard LaPiere (1934; see also Dockery & Bedeian, 1989). LaPiere asked 81 restaurants and 47 hotels if they would accept a Chinese person as a guest. Only one hotel and zero restaurants said that the Chinese would be allowed entry, while several of each were undecided. In all, 92% of hotels said that they would refuse a potential Chinese guest, along with 93% of the restaurants. LaPiere then spent a few years visiting each of the hotels and restaurants with "a young Chinese student and his wife," neither of whom knew that he was conducting a study. Of all of the businesses visited, only one did not accept the couple, and in only a handful of instances was the "reception perceptibly hesitant" on what was deemed likely to be "'racial' grounds". The reason for this result is obvious. When answering a survey, the cost of saying that Chinese customers would be refused is much smaller than the cost of actually refusing those customers. In effect, purporting to engage in discrimination when answering an anonymous questionnaire is a cheap way to express a dislike of Asians; turning away Chinese customers is an expensive one.
The extent of prejudice. Was prejudice against blacks so strong that individuals deemed the costs of discriminating to be worth it? It is difficult to tell, of course. Becker (1957) himself believed that "the extent of discrimination in the market place in this country is probably much less than in almost every other country in the world", though he still thought that it had a large effect.
In the early 20th century, the black sociologist W.E.B. DuBois (1904) surveyed 534 young black Georgians (aged 13–21) as part of his report on "Negro Crime," asking why they believed "so many Negroes" were "under the clutches of the law." The data from this survey were collated in a more readable form in a short paper by Shaun Gabbidon (2000). His figures show that, of the 438 who responded, only 19 (4.3%) blamed racial discrimination. This is the only such poll of which I'm aware that predates the end of the Second World War. Of course, touting it as incontrovertible evidence for a lack of discrimination would be a mistake, but dismissing it completely would be a much greater error.
In 1946, the National Opinion Research Center asked, "Do you think most Negroes in the United States are being treated fairly or unfairly?" Only 28% of black respondents said that they were treated fairly, and another 66% said that blacks were not treated fairly. Of the whites who responded to the same question, the results were reversed: 64% said that blacks were treated fairly, and 27% said they were treated unfairly (NORC, 1947). It should be noted that these results are similar to what was found in a recent Gallup poll, where 60% of whites said that blacks were not less likely to get a job for which they were equally qualified, compared to just 31% of blacks. Therefore, claims of widespread discrimination can be found even when there is, in fact, little or no discrimination (as currently applies).
Although attitudes do not always translate into actions, it is useful to look at the nature of prejudice. Mildred Schwartz (1967) obtained data on white attitudes toward blacks going back to the 1930s. An example is surveys asking whites if they believe that whites and blacks are equally intelligent; 42% answered yes in 1942, with the figure rising to 78% in 1956. This trend is similar for every type of question Schwartz analyzed. In the 1930s and 1940s, the majority held a prejudicial view (although in the case of intelligence they were simply correct). Yet by the '50s and '60s, the vast majority held the non-prejudicial view. The most relevant question is whether whites should receive preferential treatment (see also NORC, 1947). In 1944 and '46, fewer than half believed that whites and blacks should be treated equally by employers. Yet by 1963, over 80% said that they should.
Becker (1957, ch. 4) also noted that employees can discriminate by refusing to work with blacks. This would create a situation in which employers would be losing certain white employees when they hired blacks, which would sometimes make discriminating against equally qualified blacks rational. There are several surveys that asked white employees to what extent they were willing to work with blacks (Hazel Erskine, 1968).
In a 1942 survey, high schoolers were asked if they were willing to work with blacks "if they had an equal position to yours and worked side by side with you." In the whole sample, only 21% said that they were not, though the figure was 49% in the South.
In a 1945 poll, 25% of laborers and 19% of farmers said that they would refuse to work alongside blacks in response to a similar question.
In a different 1942 poll, 55% of respondents were willing to work alongside a black person "in an office or factory," whereas 39% were unwilling.
In 1944, 51% of respondents said that they wouldn't mind "If a Negro with the same training had the same kind of job as you at the place where you worked," whereas 43% said that they would mind.
