| Abstract | The key to testing the productivity advantage of export enterprises is to identify the learning effect and the selection effect of export. In this paper, the mechanism is divided into general learning effect, bias learning effect and selection effect. Then, the unconditional distribution of parameter - parameter correspondence method proposed by Combes et al (2012) is used to analyze the productivity distribution of export and non-export enterprises, and quantitative identification and decomposition of the export business productivity advantages. The results show that: (1) There are significant learning effects in China's export enterprises, and the learning effect of inefficient enterprises is greater. The choice effect of export does not exist, and even slight adverse selection effect exists. If bias learning effect is neglected, we will underestimate the general learning effect and the wrong choice of the direction of the effect. (2) Comparing different trade, the general effect of mixed trade is the strongest, but the reverse selection is the most obvious; the processing trade has strong biased learning effect, the reverse selection problem is not serious, but the learning effect of processing trade is negative.(3) The impact of export on productivity has shown significant heterogeneity in firms of different industries, different trade countries, different sizes, different ownership, different ages, different regions and import / export differences. (4) There is no general learning effect and selection effect in processing trade, which is the key to the paradox of China's export-productivity. However, there are biased learning effects in processing trade that inefficient enterprises can obtain learning effect in processing trade. This paper has important implications for understanding the productivity decision mechanism of China's export enterprises. |