THE SECOND GREAT MIGRATION AND PROPOSITION 14

Throughout the early twentieth century, the African American population in California was small and concentrated in a handful of census tracts designated specifically for non-Whites. The 1940 decennial census, conducted immediately before the start of the SGM, indicates that African Americans comprised less than 2% of the state population and less than 3% of the population in urban counties that would come to house the largest Black populations. Holding aside the Black population, the non-Black minority population in California in 1940 was only 2.7%, leaving the state nearly 96% White.

The SGM drastically changed this, and represents one of the largest demographic shocks to White society in contemporary American history. By 1960, California's Black population grew by over 600% to approximately 885,000. In a number of cities, the Black population exploded: Berkeley, Emeryville, Richmond, and Vallejo all saw their Black populations expand by ten percentage points or more. In Compton, the Black population grew from zero to nearly 40 percent by 1960. Figure 1 displays the cities with the highest Black population growth. This population growth strained housing in the few Black neighborhoods throughout the state, increasing demand for housing in neighboring communities (HoSang Reference HoSang2010). As the Black community grew, political elites and homeowners sought to protect White communities from what they saw as a threat to home values and neighborhood identity (Lipsitz Reference Lipsitz1996). Together, these actors maintained racial exclusion through a variety of official and unofficial policies, leading to some of the most entrenched segregation in the nation (HoSang Reference HoSang2010).

FIGURE 1.

Map of Rapidly Growing Black Cities 1940–1960

Note: City-level African American population growth 1940–1960 in Southern (panel A) and Northern (panel B) California. 98th percentile growth cities include Compton, Emeryville, Richmond, Vallejo, and Berkeley and additional 95th percentile growth cities of Pasadena, Elsinore, Menlo Park, Pittsburg. A map of the central valley, including 95th percentile growth cities of Bakersfield, Fowler, and Madera, is presented in Figure M.1.

The election of state legislator William Rumford (D) in 1949 and Governor Pat Brown (D) in 1958 aided in the passage of several anti-discrimination measures, precipitating the White backlash that culminated in Proposition 14. Real estate interests, politicians, and evangelical church leaders coordinated to collect signatures for a proposition to amend the state constitution, protecting what White residents believed was their right to discriminate. The measure, Proposition 14, passed 65 to 35 percent with overwhelming support from White Californians who, according to the CA Field Poll surveys, supported the measure by 3 to 1 (HoSang Reference HoSang2010). Less than one year after the passage of Proposition 14, the Watts Riot broke out in Los Angeles, which was one of the most destructive urban race riots in American history (Queally Reference Queally2015).

In sum, 1940 to 1965 represents a pronounced period of racial change and conflict in the American West, and key features of the SGM afford a unique opportunity to assess the causal effect of racial threat on White voting behavior. First, the residential decisions of the study group (Whites) were largely made in the absence of the treatment group (African Americans). The interpretation of findings from prior observational studies of racial threat are often marred by concerns over selection bias; however, in the case of the SGM, residential decisions by Whites were made largely without consideration of the Black composition of their own or neighboring communities. Second, the migration of African Americans into California was rapid and concentrated in a few cities, increasing our confidence that the 1964 vote preceded much of the White flight that occurred between mid-1960 to 1980 following the Watts Riots, the overturning of Proposition 14 by the Supreme Court, and school desegregation (Schneider Reference Schneider2008). In short, we treat the rapid increase in California's Black population as a racially threatening “shock” to White society and a potentially important source of White support for Proposition 14.

EMPIRICAL STRATEGY AND DATA

As the SGM involved the drastic growth of Black populations in key areas throughout the state of California, our empirical strategy centers upon analyzing the effect of spatial proximity to Black growth cities on White support for Proposition 14. Theories of racial threat are rooted in Key's (Reference Key1949) proposition that White political behavior in the American south was partly a consequence of the presence of African Americans in their communities. More recent work, however, argues that it is the in-migration and growth of an out-group that serves as a motivating shock to White political behavior (Green, Strolovitch, and Wong, Reference Green, Strolovitch and Wong1998; Hopkins Reference Hopkins2009; Newman Reference Newman2013). Following this work, and that by Enos (Reference Enos2016), we conceptualize racial threat as the motivating effect on White political behavior of drastic changes in a spatially proximate Black population. Given that theories of racial threat argue that the psychological salience of a group is a function of its size and spatial proximity (Enos Reference Enos2016), we conceptualize our “treatment” as the proximity of White voters to epicenters of Black population growth.

We constructed a dataset from historical administrative data from the U.S. Census Bureau and the Office of the Secretary of State. The data is provided at the Census place (i.e., city) level, the finest level of aggregation we could acquire from historical sources. In total, our full dataset includes voting results for 392 cities in California. Because we are primarily interested in White voting behavior, we subset the data for our analyses to cities that were 90% or greater White (n = 340) in 1960. Our dependent variable is city-level vote for Proposition 14 (mean = 65.7%, sd = 10.6%) as reported by the 1964 California Secretary of State Supplement to the Statement of the Vote. Our key independent variable is city proximity to its nearest Black growth city. To calculate this measure, we used city-level demographic estimates from the 1940 and 1960 U.S. Census files to calculate percentage point change in the Black population (mean = 1.17%, s.d. = 3.97%). Figure 1 displays the level of Black growth in cities throughout Southern and Northern California. We identified Black growth cities as those in the 98^th percentile of Black population growth, capturing cities that experienced Black population growth between 10 and 40 percentage points over the twenty-year span, constructed a matrix of Euclidean distances between the centroids of all California cities, and defined proximity as the distance in miles from the nearest Black growth city (mean = -69.8, sd = 64). For ease of interpretation, we divide this variable by 100 and multiplied by -1, so that a unit increase indicates a 100-mile increase in proximity. By using a continuous treatment indicator on non-nested data, we bypass the concern in the racial threat literature over the sensitivity of results using multilevel data to the choice of administrative boundary (Tam Cho and Baer Reference Tam Cho and Baer2011; Voss Reference Voss1996; for more discussion see Online Appendix B).

