Taking Technology to Task: The Skill Content of Technological Change in Early Twentieth Century United States
By Rowena Gray (email@example.com)
This paper presents a new picture of the labor market effects of technological change in pre-WWII United States. I show that, similar to the recent computerization episode, the electrification of the manufacturing sector led to a “hollowing out” of the skill distribution whereby workers in the middle of the distribution lost out to those at the extremes. To conduct this analysis, a new dataset detailing the task composition of occupations in the United States for the period 1880-1940 was constructed using information about the task content of over 4,000 occupations from the Dictionary of Occupational Titles (1949). This unique data was used to measure the skill content of electrification in U.S. manufacturing. OLS estimates show that electrification increased the demand for clerical, numerical, planning and people skills relative to manual skills while simultaneously reducing relative demand for the dexterity-intensive jobs which comprised the middle of the skill distribution. Thus, early twentieth century technological change was unskill-biased for blue collar tasks but skill-biased on aggregate. These results are in line with the downward trend in wage differentials within U.S. manufacturing up to 1950. To overcome any threat to the exogeneity of the electricity measure, due for example to endogenous technological change, 2 instrumental variable strategies were developed. The first uses cross-state differences in the timing of adoption of state-level utility regulation while the second exploits differences in state-level geography that encouraged the development of hydro-power generation and thus made electricity cheaper. The results from these regressions support the main conclusions of the paper.
Review by: Chris Colvin
What was the effect of electrification on the skill content of manufacturing? Rowena Gray (University of Essex) answers this important question for the case of the US using a new data source and an instrumental variables approach. Gray uses the Dictionary of Occupational Titles, a 1949 publication which describes occupations, to classify the skill content of manufacturing jobs post-electrification. Matching these descriptions with occupational data from decennial censuses and information on plant electrification from the US census of manufacturing, Gray is able to track changes in the skill content of manufacturing due to electrification for the period 1880 to 1940.
Using regression analysis, Gray shows that the most skilled blue-collar workers were displaced by machinery, i.e. electrification resulted in unskilled-biased technical change. She also shows that electrification simultaneously necessitated more clerical and supervisory work, a skill-biased change. This bimodal distributional finding is further strengthened in her robustness exercises, which instrument for electrification using cross-state differences in the timing of the adoption of utility regulation and differences in geography necessary for hydroelectric power generation. This instrumental variable approach is needed to address the concern that electrification is endogenous to the pre-existing skill levels present in state’s labour market; various skillsets may have attracted electrification, rather than the other way around.
Gray’s paper is important because previous studies have been unable to quantify the effects of electrification on the skill content of manufacturing, or at least have been unable to demonstrate that electrification has a distributional effect, that it was simultaneously unskilled- and skilled-biased. An alternative approach to instrumental variables which Gray could have employed to determine the direction of causality would have been to complement her regression analysis with detailed business histories, a method suggested recently by Randall Morck and Bernard Yeung. Where her research potentially suffers is her reliance on post-electrification occupational descriptions; her assumption that the tasks required to carry out particular jobs before and after electrification were identical may be unrealistic. Gray’s future research agenda includes using her individual-level dataset to track cohorts across censuses in order to uncover the winners and losers of electrification. Perhaps another task, which could strengthen the results of the paper discussed here, would be to repeat her analysis using a source that describes the task content of jobs pre-electrification.
Gray’s paper is part of the new working paper series of the European Historical Economics Society. The series, edited by Nikolaus Wolf (Humboldt-Universität zu Berlin), offers the society’s members the chance to disseminate their work ahead of journal submission. Other recent papers in this series include one by Geraldine David and Kim Oosterlinck (Université Libre de Bruxelles) on the effects of the War on the Belgian art market, and a paper by Giovanni Federico (EUI and University of Pisa) on market integration across oceans. This series will surely prove to be an important outlet for work-in-progress by historical economists in future; a paper disseminated in this way could be an important signal of quality versus dissemination through the working paper series of individual institutions.