A 1958 survey found that 79% of Northern whites were "willing to work next to members of the other race," while 17% were unwilling. By contrast, only 48% of Southern whites were willing.
The same poll found that 88% of Northern whites and 77% of Southern whites would work for an employer who also hired blacks.
In 1963, 31% of Southerners objected to working alongside blacks, compared to 14% of Northerners.
The same question was asked in 1965. This time, only 24% of Southerners and 7% of Northerners objected.
Another piece of evidence comes from a survey of Louisianan firms, which asked whether their employees of a given race were more or less willing to work with their employees of the other race (Wilson & Gilmore, 1943). Of the 113 employers who answered this particular question, 65% believed that there was little difference between blacks' willingness to work with whites and whites' willingness to work with blacks. This suggests that whatever employee discrimination did exist was not much more likely to come from whites.
Indeed, racial animosity went both ways. A study of 219 black World War II veterans, for example, found that 75% entered into service "with negative attitudes toward Whites" (Roberts, 1953). A recent meta-analysis examined trends in viewers' racial preferences as revealed by their reactions to commercials, going back to 1956 (Lenk et al., 2024). The authors found that, although whites in the past had a sizable anti-black bias, it was never as large as blacks' anti-white bias.
While it's difficult to know if the level of prejudice was sufficiently high in the past to cause economic discrimination, the survey evidence suggest it may well have been. The next part of the article will examine if such discrimination did, in fact, occur.
Government discrimination
Economic discrimination may have come about through government policy. Here I consider several of the most commonly cited such policies. I conclude that none can be proved to have been caused by racism, and only one may have had an impact on black incomes relative to white incomes.
Education. One form of government discrimination that is often mentioned is differences in school quality. Evidence is available from Robert Margo's Race and Schooling in the South, 1880-1950 (1990). In 1890, there was little difference in funding in a few Southern states, but large differences in others. By 1950, however, differences in most states were minimal. Other data suggests that, by the early 1970s, there was no difference in funding (Richwine, 2011; Reuben & Murray, 2008).
Between 1890 and 1950, there were large differences in some states—with whites sometimes getting as much as five times the funding per capita. Remember that for a practice or policy to qualify as discrimination: 1) it must be motivated by racial prejudice, and 2) it must reduce the economic wellbeing of blacks relative to whites.
First, it's important to look at the motivation for differential school funding. The two states that have been studied most extensively are North Carolina and Louisiana.
According to Kousser (1980), North Carolina's black–white per capita education expenditure ratio was 0.59 in 1901–05 and 0.40 in 1906–1910. It should be noted that, black school expenditure did not decrease. In fact, it increased by almost one third between 1900 and 1910. The issue, then, is how much more the funding increased for whites.
So, what, in Kousser's view, caused the expenditure ratio to decline? While it was a "Radical Republican notion" that educational spending should not vary by race, "Many North Carolina Democrats, especially after 1900, espoused another [notion] … Contending that one should receive in proportion to what one pays, these Democrats sought … to force black schools to operate on taxes raised from blacks and, by increasing the proportion of funds for schools raised at the local level, to allow whites living in rich areas to better themselves without sharing their largess with whites who lived in areas where almost everyone was comparatively poor".
Therefore, the main reason for the decline was Democrats' desire to avoid redistributing wealth from whites to blacks and from rich to poor. Kousser calculated the "black balance of payments," that is, the difference between their share of the education and their portion of taxes paid, for these periods, finding that, indeed, there was a very large decline. Yet blacks still received substantially more in education funding than they were owed based on their total tax contributions.
It is of course possible that racism played a part in the government's decision to stop redistributing white tax money to black schools (Kousser believed that it did), but what occurred between 1900 and 1910 can also be explained by lawmakers' opposition to redistribution in general.
Louisiana's story, as told by Margo (1982), is similar. While in 1890 white per capita school funding was already twice the black level, the ratio increased to over 5.5 by 1910. Margo (1982) analysed county level data for 1890 and 1910. He concluded that:
disenfranchisement reduced expenditures on black schools, enabled whites to take fuller advantage of the fiscal mechanisms outlined above, and led to a racial distribution of school spending that more closely approximated the racial distribution of the taxable resources.
