Poor housing conditions remain a major global public health challenge but direct causal evidence on their health effects is limited. We examine a rare rural housing intervention: the large-scale construction of subsidized high-quality cottages in Ireland 1883-1919. Using annual district-level data on deaths by cause, age and sex, we estimate the program's impact using difference-in-differences and triple-difference designs. Districts that built more cottages experienced mortality reductions of up to one death per thousand, driven by fewer respiratory deaths. The results are consistent with reduced indoor air pollution through improved ventilation. Cost-benefit analysis shows the scheme was highly cost-effective.

Evolutionary growth theory posits that natural selection set the stage for modern growth. I leverage micro-data from historical Germany to assess the viability of the selection mechanisms. I estimate fertility differentials and the inter-generational transmission of SES. High status couples, proxied by occupation, had 1-2 additional children, and SES was strongly heritable. To explore whether these parameters induce selection, I simulate an overlapping generation model of fertility choice and status transmission. The German parameters do not enable Clark's (2007) survival of the richest, whereas Galor and Moav's(2002) selection on quality can arise if the returns to investing in child quality are sufficiently large. Monte Carlo simulations extend the analysis beyond Germany. Survival of the richest requires exceptionally high coefficients of transmission (~0.87), and selection on quality emerges whenever returns to quality investments translate into higher fertility. Both depend on the strong heritability of the growth-complementary traits.

Studying urban development often requires within-city locations. However, the geographic limitations of census data limit historical studies in terms of granularity and coverage. This project proposes a scalable geocoding procedure to obtain within-city locations for households in U.S. cities for the 1880–1950 census. Using inferred street intersections within enumeration districts (EDs), this approach requires minimal data inputs, rather than detailed historical maps and city directories. ED centroids are assigned to households to obtain approximate locations. We geolocate 80% of all EDs across 133 cities across the six census-waves. The 1950 sample includes approximate location for 28% of the US population. When benchmarking this procedure against ED polygons, we successfully assign locations to 78% of all EDs. This compares favourably to 50% coverage for existing address-based approaches. The resulting crosswalk would enable researchers to use within-city variation to study patterns of urban development, as well as to exploit fine-grained geographic variation to answer causal questions.