Knowledge that Transforms

Empirical models for dyadic interactions between n agents often feature agent-specific parameters. Fixed-effect estimators of such models generally have bias of order n−1, which is nonnegligible relative to their standard error. Therefore, confidence sets based on the asymptotic distribution have incorrect coverage. This paper looks at models with multiplicative unobservables and fixed effects. We derive moment conditions that are free of fixed effects and use them to set up estimators that are n-consistent, asymptotically normally distributed, and asymptotically unbiased. We provide Monte Carlo evidence for a range of models. We estimate a gravity equation as an empirical illustration.

We develop an almost affine term-structure model with a closed-form solution for factor loadings in which the spot rate and the risk price are fractionally integrated processes with different integration orders. This model is used to explain two stylized facts. First, predictability of longterm excess bond returns requires sufficient volatility and persistence in the risk price. Second, the large volatility of long-term bond returns requires persistence in the spot rate. Decomposing long-term bond returns, we find that the expectations component from the level factor is more volatile than returns themselves and that the risk premium correlates negatively with level-factor innovations.

We use a tailor-made survey on a Swedish sample to investigate how individuals' relative income affects their demand for redistribution. We first document that a majority misperceive their position in the income distribution and believe that they are poorer, relative to others, than they actually are. We then inform a subsample about their true relative income and find that individuals who are richer than they initially thought demand less redistribution. This result is driven by individuals with prior right-of-center political preferences who view taxes as distortive and believe that effort, rather than luck, drives individual economic success.

We consider the dynamic factor model where the loading matrix, the dynamic factors, and the disturbances are treated as latent stochastic processes. We present empirical Bayes methods that enable the shrinkagebased estimation of the loadings and factors. We investigate the methods in a large Monte Carlo study where we evaluate the finite sample properties of the empirical Bayes methods for quadratic loss functions. Finally, we present and discuss the results of an empirical study concerning the forecasting of U.S. macroeconomic time series using our empirical Bayes methods.

Numerous studies find that criminal court judges issue racially disparate sentences, but whether these patterns reflect tastes for discrimination remains unclear. An alternative explanation is statistical discrimination, which implies that judges rely on race to predict a felon’s latent criminality in the absence of perfect information. This paper uses an empirical approach that distinguishes between taste-based and statistical discrimination. The intuition is that if the rank order of judicial incarceration rates depends on race, then this is symptomatic of taste-based discrimination. The rank-order test results imply that we cannot reject the null hypothesis of no taste-based discrimination.

This paper investigates how collusion between regulators and firms affects workplace safety using the case of China’s coal mine deaths. We argue that decentralization makes collusion more likely and that its effect is strengthened if the transaction costs of collusion are lower. These hypotheses are tested by investigating the impact of decentralization contingent on regulators’ characteristics. Exploring both decentralization and centralization reforms in the coal mine industry, we find that decentralization is correlated with an increase in coal mine death rates. Moreover, this increase in mortality is larger for the regulators with lower transaction costs (proxied by the locality of origin).

This paper develops a new approach for measuring the stringency of a major form of land use regulation, building height restrictions, and applies it to an extraordinary data set of land-lease transactions from China. Our theory shows that the elasticity of land price with respect to the floor area ratio (FAR), a building height indicator, is a measure of the regulation's stringency (the extent to which FAR is kept below the free-market level). Using a national sample, estimation allowing this elasticity to be city-specific shows variation in the stringency of FAR regulation across Chinese cities. Single-city estimation for Beijing shows that stringency varies with site characteristics.

Admission practices at high-profile universities are often criticized for undermining academic merit. Popular tests for detecting such biases suffer from omitted characteristic bias. We develop a bounds-based test to circumvent this problem. We assume that students who are better qualified on observableswould, on average, appear academically stronger to admission officers based on unobservables. This assumption reveals the sign of differences in admission standards across demographic groups that are robust to omitted characteristics. Applying our methods to admissions data from a British university, we find higher admission standards for men and slightly higher ones for private school applicants, despite equal admission success probability across gender and school background.

Using data on the age, sex, ethnicity, and criminal involvement of more than 14 million residents of all ages residing in approximately 4,000 Dutch neighborhoods, we test if an individual's criminal involvement is affected by the proportion of criminals living in his or her residential neighborhood. We develop a binomial discrete choice model for criminal involvement and estimate it on individual data. We control for both the endogeneity that may be related to unobserved neighborhood characteristics and for sorting behavior. We find significant social interaction effects, but our findings do not imply multiple equilibria or large multiplier effects.

Business decision makers are increasingly using predictive social media analytic tools in forecasting exercises but ignoring potential model uncertainty. Using data on the universe of Twitter messages, we calculate the sentiment regarding each film to understand whether these opinions affect box office opening and DVD retail sales. Our results contrasting eleven different econometric strategies including penalization methods indicate that accounting for model uncertainty can lead to large gains in forecast accuracy. While penalization methods do not outperform model averaging on forecast accuracy, evidence indicates they perform equivalently at the variable selection stage. Finally, incorporating social media data greatly improves forecast accuracy.