Date of Award

Spring 2021

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Berry, Steven


Chapter 1 explores possibilities for non-parametrically identifying peer effects. In my model, outcomes associated with individuals are an unknown function of expected peer outcomes and other individual-specific attributes. Unobserved heterogeneity in outcomes across individuals is captured by an additively separable individual-specific error. The error is mean dependent on peer group, rendering expected peer outcome an endogenous covariate. Exogenous peer attributes are absent from the individuals’ remaining covariates.\\ When the data is cross-sectional, I propose imposing a stringent but intuitive assumption on how individual outcomes depend on expected peer outcomes and remaining individual-specific attributes. Specifically, I assume one of the individual-specific characteristics indexes how strongly an individual interacts with his peers. If the index is zero, the individual is not directly affected by peer outcomes. Under this assumption, the model is identified, up to a normalization. When panel data is available, the assumption is unrequired, and the model is identified via a more traditional IV-based approach. The cross-sectional result leads to tests for whether peer effects estimates reflect forces associated with social interactions or not.\\ Chapter 2 studies how economies of scale and product differentiation affect manufacturer-supplier relationships. A model consisting of two manufacturers, each with pre-existing relationships to two separate suppliers, is analyzed. At the start, an unrelated manufacturer-supplier pair decides whether to invest in a new relationship. Based on the resulting network of manufacturer-supplier relationships, a manufacturer's input price is determined by Nash bargaining if it's related to one supplier, and a first-price auction if otherwise. When manufacturers are horizontally differentiated, hold-up of investment by neighbor manufacturers causes manufacturer-supplier network connectivity to be too low. On the other hand, overinvestment in outside option relationships causes network connectivity to be inefficiently high. Shocks to individual manufacturers or suppliers have disproportionately large (small) welfare consequences vis-a-vis ex-ante market shares when the network is under (overly) connected compared to the socially optimal network.\\ Chapter 2 also estimates a micro-founded model of firm-to-firm relationship formation, using prices, quantities and product-supplier-level network data for 2008-16 U.S. automobiles. The model incorporates the theoretical model's key elements - supplier-level economies of scale, downstream market product differentiation and relationship network contingent input pricing. To identify the model, I assume manufacturers Nash bargain with suppliers inherited from previous periods. I then exploit variation in these suppliers' quantities to identify how production costs vary with rival product output. I find, on average, main suppliers of chassis, exterior and combined inputs experience significant economies of scale. Ex-post and on average, manufacturers do not benefit from forming their chosen relationships, absent rents from outside option relationship overinvestment or compensation from other firms. In comparison, hold-up of relationship investment is less significant in affecting incentives to form relationships.