Knowledge that Transforms

We construct a new dataset containing parcel sizes and building footprints of Canadian manufacturing plants and decompose industrial density (parcel size per worker) into: crowding (floorspace per worker); building height (floorspace to building footprint); and parcel coverage (building footprint to parcel size). We find that establishments occupy parcels more densely in big cities and central locations, and that larger establishments use less land per worker. Floorspace per worker is unrelated to distance from the city centre. The estimated elasticity of substitution between land- and non-land factors is small, between 0.14 and 0.42.

Researchers commonly use jackknife-based instrumental variables estimators to eliminate the many-instruments bias of two-stage least squares. Where inference must be clustered, however, the jackknife fails to eliminate the bias. We propose a cluster-jackknife approach in which first-stage predicted values for each observation are constructed from a regression that leaves out the observation’s entire cluster. The cluster-jackknife instrumental variables estimator (CJIVE) eliminates many-instruments bias, and consistently estimates causal effects in the traditional linear model and local average treatment effects in the heterogeneous treatment effects framework. We illustrate the method in an application estimating the effects of pre-trial detention in Miami-Dade County.

This paper exploits spatial quasiexperimental variation around French departmental borders to provide novel evidence that regional borders disturb surrounding economic activity; namely, commuting flows, catchment areas, and residential density. The lack of integration of local transport networks at regional borders—leading to a 4.5 km travel distance penalty when crossing borders—explains the observed border effects. Using a spatial quantifiable equilibrium framework and accounting for construction costs, a policy simulation exercise shows that integrating local transport networks leads to a 6.36% average growth in real per capita residential income.

While investment in most sectors declines in response to a contractionary monetary policy shock, investment in the manufacturing sector increases. Using manually digitized aggregate income and balance sheet data for the universe of US manufacturing firms, I show that this increase is driven by the types of firms which are least likely to be financially constrained. A two-sector New Keynesian model with financial frictions can match these facts; unconstrained firms take advantage of the decline in the user cost of capital caused by the monetary contraction, while constrained firms are forced to cut back. Removing firm financial constraints in the model dampens the response of manufacturing output to monetary shocks by about 25%.

We study judicial in-group bias in Indian criminal courts using newly collected data on over 5 million criminal case records from 2010–2018. After classifying gender and religious identity with a neural network, we exploit quasi-random assignment of cases to judges to determine whether judges favor defendants with similar identities to themselves. In the aggregate, we estimate tight zero effects of in-group bias based on shared gender or religion, including in settings where identity may be especially salient, such as when the victim and defendant have discordant identities. Proxying caste similarity with shared last names, we find a degree of in-group bias, but only among people with rare names; its aggregate impact remains small.

We examine the use of instrumental variable (IV) methods to measure the effect of a ceteris paribus change in an endogenous variable on an ordered outcome. Specifically, we use these methods to investigate the effect of neighborhood characteristics on subjective well-being (SWB) among participants in the Moving to Opportunity (MTO) housing voucher experiment. We find that the estimated positive effect of a decrease in neighborhood poverty on SWB is sensitive to the specification of the first-stage auxiliary equation for endogenous neighborhood poverty. Our results highlight the influential role of control function restrictions in complete triangular models.

A flourishing empirical literature investigates the prevalence of p-hacking based on the distribution of p-values across studies. Interpreting results in this literature requires a careful understanding of the power of methods for detecting p-hacking. We theoretically study the implications of likely forms of p-hacking on the distribution of p-values to understand the power of tests for detecting it. Power can be low and depends crucially on the p-hacking strategy and the distribution of true effects. Combined tests for upper bounds and monotonicity and tests for continuity of the p-curve tend to have the highest power for detecting p-hacking.

This paper estimates the causal effect of scaling up access to psychotherapy. I study a 2008 reform of the Danish public health insurance, which introduced 60% coverage for psychotherapy by private practice psychologists for patients aged 18–37 diagnosed with depression or anxiety. Using administrative data from 1995–2019 and quasi-experimental methods, I show that psychotherapy coverage reduces psychiatric hospital contacts, physical health care use, and suicide attempts, but has no effect on labor market outcomes, including employment or disability pension. Still, savings from reduced use of other health care services exceed policy costs, making it both cost-reducing and welfare-improving.

Should healthcare providers strictly adhere to clinical guidelines, or is some discretion beneficial? Using bunching estimation and microdata from 619,907 patients in Chile, I examine discretion in the diagnosis of hypertension (blood pressure ≥ 140/90mmHg). I find bunching in blood pressure at 10% of clinics, where 6 to 62% of patients' test results expected just above the hypertension diagnostic threshold are instead recorded just below it, suggesting providers use their discretion to consider some tests false positives. Leveraging variation in bunching across clinics, I show that affected patients have lower cardiovascular risk, suggesting discretion improves patient classification, partially driven by heuristic thinking.

Using job posting wage data, we find a substantial decrease in the urban wage premium for occupations with high working-from-home (WFH) adoption following the COVID-19 pandemic, accompanied by an employment shift away from large cities. Based on a conceptual framework, the empirical findings suggest that WFH adoption lowered the productivity premium of large cities. A skill-level decomposition reveals that the urban wage premium decline was largely driven by reduced wage returns to interpersonal skills in large cities, suggesting that the reduced urban productivity premium was a result of weakened agglomeration economies due to decreased interpersonal interactions in large cities.