What was a pull factor for African Americans to leave the rural South and move to the urban North at the end of the nineteenth century quizlet?

1Many papers examine this selection process, which demographers often call the “healthy migrant hypothesis.” See, e.g., Halliday and Kimmitt (2008), for evidence that the healthy tend to have higher geographic mobility.

2In our conception, both schooling and migration are decisions that occur early in life, prior to labor market participation. Of course, some migration (and schooling, for that matter) occurs at older ages. However, as we document below, most migration by African Americans out of the Deep South was indeed by younger individuals.

3At most we can only observe educational attainment. Across countries there is strong negative relationship between education and mortality (as shown in Preston, 1975, and many papers that followed), and the same is true within countries. In terms of our model, this correlation could be the consequence of a positive relationship between our latent characteristic (α) and longevity, or because education improves prospects for longevity, or both. Cutler and Lleras-Muney (2010) give a good discussion of the issues involved, and provide links to the literature. Work by Black, et al. (2013) shows why identification of the effect of education on mortality is difficult.

4To see this result, set (2) and (3) to be equal and use the implicit function theorem.

5Notice, for instance, that once we assume independence, we can use Bayes rule to write out E(yM=0|C) = {[Pr(N) + Pr(C)]/Pr(C)} E(yM=0|N ∪ C) − {Pr(N)/Pr(C)} E(yM=0|N). Each element on the right-hand side of this latter equation corresponds to an easily-estimated moment: the term [Pr(N) + Pr(C)] is estimated by the mean migration rate for individuals born in non-railway towns (M̄z=0); the term E(yM=0 |N ∺ C) has expectation E(ȳM=0, z=0), i.e., mean longevity among non-migrants from non-railroad towns; and so forth.

6The merge was made possible by the Center for Medicare and Medicaid Studies and the Social Security Administration (and with extensive confidentiality protection). We are grateful to James Vaupel for his role in making these data available to us.

7We observe residence only in older age, so some individuals classified as non-migrants could be individuals who migrated and then returned to the South in old age. Analysis below suggests this is a limited phenomenon (see footnote 12).

8For example, we would not want to include Virginia, because conceptually a move from a birthplace in Northern Virginia to nearby Washington DC is quite different than the long-distance moves that were most common during the Great Migration.

9The 11 former Confederate states are Virginia, North Carolina, South Carolina, Georgia, Florida, Alabama, Mississippi, Louisiana, Tennessee, Texas, and Arkansas. (Kentucky and Missouri were officially neutral, though they were represented by stars in the Confederate flag when secessionist parts of these states joined the Confederacy in 1861.) We refer to non-Southern states as the “North” merely for convenience.

10Boustan (2010) discusses migratory patterns and gives references to the extant literature. In early work, Wright (1906) noted the emergence of these migratory streams (as discussed in Trotter, 1991).

11In contrast, individuals in our sample who remain in their home States often reside in non-metropolitan areas: 0.45 in South Carolina, 0.42 in Georgia, 0.33 in Alabama, 0.77 in Mississippi, and 0.28 in Louisiana.

12As for return migration at older age, we conduct an analysis using the 2000 Census, which asks the location of residence 5 years earlier (i.e., 1995). Our analysis was restricted to black men and women born in the Deep South who were aged 60 and older in 2000. We estimate transition probabilities conditional on residence in 1995. Among those still living in the Deep South in 1995, virtually all (over 99%) remained in the Deep South in 2000. As for individuals living in the North, 97.4% remain in the North and only 1.9% return to the Deep South, with 0.7% moving elsewhere in the South (e.g., Florida). In short, it appears that mobility rates are quite low at older ages.

13Among many contributions are Smith and Welch (1989), Maloney (1994), and Margo (1995). Collins and Wanamaker (2012) provide evidence of positive selection into migration. There are difficulties in interpretation given price differentials across locations (see Black, et al., 2013).

14Additional details are available in an appendix online.

15We have access to data through 2002. For the regression in which dependent variable is survival to age 70, we use birth cohorts, 1916–1932. For survival to age 75, we can only use birth cohorts, 1916–1927. Hence sample size is smaller in the second regression.

16Using our data we find that life expectancy at age 65 conditional on not surviving to age 75 is 4.8. Life tables from 2001 indicate that life expectancy at age 75 for blacks was approximately 10, so at age 65 life expectancy conditional on surviving to 75 was approximately 20. Given our estimate that the LATE of migration on survival to age 75 is −0.10, around a mean survival rate of approximately 0.70, migration is estimated to reduce the probability of survival to age 75 from approximately 0.75 to 0.65. Let EN be life expectancy at age 75 for those who move North and ES be the life expectancy for those remaining in the South. Given all this, the effect of migration on expected extra years (at age 65) is [0.65×(EN+10) + 0.35×4.8] – [0.75×(ES+10) + 0.25×4.8]. If for those over age 75 migration has no effect on mortality, ES=EN=10, and we have an estimated impact of −1.52 expected extra years. If we were to suppose instead that at age 75 migrants continue to have higher mortality, this number increases in absolute value. For instance, if migrants have life expectancy 1 year less than non-migrants at age 75 (EN=9.5 compared to ES=10.5), our calculation is −2.22 expected extra years.

