Zelinsky's Cultural Regions and Modernization at Home, 1880-1930
A more powerful test of Sowell's hypothesis
In the previous dispatch, we outlined a novel strategy for testing Sowell’s cultural persistence hypothesis. Sowell argued, in effect, that the cultural expansion of New England was central to the modernization process at home in the American century. Whereas New England culture was especially conducive to socioeconomic development, he argued, Southern culture had a retarding effect. In order to test this hypothesis, we obtained census microdata from 1880-1930. We constructed a novel measure of cultural persistence from surnames by computing the odds that a given surname was from a particular region. The results yielded support for Sowell’s hypothesis among whites but not blacks.
We suspected that the test was not powerful enough because of methodological problems. In particular, the census subregions are not coherent cultural regions. And perhaps throwing out data tied to surnames with a low regional signal introduced poorly understood biases into the dataset. We have redone the analysis with Wilbur Zelinsky’s cultural regions and kept the entire dataset of 3.8m observations to avoid confounding due to censoring. Otherwise we proceed as before.
Zelinsky is the greatest cultural geographer of the United States. We highly recommend his The Cultural Geography of the United States (Simon & Schuster: 1992). His major cultural regions are New England, the Midland (what the Census Bureau calls the Mid-Atlantic subregion), the Middle West, the South and the West. See the map below and the table for approximate period of settlement formation and the origins of the settlers.
As we did before, we first report our null model for socioeconomic status in early 20th century America. We find an even larger fixed-effect for sex (b = -9.7, P < 0.001) and race (b = -2.5, P < 0.001) than before (-9.0 and -2.0 respectively), suggesting that our new econometric strategy provides a more powerful test. The next figure displays the fixed-effects for decades/censuses. Upward mobility was especially rapid in the 1900s and 1920s, but modest in the 1910s, the decade of the world war.
Region fixed-effects relative to the Middle West — the all-American melting pot region — are displayed next. Unsurprisingly, New England comes out on top and the South at the bottom. There’s also a significant Midland (Pennsylvania and New York) and a Western premium. But recall that these are fixed-effects due to all sorts of particulars of the cultural region, including climate, soil quality, population history, and so on. The fixed-effects tell us nothing about the effect of culture on socioeconomic modernization. This is the null model that we have to control for in order to isolate the causal effect of cultural persistence.
We fit race-specific mixed-effects models as before. We control for decade, sex and region fixed-effects (not reported). We stratify by ancestral region. We admit random effects by ancestral region that we model as a linear function of log odds of the surname being from one of Zelinsky’s cultural regions. If Sowell’s hypothesis holds water, then we should expect a dose-response relationship between the odds of New England and Southern ancestry and higher and lower socioeconomic status respectively, after controlling for sex and region fixed-effects. We have 2.4m observations for non-Hispanic whites and slightly more than 400 thousand observations for non-Hispanic blacks. The dataset is too large to upload to my GitHub but is available upon request (if you have the cloud space for it). The next two figures document the main result.
For US-born, non-Hispanic whites, we find the expected dose-response relationship for New England and Southern ancestry. The coefficient for the log odds for New England ancestry is extremely large and highly significant (b = 0.37, P < 0.001). The coefficient for the South is negative but only marginally significant (b = -0.10, P = 0.089). In addition, both the West ancestry (b = 0.22, P = 0.045) and the Midland ancestry (b = 0.33, P < 0.001) sport large cultural persistence premia over Southern ancestry. Meanwhile, log odds of being from the Middle West sports a negative premium (b = -0.08, P < 0.001). The data is consistent with Sowell’s persistence hypothesis. It also reveals a more complex pattern of polarization due to cultural persistence.
For US-born, non-Hispanic blacks, contra to our previous results, we now find the expect dose-response relationship. The slope for the odds of Southern ancestry among blacks sports the predicted sign but is not significant (b = -0.06, P = 0.484). However, the slopes for the odds of ancestry in other regions are positive and significant. New England, Midland and Middle West cultural ancestry among blacks predicts higher socioeconomic status, with slope coefficients in the range 0.076-0.094, all significant at the 1 percent level. The slope coefficient for Western cultural ancestry among blacks is very large and significant (b = 0.24, P < 0.001). These results are somewhat consistent with Sowell’s hypothesis. But they rather suggest a more interesting pattern whereby blacks from the newly-settled West faced dramatically better opportunities across the United States. Yes, blacks with New England ancestry did better than blacks with southern ancestry. But these premia are dwarfed by the premium associated with Western ancestry for non-Hispanic blacks. Our interpretation is that blacks with Western ancestry had greater opportunities to build up social capital.
