We provide theory and evidence on the contribution of slavery wealth to Britain’s economic development prior to the abolition of slavery in 1833. We combine data on individual slaveholders from compensation records, an exogenous source of variation in slavery wealth from weather-induced shocks to mortality of the enslaved during the middle pas-sage, and a quantitative spatial model. Exogenous increases in slavery wealth reduce the agricultural employment share, increase the manufacturing employment share, raise the number of cotton mills, and increase property values. Quantifying our model, we nd that slavery wealth raises aggregate income by the equivalent of around a decade of economic growth, and increases local income in places with the greatest involvement in slavery by more than 40 percent. 

We identify negative spillovers exerted by large, successful manufacturing plants on other local production facilities in China. A short-lived alliance between the U.S.S.R. and China led to the construction of 150 "Million-Rouble plants" in the 1950s. Our identification strategy exploits the ephemeral geopolitical context and the relative position of allied and enemy airbases to isolate exogenous variation in plant location decisions. We find a boom-and-bust pattern in hosting counties: treated counties are twice as productive as control counties in 1982, but 30% less productive in 2010. The average other establishment in treated counties is unproductive, does not innovate, and charges high markups. We find that (over)specialization limits technological spillovers. This prevents the emergence of new industrial clusters and leads to a flight of entrepreneurs.

We examine the distributional consequences of trade using the Repeal of the Corn Laws and the Grain Invasion during the 19th-century. We use a newly-created dataset on population, employment by sector, property values, and poor law transfers for over 10,000 parishes in England andWales from 1801–1901. In response to this trade shock, we show that locations with high-wheat suitability experience population decline, rural-urban migration, structural transformation away from agriculture, increases in welfare transfers, and declines in property values, relative to locations with low-wheat suitability. We develop a quantitative spatial model to evaluate the aggregate economic implications of these findings. Undertaking counterfactuals for the Grain Invasion, we show that geography is an important dimension along which the distributional effects of trade occur.

This paper analyzes 630,000 paintings from 1400 onward to uncover how visual art reflects its socioeconomic context. We develop a learning algorithm to predict nine basic emotions conveyed in each painting and isolate a context effect---the emotional signal shared across artworks created in the same location and year---controlling for artist, genre, and epoch-specific influences. These emotion distributions encode subtle but meaningful information about the living standards, uncertainty, or inequality characterizing the context in which the artworks were produced. We propose this emotion-based measure, derived from historical artworks, as a novel lens to examine how societies experienced major socioeconomic transformations, including climate variability, trade dynamics, technological change, shifts in knowledge production, and political transitions.

Does industrial concentration shape the life and death of cities? We identify settlements from historical maps of England and Wales (1790-1820), isolate exogenous variation in their late 19th-century size and industrial concentration, and estimate the causal impact of size and concentration on later dynamics. Industrial concentration has a negative effect on long-run productivity—independent of industry trends and consistent with cross-industry Jacobs externalities. A spatial model quantifies the role of fundamentals, industry trends, and Jacobs externalities in shaping industry-city dynamics and isolates a new, dynamic trade-off in the design of place-based policies

This paper presents new evidence on the critical role of lower-level organizational leaders. Unlike top managers, frontline leaders are essential for implementing organizational strategies by maintaining team cohesion when shirking is profitable for workers. We study this in the context of the Union Army during the U.S. Civil War, using data on 2.2 million soldiers and tracking captains and their 100-soldier companies at weekly frequency throughout the conflict. We estimate leader fixed effects during non-combat weeks to measure leadership quality in a leader value-added framework. We validate this measure by showing that captains were not assigned based on prior unit performance or observable pre-war characteristics. High-quality leaders earned more after the war, but not before, and were more frequently recognized as good leaders in their postwar biographies. Daily event-study estimates around major battles show that better captains significantly reduced desertions in combat. Exploiting quasi-random leader turnover, we find evidence that this effect is causal. Using digitized battle maps, we rule out risk aversion as a mechanism and find instead that better leaders had higher mortality rates, consistent with a leading-by-example explanation. We also document modest learning-by-doing effects. These findings highlight the often-overlooked importance of frontline leadership, where direct supervision and interpersonal influence are strongest.

How do communication costs affect the production of new ideas and inventions? To answer this question, we study the introduction of the Uniform Penny Post in Great Britain in 1840. This reform replaced the previous system of expensive distance-based postage fees with a uniform low rate of one penny for sending letters anywhere in the country. The result was a large spatially-varied reduction in the cost of communicating across locations. We study the impact of this reform on the production of scientific knowledge using citation links constructed from a leading academic journal, the Philosophical Transactions, and the impact on the development of new technology using patent data. Our results provide quantitative causal estimates showing how a fall in communication costs can increase the rate at which scientific knowledge is exchanged and new ideas and technologies are developed. This evidence lends direct empirical support to an extensive theoretical literature in economic growth and urban economics positing that more ideas can emerge from communication between individuals. 

This paper studies the economic geography of the North American fur trade between 1670 to 1870. European hat makers used felt made from beaver pelts to produce hats, which were a luxury item for the wealthy. With the Eurasian beaver nearly extinct, Europeans turned to North America, trading beaver pelts with Indigenous communities in exchange for iron goods and textiles. We introduce a computational spatial model that integrates a biological beaver growth model with detailed transport cost calculations across North America's extensive river system, forming a "beaver gravity model". The model illustrates that the over-exploitation of beavers drove the westward expansion of the fur trade, simultaneously shedding light on the spatial dynamics of different Indigenous territories.

Economists have long used instrumental variables to evaluate the causal effect of an endogenous treatment variable on outcomes of interest. Underlying effect mechanisms are of equal interest, but IV models are not suitable to identify the causal effect of an intermediate outcome on a final outcome. This paper presents a novel procedure to identify causal mediation effects without altering the key features of the IV model. If the unobserved confounding variables that cause the treatment and the intermediate outcome are independent of confounders that cause the intermediate and final outcomes, the IV model will be partially confounded. In this case, the same instrumental variable can identify a direct treatment effect and an indirect effect mechanism. Under linearity, the model can be estimated in a standard 2SLS regression framework. We consider three applications to demonstrate our procedure and provide a test for the validity of the partial-confounding assumption.