Urban and regional distinctions for aggregating time series data

Abstract.  This article argues that using either the SIC or NAICS one-digit classifications as a method of aggregating two- and three-digit time series data can ignore important regional characteristics. We present a pairwise cointegration approach of aggregation where the aggregated sectors can vary widely across regions. By systematically constructing region-specific sectors from more detailed industries, we find that the level of agglomeration across rural and urban areas can affect the composition and number of local sectors in a region. We use the results pointing to rural/urban geographic distinctiveness to further consider the Carlino and Defina (1998, 1999 ) finding that monetary policy has disparate effects across regions in the U.S.