However, it should be noted that, in both 1890 and 1910, the variable for share of blacks who were registered to vote was non-significant despite being in the expected direction. Either way, Margo's findings are similar to Kousser's, and do not suggest that racism per se played a decisive role. Whether the disparities in funding caused a decline in the black-white income ratio is more difficult to say.
If the U-shaped pattern of racial differences in school quality over time was a major contributor to the income gap, one would expect the income gap to follow the same pattern. However, data reported in the first part of the article show that the Southern racial income gap was already large in 1880, at which time educational disparities in education were relatively small—and they only widened moderately between 1880 and 1940 (from 47% to 58%), during which time the difference in educational quality increased sharply. Furthermore, when the difference in quality declined over the next few decades, the racial income gap fell only slightly (in 1960, it was 53%). Overall, the data suggest a slight correlation between educational funding and economic outcomes, but the link is weak and unproven.
Redlining. In the 1930s, the Home Owners' Loan Corporation (HOLC), an agency established under FDR, created color-coded maps of major cities, with red areas designated as the riskiest. According to some researchers, these maps can explain practically anything. It has been claimed that they:
Gave black adults asthma (Schuyler & Wenzel, 2022).
Made black people fatter (Owens et al., 2024).
Made minority neighborhoods smell worse (deSouzza et al., 2025).
Blinded residents of poorly rated neighborhoods (Hicks et al., 2023).
Decreased participation in birdwatching (Ellis-Soto et al., 2023).
Prevented nonwhites from adopting pets (Blackwell et al., 2024).
However, the maps do not appear to have been widely circulated and they seem to have little impact on lending practices (Fishback et al., 2024; Hillier, 2003a, 2003b, 2003c; Malach, 2024). Indeed, data from Philadelphia has shown that blacks were no less likely to receive loans in the 1930s than whites (Hillier, 2003c). The essential non-existence of red-lining can be seen in the historical trends in homeownership by race, as presented by William Collins and Robert Margo (2011). As can be seen from the table, there has been no substantial change in the black-white gap in home ownership since the 1930s.
Slavery. When it comes to slavery, the first question is whether the practice was, in fact, caused by racism. While the answer may seem obvious, it is not. Every civilisation on earth has had slavery, and the reason was not always racism.
America was a slave society from the founding of the first colonies, with the majority of the slaves being white until the early 18th century. It is true that these white slaves were not permanently enslaved, and in some cases came willingly to America. However, studies of indentured servants have found that they were not treated any better than the permanently enslaved, and that they often died before their terms expired.
Don Jordan and Michael Walsh (2007) did not shy away from referring to them as slaves in their book, White Cargo: The Forgotten History of Britain's White Slaves in America. The authors write: "The lengths through which some previous writers on this subject have gone to separate out servitude from slavery seems to us to miss the point that there were, and are, different types of slavery". Other books which expressly refer to indentured servitude as slavery include They Were White and They Were Slaves: The Untold History of the Enslavement of Whites in Early America (1993), and The Irish Slaves: Slavery, indenture and Contract labor Among Irish Immigrants (2010).
What's more, the first African slaves in America were also indentured servants, as they were not permanently enslaved; nor does it appear that their status passed down to their children. It was not until the 1660s, by which point Africans had been in North America for about forty years, that any of them were permanently enslaved (Higginbotham, 1978; Morgan, 2001). The decline of white indentured servitude and rise of black permanent slavery can be explained by economics (Galenson, 1984). Whites became indentured servants to pay for their transportation to America, so when the price fell, they no longer needed to sell themselves. It wasn't necessarily racism that made black slavery so common, but the forces of supply and demand. (Racism was later used to justify slavery.)
