Cartel Screening in the Brazilian Fuel Retail Market

We aim to contribute to the challenges faced by the antitrust authorities in the identification process of anti-competitive market behavior. We propose two econometric models to select possible cases to be investigated: (i) The Markov-Switching GARCH (MSGARCH) Models; (ii) The Local Gaussian Correlation (LGC) approach. We compare both models. Our results indicate that the LGC model, based on the correlation of the resale price margin and the variability of prices, may provide a biased estimation of the likelihood that a market is practicing cartel. The MSGARCH model, based only on the log deviation of the gasoline sale price, showed a better accuracy in cartel detection