Introduction In much of the social science literature riots are simultaneously seen as high-profile examples of inter-ethnic tension but also as curiously impotent-- as having little or no effect in the longer run. They are often considered significant only because they dramatically illustrate racial tensions in the United States. I argue, however, that in the right circumstances, ethnic disturbances can have significant long term effects. We see riots cluster around particular time periods-- in particular in the period from 1917 to 1921 and from 1980 to 1993-- when the rules of race relations are in flux. Riots, in turn, help shape new rules of race relations.
Riot outcomes in general have fallen below the radar of political scientists because we tend to focus on the governmental response to the riots, and this response is generally acknowledged to be ineffectual. But the response to riots is often non- governmental, and, because these riots are primarily local events, the initial response is largely local as well. Federal intervention may occur only further downstream. In this view, racial and ethnic re-negotiation takes place first at the local level and then at the national level. The predominant image of government’s role in race relations--formed in the context of the 1960s, when states and localities were laggards being coerced by the federal government’s executive, legislative and judicial branches-- is misleading. This image, in the broader sweep of the 20th century, is the exception rather than the rule. Federal agencies rarely create any program ex nihilo -- they scavenge among state and local programs for examples of what they want, incorporating local initiatives into federal agendas. Federal intervention is critical-- but largely because it reinforces certain local alternatives at the expense of others.
This paper demonstrates two things: that at certain historical moments like those of 1917-1921, inter-racial disturbances in the United States shared common structural conditions (and are profitably analyzed as part of a common social process), and second, that these riots were ‘critical junctures’-- not because of the events themselves, but because they accelerated institutional shifts, ushering in an era of racial containment. I’ll close by briefly indicating some parallels and differences with contemporary inter-ethnic disturbances.
Civil Disturbances 1917-1921 Between 1917 and 1921 there were nine major urban civil disturbances in the United States, punctuating the background hum of more localized events--the lynchings, firebombings and assaults--that were the basic vocabulary of inter-racial violence at the time.1 It should be understood that the outbreaks of these anti-black riots were symptomatic of rising racism in both North and South. Unlike much of the violence of the late 19th century, which was directed at immigrants as well as blacks, by the early 20th century, violence was primarily targeted at African-Americans. Some scholars have suggested that anti-immigrant sentiment in this period waned as immigrants made the transition from ‘other’ to ‘white,’ and moreover, that a part of ‘becoming white’ was taking part in violence against black Americans.2
List of Cases
East St. Louis, Illinois July 2, 1917
Houston, Texas August 23, 1917
Charleston, South Carolina May 10, 1919
Washington, D.C. July 19, 1919
Chicago, Illinois July 27, 1919
Knoxville, TN August 30, 1919
Omaha, Nebraska September 28, 1919
Elaine, Arkansas October 1, 1919
Tulsa, Oklahoma May 30, 1921
These disturbances have not exactly been forgotten, but not much has been made of them either. They are recounted as part of the broader story of white racism and recalcitrance against blacks in the United States, but not as being particularly significant in and of themselves. When these riots have been remembered, it has usually been as a string of case studies, each illustrating a unique confluence of events, albeit resulting in similar anti-black violence. This is somewhat understandable. Apart from their timing, the nine civil disturbances between 1917 and 1921 seem, on the face of it, to have little in common. They are widely dispersed geographically, some in the midwest, some in the west, the south or border states. The historical accounts have gravitated toward the particularities of the immediate events surrounding the disturbances. For example, Williams and Willliams, who chronicle several of the disturbances (Knoxville, Elaine AK, Tulsa and Chicago) in their Anatomy of Four Race Riots (1972), proceed to dissect these in minute hour-by-hour detail, without providing any real explanation of why these riots took place when and where they did. Elsewhere, we learn, for instance, that the riot in Elaine began over a share-cropping dispute, that Tulsa’s began with a gathering outside the courthouse following arrest of an alleged black rapist, that Chicago’s 1919 riot began after the drowning of a black boy at an all-white beach on Lake Michigan. D.C.’s riot began after the newspapers began printing lurid stories on black crime, and Atlanta’s followed clashes by white sailors and demobilized black soldiers on leave (Shapiro 1988).