Note that black slaves were not only owned by whites. The current best estimates for 1830 suggest that blacks and whites had a similar propensity for owning slaves—Southern white households were perhaps three times as likely to own slaves, but this difference may be largely explained by differences in wealth and undercounting of black slaveowners (Lightner & Ragen, 2005). In Charleston, South Carolina, for which the data are exceptionally good, between half and two thirds of free black households owned slaves (Koger, 1985; see image below). It has been argued that black slave owners were typically "benevolent," simply buying family or friends in order to save them from other slave owners. Yet there is a lot of research contradicting this claim (see Koger, 2006; Graber, 2022, ch. 7).
Another key fact to mention is that comparatively few slaves were brought to the US during its history—about 500,000, representing 4% of the total African slave traffic to the Americas (Curtin, 1969; Eltis, 2008). Almost all the growth in the slave population occurred through natural increase (Hacker, 2020). Because Africa at the time still had a form of slavery in which the status was both permanent and inherited, it makes sense that Africans were the first group to be enslaved in this manner in the US.
The second question is whether slavery had a large effect on blacks' incomes relative to whites'. The effect of slavery on wealth was obviously large—as shown by blacks' rapid accumulation of wealth after slavery ended, though this did not last long (Derenoncourt et al., 2022; Higgs, 1984; Margo, 1984a; see also the homeownership data above).
Yet income and wealth are different, and there seems to have been almost no growth in black income between 1860 and 1900—if in-kind payments to slaves are counted as wages (O'Brien, 2000). In addition, blacks who were born as slaves were only slightly lower in occupational status in 1880, after controlling for region (Sacerdote, 2005).
Overall, slavery does not seem to meet either criterion for a practice to qualify as economic discrimination. It is not possible to say for certain that it was primarily motivated by prejudice, and the evidence suggests it had little to no effect on blacks' incomes. (This is obviously not a defence of slavery.)
Individual discrimination
Economic discrimination may also occur at the level of individual firms or industries. Each section below considers a specific type of labor in a certain time period. Once again, I conclude that evidence for substantial economic discrimination against blacks is weak.
Southern agricultural workers, 1880-1910s. Robert Higgs (1978) drew attention to a little-known agricultural survey conducted in 1887 by the state of North Carolina. The survey included this question: "Any difference [in average wages paid] to white and colored [laborers]?" In each county, one landowner and one laborer or tenant was interviewed, for a total of 95 responses from each type of respondent. Of the landlords, 94 (99%) said that blacks got the same wages as whites, while only one said they did not. The tenants and laborers were less unanimous, but most (77, or 81%) believed that there was no racial difference in wages.
In an earlier paper, Higgs (1972) examined data collected by the US Department of Agriculture in 1898–99.
Without taking into account room and board, Higgs found that blacks' wages were just 9% lower than whites' wages during the normal season, and 8% lower during the harvest season. He hypothesized that, if room and board were taken into account, the wage difference would become smaller still, or perhaps even disappear entirely, as the author of the report from which the data came noted that blacks tended to get more in-kind benefits.
Higgs' findings were replicated by Kenneth Ng and Nancy Virts (1993), who found that the ratio of black to white Southern rural labor income was 0.89 in 1880 when only those above the age of fifteen were included in the population estimates.
Small (or perhaps non-existent) wage differences were also found by Stephen DeCanio (1974) in a detailed study of racial productivity differences on Southern farms in the late 19th to early 20th century. He began by examining contemporary literature, finding mixed opinions: while many believed that blacks made for less productive laborers, others believed that they were equal or superior to whites. Next, DeCanio analyzed census data from several decades and observed no consistent relationship between racial composition and productivity. For the South as a whole, in the years 1880 to 1910, he found (after controlling for numerous variables that differ between counties) that white cotton farmers were most productive, followed by black cotton farmers, and that black and white non-cotton farmers were about equally productive. The overall difference favored whites, but DeCanio argued that this was largely explained by whites tending to farm better land in a given county.
Intelligence is a fairly weak predictor of income among farmers (Perry & Schreiner, 1990), and tools used for farming in the South were primitive, even by the standards of the time (Fergeler, 1998), which likely made intelligence even less important. Indeed, it has been found that education (a proxy for IQ) is much less correlated with farmer productivity in undeveloped locations (Lockheed et al., 1980). Therefore, it is not completely implausible that blacks were about as productive as whites.
