In this paper we construct exchange rate indexes for the nine U.S. Bureau o
f the Census regions. The results indicate there are nontrivial differences
between them and an identically created overall U.S. exchange rate index.
The national index is cointegrated with only two of these regional indexes,
and in a Granger sense, it is causing one regional index. In addition, our
results indicate that two of the regional exchange rate indexes are interc
hangeable with the national index. These results show that researchers inve
stigating how exchange rate changes affect regional exports or regional eco
nomic growth should be cautious in making inferences based upon a national
exchange rate index.