It’s not that these histories aren’t interesting. These were, after all, some of the bloodiest riots in American history-- 38 people died in the Chicago riots alone (23 blacks and 15 whites) and more than 500 persons were injured. Though the riots usually began with whites targeting blacks; blacks were hardly passive victims; there were a considerable number of deaths among whites as well, as blacks mobilized in self-defense. Both blacks and whites used rifles, pistols, knives, and firebombs. Both groups tested new technologies in these riots; automobiles were used, for instance, in Washington D.C and Chicago to drive up and down streets while their occupants fired at bystanders. In Tulsa, there were rumors of airplanes being used to track the movement of blacks, and even to hurl down bombs. There were a lot of rumors in general: of women and children murdered; of black rapes and white looting, of rings of black gun-smugglers in East St. Louis and of Mexicans assisting blacks in the manufacture of bombs in Chicago (Commission 1968: 21); and of all kinds of secret plots and conspiracies against both whites and blacks. This all took place against a backdrop of political corruption and police incompetence; of racist unions, and a sensationalist press.
The stories of these riots make for vivid reading, but what is strikingly absent, from a social scientists’ point of view, is any attempt to draw comparisons among these cities, or to similar cities that did not experience racial disturbances. The causes of each event are seen as essentially non-replicable-- though the results are acknowledged as playing a part in the tapestry of unequal race relations being woven in the United States. The best accounts of the period--many of them written about Chicago’s 1919 race riot-- do point out structural factors underlying the disturbances. The Chicago Commission on Race Relation’s report The Negro in Chicago , published only three years after the riot, gathered extensive demographic, economic and fieldwork data to make the argument that the riots had three main causes: the migration of blacks from the South, the problem of housing for blacks in Chicago, and tense relations at work (Chicago Commission 1968). This explanatory framework has been borrowed by historians to explain the underpinnings of race relations not only in Chicago but in other cities of the time. Accounts of Chicago (Spear 1969, Tuttle 1972, and Hirsch 1983), Detroit (Sugrue 1996), Cleveland (Kusmer 1976) and Springfield, IL (Senechal 1990) have all borrowed from and elaborated on the suggestions of the Chicago Commission’s report. But while there have been a number of excellent case histories, historians have not attempted a comparative analysis of riot antecedents. It isn’t clear, therefore, whether the structural preconditions apply across riot cities, or really distinguish them from non-riot cities.
There has been only one quantitative effort at analyzing possible commonalities among the riots of the period: Lieberson and Silverman’s 1965 study of paired cases of riot and non-riot cities from 1913 to 1963. The study has the advantage of looking at both riot and non-riot cities across a wider pool of cities, but is limited by its methodology, which relies on a simple comparison across cases.3 They concluded that nothing in particular distinguished riot from non-riot cities. Why haven’t more of these studies been done? Part of the answer has to do with the long shadow of urban riots in the 1960s. A lengthy list of authors tried and failed to come up with any meaningful quantitative comparisons of underlying factors to the riots of the time.4 Their negative findings cast a pall on the use of quantitative tools to explore riots more generally. But the other part of the answer has to do simply with the difficulty of getting adequate data to test any hypotheses.
The data presented here were originally collected by Steven Ruggles and his collaborators at the University of Minnesota in 1997 as part of the Integrated Public Use Microdata Series or IPUMS. IPUMS is a database of census data from 1850 to the present (Ruggles and Sobek 1997). The data is a random sampling of census records, not the complete census, but it’s the only machine-readable collection of census material that is coded to allow comparison across years. For the 1910 census, for instance, the project took a sample of one in every two hundred and fifty records. That doesn’t sound like much, but realize that even for the subsample of cities over 25,000 which I focused on, there were still about 113,000 individual records.5 For 1920, which had a random sample of one in every two hundred individuals, there were almost 173,000 records.6 These individual records hold person and household data, including information on race, gender, ethnicity, citizenship, (and for immigrants, year of arrival and proficiency in English), city of residence, home ownership, occupation, education levels and literacy. By aggregating these data to the city level, we can see if there is something particular about the socio-economic conditions of cities that might have led to a riot outcome. These data, then, allow us to take another look at the hypotheses generated by historians and sociologists regarding the structural conditions underlying inter-racial disturbances.
