Export intensity and productivity

To identify the export premium, researchers typically estimate a fixed difference in productivity between exporters and non-exporters, ignoring how productivity can differ between firms exporting at different intensities. Using World Bank Enterprise Survey data of emerging markets, we do a systematic analysis of the productivity and export relationship across the spectrum of export intensity. We find that pure exporters, defined as firms exporting 100% of their sales,  are distinctly lower in productivity, on average, than regular exporters i.e., firms that both sell domestically and export, and that among regular exporters,  there is a U relationship between productivity and export intensity. Firms exporting at low intensities -i.e. those selling 10% of output or less, and high intensities- i.e. those selling 90% of output or more, are significantly more productive than their non-exporting and pure exporting neighbors, indicating that participating in a second market even at a marginal level implies distinctly different firm productivity. We theoretically explain observed behavior by introducing heterogeneous costs for serving both domestic and export markets into a Melitz model. A simulation of the model yields estimates of export intensity patterns similar to that observed in real world data. The consistency between the empirical results and the model provides support for a modification of the self-selection hypothesis. While standard models assume low costs for selling domestically and high costs for exporting, the wide-spread presence of relatively less productive pure exporters in emerging markets suggests that the reverse can also be true. Thus the self-selection of productive firms should be into both export and domestic markets, rather than just exporting.