Empirical Studies on Auto Recall

This dissertation studies the distinct aspects of our quality-related automotive datasets across three chapters: Chapter 1 studies defect rates as independent variables to endogenous recalls decisions with a structural model, Chapter 2 studies supply chain distance as independent variables to endogenous defect rates with a selection-corrected model, Chapter 3 studies defect characteristics as independent variables to endogenous recalls rates with a longitudinal sample.;Chapter 1 studies whether automakers are averse to consumer complaints or to the government recalls they attract. (We investigate how consumer complaints influence government and manufacturer recalls.) We study this question with 8,439 manufacturer recalls, 5,685 government recalls, and 976,062 SaferCar complaints. The government's SaferCar program enables consumers to report quality issues. Mining the text of these reports, we generate 48 dynamic defect variables, such as Crash, Fire, and Injury. We find our variables significantly predictive of the recall decisions.;We model the agents' joint recall decisions as an asymmetric dynamic discrete choice game. We estimate each agent's underlying recall, preemption, and defect cost parameters. Our structural estimates reveal the relative cost of receiving a specific type of a consumer complaint---for instance, one that involves a fire and an injury claim---with respect to a voluntary recall and a mandated government recall.;Our structural estimates imply (i) there is a little overlap in the sets of products the agents recall and in the types of complaints they respond to, (ii) the cost of a recall does not depend on who initiates it, and (iii) auto manufacturers recall faulty products to avoid receiving defect reports but not to preempt anticipated government recalls. Our structural model is useful for managerial implications and regulatory policies: it enables us to perform counterfactual experiments on the these relative cost estimates, measuring their impact on the agents' recall behaviors.;Chapter 2 explores the effect of supply chain proximity on product quality. Specifically, we study whether an increased dispersion between supply chain players causes more quality issues. To study this relationship, we merge four independent data sources from the automotive industry, collecting (i) auto component failure rates, (ii) upstream component factory locations, (iii) downstream assembly plant locations, and (iv) product-level links connecting the upstream and downstream factories.;Combining these datasets yields a large supply chain sample: we detail the flow of 27,807 products through 529 supplier factories and 175 assembly plants. We find that when the distance increases by 100 kilometers, the defect rate of a component increases by 0.21 percent on average. Further, we find that supply chain distance is more detrimental to product quality when manufacturers produce new model generations, high-end products, components with more complex configurations, and also when the manufacturers source from suppliers who invest relatively little in research and development.;Therefore, we identify an important driver of quality and coordination problems in the auto industry: geographic distance spanned by the upstream suppliers and downstream manufacturers. Our study also reveals managerial suggestions for when supply chain distance is more harmful for auto firms.;Chapter 3 investigates how a broad collection of defect signal patterns and characteristics trigger manufacturer-initiated and government-initiated auto recalls. We analyze nearly one million defect reports from NHTSA to study the factors that spur voluntary and mandated recalls. The defect reports data is comprehensive: it comes from a large set of distinct sources such as consumer complaints, insurance companies, and defect petitions. Using the defect variables and statements, we construct a rich set of independent defect vectors. We then study how the government and manufacturers respond to these noisy signals: the defect rates show a large amount of volatility over time.;We estimate the hazard rates of government and manufacturer recalls with respect to our defect signals. We find that defect report sources, distinct failure-related and product-related keywords in the defect statements, short-term and long-term defect volatilities, historical defect accumulations, defect reports from the previous quarters, defect spillovers from similar auto models, and product-level characteristics influence the government and the manufacturers in statistically different ways. Put differently, we find that the government and automaker recall rates change differently with our noisy defect variables.;Also, we find that the hazard rates of both government and manufacturer recalls to the number of defects decrease with the prospects of the other agents' recalls for the same product. Hence, we estimate an insignificant regulatory deterrence effect. Our results have important implications for both auto quality managers and regulatory policymakers: we reverse-engineer a large cluster of underlying defect types that influence government-initiated and manufacturer-initiated recalls. (Abstract shortened by ProQuest.).