Basically the data can be used to confirm whether the historical accounts of the urban disturbances-- as being sparked by contestation over housing, jobs and public space between whites and recent black migrants from the South--are correct. Three hypotheses can be tested separately:
The first is whether it these riots were related to demographic shifts, and in particular, the urbanization of Southern blacks, who were moving in increasing numbers into cities across the South, North and Mid-West. Black migration from the South to the North and West had been increasing steadily since the 1880s, but the Great Migration really began after 1910. The migration figures tend to be a little sketchy, but between 1910 and 1920 about 500,000 or more blacks moved out of the South, with most of the migration occurring after 1916 (thirteen southern states had an absolute loss of black population).7 The reasons for the migration are various: the devastation of the cotton crop in the South and resulting tensions within the sharecropping system certainly contributed, as did the increase in black lynchings. The outbreak of war in Europe, added to the effects of anti-immigrant legislation, meant the regular Northern labor supply was cut off (Chicago Commission 1968: 79). After the U.S. entered the First World War, an additional one million men were taken out of the labor force (Johnson and Campbell 1981: 71). A wartime economy meant a tremendous demand for labor, and for blacks, a once-in-a-lifetime opportunity.8 Employers who never would have considered blacks for semi-skilled manufacturing jobs before then began not only hiring but actively recruiting black migrants. Within a very short period of time African-American populations in Northern and Midwestern cities doubled, tripled and quadrupled.9
The second hypothesis is that labor competition was increasing between blacks and whites in this period, since blacks were moving increasingly into both skilled and unskilled positions previously occupied only by whites. This period was characterized by increasingly bitter accusations that blacks were being recruited by Northern employers as strikebreakers. In East St. Louis, after strikes in 1916 at meat-packing plants and in 1917 at the Aluminum Ore Company, there was considerable resentment on the part of white workers about the hiring of non-unionized blacks (Rudwick 1964: 18, 27; Chicago Commission 1968: 74).10 In the Chicago Stockyards strike of 1904 and the teamsters strike of 1905 employers used nonunion black workers as strikebreakers (Trotter 1993: 63; Spear 1969:36). 11 The fact that white immigrants also occasionally played the role of scabs and that blacks were discouraged from joining predominantly white unions was conveniently overlooked (Chicago Commission 1968 : 419-420; Kusmer 1976: 67).12 Susan Olzak uses time-series and event-history analysis in her study of ethnic competition and conflict between 1877 and 1914 to reach similar conclusions. Collecting information on inter-ethnic conflict across 77 cities, she found, in particular, that as occupational segregation decreased and union formation accelerated, inter-ethnic conflict increased. Increasing contact and competition in the workplace, she argued, leads to overt conflict (Olzak 1992).
The third hypothesis is that there was competition over housing and public space.13 Residents of middle and upper class neighborhoods, who had the resources to move further out to the city’s edge or beyond, tended to give way in the face of black middle class incursions, while working class whites, who were irrevocably tied to their investments in their homes, were more likely to stay and resist, if at all possible. In Cleveland, for example, historian Kenneth Kusmer describes how the areas of black settlement were bounded in some areas by native born settlements, in other areas by neighborhoods of Russian Jews, and elsewhere by other Central European immigrants. As African Americans sought housing, native-born whites in adjacent neighborhoods moved readily to outlying areas. Russian Jews, who were largely renters, moved relatively quickly as well. It was the second-generation immigrant neighborhoods, with fewer resources than the native born but higher rates of home ownership than Russian Jews, which held fast and at times violently resisted black migration into their neighborhoods (Kusmer 1976: 170-171). Immigrants in these neighborhoods, “[h]aving raised themselves above poverty, acquired a small home (with perhaps a large mortgage as well) and attained a modest level of income.... were fearful of association with any group bearing the stigma of low status” (Kusmer 1976: 171). There were similar processes at work in other cities: In Chicago’s 1919 race riot, historian Dominic Pacyga points out that more blacks were killed in Irish neighborhoods to the east of the ‘black belt’, where middle-class African Americans were competing for housing, than to the west, where blacks crossed Polish and Italian neighborhoods on the way to work in the meat-packing district (Pacyga 1997). As black neighbors moved in, threatening to devalue the real-estate investments of the Irish lower-middle class, the goal became the removal of African-Americans from the neighborhood (Pacyga 1997: 205; see also Grossman 1989: 175).
To test these hypotheses the 1910 and 1920 individual records are aggregated to the city level, so that instead of 113,000 cases for 1910 there are 230 cases for cities having over 25,000 in population; the individual cases are likewise aggregated 1920 data. Merging the 1910 and 1920 data (dropping those cities that appear in one sample but not the other) there is a final count of 171 cases-- cities with a population over 25,000 present in both 1910 and 1920. Out of these 171 cities, six had civil disturbances in this period--East St. Louis, Houston, Omaha, Chicago, Knoxville and Washington D.C. (three other cases--Charleston, Tulsa, and Elaine, Arkansas--were dropped for lack of data). The independent variables include measures of migration, labor force participation and home ownership by race and immigrant status; the dependent variable simply indicates a ‘riot/ ‘no riot’ outcome. I then use a multivariate logistical regression analysis to test which variables contribute to an outcome of ‘riot.’