In order to test the robustness of our results, we also look at occupational prestige. We begin, as with socioeconomic status, with the null model that includes fixed-effects for sex, race, decade and region.
We find a very large effect for sex (b = -142.0, P < 0.001), suggesting that women were almost entirely locked out of prestigious occupations. Surprisingly, we find that the race fixed-effect vanishes (b = -0.2, P = 0.468). In order to make sure this surprising result is robust, we compute mean occupational scores by race. We find a significant difference in means (d = 7.0, StdErr = 0.23). So blacks did indeed have lower occupational prestige on average. But the racial bias vanishes once we control for region fixed-effects, likely due to the fact that, in 1880-1930, most blacks still resided in the South.
We find the same diachronic pattern as we did with socioeconomic status: the gains in occupational prestige were dramatic in the 1920s, and especially the 1900s. In the 1910s, however, occupational prestige actually declined slightly.
The region fixed-effects are somewhat different. Whereas the Midland and the West had comparable fixed-effects for socioeconomic status, there emerges a very significant gap between the two in occupational prestige. The reason for this intriguing pattern is not entirely clear. On the other hand, the pattern of New England in the lead and the South bringing up the rear is the same as the one we found for socioeconomic status.
Controlling for these fixed-effects, we again examine the effect of ancestry in cultural regions by looking at the slope coefficients of the odds of the surname being from one of Zelinsky’s regions.
We first look at US-born, non-Hispanic whites. The largest premium is again found to be associated with New England ancestry (b = 2.0, P < 0.001), while that for the South is statistically indistinguishable from zero (b = 0.5, P = 0.343). We find a positive premium associated with Western ancestry, although it is only marginally significant (b = 1.7, P = 0.060), possibly due to fewer surnames being associated with the newly-settled Western regions. We also find a positive premium for Midland ancestry (b = 1.6, P < 0.001). Intriguingly, we find a significant negative premium associated with Middle West ancestry (b = -1.6, P < 0.001), perhaps because of the region was predominantly rural and afforded fewer opportunities for building up social capital. Recall that, in the latter half of the 19th century, the Middle West emerged as the largest beef producing region in the world. It is not surprising that the sons of cattle farmers and meat packers fared worse on the occupational prestige scale in the early 20th century.
We now look at US-born, non-Hispanic blacks. The coefficient for the odds of Southern ancestry bears the expected sign but is statistically indistinguishable from zero (b = -0.04, P = 0.971). Meanwhile, blacks with ancestry in other regions sport significantly positive occupational prestige premia. New England blacks do not stand out from the crowd (b = 1.2, P = 0.007). For instance, the premium for Midland ancestry is higher (b = 1.6, P < 0.001), and that for the Middle West (b = 1.2, P < 0.001) is comparable to New England ancestry among blacks. Once again, we find that blacks with Western ancestry enjoyed much higher occupational prestige (b = 2.4, P = 0.027).
For both blacks and whites then, the results for occupational prestige are largely consistent with those we found for socioeconomic status.
We have established that, consistent with Sowell’s hypothesis, and after controlling for region fixed-effects, New England and Southern cultural ancestry predicts respectively higher and lower socioeconomic status and occupational prestige for both blacks and whites in 1880-1930.
For whites, we also found that cultural ancestry in the Middle West predicts lower socioeconomic status and occupational prestige, not just relative to New England ancestry, but also Midland and Western ancestry. This effect is more pronounced for occupational prestige than socioeconomic status, suggesting that it may be due to specific occupational structure in the beef and dairy producing Middle West.
For blacks, we also found that ancestry in the newly-settled Western regions predicts both higher socioeconomic status and occupational prestige relative to ancestry in all other cultural regions. The greater opportunities and freedoms out in the West were perhaps especially conducive to social capital formation among African-Americans.
These results suggest a significant influence of cultural factors in the combined and uneven development of the United States in the crucial period of the second industrial revolution, 1880-1930. Some always held better cultural cards than others. Socio-cultural advantages were amplified during modernization at home in the American century.
Prof. Zelinsky died in 2013, but David Hackett Fischer is alive, and is an emeritus professor at Brandeis. You should send these recent posts to him, and see if he has any comments.