Do More Productive Firms Locate New Factories in More Productive Locations? An empirical analysis based on panel data from Japan's Census of Manufactures

Using a Melitz-style model of heterogeneous firms, Baldwin and Okubo (2006) recently presented a theoretical model in which self-sorting occurs and more productive factories choose to locate in more productive areas. The model suggests that firm-specific factors and regional factors affect each other through the endogeneity of location decisions. However, to date there have been few studies empirically testing this issue. Against this background, our aim is to examine the relationship between firms and location-specific factors in location decisions using factory-level panel data from Japan's Census of Manufactures. We begin by estimating how much of the differences in factories' TFP levels can be explained by both firm and location effects. The estimation results show that both effects have a significant impact on the productivity level of a factory, and that the firm effects are more important than the location effects. We also find a statistically significant negative correlation between firm effects and location effects, and investigate what causes this relationship. One potential explanation is that more productive firms may tend to set up new factories in less productive locations such as rural areas, where factor prices such as land prices and wage rates are usually low, in order to benefit from low factor prices. To examine this issue, we estimate a mixed logit model of location choice. The results indicate that more productive firms indeed tend to set up new factories in low-productivity locations, which is consistent with our hypothesis.