At the national level, trade-weighted real exchange rates and foreign incomes significantly impact exports. At the subnational level, these variables are generally insignificant in the few studies that include them. We argue that the standard use of national trade weights in the construction of subnational trade-weighted averages of exchange rates and foreign incomes is inappropriate, and we then construct these variables using state-specific trade weights. In our panel data analysis of state-level manufacturing exports, these variables enter significantly and with the expected signs. Also, their out-of-sample forecasting ability is significantly better than that of the national tradeweighted variables.