Measures of cities’ population, black population, and foreign-born population in 1910 are all used as controls. The population change hypothesis is tested by measures of change in city population, change in black population, and change in foreign-born population over the decade 1910-1920. I expected the greater the change in a city’s population, the greater the chance of race riot. The labor competition hypothesis is tested by change in numbers of blacks and whites holding jobs in three occupational categories, “skilled labor,” “service jobs,” and “unskilled labor.” I expected the greater the change in these occupational categories, the greater the chance of riot. Finally, with housing competition, I expected that as ownership increased, so would the chance of riot.
Logistical Regression Effects of Population Change, Labor Composition,
and Housing Ownership on Race Riots in Chicago 1910-1920 B SE
Constant -9.7995 ** 4.1210
City Population 1910 -.000047 * .0000256
Black Population 1910 .0001 .0000753
Foreign-Born Population 1910 .0000874 .0000561
Change in City Population 1910-20 .0001 * .0000788
Change in Black Population 1910-20 -.0006 .0005
Change in Foreign-Born Population 1910-20 -.0001 .0001
Change in Black Unskilled Labor .0002 .0009
Change in Blacks in Service Jobs .0027 .0019
Change in Black Skilled Labor .0066 * .0039
Change in White Unskilled Labor -.0004 .0003
Change in Whites in Service Jobs .0006 .0006
Change in White Skilled Labor -.0012 * .0006
Change in Black Ownership .0002 .0003
Change in White Ownership .0002 * .0000947
Chi Square 31.085
Degrees of Freedom 14
Most of the variables are self explanatory, except for “skilled labor”, “service jobs” and “unskilled labor” which are listed in the chart below:
Table 3 Skilled Workers
Cement and concrete finishers
Compositors and typesetters
Cranesmen, derrickmen and hoistmen
Decorators and window dressers
Electroypers and stereotypers
Excavating, grading and road machinery operators
Forgemen and hammermen
Heat treaters, annealers, temperers
Inspectors, scalers and graders, lumber
Jewelers, watchmakers, goldsmiths, and silversmiths
Mechanics and repairmen, airplane, automobile, office machine, and railroad
Mechanics and repairmen
Millers, grain, flour, feed, etc.
Motion picture projectionists
Opticians and lens grinders and polishers
Painters, construction and maintenance
Pattern and model makers
Photoengravers and lithographers
Piano and organ tuners and repairmen
Plumbers and pipe fitters
Pressmen and plate printers, printing
Rollers and roll hands, metal
Roofers and slaters
Shoemakers and shoe repairers
Stone cutters and stone carvers
Structural metal workers
Tailors and tailoresses
Tinsmiths, coppersmiths, and sheet metal workers
Tool makers, die makers and setters
Craftsmen and kindred workers
Members of the armed services
Attendants, hospital and other institution
Attendants, professional and personal service
Attendants, recreational and amusement
Barbers, beauticians and manicurists
Boarding and lodging house keepers
Charwomen and cleansers
Cooks, except private household
Counter and fountain workers
Firemen, fire protection
Guards, watchmen, and doorkeepers
Housekeepers and stewards, except private household
Janitors and sextons
Marshals and constables
Policemen and detectives
Sheriffs and baliffs
Ushers, recreation and amusement
Waiters and waitresses
Watchmen (crossing) and bridge tenders
Service workers, except private household
Unskilled Labor Fishermen and oystermen
Garage laborers, car washers and greasers
Gardeners, except farm, and groundskeepers
Longshoremen and stevedores
Lumbermen, raftsmen and woodchoppers
Some of the changes in the variables in the model are in fact significant. The significance levels are not exactly overwhelming, but the findings do seem to provide some support for all three hypotheses:
1) population change: holding city population, black population and foreign born constant, we see that smaller cities were more likely to experience civil disturbances in this time period. Change in population is significant, which bolsters the migration thesis, though black migration is not significant.14 Basically, the greater the change in population, the more likely a city was to experience a disturbance in the time period. Estimating from the odds ratio, for every increase of 10,000 persons in a city’s population, there was corresponding 1% increase in the change of riot.
2) labor competition: the main locus of competition between blacks and whites is in the area of skilled labor, rather than unskilled labor or service jobs. This helps confirm a version of the labor competition thesis, with an emphasis on upwardly mobile blacks, rather than those at the bottom rungs of the ladder, the common laborer. As black skilled labor increased, or white skilled labor decreased, the more likely a city was to have a riot. For every increase of 10,000 black skilled laborers, there was a 6.7% increase in the chances of rioting; a corresponding decrease in white skilled labor meant a 1.2% increase in likelihood of a disturbance.
