| Summary: | This thesis comprises three related essays in microeconometrics. Motivated by trends in prices, purity and supply disruption in the US markets for cocaine and heroin over the past thirty years, the first essay conducts an applied theoretical and empirical analysis of the impact of seizures on retail market conditions. The main finding is that rising supply disruption may be directly responsible for falling prices. This occurs since seizures raise the cost of wholesale narcotics, incentivising sellers to dilute their product. This leads to declines in purity, reducing demand and causing prices to fall. The second and third essays make a greater methodological contribution, focussing on identification of spillover (or peer) effects. Each seeks to address a shortcoming of existing identification strategies. The second essay studies identification of spillover effects in the absence of network data, using panel data instead. Identification conditions are derived under which both spillovers, and the underlying network, are either partially or fully identified. The approach is applied to study research and development (R&D) spillovers between US firms, finding evidence of underinvestment in R&D. The third essay studies identification of spillover effects if the researcher does not observe a source of strictly exogenous variation in the outcomes. Instead, variation in the variance and covariance of the outcomes over the network is exploited. The essay derives testable identification conditions, which hold for generic networks and fail only in special cases. The method is applied to study spillovers in education, finding evidence of strong, positive peer effects in mathematics test scores for first year kindergarten students in the US.
|
|---|
