| This thesis examines the utilization and retirement choice of kilns by cement plants. A kiln is the most important piece of equipment in a cement plant and the number of kilns in the plant determines the plant's capacity. The analysis takes into account (a) the discreteness in kiln utilization and retirement choices and (b) that these choices depend on a firm's belief about the stochastic processes of the exogenous variables such as product and input prices.; These decisions are analyzed using stochastic dynamic programming techniques. In this thesis, semiparametric estimates are obtained for as many parameters of the decision rules as possible.; Specifically, plant level data on fuel and electricity consumption are used to estimate fuel and electricity coefficients by nonlinear and linear regression respectively. Estimates are also obtained in the presence of plant-specific effects.; Additional parameters of the utilization and retirement decision rules are obtained from a static model by "maximum score". These estimates are compared to the logit maximum likelihood estimates. The logit model is tested by nonparametrically estimating the probability of nonutilization. It is concluded that the logistic distribution may be inappropriate for the static model. Kiln-specific effects are tested by comparing the maximum score estimates to the results of a "conditional maximum score" estimator. Simulations of our model suggests that kiln utilization is more sensitive to changes in fuel and raw material prices than to other input prices.; Lastly, a program is developed for estimating the remaining parameters, using the "nested fixed point maximum likelihood" procedure.; This thesis has attempted to obtain reliable structural estimates of kiln utilization and retirement decision rules. The methodology can be potentially applied to analyze capital utilization and retirement, in general, at the micro level; this would be important for prediction purposes, for analysis of investment policy and economic growth. |
