ST13 The Lag-o-Matic: An Improved Method for Selecting Lag Structure of Multiple Predictor Variables in the Absence of Theory     Contributed

Dr. David Sharp Division of Business University of Southern Mississippi ,Stephen Finnigan
U.S. Department of State
Abstract: Theory generally determines the appropriate lag of each independent (i.e., predictor) variable in a time series regression model. But, what does one do in the absence of theory? Some rely upon correlations or stepwise techniques to determine best fit betwe en the dependent (i.e., response) variable and alternative lagged values of each predictor variable, one by one. Regardless of scheme, variable-by-variable lag selection without theory can be a daunting task. Moreover, when the lagged predictors are regres sed together, unforeseeable interactions can lead to bizarre results, even when time series are pre-whitened with minimal multicollinearity. Such results force additional trips back to the drawing board to find just the right combination of predictor lags. We introduce the Lag-o-Matic, our timesaving SAS program that eliminates many of the hassles associated with lag selection in the face of uncertainty. The Lag-o-Matic automatically (1) lags the predictor variables over a user-defined range; (2) runs regre ssions for all possible lag permutations (i.e., 1,1,1; 1,1,2; 1,2,1, etc.); and (3) restricts results according to user-defined selection criteria (e.g., face validity, significant t-tests, R-sq, etc.). Output contains a short list of models from which the researcher can make quick comparisons and choices. We contend that, while most time series researchers will realize obvious gains from such a program, forecasters will especially appreciate the ease and speed with which the program can produce easy-to-exp lain forecasts intended for lay audiences.

Biography:
David C. Sharp is an Assistant Professor of Economics at the University of Southern Mississippi-Gulf Coast. He holds a Ph.D. in Economics from the University of Memphis. Dr. Sharp's primary interests are in microeconomics and applied econometrics, and he h as published his findings in a number of academic journals, including Journal of Economics Issues, Journal of Forensic Economics, Review of Social Economy, Journal of Family and Economic Issues, and Applied Economics Letters.

Stephen M. Finnigan is employed by the U.S. Department of State in Washington, DC. He holds a MSc. in Economics from the London School of Economics and Political Science and a B.A. in Economics from Willamette University. Formerly an associate economist fo r a consulting firm in the Washington, DC area, Mr. Finnigan has developed methodologies and conducted econometric analyses on a wide-range of projects.