A machine learning approach to improving occupational income scores

Abstract

Historical studies of labor markets frequently suffer from a lack of data on individual income. The occupational income score (OCCSCORE) is often used as an alternative measure of labor market outcomes. Using modern Census data, we find that the use of OCCSCORE biases results towards zero and can frequently result in statistically significant coefficients of the wrong sign. We use a machine learning approach to construct a new adjusted score based on industry, occupation, and individual demographics. Our alternative score substantially outperforms OCCSCORE in both modern and historical contexts. We illustrate our approach by estimating racial and gender earnings gaps in the 1915 Iowa State Census and intergenerational mobility elasticities using linked data from the 1850-1910 Censuses.JEL codes: C21, J71, N32

Ethnic enclaves and immigrant outcomes: Norwegian immigrants during the Age of Mass Migration

“…Occscore refers to the IPUMS occscore based on 1950 median occupational earnings. Lido refers to age/race/state-adjusted occscores from Saavedra and Twinam (2018). The last two rows replace farm income with county-level farm income estimated from the 1920 Census of Agriculture.…”

El Sueño Americano? The Generational Progress of Mexican Americans Prior to World War II

“…11 Also following Collins and Wanamaker 2017, we estimate self-employed earnings based off of information in the 1960 Census, and further increase farmer and farm laborer income to reflect in-kind benefits (see Online Appendix C). In addition to these imputed income scores, we find it valuable to also look directly at occupational categories that do not rely on the assumptions we make in the calculation of the income scores and that will not suffer from some of the documented issues with using occupation-based income scores (Saavedra and Twinam 2019;Inwood, Minns, and Summerfield 2019). Figure 2 shows that the occupational gaps between first-generation Mexicans and non-Mexican whites were large.…”

Working Their Way Up? US Immigrants' Changing Labor Market Assimilation in the Age of Mass Migration

“…For example, if farmers are 20 percent of the work force and range between the 50th and 70th percentile, we assign them a rank of 0.6. 32 This property of occupational scores and rankings has led to criticism by Inwood, Minns, and Summerfield (2019) and Saavedra and Twinam (2020).…”