Extending Dynamic Factor Analysis to Panel Data: An Application on Benchmarking Index for Water Utilities

• Activity Number: 475
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Nikolaos Zirogiannis*+ and Yorghos Tripodis and Joe Moffitt and Alexander Danilenko
• Companies: University of Massachusetts at Amherst and Boston University and University of Massachusetts at Amherst and World Bank
• Keywords: Dynamic Factor Analysis ; Panel Data ; EM algorithm ; Leading indicator ; Water utilities

Abstract

We develop a dynamic factor analysis model that can be applied on panel data. While dynamic factor analysis is a widely used statistical technique, there is limited research regarding its application on panel data. In simple dynamic factor analysis, we have multiple variables of the same subject, such as a company or a country measured at each time period. In a more general panel data setting, we have several variables of multiple subjects, such as a company, nested within a higher level variable, such as a country, measured at each time period. This creates an extra element of complexity, since we need to account for between company variability, as well as within and between country variability. In this paper, we propose a simple solution to this problem by using the EM algorithm to estimate factor loadings for panel data. We explore different correlation structures for the within-country variability and propose a model selection methodology. The model is then applied on a dataset of various countries, each with multiple water utilities. The goal is to create a leading benchmarking index that will be able to predict the inability of a utility to cope with external stresses.