Gavin Wright (1986) collected data on farm operators, and found a large racial disparity in the South Atlantic region, but a much smaller difference in the South Central states. The issue with such comparisons is that the term "farm operator" could refer to several different types of labor: owners, overseers, and either cash or share tenants. Wright's analysis shows that blacks were much less likely to be owners or managers, and more likely to be tenants who were, of course, paid less.
When whites and blacks worked as simple agricultural laborers in the late 19th century, they tended to receive similar wages. This is plausible, given that their productivity was similar—perhaps with a slight white advantage. And it is inconsistent with substantial economic discrimination aginst blacks. In a contemporary survey, most respondents indicated that there was no difference in wages between blacks and whites in North Carolina.
What about sharecroppers? Lee Alston and Kyle Kauffman (2011) analyzed data from a 1911 survey of 39 Georgian plantations. Of the total sharecroppers on these plantations, 325 were black and 48 were white. Alston and Kauffman found that blacks and whites made almost the same incomes, with large overlap in confidence intervals. Their conclusion was as follows:
[W]e found that black and white sharecroppers received equal incomes. This result … indicates the prevalence of competitive markets for labor on the lower rungs of the agricultural ladder
Virginia, 1900-1926. Higgs (1977b) analyzed wage data from about 500 labor contracts in 1900 and about 1,500 in 1909 from Virginian firms. In total, over 19,000 workers were included in these contracts. When he limited the sample to firms whose workforces were racially integrated, his results were as follows:
Most firms paid blacks and whites the same wages when the position was unskilled, while blacks were typically paid somewhat less in skilled positions. These data do not suport the discrimination hypothesis for two reasons: 1) as the position becomes more skilled, qualified blacks become less replaceable, leading to higher costs of discrimination (Becker, 1957), and 2) because discrimination can only occur when a superior black is paid the same or less than an inferior white, and superior blacks must have been more common in less skilled occupations.
Warren Whatley and Gavin Wright (1994) obtained data from the Bureau of Labor for the state of Virginia, for the years 1908, 1909 and 1926. A major advantage of these data is that they refer to all workers, not just for those from integrated firms. Regarding the 1908 data, the authors note that:
The white distribution is bimodal, the black distribution unimodal; but the peak for the black distribution is at virtually the same wage level as the left-handed peak for the white distribution. If we interpret the first peak as the 'unskilled' market and the second peak as the 'skilled' market, then we may say that despite the prevalence of racial segregation, unskilled wages were effectively not differentiated by race. Apparent racial differences were, in reality, skill differentials
Indeed, looking at the figure for 1909, there is virtually no difference between blacks' and whites' wages in unskilled occupations.
Although the authors do not report the exact averages, the graph suggests that wages for white unskilled workers were at least 50% higher. Whatley and Wright explain the change between these two dates as follows:
When we try to identify the reasons for this divergence, many proximate factors can be listed: the stagnant world market for cotton during the 1920s had a disproportionate effect on blacks, as did the depressed conditions in lumber and saw milling; real wages in cotton textiles, on the other hand, rose to a relatively high plateau during the wartime boom, and proved resistant to downward adjustment for a number of reasons.
This explanation does not fit with any hypothesis involving racial discrimination, as it is very unlikely that white attitudes toward blacks became less favorable at this time.1
Public school teachers, 1890-1950. In Race and Schooling in the South, 1880-1950, Robert Margo (1990) reports the following figures for racial wage gaps among teachers in the South:
As with black–white ratios in per pupil education spending, the pattern is U-shaped. In 1890, there was very little difference in pay, but in 1910 and 1936 black teachers only made about half of what white teachers made. By 1950, however, the difference in pay was small in five of seven states (Mississippi is the exception). The similarity between this pattern and the pattern in per pupil education spending suggests that the former is partly explained by changes in the funding received by the district at which the teachers worked.
