Identifying Best-Practices: A Monte Carlo Evaluation of Quantile Regression

Abstract

The area of best practice research has only rcccntly begun to embrace statistically based comparisons as a basis for ying recommended practices. In part motivated by growing interest in performance measurement activities these new hes hold significant potential for improving out ability to identify and utilize true best practices. Unfortunately, little been done to study how to apply statistical methods to this task appropriately. In this paper a Monte Carlo evaluation is loped to demonstrate that how Quantile Regression methods can be used to identify best practice. After a brief literature and a summary of the Quantile Regression technique, the paper develops a specific monte carlo simulation design statistical situations with varying numbers of high, medium and low performing organizations. Next, we apply quan regression to the simulated data and attempts to develop some reasonable guidance about how to apply quantile rcgres to real world data. The results demonstrate that quantile regression can accurately estimate different models for different types of organizations (e.g. high and low performing) and should be considered as an effective tool for the empirical study of practices when samples of similar organizations are available. As an order based statistical estimation approach, it also the virtue of being more robust than typical moment approaches. This paper is based on a working paper developed at The Maxwell School in June 2001, which was partially funded by doctoral research grant from the Dean of The Maxwell School and the Center for Technology and Information Policy.