Discrete choice models (DCMs) are used to forecast demand in a variety of engineering, marketing, and policy contexts, and understanding the uncertainty associated with model forecasts is crucial to inform decision-making. This thesis evaluates the suitability of DCMs for forecasting automotive demand. The entire scope of this investigation is too broad to be covered here, but I explore several elements with a focus on three themes: defining how to measure forecast accuracy, comparing model specifications and forecasting methods in terms of prediction accuracy, and comparing the implications of model specifications and forecasting methods on vehicle design. Specifically I address several questions regarding the accuracy and uncertainty of market share predictions resulting from choice of utility function and structural specification, estimation method, and data structure assumptions. I1 compare more than 9,000 models based on those used in peer-reviewed literature and academic and government studies. Firstly, I find that including more model covariates generally improves predictive accuracy, but that the form those covariates take in the utility function is less important. Secondly, better model fit correlates well with better predictive accuracy; however, the models I construct— representative of those in extant literature— exhibit substantial prediction error stemming largely from limited model fit due to unobserved attributes. Lastly, accuracy of predictions in existing markets is neither a necessary nor sufficient condition for use in design. Much of the econometrics literature on vehicle market modeling has presumed that biased coefficients make for bad models. For purely predictive purposes, the drawbacks of potentially mitigating bias using generalized method of moments estimation coupled with instrumental variables outweigh the expected benefits in the experiments conducted in this dissertation. The risk of specifying invalid instruments is high, and my results suggest that the instruments frequently used in the automotive demand literature are likely invalid. Furthermore, biased coefficients are not necessarily bad for maximizing the predictive power of the model. Bias can even aid predictions by implicitly capturing persistent unobserved effects in some circumstances. Including alternative specific constants (ASCs) in DCM utility functions improves model fit but not necessarily forecast accuracy. For frequentist estimated models all tested methods of forecasting ASCs improved share predictions of the whole midsize sedan market over excluding ASC in predictions, but only one method results in improved long term new vehicle, or entrant, forecasts. As seen in a synthetic data study, assuming an incorrect relationship between observed attributes and the ASC for forecasting risks making worse forecasts than would be made by a model that excludes ASCs entirely. Treating the ASCs as model parameters with full distributions of uncertainty via Bayesian estimation is more robust to selection of ASC forecasting method and less reliant on persistent market structures, however it comes at increased computational cost. Additionally, the best long term forecasts are made by the frequentist model that treats ASCs as calibration constants fit to the model post estimation of other parameters.

Reducing automobile-based gasoline consumption has been a major U.S. public policy issue recently. A key driving force behind policymakers' desire is the concern of environmental externalities and national security. Currently, there are three public policies towards reducing automobile gasoline consumption: raising federal gasoline tax, raising the Corporate Average Fuel Economy (CAFE) Standards and vehicle scrappage subsidies of government to retirement of old vehicles. My research studies the effectiveness of these policies in the United States. Among all polices, economists often argue that higher gasoline tax would be more effective in improving fuel economy efficiency. In my first chapter, I ask how gasoline prices influence households' automobile replacement decisions and thus market fuel economy efficiency, which is measured by average mileage per gallon in a city. I specify and estimate a structural dynamic model of consumer preference for new and used vehicles following the methodology proposed by Gowrisankaran and Rysman (2009). Since gasoline costs accounts for 65% of total operating costs, the current and future gasoline price must need to be taken into consideration for rational forward-looking consumers when they are making vehicle choices. Besides, the replacement decision for vehicles is dynamic as well: facing depreciation as the automobile ages and the improving features for new products, consumers need to decide whether to replace the vehicle in the current period or later. Therefore, a dynamic model of consumer choice would be crucial to correct policy evaluation of fuel economy efficiency, while previous literature fails to consider the dynamics. By taking dynamics into consideration, I am able to capture the inherent dynamic nature of a forward-looking consumer's decision, with rational expectation on the evolution of vehicle attributes and retail gasoline prices. I estimate the model using a rich dataset combing vehicle registration data on different cities, vehicle characteristic data, average gasoline price, etc. Although a high gasoline tax is never put in practice in the U.S. and may not be political feasible, I further conduct an experiment of raising gasoline tax to test how fuel economy efficiency is affected based on my model estimates. Experiments suggest that keeping a $4 gasoline price would result in a steady trend for a city's fleet fuel efficiency increase, while doubling current rate will only increase fuel efficiency in the first several years, but experience drops over time. The Corporate Average Fuel Economy (CAFE) are regulations in the United States that intended to improve the average fuel economy of cars and light trucks sold in the US. However, it is long been realized that with a more fuel efficiency car, consumers may be induced to drive more which partially offsets the original energy saving by the policy. Therefore, to assess the effectiveness of CAFE standards, it is crucial to ask: how fuel economy efficiency, which is measured by mileage per gallon (MPG), affects households' vehicle mileage traveled and its distribution. In my second chapter, I answer the question by estimating a structural model for joint determination