3) housing: home ownership tests the housing competition thesis. The higher the number of white homeowners, the greater the chances of an urban disturbance. This provides some confirmation to the notion that it was the resistance of white homeowners to the increasing movement of blacks into formerly all-white residential neighborhoods that helped contribute to the civil disturbances of the period. For every 10,000 new white homeowners, there was a 2% increase in the chances of rioting.
The purpose of the regression is to provide support for the contention that the riots are comparable; that they share some common structural characteristics. So while the evidence is not incontrovertible, it is significant. The regressions above suggest, contrary to much of the earlier research on race riots in the United States, that in at least some circumstances riots do, in fact, share common structural characteristics. In particular, the evidence supports historians’ contentions that between 1917 and 1921 population change, and consequently competition over housing and jobs, set the stage for inter-racial violence. This still leaves open the question: So what? What do these riots contribute to our understanding of American race relations, other than to serve as markers of the high-points of inter-ethnic violence? Why do these riots matter ?
Scholars looking for answers to similar questions after the period of racial rioting in the 1960s concluded that while riots had psychological and ideological consequences for the black community, the riots themselves elicited a limited institutional response. In general, studies found that riot commission reports were almost universally ignored, that rioting had at best moderate effects on programmatic spending by government agencies, and that the clearest beneficiaries of the rioting were law enforcement agencies (see Button 1989; Button 1978; Lipsky and Olson 1977; Platt 1971). The conclusion, on the whole, was that riots didn’t seem to have much of an effect at all. It seems that social scientists’ interest in riots was simply a reaction to particularly horrific occurrences of inter-ethnic violence, and an attempt to understand how to avoid such incidents in the future.
This response is singularly unsatisfactory. I think we have to ask a different question: Why is it that even though black migration accelerated in the 1920s, with a million more Southern blacks migrating to urban areas, most of them to northern cities, there were no further inter-racial explosions in that decade? If these earlier riots were post-war phenomena, triggered by recession, why wasn’t there more tension after 1929? The next large-scale racial disturbances were in Harlem in 1935 and then Detroit in 1943. The relative absence of large-scale violence after 1921 indicates that these earlier riots may have been something of a turning point in urban race relations. The question then is what kind of turning point were they?
One possible approach to this question is available from the historical institutionalists. Steven Krasner, who writes primarily about international relations, posits a view of history as a series of ‘critical junctures’ when periods of institutional stability are disrupted by intermittent exogenous shocks.15 Institutions are thrown off-balance by these events and either adapt through radical change, or collapse, spurring the rise of new institutional arrangements. However, this description of order, breakdown of order, and re-establishment of order was not what happened in 1917-1921. Cities experiencing these riots did not see a massive breakdown and rebirth of institutions.
Nonetheless these early riots may have been less revolutionary ‘critical junctures,’ not challenging the stability of the state but still acting as a watershed in aspects of the state’s institutional development. I argue that just as demographic and economic competition contributed to riots, the riots themselves contributed to the rise of institutions which addressed these conflicts, and contained them.16 These racial disturbances left behind an ‘institutional legacy’-- that is, they accelerated some local organizational and institutional outcomes at the expense of others. In addition, this legacy left a more permanent mark on national inter-racial interactions.17 Crises should be seen as simply an opening for institutional change. The opening itself does not guarantee any particular institutional outcome, or that any outcome will persist. As political scientists Collier and Collier find in their study of regime transitions in Latin America, the persistence and stability of institutional legacies is not automatic, but “rather, is perpetuated through ongoing ... political processes” (Collier and Collier 1991: 31).18 Institutions are not invulnerable to change, to decay, or to challenge.19 “In analyzing the legacy of the critical juncture,” Collier and Collier note, “it is important to recognize that no legacy lasts forever.” Some legacies, ‘self-destruct,’ by producing political dynamics that mitigate against the formation of stable institutional patterns (Collier and Collier 1991: 34). What allows an institution to persist is how it meshes with other institutional processes already underway.
So what does the historical record tell us about the aftermath of the 1917-1921 riots? As late as 1910 blacks were less segregated in midwestern and northern cities than some recent immigrant groups, notably Italians, Russians and Romanians (Lieberson 1980; Taueber and Taeuber 1965: 235-238; Kusmer 1976: 43, 164). This changed between 1910 and 1920, as immigrant groups became less residentially distinct, and the black population, increasing in numbers, was channeled into neighborhoods with existing African American populations.20 As African Americans began increasing in numbers in urban areas, whites in northern cities began experimenting with ways to minimize contact between black and white populations, particularly by keeping blacks in restricted residential areas. By the 1920s efforts to maintain racial boundaries had settled on private restrictive agreements or covenants. These were contracts not to sell, rent or lease property to minority groups, usually blacks (but also Jews and Asians) either among individuals or between individuals and an interested third party like a developer, real estate board, or neighborhood improvement agency. As a form of private contract covenants were legally enforceable in court, and in fact, were upheld by the Supreme Court.
