Luther Yap & Yuci Zhou, Princeton University

Sep 21, 12:00 pm1:20 pm


Event Description

Valid and Shorter Confidence Intervals for IV when F is small- Luther Yap


To address the large-sample inference problem caused by possibly weak instruments in the single variable just-identified IV model, applied research typically adopts the approach of Stock and Yogo (2005) -- t-ratio-based inference with first-stage-F-dependent critical values. To accommodate small values of F, we extend this ``F-based" approach to its logical conclusion by incorporating the full set of sufficient statistics for the model, resulting in a new inference procedure that we call VtF. We show that VtF confidence intervals are generally expected to be shorter than competing ``robust to weak instrument" intervals, including those from the recommended benchmark of Anderson and Rubin (1949) (AR). In a sample of 89 specifications from 10 recent empirical studies drawn from five general interest journals, VtF intervals are shorter than AR intervals 100 percent of the time, and are even shorter than the (invalid) intervals 30 percent of the time.


Public Procurement in the Presence of Private Market Spillovers: Evidence from WIC - Yuci Zhou


Governments frequently provide in-kind transfers of goods to low-income households. In doing so, a central tension the government faces is between increasing consumer surplus for program participants and containing costs. A less frequently discussed trade-off is between public procurement of these goods and consumer surplus in the private market. We study WIC’s procurement of infant formula wherein state WIC agencies award a sole source contract to a manufacturer. WIC regulations ensure availability of WIC-eligible formula for beneficiaries in stores, but in turn the market power of WIC brands in the private market increases substantially. We provide suggestive evidence that the market power effect is mainly generated by differences in availability of WIC and non-WIC brands in stores. Welfare implications of this cost containment mechanism and counterfactual mechanisms are discussed.