A Short Exercise to Assess the Consequences of Temporal and Spatial Aggregation on the Observed Spatial Interactions

Since our observation of the regional economy depends on the scale of temporal/spatial units, even under the same underlying disaggregated level data generating process, we can encounter different neighborhood effects or spillovers. In this paper, the amount of spillover is defined by the forecast error variance decomposition (FEVD), and the direction of spillover is defined by the long-run sign of the cumulative impulse response function (CIRF). From an exercise using a constructed regional economic system, the size of spillover was found to decrease with spatial aggregation in a multi-level structure regional economic system. However, no monotonic trend was found in terms of the relative portion of positive/negative spillovers. In addition, the results from the real world data using different levels of aggregations, and the results drawn from the exercise on the constructed regional economic system are compared. From this comparison, a multi-level structure model which assumes the existence of higher level common factor affecting the regional units was found to concord with logical experiments conducted over the constructed regional economic system.