Firm heterogeneity in productivity across Europe. What explains what?

Abstract

There is a substantial heterogeneity in productivity when comparing individual firms. However, even when heterogeneity is found, some questions still remain unaddressed. For instance, when focusing on EU nothing is known about the importance of firms' heterogeneity compared with that of location. This is a point to be addressed on empirical grounds: location is expected to affect firms, but there is no evidence quantifying the magnitude of these two effects across Europe. How much the difference is due to individual heterogeneity and how much it is a result of territorial influences? We depart from these arguments and contributes to the issue of EU TFP divide by questioning if differences in TFP levels depend on firms and regional specific effects. In other words, does location matter in understanding TFP regional disparities across Europe? If it does, how much of TFP variability is the result of being located in a region instead of in another. And, what about country-effects? In order to answer to these questions, we proceed by using data of firms operating in the seven European countries comprised in the EFIGE dataset (Austria, France, Germany, Hungary, Italy, Spain, United Kingdom, henceforth, EU7-EFIGE countries). In this respect, when focusing on these countries, the role of being located in different regions will be investigated, net of the country-effect. Furthermore, a deep-analysis on the impact of regionalism within a given country, will be made by focusing on France, Italy and Spain. The key variable of the study is the TFP, which has been calculated at firm level by Bruegel for 2008 by employing the Levinsohn and Petrin (2003) approach. The empirical setting we propose is consistent with the type of analysis we carry out. Indeed, in order to explain the role of different factor in explaining firms' TFP, we consider the multilevel approach. This model allows us to evaluate whether and to what extent space matters in determining firms' performance. In fact, multilevel regressions combine different levels of data aggregation and relate them in ways that render the simultaneous existence of distinct level-one and level-two equations explicit. After having found high TFP heterogeneity across firms and regions, we confirm that firm-specific characteristics greatly affect individual TFP. They dominate to location. Another evidence regards the regional effect. It is high when estimations disregard the country-effects: in such a case, location across EU regions explains about 15.2% of the firms differences in TFP. After controlling for country-effects, we find that about 95.3% of the variance in European firms' TFP is due to firms' characteristics and 4.7% is ascribable to regionalism. These proportions slightly differ when considering the case of France, Italy and Spain and when regressions attempt to capture the role of sectoral membership.