What knowledge do firms need to survive? In this paper we use data on the work history of all formal workers in Brazil to measure the industry-specific, occupation-specific, & location-specific knowledge that workers bring into new firms: pnas.org/content/115/50… (thread 1/) 
We measure these three knowledge channels by looking at workers that switched jobs, focusing on changes among related and unrelated industries, occupations, and locations. 2/
For instance, a worker moving between related industries (banking & insurance) should carry more industry-specific knowledge than a worker moving between two unrelated industries (e.g. banking and soybean farming) 3/
We then look at the effect of these three knowledge flow networks in the probability that new firms survive or grow. We used a variety of statistical models: a fixed-effect panel regression, a cox-proportional hazards model, and a new type of Bartik instrument in a 2SLS. 4/
We also separated among pioneer firms (firms operating in an industry that was not present in a location) & non-pioneer firms (firms operating in an industry that was present in a location). 5/
For pioneers, we found that survival increases when they hire workers with industry & location specific knowledge, but not with occupation-specific knowledge. This is after controlling for schooling, firm size, wages, & industry, location, & time fixed-effects 6/
This finding is confirmed by a cox proportional hazards model, and by a Bartik 2SLS, which uses national level fluctuations in an industry as an exogenous shock in the local labor supply of related industries. 7/
We also compared the survival rates of pioneers & non-pioneers (in the Appendix), finding that the effects of industry-specific knowledge are more pronounced for pioneers. Occupation-specific knowledge appears to play a minor role for non-pioneers 8/
pnas.org/content/pnas/s…
pnas.org/content/pnas/s…
These results are interesting for a few reasons. First, they are helping us understand the role of relatedness in diversification at a more micro-level. 9/
The fact that regions are more likely to enter related economic activities is well known, but here, we can see exactly who entered, who was hired, & the experience each worker had. By splitting experience into 3 channels we can finally see inside relatedness' "black-box."10/
The results also suggest that industry-specific knowledge may be less transferable, more valuable (it enhances firm survival), and maybe more tacit than occupation-specific knowledge. 11/
More importantly (to me), is the fact that this paper represents the debut of @Cristian_jf as first author in a high impact factor journal. Cristian Jara is a fantastic scholar who deserves all the credit & will soon complete his PhD in my group. Keep him on your radar! 12/
The paper was also a collaboration between three members of @CollectiveMIT (Cristian Jara, Bogang Ju, and I) & @HarvardEcon (Ed Glaeser). It shows that by collaborating across disciplines, we can create work combining the best of different academic worlds. /End_Thread
