This dissertation examines the finished-goods inventory in the U.S. automobile industry using
several econometric methods -- including OLS, the Bayesian approach, the variable mean
response estimation, and the simultaneous equation system. The research starts with the classical
economic order quantity (EOQ) model. The EOQ model is an important, well-known, and
widely-taught inventory management model found in most Operations Management textbooks.
The design of the classical EOQ model is based on several fundamental assumptions. The
extension of the EOQ model has been doing by testing and adjusting the assumptions. The goal
of this dissertation is to investigate inventory problems arising from violations of the rather
unrealistic assumptions underlying the EOQ model, such as the constant demand.
In this research, five models are developed and applied empirically to the U.S.
automotive industry. The objectives of the models are to answer the following research questions:
(1) We would like to test the EOQ model empirically under its assumptions. Can the EOQ model
stay consistent within its specifications? (2) Should the demand in the EOQ model be random
rather than constant? (3) If the demand is random, is demand stochastic and dynamic in terms of
the coefficient of demand uncertainty when incorporated with the EOQ variables? (4) Is the
demand endogenous rather than exogenous? If the demand is endogenous, are the demand and
the order quantity in the EOQ model correlated and jointly determined? (5) Do the service rate
effect and demand-stimulation effect exist simultaneously in U.S. auto dealerships? Overall, the
empirical results demonstrate many avenues for considering new inventory policies to
incorporate with the classical EOQ model. From the specifications, the proposed methodologies
and models can provide inventory decision-makers with different managerial implications and
strategic improvements. Finally, this research should contribute to the literature on the
methodological and empirical fronts.