IMPROVING LOCAL MANUFACTURING EMPLOYMENT FORECASTS USING COINTEGRATION ANALYSIS

Procedures for tracking and forecasting economic conditions in regiona l economies have evolved significantly over the last 30 years. Much of this evolution has followed developments in macroeconomics, where tec hniques for tracking/forecasting key economic variables have tended to originate. This technique adoption and adaptation process continues t oday, as developments in the modeling of cointegrated macroeconomic ti me series have begun to appear in the regional modeling and forecastin g literature. This paper presents an effort at modeling a segment of a regional economy using the cointegration testing procedures suggested by Johansen and Jusilius (1990) to develop a forecasting model for ma nufacturing employment in Milwaukee, WI. The paper demonstrates how Ve ctor Error Correction (VEC) modeling can lead to gains in the accuracy of local manufacturing employment forecasts relative to more traditio nal VAR models in either levels or first-differenced form, In the proc ess, it demonstrates procedures for developing a relatively simple VEC model that reveals something about the structure of the local manufac turing sector, including possible linkages to the national economy. Th is information can assist local policy makers in anticipating and adap ting to business cycle-related fluctuations in this critical sector of the local economy.