In an earlier paper, Margo (1984b) used data from Florida, North Carolina and Louisiana to see how much of the wage differences from that period could be explained by variables from the Census and from government reports on education. One of the variables from government reports was the portion of teachers who received a first-grade teaching certificate (indicating a high test score). Margo's found that these variables could only explain about 10-20% of the difference in wages. He (1990) also studied the black–white gap in teacher salaries using data from 1940. He found that age and years of schooling could account for just 14% of the gap, leaving a black-white ratio of 0.54.2
Both of these studies by Margo do provide evidence for discrimination. However, caution is warranted. In the first study, the most important variable was the portion of teachers in the county receiving first-grade certificates, but it is unknown how well this variable captured actual teaching ability. What's more, matching blacks and whites on years of schooling certainly did not match them on skills associated with education—many later studies show that the racial IQ gap among those enrolled in high school or college in intelligence is the same as it is in the general population (see Shuey, 1966, ch. 4 for a meta-analysis). Margo's later study only controlled for two variables.
Bradley and Mary Hansen (2006) used individual and county level data from Virginia for 1905-06 and 1906-07—mostly to see whether there were differing returns to education. They found some differences, though in my opinion they are overstated. Here is their main table:
Aside from special certificates, which were held by relatively few teachers (just under 4.2%), and emergency certificates, which were also relatively rare (14.8%), certificates were similarly lucrative for both races. This makes sense, as they were objective measures of ability (however well they captured it). On the other hand, schooling had much lower returns for blacks, as one would expect from the fact that black high school and college and students are less intelligent than similarly educated whites. The lower return for education has been found at the national level, and should not be taken as unambiguous evidence of discrimination. As noted above, discrimination should decrease with education, as it is more costly to discriminate against higher-skilled blacks.
Another explanation for the racial wage gap is the difference in school attendance, which has the same U-shaped pattern as wages and funding. According to Margo (1990), the black–white ratio in the number of days in the school year was about 1 in both 1890 and 1950, but was lower in the intervening years. In 1910, the ratio ranged from 0.49 in Louisiana to 0.91 in Mississippi. And in 1935, it ranged from 0.73 to 1.0. This information can be used in conjunction with Margo's estimated wage differences that were unexplained by the observed variables to get a daily wage ratio. The Floridian black-white ratio in school year length was 0.81 in 1910; dividing the previous estimate (0.57) by this number gives a daily wage ratio of 0.70 for Florida. Completing the same process with the other two states gives the results of 0.86 for Louisiana and 0.72 for North Carolina.
Probably the best estimate of the black–white wage ratio after accounting for known influences, then, is between 0.7 and 0.8, leaving the rest to be accounted for by discrimination (and perhaps unobserved differences in human capital). Although this gap is not proof of racial discrimination, it would not be surprising if black teachers did face such discrimination. This is because the teachers included in the studies taught in public schools, where the government had a monopoly on the hiring process (a point made by Margo, 1984b). Because monopolies (especially government ones) do not need to worry about efficiency, the costs of discrimination are lower (as discussed at length by Becker). When interpreting these data, it is important to stress that there was no difference in 1890, which is rather difficult for the discrimination hypothesis to explain, as there is no reason to believe that racism began rapidly increasing after 1890.
Ford autoworkers, 1918-1947. Cristopher Foote and colleagues (2003) analysed data from the historical records kept by Ford on over 4,000 workers between 1918 and 1947. They found that, while blacks and whites had approximately the same wages when controlling for several variables, blacks were much more likely to be employed in the foundry, where work was more dangerous. Because standard economic theory suggests that such labor would command a premium (e.g., Friedman, 1962), this could be taken as evidence that Ford did discriminate against blacks.
Unfortunately, Foote and colleagues do not provide an estimate of the premium paid to white foundry workers, but a simple equation can be used to estimate that it must have been about 3.75%.3 Because 41% of blacks were employed at the foundry, the average black worker would have received 1.54% (41% of 3.75%) higher wages if the foundry premium were applied. It appears, then, that whatever disparity did exist was small.
