Firms facing complex objectives often decompose the problems they face, delegating different parts of the decision to distinct subunits. Using comprehensive data and internal models from a large U.S. airline, we establish that airline pricing is not well approximated by a model of the firm as a unitary decision maker. We show that observed prices, however, can be rationalized by accounting for organizational structure and for the decisions by departments that are tasked with supplying inputs to the observed pricing heuristic. Simulating the prices the firm would charge if it were a rational, unitary decision maker results in lower welfare than we estimate under observed practices. Finally, we discuss why counterfactual estimates of welfare and market power may be biased if prices are set through decomposition, but we instead assume that they are set by unitary decision makers.

We study the lifetime effects of the first and largest American youth employment and training program in the United States-the Civilian Conservation Corps (CCC), 1933-1942. We match newly digitized enrollee records to census, World War II enlistment, Social Security, and death records. We find that longer service in the CCC led to improvements in height, health status, longevity, geographic mobility, and lifetime earnings but did not improve short-term labor market outcomes, including employment and wages. We address potential selection into CCC duration using several approaches, most importantly two newly developed control-function approaches that leverage unbiased estimates of the short-term effects of a randomized controlled trial of Job Corps (the modern version of the CCC). Our findings suggest that short- and medium-term evaluations of employment programs underestimate effects because they fail to capture lifetime effects and often ignore or underestimate health and longevity benefits that increase in magnitude at later ages.

In the 1960s, two landmark statutes-the Equal Pay and Civil Rights Acts-targeted the long-standing practice of employment discrimination against U.S. women. For the next 15 years, the gender gap in median earnings among full-time, full-year workers changed little, leading many scholars to conclude that the legislation was ineffectual. This article revisits this conclusion using two research designs, which leverage (i) cross-state variation in preexisting state equal pay laws and (ii) variation in the 1960 gender gap across occupation-industry-state-group cells to capture differences in the legislation's incidence. Both designs suggest that federal antidiscrimination legislation led to striking gains in women's relative wages, which were concentrated among below-median wage earners. These wage gains offset preexisting labor market forces, which worked to depress women's relative pay growth, resulting in the apparent stability of the gender gap at the median and mean in the 1960s and 1970s. The data show little evidence of short-term changes in women's employment but suggest that firms reduced their hiring and promotion of women in the medium to long term. The historical record points to the key role of the Equal Pay Act in driving these changes

While hypothesis testing is a highly formalized activity, hypothesis generation remains largely informal. We propose a systematic procedure to generate novel hypotheses about human behavior, which uses the capacity of machine learning algorithms to notice patterns people might not. We illustrate the procedure with a concrete application: judge decisions about whom to jail. We begin with a striking fact: the defendant’s face alone matters greatly for the judge’s jailing decision. In fact, an algorithm given only the pixels in the defendant’s mug shot accounts for up to half of the predictable variation. We develop a procedure that allows human subjects to interact with this black-box algorithm to produce hypotheses about what in the face influences judge decisions. The procedure generates hypotheses that are both interpretable and novel: they are not explained by demographics (e.g., race) or existing psychology research, nor are they already known (even if tacitly) to people or experts. Though these results are specific, our procedure is general. It provides a way to produce novel, interpretable hypotheses from any high-dimensional data set (e.g., cell phones, satellites, online behavior, news headlines, corporate filings, and high-frequency time series). A central tenet of our article is that hypothesis generation is a valuable activity, and we hope this encourages future work in this largely “prescientific” stage of science

The Neolithic revolution saw the independent development of agriculture among at least seven unconnected hunter-gatherer populations. I propose that the rapid spread of agricultural techniques resulted from increased climatic seasonality causing hunter-gatherers to adopt a sedentary lifestyle and store food for the season of scarcity. Their newfound sedentary lifestyle and storage habits facilitated the invention of agriculture. I present a model and support it with global climate data and Neolithic adoption dates, showing that greater seasonality increased the likelihood of agriculture’s invention and its speed of adoption by neighbors. This study suggests that seasonality patterns played a dominant role in determining our species’ transition to farming.

This article studies the long-run effects of slavery and restrictive Jim Crow institutions on Black Americans’ economic outcomes. We track individual-level census records of each Black family from 1850 to 1940 and extend our analysis to neighborhood-level outcomes in 2000 and surname-based outcomes in 2023. We show that Black families whose ancestors were enslaved until the Civil War have considerably lower education, income, and wealth than Black families whose ancestors were free before the Civil War. The disparities between the two groups have persisted substantially because most families enslaved until the Civil War lived in states with strict Jim Crow regimes after slavery ended. In a regression discontinuity design based on ancestors’ enslavement locations, we show that Jim Crow institutions sharply reduced Black families’ economic progress in the long run.

We develop empirical tools for studying discrimination in multi-phase systems, and apply them to the setting of foster care placement by child protective services. Leveraging the quasi-random assignment of two sets of decision-makers-initial hotline call screeners and subsequent investigators-we study how unwarranted racial disparities arise and propagate through this system. Using a sample of over 200,000 maltreatment allegations, we find that calls involving Black children are 55% more likely to result in foster care placement than calls involving white children with the same potential for future maltreatment in the home. Call screeners account for up to 19% of this unwarranted disparity, with the remainder due to investigators. Unwarranted disparity is concentrated in cases with potential for future maltreatment, suggesting that white children may be harmed by "under-placement" in high-risk situations.

We construct measures of industry performance and welfare in the U.S. automobile market from 1980 to 2018. We estimate a demand model using product-level data on market shares, prices, and attributes, and consumer-level data on demographics, purchases, and stated second choices. We estimate marginal costs assuming Nash-Bertrand pricing. We relate trends in consumer welfare and markups to trends in market structure and the composition of products. Although real prices rose, we find that markups decreased substantially, and the fraction of total surplus accruing to consumers increased. Consumer welfare increased over time due to improved product quality and improved production technology.

Decision makers, such as doctors, judges, and managers, make consequential choices based on predictions of unknown outcomes. Do these decision makers make systematic prediction mistakes based on the available information? If so, in what ways are their predictions systematically biased? In this article, I characterize conditions under which systematic prediction mistakes can be identified in empirical settings such as hiring, medical diagnosis, and pretrial release. I derive a statistical test for whether the decision maker makes systematic prediction mistakes under these assumptions and provide methods for estimating the ways the decision maker’s predictions are systematically biased. I analyze the pretrial release decisions of judges in New York City, estimating that at least 20% of judges make systematic prediction mistakes about misconduct risk given defendant characteristics. Motivated by this analysis, I estimate the effects of replacing judges with algorithmic decision rules and find that replacing judges with algorithms where systematic prediction mistakes occur dominates the status quo.

We propose and implement a new approach that allows us to estimate income-specific changes in household welfare in contexts where well-measured prices are not available for important subsets of consumption. Using rich but widely available expenditure survey microdata, we show that we can recover income-specific equivalent and compensating variations from horizontal shifts in what we call “relative Engel curves”—as long as preferences fall within the broad quasi-separable class (Gorman 1970, 1976). Our approach is flexible enough to allow for nonparametric estimation at each point of the income distribution. We apply the methodology to estimate inflation and welfare changes in rural India between 1987 and 2000. Our estimates reveal that lower rates of inflation for the rich erased the real income convergence found in the existing literature that uses the subset of consumption with well-measured prices to calculate inflation.