17As yet another alternative, we tried a cruder location variable—the fraction of people within one’s county who are born in a railway town—which then allows us to include people for whom we have county of birth but not birth town. As noted above, in baseline regressions we exclude about one quarter of our sample because we lack exact town of birth. When we use a county-based strategy, our sample size increases to 960,552 for the regressions in Panel A and to 673,356 for Panel B. For these samples, OLS estimates are nearly identical to those in column (1); the 2SLS estimate (with standard error) for “survival to age 70” is −0.096 (0.032); and the 2SLS estimate for “survival to age 75” is −0.155 (0.046). Of course this is a different LATE—one generated using an identification strategy that corresponds less well to the theory we presented above. Still, qualitative inferences are similar.

18Our sample of individuals born in smaller towns only eliminates those born in New Orleans, LA; Atlanta, GA; Birmingham, AL; Savannah, GA; Shreveport, LA; Charleston, SC; Montgomery, AL; Mobile, AL; Spartanburg, SC; Macon, GA; Columbia, SC; Greenville, SC; Augusta, GA; Jackson, MS; Sumter, SC; Columbus, GA; Baton Rouge, LA; Meridian, MS; Tuscaloosa, AL; and Selma, AL.

19Wald estimators (see (6) above) can be calculated directly from these means; they are 0.613 – 0.680 = −0.067 for men, and 0.711 – 0.819 = −0.108 for women. These differ slightly from the estimates presented in Table 4 because those estimates are from regressions that also include cohort and State indicator variables.

20To give two recent examples, Atack and Margo (2011) show that the arrival of rail transportation increased farmland value in the American Midwest in the mid nineteenth century, and Donaldson’s (forthcoming) work suggests that railroads increased local real income levels in nineteenth and early twentieth century India.

21The idea that early-life economic conditions play a crucial role for shaping long-run health has been developed in an important literature, e.g., Barker (1990), and Fogel (2004). Importantly, for our study, Preston, Hill, and Drevenstedt (1998) provide evidence that among African Americans, an unhealthy childhood environment was associated with a reduced probability of survival at every age, up to age 85.

22We tried this exercise also for survival to age 75, estimated with 1,052,479 observations. Here the “railroad town effect” is similarly very small, 0.0018, with a clustered s.e. of 0.0013 (not statistically significant).

23This follows from observing that in model (9), plim(bIV) = β + plim{(δ̄z=1 − δ̄z=0)/(M̄z=1 – M̄z=0)}. The denominator of the adjustment term is calculated using statistics from column (3) of Table 3.

24Many black agricultural workers in the South were in sharecropping arrangements. As is clear from discussions such as Braverman and Stiglitz (1982), an increase in farmgate crop prices (surely plausible when rail transportation arrives) needn’t increase sharecrop income.

25Unfortunately, the Vital Statistics death data do not record state of birth for other years of interest to us, e.g., 1950 and 1970.

26For instance, we estimate that from age 65 to 70, the effect of migration on women’s mortality is 0.05, from a base of 0.10, and the effect on men’s mortality is 0.07, from a base of 0.18.

27Gibson, et al. (forthcoming) do not argue, though, that migration therefore reduces overall well-being among Tongans, because migration increases income (McKenzie, Gibson, and Stillman, 2010), and may also have contributed to improved mental health (Stillman, McKenzie, and Gibson, 2009).

28Importantly, an individual can have COPD or cirrhosis listed as a present condition (as analyzed in Table 7), but have some other more-common primary cause of death, such as cardiovascular disease or cancer of respiratory and intrathoracic organs.

29For each cause of death we formed 95 percent confidence intervals, using a bootstrap procedure (with 1000 replications). For chronic liver disease and cirrhosis, the index confidence intervals were 1.38–1.65 for men and 1.27–1.56 for women. For cancer of respiratory and intrathoracic organs, the corresponding index confidence intervals were 1.03–1.08 for men and 1.29–1.38 for women.

30Along these same lines, Higgs (1977) points to a U.S. Census (1918) report in which Census officials noted that most of the black population of the time lived outside the largely urban “registration area” for which mortality data were available, but then speculated that “it is highly probable that mortality is much lower in this rural element than it is in the population of the registration area, which is largely urban and largely a migrant population.” These demographers express a concern that black migrants from the South were “subjected to conditions similar in some respects to those encountered by the foreign immigrant, and the difficulties of adjustment to these conditions may be reflected in the higher mortalities from such causes as tuberculosis and pneumonia” (p. 314). Cutler and Miller (2005) show that from 1900 to 1936 there was a sharp drop in percentage of deaths due to infectious diseases in major U.S. cities, and argue more generally that the “urban mortality penalty” largely disappeared during this period, thanks to clean water technologies.

31Black, et al. (2013) express concerns about comparing earnings across local markets with differing prices. Also, Foote, Whatley, and Wright (2003) note that even in jobs in which in whites and blacks were similarly paid in the North, black workers were sometimes disproportionately assigned particularly unpleasant and dangerous work.

The authors declare that they have no relevant or material financial interests that relate to the research described in this paper.

Page 2

State of Residence in Adulthood, African Americans Born in the Deep South, 1916–1932