We use novel quasi-experimental variation to first test whether firm-specific demand shocks impact wages, then to disentangle predictions coming from wage bargaining and firm upward labor supply (wage posting). We use a unique institutional feature of public procurement auctions in Brazil: the moment in which the auction ends is random. Under this setting, for close auctions in which firms are constantly outbidding each other by incremental amounts, winner and runner-up are as good as randomly assigned. Using this first variation, we find that winning a government contract increases wages. In addition, contract value is higher for auctions that (randomly) end earlier. We use the two sources of exogenous variation combined to disentangle the effect on wages that comes from changes in firm size (wage posting) and the part that comes from changes in contract value holding size constant (bargaining). We find direct evidence of bargaining.
[go to paper]

How does a boss' ability affect their workers' wages? If a high-ability boss induces higher productivity, this effect should be positive. On the other hand, if a good boss is someone who decreases their subordinates' wages, this effect would be negative. I propose an estimator inspired by the peer effects literature, and estimate the effect of bosses' types on other workers' wages in three types of local labour markets. In small town and rural labour markets and in labour markets defined by knowledge-intensive skills, the estimated effect is negative and statistically significant. In the market for low-skilled service sector workers, instead, the effect is small and not statistically significant.
[go to paper]

We use quasi-random allocation of government contracts as exogenous variation on job creation. We then evaluate whether workers who have connections in firms that are hiring are more likely to (i) change jobs, and (ii) receive higher wages without moving. This is a direct test of exposure to information on job creation.

How can one estimate product complementarity using a large administrative data set? In this research I use data from government contracts that result from procurement auctions to estimate the complementarity between different products from the producer's point of view. A network can be drawn where nodes are products and edges are relative frequencies of products being sold by the same firm. Then, edge weights - how common we observe two products being sold by the same firm - can be thought of as the complementarity between two products. Additionally, by exploring the order in which auctions happen, I can explore the progression of firms over the product space: if it is common for one product to be sold before the other, we can think of the first as a possible steppingstone towards the latter.