PurposeThis paper aims to assess all the Brazilian electric power transmission line auctions occurred between 1999 and 2017.Design/methodology/approachA copula-based Roy/endogenous switching regression model is used. The suitability of this model is twofold: it takes into account the selection bias problem involving auctions data and it allows more flexibility in modeling the joint distribution between the unobserved components of the selection and outcome equations; thus, normal distribution assumptions are not needed.FindingsThe main results suggest that stated-owned companies have the highest probability of winning an auction, and there is a non-competitive behavior among the players in the auction. The results also suggest some departure from joint normality in the data.Originality/valueThe copula-based sample selection approach used in this paper is consistent under non-normality and allows one to address different types of nonlinearities in the data such as asymmetry and heavy tails.