This paper has two objectives. First, I evaluate the role of labor turnover as a mechanism for transfer of efficiency-enhancing knowledge. Second, I examine the extent to which worker flows can explain evidence on agglomeration advantages. In order to do so, I use a matched worker-firm dataset for Veneto Region of Italy (Card, Devicienti, Maida, 2010). Employing the method in Abowd, Creecy and Kramarz (2002), I estimate wage equations where both firm and worker effects can be identified and I define good firms as the high-wage-firms (HWFs), i.e. those establishments with top values of the estimated firm effects. Then I construct firm-specific measures for the share of workers in Venetian firms with experience gained at good firms. By using this measure within a productivity regression framework I can evaluate whether employees trained at good firms who later join other local firms bring with them some of the knowledge that they acquired.
     If labor mobility is to act as a channel for agglomeration advantages, we would imagine to observe the following. First, HWFs should have a firm-specific advantage that could be the basis for knowledge transfer. Second, non-HWFs that hire workers with previous experience from HWFs should benefit in terms of increased productivity. Third, the probability of hiring from a good firm should be higher in localities with a higher share of good firms (i.e. firm location should matter)
     In this paper, I use Social Security earnings records for employees and balance sheet data and location information for employers in order to evaluate the evidence on all four points for Veneto manufacturing during the 1990s. As a first exercise to assess the potential for knowledge transfer in the region, I look for evidence of a HWF advantage by estimating equation using the detailed financial information at our disposal. I show that the HWFs are more productive and have higher capital (in particular intangible) per worker.
     I progress to examine the extent to which non-HWFs benefit from hiring workers from HWFs. I enter annual firm-level measures of the share of workers with recent HWFexperience in a Cobb-Douglas production function. I find that non-HWFs which hire workers with previous experience from HWFs benefit in terms of increased productivity. In the sample of firms that hire from HWFs during the period, I find that a one standard deviation increase in the share of workers from HWFs increases productivity by 0.89 percent. These results are not likely to be driven by selection of workers, nor unobservable productivity shocks.
     Exploiting information on the location of firms, I show that for a non-HWF the probability of hiring a worker with HWF experience is increasing in the share of HWFs in the local labor market where the non-HWF is located. Put it differently, firm location matters. My results are then consistent with the model in Combes and Duranton (2005) where firms that cluster in the same local labor market benefit from better access to workers whose knowledge enhances efficiency.