This paper tries to identify the different channels through which natural disasters affect exports of agricultural products in developing countries. It begins by presenting a simple theory set-up that highlights the different mechanisms at work. It then takes some predictions of this theory to the test. Matching different sets of disaster variables (occurrence and intensity) from EM-DAT and GeoMet datasets with trade data at the 6 digit-HS level, our first estimate point to a negative but statistically non-robust relation between disasters and agricultural exports. Following our theory set-up, we attribute this result to mixing three confounding effects with different magnitudes and opposite signs on trade. Using other sources of data, we could then identify two of the effects: a negative and statistically significant effect of disasters on exports when they occur in rural areas and at growing seasons times; and a positive and (very) robust relation with exports towards culturally close partners and where an important diaspora is settled. This points to show that disasters are redistributing exports across partners. However, the ’solidarity’- consistent effect does not seem to last over time. All in all, notably due to the limited physical impact of most of the disasters over time and space and thanks to the pain relief provided by culturally close importers, natural disasters do not appear to make small developing countries suffer that much economically.