Frankel and Romer (1999) proposed an instrument variable for trade intensity to robustly assess the causal impact of international trade on standards of living. The instrument is based on OLS estimates of the gravity equation and has been widely used in the literature. In this note I show that the instrument is unrelated to income in the mid-2000s. Re-estimating the gravity equation using PPML, I show that the resulting instrument is strongly related to GDP per capita, but weakly correlated with trade, suggesting that what is captured may be a direct link between geography and income.