In 1880, only seven states set a minimum age for factory employment. By 1910, all but four had a minimum age. In this paper, I use newly-collected manufacturing and opinion survey data from North Carolina to understand the underlying political economy. First, I document substantial heterogeneity in the use of child labor, even among cotton mills, and show that new technologies such as electricity did not reduce the demand for child labor. Second, by estimating the relative labor productivies of children and men, I show that children were underpaid, helping explain manufacturer opposition to regulation. Third, I show that lower-income individuals, rather than those directly competing with children, were more likely to oppose regulation. Fourth, I show that the regulations on child labor, as enacted, had little effect on manufacturers, and that child labor was falling prior to enactment. Together, these results suggest that child labor regulation resulted from, rather than caused, the decline in child labor force participation.

During the early 1900s, manufacturers replaced steam- and water-power with electricity.  In this paper, I use a newly-collected establishment-level dataset to answer three sets of questions.  First, I show that factory electrification led to large and rapid increases in productivity.  Second, I show that electrification increased the wages paid by manufacturers but also widened the within-plant wage distribution.  Third, I show that electricity diffusion was delayed by the costs of adoption.

I collect a new establishment-level dataset from reports published by the North Carolina Department of Labor between 1905 and 1926.  These reports include detailed factory-specific information on revenue, employment, capital, and energy use.  Because the original records from the national Census of Manufacturers have been lost, prior research on electrification has relied on aggregate data at the industry- or location-level.  Several features of this dataset make it particularly useful.  First, the use of establishment-level data allows me to isolate the effect of electrification where studies using aggregate data conflate changes on the intensive- and extensive-margin.  Second, because I am able to link factories across reports, I can describe factors which predict factory electrification.  Factory-level panel data with this level of detail and frequency is quite rare in economic history.

In my baseline estimates, I find that electricity adoption increased factory TFP in the textile sector by ten percent relative to plants that did not electrify.  Electrified factories adopted new production processes that were more capital- and energy-intensive.  I address the endogeneity of electrification two ways.  First, I control for establishment fixed-effects to capture factory-specific features that may drive electricity adoption.  Second, I instrument for electrification using local variation in access to electricity.   These changes support narrative evidence that electricity adoption allowed factories to reorganize production by removing the need to structure factories around a steam engine or waterwheel.

Turning to the effects of electrification on workers, I find that electrified factories paid higher average wages.  However, within factories, I find that electrification increased the top wages of men and women but not the bottom wages, widening the pay dispersion within plants. I also find an increase in the number of machines operated by each worker, suggesting that increased wages served partly to compensate workers for an increase in work-intensity.  This evidence points to electrification as an early case of skill-biased technological change.

Although a growing literature in economic history has discussed the role of the corporation and other organizational forms, the scarcity of historical firm-level data has made it difficult to study how firms choose their form.  I draw on records from the North Carolina Department of Labor Report for 1915 which provides factory-level data for over 3800 plants and describes whether each plant is owned by a corporation, partnership, or individual.  I find that factories owned by corporations are larger, more productive, more capital-intensive, and more likely to operate on a year-round basis.  Factories owned by a corporation are more likely to survive than others until at least 1926.  Differences in the amount of capital invested and industry explain much, though not all, of the apparent differences in productivity.  Corporations accounted for less than half of the factories through the eighth decile of capital invested, suggesting that even though access to the corporate form was not legally restricted in this time and place, the costs of incorporation outweighed the benefits for most small- and medium-sized enterprises.

This paper uses job titles, as reported to the Census, to create a measure of “new work”, identifying job categories that first appeared during the latter half of the twentieth century.  I show that new employment was a substantial and growing feature of the economy – more than 6% of men in 1900 held jobs that did not exist in 1850.  Urban, white, and literate workers were more likely to hold these new jobs.  A majority of new jobs served one of three functions: using new technologies, reducing transaction costs, or providing consumer services.  I document the path of new job diffusion, showing that new employment followed a S-curve similar to other cases of technology diffusion.  I also highlight similarities between the historical creation of new jobs and these trends in the modern era, emphasizing the role of cities and education in both periods.

In this paper, I construct a new block-level measure of segregation in Southern cities in the early twentieth century by collecting data from city street directories which include information on the race of each household. This measure has several benefits relative to alternative Census-based measures: it is available on an annual basis, making it possible to document changes at a higher frequency than the decennial Census, and it is constructed at a more-precise level than measures based on city wards or Census tracts.  As a proof of concept, I collect this data for Charlotte, North Carolina, but will continue data collection for other southern cities.  I find substantial block-level segregation: around 80% of blocks in 1910 were occupied entirely by members of one race.  Next steps include collecting data for additional cities and linking blocks through time to understand the process of block transition.