We gathered a number of additional control variables at the census tract level and merged them with our dataset via a weighted spatial join. We obtained 1964 voter registration figures for cities from the Berkeley School of Law Center for Research and control for city-level percent Democrat (of registered) to rule out partisanship as a confounder. Measures of population density control for variation in geographic and population size of each city. As poorer and more racially conservative Whites might be more likely to live adjacent to high Black growth cities, we include controls for median income, home ownership, and unemployment (descriptive statistics are included in Online Appendix C).

ROBUSTNESS CHECKS

We demonstrate in the Online Appendix that our results hold when using beta regression (Table D.1) and when including county-level fixed effects (Table D.2). Given our interest in the behavior of White voters, we demonstrate that our results hold when further restricting the analysis to cities that were 95% or greater White in 1960 (Table E.1), and when employing an ecological inference model to estimate support for Proposition 14 among White voters (Table F.1). Alternatively, our results hold when lifting the percent White sample restrictions (Table E.1). Additionally, our results hold when using different thresholds to define a “treatment” city (Table G.1) and when measuring proximity to Black growth cities using driving distances and times (Table G.2). To account for possible post-treatment bias, we demonstrate that our results hold when replacing our 1960 control variables with pre-treatment (i.e., before the SGM) variables derived from the 1940 Census (Table H.1). While column 1 of Table 1 demonstrates that our results hold when analyzing all predominantly White California cities, to further ensure that our results are not driven by the mid-to-large cities covered by the 1940 and 1960 Censuses, we demonstrate that the effect of proximity holds when analyzing the n = 187 smaller-sized cities not covered by these censuses (Table E.1). Next, as placebo tests, we demonstrate that the positive effect of proximity to Black growth cities is restricted to Proposition 14 and not observed when analyzing race-neutral propositions (Table I.1). Lastly, we uncover complementary results to those presented in Table 1 when analyzing individual-level survey data estimating White support for Proposition 14 as a function of Black population growth in respondents’ counties of residence (Table J.1).

While the SGM and the 1964 election offer a case where selection bias is substantially reduced, such concern is not entirely removed. Black residents did not settle at random in California cities. Further, while “White flight” from California cities was most pronounced between mid-1960 to 1980 (Schneider Reference Schneider2008), it is possible that substantial residential sorting occurred between 1940 to 1964. One method for addressing this possibility is to re-analyze our model among targeted subsamples of the data.

First, we explore whether our results hold when looking at White cities with higher levels of White residential tenure. The 1960 decennial census includes data on when individuals moved into their residence. Using this data, we can restrict our analysis to White cities where a higher rate of residents reported having moved in before 1940 (i.e., before the start of the SGM). The first row of results in Figure 3 (full results available in Table K.1) demonstrates that the effect of proximity to Black growth cities holds (p < 0.01) when looking at above-median tenure cities. This result is critical, as it indicates that when conducting a test reducing White residential selection bias, the estimated effect of proximity remains positive and statistically significant.

FIGURE 3.

Reanalysis by Residential Tenure, White Growth, and Housing Markets

Note: OLS regression coefficients and heteroskedastic robust standard errors of proximity to Black growth cities. 90% or greater White cities. The first panel splits the sample at median percent of city residents who moved into their residence prior to 1940, the second splits the sample at median White growth (1940-1960), and the final controls for proximity to cities with the largest drop in housing availability or increase in home values. Full model results available in Appendix Table K.1.

Second, we can assess whether our results hold when looking at majority-White cities with lower levels of White population growth between 1940–1960. Racially conservative Whites residing in cities experiencing Black in-migration may have fled to adjacent all-White cities, taking their racially threatened attitudes with them. Such a process could have induced the findings we observe, suggesting they are less due to the activation of racial threat among Whites residing in proximity to Black growth cities and due instead to the migration of racially threatened Whites to neighboring cities. While we find suggestive evidence that White populations contracted the most in cities within five miles of Black growth cities, we find no evidence that White population growth disproportionately occurred within neighboring cities five or more miles away from Black growth cities (see Figure L.1). Moreover, the results presented in Figure 3 belie this concern: rather than being endemic or more pronounced in White cities experiencing high White population growth, we find the effect of proximity to Black growth cities is stronger in White cities with below median White growth (range: [−17.2, −0.003], mean = −2.2).

It is also possible that “redlining” (Rothstein Reference Rothstein2017) forced African Americans to settle in neighborhoods deemed less desirable, which may have contained poorer and more racially conservative Whites. Such possibility could explain the relationship we observe between proximity to Black growth cities and White support for Proposition 14. This possibility is not suggested by the data, as proximity to Black growth cities is not strongly correlated with pretreatment indicators of 1940 socioeconomic standing, such as median home values (r = 0.10), homeownership rates (r = 0.15), or unemployment rates (r = 0.01). Another concern is that our results are due to housing competition. It is possible that proximity to Black growth cities is capturing the effect of proximity to areas experiencing increased competition for housing. The bottom panel of Figure 3 display the coefficient for proximity to Black growth cities when adding a control variable for proximity to cities with the most drastic (95^th percentile) contraction in available housing units (i.e., vacant units for sale or rent) between 1940–1960, or the most drastic increases in home values between 1940–1960. We find the effect of proximity holds in both models (p < 0.001), indicating that proximity to Black growth cities remains positive and significant when holding constant proximity to areas manifesting symptoms of housing competition.