Using Hungarian firm-level export data, we show that about one third of firm-destination and about one half of firm-product-destination export spells are short-lived, or temporary, in each year. This is inconsistent with theories where comparative advantage is stable and market entry costs are sunk. We show how endogenous choice between variable and sunk cost trade technologies can explain the empirical importance and some characteristics of temporary trade. The predictions of our model are borne out by the data; the likelihood of temporary trade rises with lower productivity, further location and larger GDP of destination countries.