Essays on applied production analysis.

Abstract

The first essay is concerned with the existence of 'psychological barriers' in the stock market. The last two essays in this dissertation deal with the issue of recovering technological parameters that characterize production possibilities given economic data. The first essay deals with the issue of the existence of psychological barriers in the stock market. The popular press attaches particular significance to certain numerical values of the Dow-Jones index (DJIA). These magic numbers are referred to as 'resistance levels' or 'psychological barriers.' We analyze 38 years of closing values of the DJIA, focusing on the behavior of the individual digits that combined form the DJIA in the hope that we might discover evidence in favor of the existence of psychological barriers in the stock market. We are able to rationalize what at first seemed peculiar and we fail to find any such evidence. Our conclusion is that, contrary to initial impressions, there is little if any predictive power in the closing digits of the DJIA. In the second essay, linear programming--i.e. data envelopment analysis--techniques are used to get individual efficiency ratings for Spanish hospitals. Public hospitals seem to be less efficient than non-public hospitals. However, the intensity of the inefficiency, when it exists, is greater for the non-public hospitals in some key categories. The presence teaching activities doesn't seem to have any effect on the efficiency ratings. We also provide estimates of the potential savings in inputs and increases in outputs which could be obtained if the inefficient hospitals chose to use the efficient production plans. In the third essay, the use of switching regression techniques in an activity analysis framework is investigated. Maximum likelihood-based methods are proposed and different switching specifications are discussed. The viability of these newly proposed techniques is established. The methods developed combine the advantages of the two major approaches to frontier estimation: the functional flexibility of the linear programming--nonparametric and nonstatistical--approach and the statistical nature of the econometric--both parametric and statistical--approach. This combination comes at the expense of some analytical complexity.