Place-based policies often aim to improve local economic opportunity and at large scale, trigger household migration that alters the peer composition of neighborhoods (1) directly targeted and (2) indirectly affected through migration. Aside from the immediate impact of the policy, general equilibrium (GE) changes in peer composition are also important determinants of economic mobility—and create winners and losers. I study these equilibrium effects in the context of the interstate highway system, a transformative place-based policy for U.S. cities. I employ novel measures of intergenerational mobility for the near universe of 57 million children born between 1964 to 1979, constructed using machine learning methods and historical IRS tax forms. I find areas with commuting access improvements from highway construction experienced increases in average income and inflows of higher-educated, higher-occupational status, and White households. With detailed income and location for 1974 to 2018, I extend the movers design to find that both Black and White children benefit from growing up in neighborhoods (tracts) with greater average income and higher status peers. In areas with lower access improvements, which experience outflows of high-status peers, children subsequently face declines in economic opportunity. I incorporate these GE forces into a spatial equilibrium framework to quantify the aggregate consequences of the interstate system on intergenerational mobility by race.

We study the channels through which changes in local economic conditions during early childhood affect long-run outcomes for children from differing economic backgrounds. We exploit geographic variation across counties in the decline of manufacturing employment during the 1979 to 1984 period with microdata from the Longitudinal Business Database. To assess the exogeneity of local labor market shocks, we construct additional shocks by combining industry-level energy intensity with spikes in oil prices as a result of the 1979 energy crisis. With administrative and survey data that trace the full trajectory of the children’s lives, we measure how these local changes impact educational attainment, income, and the quality of the firm of employment in the modern day. We explore how migration of parents away from counties experiencing declines and changes in parental income during childhood are central mediators for our findings.

We construct novel parent-child linkages between the universe of parent tax filers in IRS 1040 forms in 1974 and 1979 and the universe of children from the Census Numident in the cohorts of 1964 to 1979. Variables used for matching are parent names of children and names of parent tax filers, which are obtained from a restricted name file provided by the Social Security Administration. Applying name-matching techniques that incorporate supervised learning methods, we flexibly compare parent names and disambiguate parent-to-parent matches. To feasibly conduct the matching for a large set of comparisons, we employ parallel computing on Amazon Web Services. This report documents the iterative process for identifying matches and the algorithm that is used for assessing the likelihood of a match. We provide match rates for different demographic groups and validate the accuracy of the linkages.

We expand the coverage of Protected Identification Keys (PIKs) for the universe of children in the cohorts of 1930 to 1940 in the 1940 Full Count Census. Parents and children are recorded together in the 1940 Census, so we assign three sets of names to link children in the 1940 Census to children in the Census Numident: father’s name, mother’s name, and child’s name. Location of birth and year of birth are additional matching variables. We document the iterative process for matching children and our approach to addressing name changes for women.  The matching is conducted with parallel computing on Amazon Web Services for men and women separately. With these linkages, we measure intergenerational mobility using IRS 1040 forms in 1974 and 1979 that report income at mid-life for these cohorts. Finally, we compare our linkages to previously constructed PIK linkages to compute how many additional matches are recovered using our machine learning algorithm and verify the accuracy of the links. Match rates are reported by gender and race.