The usual story explaining the spread of restrictive covenants is that they were a response to the Supreme Court’s ban on racial zoning (Spear 1969; Vose 1959: 5, 9; Weaver 1948: 231). Beginning in 1910, several Southern cities had used municipal zoning ordinances to prescribe separate zones for blacks and whites; by 1915 city councils in Baltimore, Richmond, Winston-Salem, Louisville, and Birmingham had enacted segregation ordinances.21 Dallas followed in 1916, and St. Louis held a popular referendum whose results showed the public two-to-one in favor (Vose 1959: 51). In 1917 Chicago’s Real Estate Board not only proposed explicit housing segregation by race, but also petitioned the city council to pass an ordinance prohibiting further migration of blacks to Chicago until such a time as the city could work out ‘reasonable restrictions’ sufficient to ‘prevent lawlessness, destruction of values and property, and loss of life’” (Grossman 1989: 174; Helper 1969: 226).22 At the time it seemed that legislated racial zoning had the potential to spread throughout the country, but as it turned out, the issue became the NAACP’s first major legal victory. The NAACP and its allies took the issue to the Supreme Court, where zoning of this sort was overruled in 1917 in Buchanan v. Warely. Though the Court was finding elsewhere that the state could not interfere with private racism, neither could it tolerate the state’s enforcement of racial zoning. These ordinances, the Court found, interfered with individuals’ property rights.23 Restrictive covenants began to be more widely used just around the time that the Supreme Court ruled against racial municipal zoning in Buchanan v. Warely. So it might seem reasonable, as a number of savvy observers have assumed, that covenants were a response to the Court’s decision. What observers have not asked is why, if covenants were replacing zoning as the main tool to restrict black residential choices, did covenants appear first in cities that had no history of zoning? If covenants were simply substituting for zoning, then covenants should have appeared first in cities which had already had experience with zoning; that is, cities mostly in the South or in the border states. All the evidence points, however, to the appearance of racial restrictive covenants not in these cities but in municipalities in the North and West.24 Cities in these regions implemented racial restrictive covenants not as part of a continuing campaign of segregation in all forms of public life (as zoning had been in the South),25 but rather as a response to the rapid demographic shifts occurring in urban areas, awareness of which was dramatically heightened by the racial disturbances of the period. Moreover, racially restrictive covenants became the primary tool to maintain segregated neighborhoods only after the racial disturbances of 1917-1921.
Figure 1: chart of racial restrictive covenants in Chicago and St. Louis26
The timing of the implementation of race restrictive covenants seems to support this interpretation of events. Long and Johnson’s data, for instance, on the formation of race restrictive covenants in Chicago and St. Louis indicate that race-specific deed restrictions in St. Louis sharply increased in number between 1915 and 1919, around the time of the East St. Louis disturbances occurring in 1917 (just across the Mississippi River). Chicago’s covenant campaigns didn’t begin until the 1920s, in the aftermath of that city’s riots in 1919. In both cities, once introduced, covenants spread quickly (Long and Johnson 1947). The difference in the timing of the appearance of racial covenants, and the coincidence of this timing with the race riots in each city provides evidence of the independent effect of the riots on the introduction of restrictive covenants.27 By the late 1920s race restrictive covenants had spread both within and among cities, particularly in the North and West.28 In theory, blacks in northern cities continued to have free access to housing, but in practice their choices were increasingly limited by both formal and informal barriers. The usual pattern was to have concentrations of these restrictive agreements in areas around black neighborhoods, relatively sparse coverage elsewhere in the city, and much more widespread application in newer suburban developments.29 Chicago, for instance, followed this pattern; by 1947 covenants covered about half of the city’s residential sections.30 Figure 2: map of racial restrictive covenants in Chicago31
With black neighborhoods hemmed on all sides, either by industrial zones, covenanted neighborhoods or Lake Michigan, the area occupied by blacks in Chicago remained substantially the same from the 1920s to the 1940s, when covenants were finally ruled unconstitutional.
