The Tulsa Race Massacre of 1921 was over in less than 24 hours, but the damage that the city’s Black citizens suffered went on for decades. Indeed, the full magnitude of the community’s economic loss is still coming into focus even on the centennial of the event—in part because new digital tools allow scholars to mine census records for data about its aftermath.
On May 31, 1921, a mob of more than 1,000 white men descended on the jail where a Black teenager was being held on suspicion of assaulting a young white woman. In response, more than 50 Black men came to the aid of the police defending the jail. When these outnumbered Black men retreated to Greenwood, a predominantly Black neighborhood, the white mob attacked its residents and burned their homes, businesses, churches, school, and hospital. Despite efforts by local authorities to cover up these events—whose death toll is now estimated at nearly 300—journalists, historians, and locals have managed to document what happened to prominent Black Tulsans and the institutions that they owned. The entrepreneurs John and Loula Williams lost the Dreamland Theatre. The Gurley Hotel, owned by Ottowa and Emma Gurley, was destroyed as well.
But until recently, the systemic impact of the massacre over time—particularly its economic impact—was difficult to track. How can researchers measure the effects of racist violence on the culture and institutions of a place years after the violence itself? To what extent did the massacre set Tulsa’s Black residents, who by many measures appeared to be thriving in 1920, on a fundamentally different economic and social path?
The two of us have found some answers through a detailed analysis of census records from before and after the massacre. In those records we find the outlines of individual stories, such as those of Gibson Van Dyke and Roy Drew, that point to the economic decline experienced by Black residents who remained in Tulsa after the massacre. Van Dyke, a Black 36-year-old, managed a rooming house in Tulsa in 1920. He and his wife, Bertha, lived on Archer Street—the commercial center of Greenwood, which was heavily damaged by the massacre. Gibson and Bertha were one of the many Black families who remained in Tulsa. A decade later, Gibson was no longer a manager. He was employed as a low-wage, unskilled laborer in the construction industry. The Van Dykes’ trajectory contrasts with that of Roy Drew, a Black Tulsan who in 1920 was a wage worker in a drugstore. In 1930, he appeared in census records for Kansas City, Kansas, where he owned and operated his own pharmacy.
Census data also tell a broader story. The 1920 complete-count census provides a comprehensive picture of Tulsa just one year before the massacre: an agricultural hub and oil boomtown that local boosters called “Magic City.” Tulsa was distinctive because both white and Black residents shared in its prosperity. Greenwood became known as Black Wall Street, not for its financial institutions but for its wealth. A distinct community of prosperous Black Americans was a rarity in 1920. (We should note that analyzing historical data presents certain challenges. For example, early 20th-century censuses asked individuals for their occupation but not their earnings. A standard approach among economic historians—which we have used in our study of Tulsa—is to impute earnings for each individual based on established sources of data on the average earnings for specific occupations.) School-attendance rates were higher in Tulsa than in Black communities in similar cities in the region. Greenwood had the makings of a robust middle class—until what some historians regard as the worst single outburst of racist violence in American history.
Data from subsequent censuses show a striking contrast to the rising Black prosperity of 1910s Tulsa. Earnings steadily declined for Black household heads in Tulsa from 1920 to 1930 to 1940. How do we know that this decrease was due to the massacre and not simply changing local or national conditions? History is not a laboratory, with controlled experiments. But in a control group of similar cities—such as Oklahoma City; Wichita, Kansas; and Little Rock, Arkansas—the earnings of Black residents remained roughly constant over the same 20-year period, despite the ravages of the Great Depression. In 1920, wages of Black household heads were 9 percent higher in Tulsa than in our set of 14 control cities; by 1940, they were 7 percent lower. Meanwhile, white Tulsans’ wages followed the same pattern as in the control cities—roughly constant from 1920 to 1930 before modestly declining from 1930 to 1940. And throughout the 20 years, white wages were higher in Tulsa than in the control cities.