of vehicle fuel efficiency choice and vehicle mileage traveled each year with a detailed micro-level data of National Household Travel Survey 2001. I further study the distributional effects on vehicle miles of fuel efficiency using instrumental quantile regression. Comparison on results and tests of weak instruments between my method and literature suggest that my model and choice of instruments provide consistent estimates, while using choice probabilities as instruments is not valid. My results support some earlier findings of rebound effects with a more precise quantitative estimation. In addition, I find new evidence that costs associated with raising CAFE standards vary across different quantiles of annual mileage driven and are especially high for those with below-average vehicle mileage driven. These findings also provide rationale in support of a tax on mileage, which is more effective in reducing gasoline consumptions, comparing to the costs of CAFE standards. My third chapter focus on 2009 CARS Program (Cash-for-Clunker). The 2009 CARS program attempted to boost the sale of new fuel efficient vehicles to replace old gas guzzlers. The program established a two-tier incentive system depending on whether buyers purchased a passenger vehicle or an SUV. The result is that many of the new purchased vehicles are indeed SUVs. The CARS program collected information about the old scrapped vehicles and linked it to the actual purchase of the new vehicles. It is thus possible to analyze the effect of preference inertia in choices by comparing the characteristics of old and new vehicles. The fact that effective prices that consumers face are determined by the mileage class of the old car also allows us to evaluate the distribution of valuation trade-offs between mileage and other characteristics such as size, performance, and vehicle class. My findings suggest that the 2009 Cash-for-Clunker is not very effective in terms of affecting consumers' choice of SUVs and big cars. For transactions under the program, consumers still prefer SUVs and large cars. The extra $1000 rebates actually increase consumers' tastes towards SUVs. / text

Submitted by Carlos Aurélio Bustamante Gonçalves (carlos.a.b.goncalves@gmail.com) on 2016-12-01T22:33:16Z No. of bitstreams: 1 carlos gonçalves - Dissertação - o mercado de caminhões no brasil -01122016.pdf: 3390001 bytes, checksum: df0eedc2484f2a8784dda76039181601 (MD5) / Rejected by Fabiana da Silva Segura (fabiana.segura@fgv.br), reason: Boa Noite, Prezado Carlos Peço corrigir alguns itens de formatação conforme segue: - Excluir o acento do nome Getulio nas páginas - Nome deve ser em maiúsculo (alternar, nas páginas que tiver o nome) - Titulo também em fonte maiúscula (alterar nas páginas que contém o título) - No rodapé permanece somente São Paulo - excluir o - SP - Linha de Pesquisa: Finanças e Economia de Empresas, alterar nas páginas que contém a lnha - Excluir na contra capa abaixo no nome do orientado FGV - EAESP - Folha de Assinaturas, alterar a linha de pesquisa e colocar a data de aprovação: 17/11/2016 Peço proceder com as alterações e submeter o trabalho novamente on 2016-12-01T23:28:13Z (GMT) / Submitted by Carlos Aurélio Bustamante Gonçalves (carlos.a.b.goncalves@gmail.com) on 2016-12-02T02:34:57Z No. of bitstreams: 1 carlos gonçalves - Dissertação - o mercado de caminhões no brasil -02122016.pdf: 3391454 bytes, checksum: 7cdcde7e62ecc28be15ec876866cca96 (MD5) / Approved for entry into archive by Fabiana da Silva Segura (fabiana.segura@fgv.br) on 2016-12-02T14:03:17Z (GMT) No. of bitstreams: 1 carlos gonçalves - Dissertação - o mercado de caminhões no brasil -02122016.pdf: 3391454 bytes, checksum: 7cdcde7e62ecc28be15ec876866cca96 (MD5) / Made available in DSpace on 2016-12-02T18:40:07Z (GMT). No. of bitstreams: 1 carlos gonçalves - Dissertação - o mercado de caminhões no brasil -02122016.pdf: 3391454 bytes, checksum: 7cdcde7e62ecc28be15ec876866cca96 (MD5) Previous issue date: 2016-11-17 / Este estudo trata do comportamento da demanda por caminhões novos no Brasil no período de 1996 a 2015 e da investigação dos fatores que a influenciam. Tal questão é relevante devido à escassez de estudos acerca deste tema, ainda que se trate do modo historicamente predominante de transporte de carga no país. O objetivo de pesquisa é determinar e quantificar os fatores que provocam o aumento ou diminuição das vendas de caminhão no Brasil. Para atingir este objetivo, foram construídos modelos econométricos a partir de dados secundários. Proxies utilizadas em outros modelos de demanda automotiva foram confirmadas e refinadas, enquanto novas proxies foram introduzidas com sucesso. Quanto aos resultados, este estudo inovou ao identificar três tipos de determinantes, e ao detalhar os efeitos e defasagens de suas influências: variáveis relacionadas especificamente ao mercado de caminhões, variáveis relacionadas ao PIB e variáveis relacionadas à confiança do comprador. Adicionalmente, realizou-se uma verificação da causalidade entre crédito e vendas, com o surpreendente resultado de que a influência ocorre no sentido de vendas para crédito. Com estes resultados, este estudo dissemina o conhecimento a respeito do comportamento do mercado a toda a cadeia produtiva, melhorando a qualidade das decisões e proporcionando aumento da eficiência para o sistema como um todo. / This is a study on the demand for new trucks in Brazil from 1996 to 2015, and an investigation on the factors that influence it. This topic is relevant due to the scarcity of studies concerning the subject, and due to the overwhelming domination of trucks in cargo transport in Brazil. It aims identify and quantify the variables that drive the sales of trucks. To reach this goal, econometric models were constructed based on secondary data. Variables usually adopted in other studies on automotive demand were confirmed and even refined and new variables were successfully introduced. This study innovates by identifying three groups of variables, and by detailing the effects and the lags of their influence: variables specific to the truck market, variables related to GDP, and variables related to the decision maker’s confidence. Additionally, a causality analysis involving credit and truck sales was performed, unexpectedly resulting in sales causing credit. Through these results, this study disseminates knowledge about the behavior of the truck market to the entire productive chain, contributing to evolve the quality of decisions and the efficiency of the entire system.