Frankel and Romer (1999), hereafter FR, proposed an instrument variable for trade intensity to assess robustly the causal impact of international trade on income per person. They generated the instrument by estimating a gravity equation with only exogenous, geography-related explanatory variables on a cross-section from 1985 using ordinary least squares (OLS). This paper revisits the FR study using a new estimation strategy, the Poisson maximum likelihood estimator (PPML), for creating the instrument for 1985. Next, I repeat the IV regressions for 2005 using both OLS and PPML for estimating the instruments. I find that the IV regressions are sensitive to the period, the sample size and the estimation strategy on which the instrument is estimated. OLS based instruments are not significant in IV regressions for 2005, while PPML-based instruments are statistically and economically significant and robust to time, but do not convincingly pass tests for weak instruments.