ÌÇÐÄvlog¹ÙÍø

Understanding the differences between rich and poor places is complicated by the fact that places differ from each other in numerous ways. In this paper, we show how a dimension reduction algorithm can unveil hidden patterns in US census data and consistently yield useful insights into the type of economic activities that separate rich and poor states over 160 years of development history. Moreover, we find this approach has a unique ability to shed light on the dynamics of evolving landscapes and changes in relevance of particular types of activities, such as the shift from manufacturing to high skill services that occurred in the US over the last 40 years. Our results have important implications for the decline of the rustbelt and the reversal of US regional income convergence from 1980 onwards.