Mail-order catalogs pioneered remote-order retail decades before the internet existed, and were in many ways the 19th-century equivalent of Amazon. This project asks whether improvements in postal infrastructure, by enabling competition from mail-order houses, displaced the rural general stores that had been shielded by the geographic isolation of their customers. I exploit two postal policy changes that made catalog purchases more accessible to rural households: the staggered roll-out of Rural Free Delivery in Iowa between 1900 and 1907, which brought mail to the farm gate, and the nationwide increases in the parcel post weight limit in 1913, 1914 and 1918, which extended the range of merchandise the mails could carry. The competition effect is identified by comparing general stores with perishable-goods stores, which faced the same postal reorganization but lay outside the catalog trade. Using data on Iowa retailers digitized from the Dun and Bradstreet Reference Book for 1899 to 1924, I find no evidence that Rural Free Delivery reduced the number, net worth, survival or size of general stores. The sector did contract over the decade following the weight limit increases, but these estimates rest on a stronger identifying assumption and I read them as suggestive. The results indicate that the erosion of geographic protection was a slower process than the alarm of contemporary merchants suggests.

Notes: Each dot is the estimated gap in total value (log net worth) between general stores and perishable-goods stores in the same towns, relative to 1912.   The unit of observation is a town–store-type–year cell. Bars show 95% confidence intervals. Dashed lines mark the end of the rural mail rollout (blue, 1908) and the three increases in the parcel weight limit (red, 1913, 1914, 1918). The two store types moved roughly in step before 1913. Once heavier parcels became mailable, the value of general stores fell steadily.

Historical U.S. censuses record where people lived but not where they worked, so how the geography of employment changed during the rise of the modern American city is largely unobserved. Using newly digitized and geocoded annual city directories, which record the home and workplace addresses of individuals and the locations of firms, I study how the spatial distribution of residents, jobs, and firms in Manhattan changed between 1896 and 1910, a period spanning the construction and opening of New York's first subway line. I show that directory-listed residents shifted steadily north, while employment remained concentrated at the southern tip. Commutes lengthened accordingly: the median commute rose from about 3.0 kilometers in 1896 to a peak of 3.8 kilometers in 1906, with most of the increase occurring after the subway opened. The lengthening was concentrated among office workers, whose jobs stayed downtown as their homes moved north, while most workers in local-serving businesses continued to live a few blocks from their workplaces. An event-study design comparing grid cells near subway stations with cells farther away also suggests that the subway deepened the division of the island into a commercial core and a residential periphery.

Access to urban public transportation may increase job accessibility and affect the economic outcomes of residents. This project exploits the opening of Manhattan's first subway line in 1904, and investigates whether proximity to public transportation improved the labor market outcomes of urban residents in the pre-automobile era. To address the endogeneity of subway placement, I construct an instrumental variable based on the hypothetical least-cost subway route. Using linked US Censuses, I find that male Manhattan residents living close to future subway stations in 1900 had better occupational outcomes and fewer weeks of unemployment in 1910. The observed effect is unique to male workers. The results are in part explained by decreased commuting time to the city center and a reduced likelihood of migrating away from Manhattan. 

Notes: Distance decay of the subway effect on occupational upgrading.  The figure plots the coefficients from an OLS regression of an occupational-upgrading indicator (1900–1910) on 100-metre distance-band dummies, with residents more than 1.5 km from any future station as the reference group. The regression controls for individual, household and neighbourhood characteristics, including 1900 occupational income, and includes ward fixed effects. Bars show 95% confidence intervals with standard errors clustered at the enumeration-district level.