Historical scholarship is being revolutionized by big data. Massive, digitized data sets are suited to answering fundamental questions about communities and societies in ways that individual narratives cannot. The U.S. census in particular is an invaluable record of the economics of the past, particularly as it pertains to the lives of ordinary people. Decennial censuses are unsealed after 72 years, and the digitization of those records allows us to drill into specific places and observe essentially every individual living there, providing comprehensive information on age, race, family, housing, and occupation. Although the census has always been a key source for understanding individual-level conditions in specific places, machine-learning techniques allow researchers to match a significant number of individuals across census years—whether they remain in the same community or move in the hope of making a better life somewhere else. (One of us has previously used linked census data to explore patterns of migration and social mobility in Victorian Britain and the early-20th-century United States, most recently including a study of the Dust Bowl migration in the 1930s.)
By some archival accounts, Greenwood was physically restored within a few years of the burning. Many buildings were rebuilt in the same location and in the same style as the originals, giving an impression of a return to normal. One local historian even wrote that by 1930, “everything was more prosperous than before.” However, if the economic and cultural prosperity of the 1910s had real momentum, a rebuilt Tulsa in 1930 shouldn’t look like it did in 1920; Greenwood in 1930 should have been even more vibrant and more prosperous than it was before the massacre. But while the physical environment arguably shows a Tulsa repaired, the data reveal deep scars that lasted at least two decades after the massacre.
The damage was evident not only in earnings but in other measures of social well-being. In 1920, Black Tulsans were more likely to own their own home, to be married, and to be employed than were Black residents of comparable cities. In 1930 and 1940, they had fallen behind on all measures. By contrast, white Tulsans maintained their standing relative to white residents of the other cities. Using standard economic-modeling techniques, we can estimate the specific effect of the massacre on Black Tulsans by controlling for trends among white Tulsans and Black residents of the comparison cities. Our model finds that by 1940, the likelihood of a Black Tulsan being married was 6 percent lower, the likelihood of being employed was 9 percent lower, and average earnings for employed Black Tulsans were 12 percent lower than if the massacre had never happened.
The massacre changed the social structure of the city. At least some prosperous Black Tulsa residents left. Ottowa and Emma Gurley, for example, moved to Los Angeles, where they opened a new hotel. But how prevalent was this response?
Comparing movers and stayers in this way reveals that Black Tulsa men with high-paying jobs were more likely to move from their home community after 1920, in comparison with both their white counterparts in Tulsa and their Black counterparts in the control cities. This outflow represents a dramatic shift for Tulsa, which, prior to 1920, had been a magnet for Black migrants.
These techniques also allow us to observe effects across generations. Our research linked children living with their parents in 1920 with their own adult census record in 1940. The massacre and its aftermath had profoundly different effects on Black children whose families left Tulsa than on those who remained. When we controlled for other factors, the choice to leave Tulsa after 1920 increased average Black adult earnings in 1940 by an astonishing 88 percent. In comparison, the migration effect for white Tulsan children was small and negative.
Our data vividly illustrate the diverging fortunes of those who left and those who stayed. For example, a Black man named Clarence Knight was a tailor in Greenwood in 1920. His wife, Daisy, worked as a housekeeper. The Knight family remained in Tulsa after the massacre. Their son, Ralph, dropped out of high school and in 1940 was a 28-year-old bellman at a local hotel. Brighter possibilities awaited many Black people who left. In 1920, Sylvester Holman, who lived in Tulsa with his wife, Annie, and their children, was working as a porter in a dry-goods store. Twenty years later, one of his sons, Clarence, appeared in census records in Nashville, Tennessee. He had graduated from high school and achieved a white-collar career as an insurance agent.
The scars from the 1921 race massacre run deep for Tulsa and its residents who experienced it firsthand. The data capture something important about the long-term cost of racist violence in America: In just 24 hours, the promising trajectory of an entire community can be significantly altered in ways that last for generations.