It's possible that even this difference could be partially explained by a greater tolerance of heat by blacks (see Foote et al., 2003). In a 1943 survey of New Orleans employers, 53% believed that blacks were better able to tolerate heat (Wilson & Gilmore, 1943). Perhaps more important were characteristics like intelligence, which will not have been fully captured by variables like education. No matter how the data is analysed, discrimination could not have been particularly great.
Manufacturing workers, 1936-1960s. Higgs (1989) analyzed data from 1936 and 1937 on wages paid to whites and blacks working at factories that hired members of both races. In each year, there were roughly 900 factories surveyed from the North and 500 from the South. His results were as follows:
Combining data from both years, it can be seen that in the North 8% of plants paid whites more, while 5% paid blacks more. On the other hand, 28% of Southern plants paid whites more, while only 3% paid blacks more. Although this itself does not prove that discrimination reduced blacks' wages, it suggests that discrimination may have been a significant factor in the South.
A much clearer conclusion can be drawn from the data reported by Leonard (1984) on over 12.5 million manufacturing workers in 1966. He states that the non-white workers were ".68 times as productive as their white counterparts," while "the earnings ratio of black to white males" was 0.69. So their difference in pay was almost exactly equal to their difference in productivity (if anything, they were paid more than they should have been).
The entire American economy, late 1930s. The earliest national survey of wages that includes information on race was conducted in 1937, with results reported by Whatley and Wright (1994):
Among unskilled workers, there was little difference in wages in the North (indeed, blacks made slightly more), whereas the difference was large in the South. Whatley and Wright warn that interpreting these results as evidence of racial discrimination in the South would be unjustified, given that data for Virginia (see above) did not show a racial wage gap three decades before this survey. Instead, these results indicate that industries where blacks made up most of the labor force declined in importance in the South in the intervening years.
A study whose design is able to account for confounders is therefore of interest. Mill and Stein (2016) examined Census income data for mixed race Americans in 1939. They tracked siblings who were recorded as "mulattos" in the 1910 Census. In some cases, one of the mulatto siblings was labeled as white in the 1940 Census while the other was labeled as black. These cases served as the main sample used in the study, as any differences between the siblings must have been due to discrimination.
As mentioned above, labor income is much more relevant than total income, as the latter includes things that racism cannot possibly affect. Controlling for education, as in the fifth column, is not appropriate, as any differences in education may have arisen through racial discrimination. Therefore, the best estimate from this study is that blacks' income would have been roughly 14% higher (e0.13) if they were indistinguishable from whites. Yet this figure should not be given much weight, as the coefficient is not statistically significant.
Conclusion
By reviewing the published literature, this article has shown that in various occupations at various times and places, there was little or no economic discrimination against blacks—either because there was no wage differential in the first place, or because adjusting for relevant factors reduced it to insignificance. In some cases, the data was more ambiguous, but there was never any clear evidence for substantial discrimination.
Claims of substantial economic discrimination against blacks are therefore unconvincing, especially when considering that there has not been any major improvement in black income, relative to white income, over the last 80 years—during which time prejudice has largely disappeared.
Many of the authors cited in this article maintained that discrimination was the main cause of racial income differentials in the 19th and 20th centuries. Some notable exceptions are Robert Higgs, Gavin Wright and Stephen DeCanio. Their views were more complicated, but mostly leaned toward discrimination playing a smaller role than market forces.
How is it possible for me to conclude that racial discrimination was a relatively small factor, despite relying primarily on the work of those who strongly disagree with that view? The answer surely lies in historians' liberal bias, which was particularly strong in the 1970s, when left-wing radicals came to dominate the field (see Graber, 2019). Some authors even analyzed their own data incorrectly: Mill and Stein (2016) took at face value an estimate that was statistically insignificant; Foote and colleagues (2003) did not consider that the foundry premium, if applied equally, would only raise black wages by 1.54%; and Margo (1984b, 1990) did not adjust his black-white teacher wage ratios by the number of days they taught per year.
The fact that such errors seemed to have always favoured the discrimination hypothesis, and were made by researchers whose analyses were otherwise intricate, suggests that liberal bias played a large role in their conclusions.