Sunday, December 22, 2019
ABSTRACT: In 2015, a supposed bid-rigging cartel that operated in the Brazilian implantable cardiac devices market was announced and public authorities began to investigate it. This paper evaluates if there is systematic correlation between the bids that are placed by competitors in the sealed phase of procurement auctions, which is a situation that may suggest coordinated and fraudulent behaviour. By applying Moran’s I statistic to the residuals of controlled bid regressions and using a novel and public database, we show that the bids that were placed by the investigated companies have positive and statistically significant autocorrelation. In addition, when we separate the data into two subperiods, namely, the period in which the cartel probably existed (2005-2015) and the period in which the cartel probably did not exist due to the conclusion of a leniency agreement (2015-2017), the Moran’s I statistic only points to autocorrelation in the first sub-sample. Our result has remained robust when we eliminate transitional periods and use alternative economic screens.
The main advantage of the economic screen based on Moran’s I statistic is its low data requirements and computational and statistical simplicity. In addition, the screen is versatile and can be applied to any type of market where public procurement occurs using sealed auctions. However, Moran’s I statistic requires prior knowledge of the identity of the companies that may form the bid-rigging cartel. Without information from documentary evidence, denunciations or leniency agreements, it becomes more difficult to construct the bidding matrix and to apply the screen. A partial solution to this disadvantage is to apply the economic screen to combinations of all sets of companies or apply it only to those with the largest market shares or to the most frequent bidders. Finally, another shortcoming of our economic screen is the possibility of finding the existence of a bid-rigging cartel when one does not truly exist (false positives). This can occur when bids are correlated due to the existence of unobserved variables that influence the placed bids. Therefore, our screen cannot be used as isolated and definitive proof of the existence of a bid-rigging scheme and it is necessary to collect additional documentary evidence.