Competitive and Complementary Relationship between Regional Economies: A Study of the Great Lake States

This paper uses a multi-level dynamic factor model suggested by Bai and Wang (2012) to identify the spatio-temporal dynamics of regional business cycles, focusing on six Great Lakes states, Illinois, Indiana, Michigan, Minnesota, Ohio and Wisconsin. The identification scheme suggested by Bai and Wang (2012) enables separate identification of the shock common to the Great Lakes region and the individual shock to each region as well as to assess the interactions between those shocks. The model is estimated using the Gibbs Sampling algorithm. Four monthly time-series variables for each state for the period of January 2003 to September 2012 were organized to implement a dual layer structure dynamic factor model for the Great Lakes region. The main advantage of using this multi-level structure model is that it is possible to assess the effect of a shock originating in one particular region on the other regions separately from the region common shock. In contrast, a single level structural model does not separate the region common shock from the region specific shock. By separating the common shock and the individual shocks, this estimation strategy prevents the possible misunderstanding or overestimation of regional interdependency induced from the co-movements of regional business cycles. Since each region is exposed to the region common shock, the degree of co-movement of each region’s business cycle is strong, possibly exaggerating or biasing the effect of region specific shocks. The simulation results show that incorporating the multi-level structure in a regional dynamic factor model significantly alters the regional interdependency relationship induced from the single level structure model. The variance decomposition shows that much of the region specific business activities can be explained by the region common shock, and the impulse response function shows that the effect of region specific shock is exaggerated or biased in the single level structure model.