A COMPARATIVE-ANALYSIS OF TIMES SERIES APPROACHES TO MODELING INTERSECTORAL AND INTERCOUNTY EMPLOYMENT LINKAGES IN RURAL REGIONAL LABOR-MARKETS

The performance of five criteria for identifying significant interregional labor market linkages is compared. The criteria suggest differential model specifications based on the implicit tradeoff between parsimony and specification bias. Models were specified using: Akaike's final prediction error; Schwarz's Bayesian information criterion; an input-output model; Aoki's time series algorithm; and a combined input-output/time series approach. Models of the changes in monthly employment for twelve industries in six regions in northeastern Utah were specified using each criterion. Models based on Schwarz's criterion and the input-output criterion were preferred to those based on other criteria.