Working Papers

Can Government Relief Spending Compensate for Recession-Related Longevity Losses? Evidence from the Great Depression and the New Deal - with Ariadna Jou

This study explores the effect of government relief spending on longevity by examining county-level variation in New Deal spending and linking people in the 1930 and 1940 censuses to their death dates as recorded on FamilySearch, a large public wiki-style family tree.

View the most recent draft here.

Uneven Bars? Looking for Evidence of Environmental Microaggression Theory in NCAA Women’s Gymnastics - with Seth Cannon

This study utilizes a unique dataset of scores from NCAA women’s gymnastics meets to examine whether Black gymnasts experience a performance decline when performing at Brigham Young University, a university with a history of racial controversy and whose namesake was vocal in support of anti-Black policies in the mid-1800s.

View the most recent draft here.

The Effect of WWII Expenditures on the Longevity of the Black Population - with Adriana Lleras-Muney and Joe Price

This study exploits cross-sectional variation in WWII wartime production contracts distributed at the metropolitan district level that were attached to an anti-discrimination order (Executive Order 8802) to estimate the effect of government spending on the average lifespan of Black people in those metro districts.

Working draft available soon.

Works in Progress

Investigating High Variance in Black Americans' Age Reporting across the 1900-1940 U.S. Censuses - with Joe Price

Many studies that investigate data accuracy in the publicly available U.S. full-count censuses have made note of a gap in quality between the reported data for White and Black Americans. Using the newly-published Census Tree dataset to link individuals across the five publicly available 20th century full-count Censuses from IPUMS (1900-1940), we quantify this gap in quality of birth year reporting across race groups by examining the difference between the maximum and minimum recorded birth years for White or Black U.S. born individuals. We find that this gap is nearly twice as large on average for Black Americans than it is for White Americans after controlling for place of birth and gender, which has important implications for blocking methods used in record linking algorithms.

Using Computer Vision and Machine Learning to Identify and Describe the 1900s U.S. Census Enumerators - with Joe Price

This project will focus on describing the individuals who enumerated historic U.S. Censuses by automatically transcribing their names using computer vision and finding them in corresponding census records. The end goal is to produce a dataset attaching every enumeration district to a single enumerator and their characteristics that can be used in future research on census record quality.