Successful export growth and diversification require not only entry into new export products and markets but also the survival and growth of export flows. For a cross-country dataset of product-level bilateral export flows, exporting is found to be a perilous activity, especially in low-income countries. Unobserved individual heterogeneity in product-level export flow data prevails even when a wide range of observed country and product characteristics are controlled for. This questions previous studies that used the Cox proportional hazards model to analyze export survival. Following Meyer (1990), a Prentice-Gloeckler (1978) model is estimated, amended with a gamma mixture distribution summarizing unobserved individual heterogeneity. The empirical results confirm the significance of a range of product-as well as country-specific factors in determining the survival of new export flows. Important for policy-making is the finding of the value of learning-by-doing for export survival: experience with exporting the same product to other markets or different products to the same market is found to strongly increase the chance of export survival. A better understanding of such learning effects could substantially improve the effectiveness of export promotion strategies. Export survival, Cox proportional hazard, low-income countries. JEL codes: F10, F14, C41
