An Empirical Comparison of Constrained Optimization Methods for Benchmarking Economic Time Series

• Activity Number: 300
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Irene Brown*+
• Companies: U.S. Census Bureau
• Keywords: Benchmarking ; Constrained Optimization ; Time Series ; Nonlinear Programming

Abstract

Economic Programs at the U.S. Census Bureau use the Causey-Trager method for benchmarking monthly and quarterly series to annual series and to the Economic Census every five years. This procedure uses an iterative, nonlinear constrained optimization technique known as steepest (gradient) descent to obtain the benchmarked series. Many practitioners prefer and predominately use conjugate gradient descent instead. This paper investigates how the techniques differ in terms of convergence (number of iterations to obtain the minimum), value of the obtained objective function, level estimates of the benchmarked series, and preservation of the original series period-to-period changes. We found that conjugate gradient descent generally converged faster and achieved a smaller objective function, but the resulting benchmarked series were very similar.