This paper studies the long-run consequences of developing without industrialization. Countries with higher historical peaks in manufacturing employment grow faster, accumulate more tertiary education, experience faster fertility declines, and develop more skill-intensive service sectors. The same pattern appears within the United States: counties with larger manufacturing booms later developed service occupations with higher educational content. I interpret these facts through a unified growth model with structural change, endogenous fertility, education, and heterogeneous producer services. In the model, manufacturing creates demand for high-skill organizational tasks that build the skill and knowledge base of the service economy that follows. Calibrated to the United States over 1850--2019, the model explains 31 percent of cross-country log GDP per capita gaps, 60 percent of GDP per capita growth differences, 53 percent of tertiary education differences, and 39 percent of cumulative population growth differences. Countries can move into services without industrializing, but the service economy they build is less skill-intensive and grows more slowly.

This paper uses new data on copyright registration title pages from the Library of Congress (LOC) to analyze the intellectual and cultural development of the United States over 1790–1870. I construct national time series of book production over this period which show an uptake in per-capita terms in 1830, well before the start of the Second Industrial Revolution and the era of "knowledge based progress". Matching authors to locations (at the county level) using declassified census data reveals that the spatial distribution of intellectual production in the early 19th century is strongly correlated with inventive activity over 1860–1940 and the evolution of the manufacturing sector. Identification is based on a shift-share type instrument exploiting the large internal migration patterns occurring in this time period.

We study how technological innovations reshaped U.S. labor markets between 1860 and 1950 by combining over 20 million digitized newspaper pages from the Library of Congress with occupational data from the U.S. Census starting in 1860. Using equipment advertisements, we track the arrival of 123 major technologies across time and space, and measure local labor market responses through job ads and shifts in occupational structure. We find that new technologies trigger the rapid decline of some occupations (substitution) and the gradual rise of others (scaling). While short-run disruptions are uneven, the long-run effects are broadly positive: wages, employment, and productivity tend to rise in counties that adapt more quickly to changing labor demand. A half-century later, these counties are richer, have lower unemployment, and more polarized labor markets. Our findings offer historical perspective on modern fears of automation-driven job loss, suggesting that technological disruptions have historically reallocated labor rather than replaced it.

This paper studies how labor market power affects firm dynamics and aggregate productivity. We build a dynamic model of neoclassical monopsony with occupational choice, firm growth, and productivity-enhancing technology adoption. Labor market power lowers efficiency and leads to aggregate output losses by distorting the allocation of labor, entrepreneurship, and innovation decisions. The model is consistent with cross-country evidence of higher life-cycle firm growth and higher productivity investment in more competitive labor markets and can explain up to 42% of differences in income per capita across countries. We find that about one-third of the losses are attributable to a distorted selection of entrepreneurs and a lack of innovation, suggesting that efficiency losses may be greater than those estimated by previous studies.