Statistical Methods for Detecting Cartels

In a joint research project, David Imhof (Swiss Competition Commission and Ph.D. candidate at the Department of Economics) and Prof. Martin Huber applied so-called machine learning algorithms to detect illegal cartels emerging in public tenders of the Swiss construction sector. The method aims to learn from specific patterns in the bids of firms participating in tenders whether a cartel is likely in place or not. To this end, the authors constructed specific statistical indicators or “screens” related to the distribution of bids and assessed how well these screens predict the presence or absence of cartels in Swiss construction data. The machine learning approach was found to correctly predict the presence or absence of cartels in more than 80% of public tenders. It therefore appears to be a potentially valuable tool for competition agencies to detect and fight cartels. Further information on the method and the results can be found in the SES Working Paper 494: “Statistical Methods for Detecting Cartels” by Martin Huber and David Imhof.

For more information on the topic, see