The rise of restrictive covenants is hardly the whole story. There were alternative responses to the riots which were less successful. For example, the 1922 Chicago Commission studying the aftermath of the riot suggested that one solution to the underlying tensions would be for the public sector to become more involved in providing housing for new black migrants and to encourage the dispersal of African Americans through the city (Chicago Commission 1968: 645). This clearly did not happen. On the other hand, the spread of the Ku Klux Klan in northern and mid-western cities in the early 1920s in response to immigration and black migration failed as well.32
Figure 3: map of spread of KKK in Chicago33
This mobilization peaked after only a few years in the face of immigrant and black hostility to the movement, and the failure of the Klan’s candidates at the ballot box (Jackson 1992 : 126, 142). Crises spawn a wide range of institutional and organizational responses, some of which persist, many of which do not. So why were racially restrictive covenants and their cousins--racial steering, block-busting, and redlining-- more successful, institutionally, then their alternatives in the aftermath of the riots?34 The answer, I think, is that the institutional legacies of critical junctures are diminished or increased by their interactions with broader institutions which are going through their own processes of change.
In the 1910s and 20s responses to the perceived incursion of blacks were primarily local, and not only local but private. The federal government was largely absent from the picture. While the federal judiciary upheld local racial restrictions, the federal government simply did not have the capacity to meddle in any kind of detail in municipal affairs.
Figure 4: chart of federal transfers to the states 1912-192735
This is reflected in the figures for federal transfers to the states-- before 1921 transfers were minimal. The passage of the Federal Highway Act that year saw the first significant flow of funds to localities--and was opposed by urban representatives, who saw it primarily as a boondoggle benefiting rural counties (MacDonald 1928: 239-240). Metropolitan governments, for their part, did not have the inclination, or the resources, to address the structural issues underlying civil disturbances.
As a result, restrictive covenants and the like were initially sponsored not by the public sector but rather by private institutions. Real estate boards and their allies, particularly neighborhood and home owners’ associations, endorsed restrictive covenants and mobilized campaigns to collect the necessary signatures for their creation (Philpott 1978: 193-196; Long and Johnson 1947: 38).36 Local real estate agents were enjoined by the National Association of Real Estate Boards (NAREB) to avoid “infiltration of inharmonious elements” into neighborhoods. Article 34 of NAREB’s code of ethics stated that “A realtor should never be instrumental in introducing into a neighborhood ... members of any race or nationality... whose presence will clearly be detrimental to property values in that neighborhood” (Long and Johnson 1947: 58).37 Real estate boards could, and did, expel or ostracize members who did not keep to the code (Helper 1969: 23, 190, 228, 230). The fact that Chicago was home to the National Association of Real Estate Boards was instrumental not only in the local realtors’ enthusiastic support for covenants, but in the propagation of covenants elsewhere. In 1927 NAREB drafted a standard restrictive covenant document which was then shared with local real estate boards around the country (Plotkin 1998; Helper 1969: 229-30), and encouraged local boards to form home owner’s associations to sign onto covenant agreements.38 Together these loose confederations of national and local associations did much of the work of enforcing patterns of racial residential segregation which kept African Americans residents confined to deteriorating urban neighborhoods.
Restrictive covenants succeeded in part because they were in step with the segregationist notions of the time, but their implementation and diffusion was by no means preordained. Like other responses to the riots, racially restrictive covenants ran into challenges from very early on. The NAACP began contesting covenants almost as soon as they began to be implemented--though they survived a Supreme Court decision in 1926 and weren’t overturned until 1948.39 If racial restrictions on residential housing had relied solely on local efforts, it’s likely that residential desegregation would have begun to take place sooner than it did. Some scholars have pointed out that even at their height racially restrictive covenants were not always effective, particularly as demographic pressures simply became irresistible in the 1940s (Hirsh 1983: 16, 30; Weaver 1948: 212, 234).40 These racialized institutions would not have been able to sustain themselves if not for the backing of local institutions like real estate boards, and later, the federal institutions which played an increasingly important role in the housing market after 1932.
Local privatist responses were inherently vulnerable to challenge. While they originated in the absence of a strong federal presence, they persisted because of that presence. With the Depression, the federal government began to increase its involvement in local affairs, not least in the area of housing. This involvement began with the establishment of the Home Owners Loan Corporation (HOLC) in 1933, which was designed to reduce mortgage foreclosures.41 HOLC single-handedly established the pattern for long-term mortgage loans. In doing so, it had to make predictions about how housing covered would fare over the life-time of the loan. Housing appraisals became increasingly systematized, rating not only the structural integrity of the homes themselves but the neighborhoods surrounding them. Areas were rated, with the ratings privileging homogeneous native-born white collar neighborhoods, and downgrading older, mixed-use, ethnic or black neighborhoods (Jackson 1980: 423). As a result, African-American neighborhoods were systematically deprived of mortgage and other lending. The patterns established under HOLC persisted after it was folded into the Federal Housing Administration (FHA). “The FHA started its career by accepting prevalent real estate doctrine that nonwhites should be kept out of white neighborhoods in order to protect property values” (McEntire 1960). The FHA’s 1938 Underwriting Manual instructed appraisers that “if a neighborhood is to retain stability, it is necessary that properties shall continue to be occupied by the same social and racial classes.” Appraisers were to predict “the probability of the location being invaded by... incompatible racial and social groups.”42 To ‘preserve’ neighborhood character, FHA officials were enjoined to uphold racial restrictive covenants. The Manual openly recommended “subdivision regulations and suitable restrictive covenants” that would be “superior to any mortgage” (Jackson 1985: 208).43 Federal intervention encouraged segregated housing by reinforcing local discriminatory practices, among them race restrictive covenants.
