The rise in the availability of remotely sensed weather data has resulted in economists predicting different outcomes using rainfall as an explanatory or instrumental variable (IV). We analyze 174 papers to identify common rainfall metrics used as an instrument and show the extent of their ad hoc use in predicting a range of outcomes. We use agricultural productivity as a case study to examine the suitability of using different rainfall metrics as an IV. To that extent, we test the predictive power of the 14 most common rainfall metrics in the economics literature, calculated through six remote sensing products across six countries, on agricultural productivity. We find a large amount of heterogeneity in the performance of rainfall metrics. We also find concerning evidence about the validity of using rainfall metrics as an instrument, especially regarding possible exclusion restriction violations and weak instrument problems. Our findings emphasize the need for researchers to carefully select and justify their use of a particular rainfall metric to improve the reliability of their analysis.