We use a new dataset to estimate the impact of temperature on economic activity at a more geographically and temporally disaggregated level than the existing literature. Analyzing 30-kilometer grid cells at a monthly frequency, temperature has a negative, highly statistically significant, and quantitatively large effect on output: a 1 °C increase in monthly temperature is associated with a 0.77 percent reduction in nighttime lights, a proxy for local economic activity. The effects of even a temporary increase in temperature persist for almost one year after the shock. Increases in temperature have an especially large, negative impact on growth in poorer countries, indicating that they are more vulnerable to the impact of climate change.