The story of the federal government’s role in promoting segregated housing has been told extensively elsewhere (Jackson 1988; McEntire 1960; Abrams 1955; Weaver 1948).44 The point I want to emphasize here is the link between local and federal institutions. As the federal government played a greater role in local affairs during the Depression, it did not step in and invent entirely new institutional structures and practices. More often it melded its new role with existing practices. In the case of housing this meant adopting the standard practices of local real estate boards and banks, particularly in respecting and upholding local racially restrictive covenants and racial steering. As HOLC and then the FHA got underway, they were staffed at the federal and local levels by former real estate professionals who operated under and accepted the profession’s racial strictures (Weaver 1948: 72), who then transferred these norms into federal practices. For example, Frederick Babcock a prominent figure in real estate appraisal, who wrote a widely used appraising text The Valuation of Real Estate (1932) advocating the enforcement of racially segregated neighborhoods,45 went on to write the FHA’s 1938 Underwriting Manual, which incorporated similar assumptions of the desirability of neighborhood homogeneity and the pernicious influence of black home ownership on property values (Helper 1969: 197, 202). In this and other respects, the FHA’s institutional racism was a reflection of local attitudes and policies.
Historian Kenneth Jackson writes that the lasting damage done by the national government was that it “put its seal of approval on ethnic and racial discrimination” and that “more seriously, Washington’s actions were later picked up by private interests, so that banks and savings and loan institutions institutionalized the practice of denying mortgages ‘solely because of the geographical location of the property’” (Jackson 1985: 217), that is, redlining. While the role of the federal government in reinforcing racially segregated housing in urban areas was certainly crucial in extending the half-life of race restrictive covenants and other local segregatory tools, the historical record shows that influence between local and federal institutions ran both ways-- the federal government first adopted local racial institutions, and in doing so, reinforced them.46
What do these earlier riots have do with us now, today? I’ll mention a couple of points briefly here. The major riots which have taken place in the contemporary period are, like those in the earlier period of the century, competitive riots. Urban ethnic disturbances, from 1980 to the present, share many of the same characteristics as the earlier riots from 1917-1921. They too are characterized by rapid urban demographic shifts and competition among new and old residents over scarce resources, occurring in the context of an economic downturn (and the general shift of resources away from cities) (Jones-Correa 1998).
In the contemporary case, the migrants are new immigrants arriving to the U.S. after 1965 (many of them since 1980) into particular regions, and to particular cities within those regions. One of the characteristics of the post-1965 immigration wave is that it is a very concentrated, targeted migration.
Figure 5: chart of immigrants 1985-1990
Figure 6: chart of foreign born by state
This immigrant wave is bifurcated between the highly educated and those with relatively few skills. Those with less skills tend to settle in areas where they can afford housing, which means settling in areas in or adjacent to poorer black neighborhoods. The black middle class, having more resources, has been moving out of these neighborhoods since the late 1960s, and this trend only accelerated through the 1970s and 1980s. Those left in these neighborhoods are the those with the fewest resources. The convergence of two populations with very few resources and overlapping occupational and residential niches leads to friction, and occasionally leads to rioting. Four of the top ten immigrant-receiving urban areas had riots in the 1980s and 1990s--Los Angeles, New York, Miami and Washington D.C.
The hypotheses, then, are similar to those for the 1910-1920s period, positing that population change, job competition and residential displacement contribute to riot occurrence. We can test these hypotheses using, again, the IPUMS census data. For 1980 there are 515,201 records for individuals residing in cities with populations of 100,000 or more; for 1990 there are 438,482 such records. Aggregating these samples to the city level (134 cities with populations of 100,000 or more for 1980 and 135 in 1990), and then merging the files, we end up with 103 cases: those cities with populations of 100,000 or more included in both the 1980 and 1990 samples.