VoxEU Column Competition Policy

Cartels uncovered

Should cartels be regulated? This column outlines a new economic toolkit that models the creation of cartels and compares these predictions with real-life observations. Focusing on forty years of postwar data from Finland, this column finds that once cartels are formed, they are long-lasting. If unregulated the amount of cartels will only increase.

Competition policy matters. The US government, as well as the EU and its member states, spend large amounts on shaping and executing competition policy – known as antitrust in the US. The private sector spends correspondingly large amounts in shaping the outcome and participating in the law battles.

But strangely enough, there is no assessment of the effectiveness of competition policy. Recent experiences from the introduction of “leniency policies” – basically the EU’s attempt to put cartel members in a prisoner’s dilemma situation – suggest that competition policy has been successful in detecting cartels. Despite this, there is little understanding of how large the problem of illegal anticompetitive practices is under the current regimes, or how things would differ if some alternative policy was introduced.

A lack of tools

A key reason for this state of affairs is that we lack tools to study these questions. This is not because economists haven’t been aware of the importance of the questions, but has more to do with the fact that the limited amount of data we have available on cartels is generated in a very peculiar way, namely through some form of competition policy. As an example, consider the situation that existed in many European countries up to the Second World War: Cartels were legal, sometimes even encouraged. In such circumstances it was next to impossible to know how many cartels there were, as there was no reason to monitor them. In the current regime, observing cartels is difficult for exactly the opposite reason – because they are illegal and punishments are severe, cartels try to hide as well as they can.

Economists have started to address these questions with renewed vigour. In their two papers, Joseph Harrington and Myong-Hun Chang (Harrington and Chang 2009, Chang and Harrington 2009) have built dynamic game-theoretic models where cartels are born and where they die. Their models allow for different types of competition policies, including the modern ones with detection, leniency, and punishment. These models generate predictions on how the number of cartels would develop under different policies and under different parameter values that govern the birth, stability and death of cartels. At the same time Miller (2009) has asked what the introduction of leniency in the US can say about the prevalence of cartels.

In recent research (Hyytinen et al. 2010), we try to tackle the issues head-on. The objective of our research is two-fold. First, we aim at building a toolkit that will enable economists to answer questions of the following sort: How many cartels are there? How many would there be if competition policy was different – if there was no leniency, or the fines were higher? Second, we use the toolkit to study Finnish cartels from a period – from 1950 to 1990 – when cartels were legal. The reason to study such a legal cartel regime is that it allows us to understand the counterfactual world of no modern competition policy. Such an understanding is an essential ingredient when one wants to assess the efficiency of modern competition policy. The second reason to use this particular data set is the wealth of information we have access to, because Finland (like many other countries over some period), had a Registry for legal cartels. While being a great source of information, the Registry is incomplete. Not all cartels were registered, and those that were, were not necessarily in the Registry for their whole lifespan.

The tool we build has two key parts:

  • First, an economic model of cartel behaviour. While a number of models could be used, we build on the work of Harrington and Chang because in their models, cartels are “really” born, and they really “die”.
  • Second, we use a statistical tool called Hidden Markov Models.

Hidden Markov Models, while widely used in a number of fields including computer science, speech recognition and gene sequencing, have been rarely used in economics. The models have two parts. The hidden part models the world as it really is – in our case, an industry being in a cartel, or not being in a cartel, or being exposed by the competition authority, or a whistle-blower having just blown the whistle. The second part is what we – the outside observers – observe, and how this is linked to what the world is really like. These observations include also the possibility of observing nothing, i.e. not knowing whether a particular industry had or did not have a cartel in a particular period of time.

Matching these two ingredients, a theoretical model of cartel behaviour, and a statistical model of how the data is generated, yields our tool. This tool can be used to analyse the kind of data modern day competition authority actions generate. Most of the time we do not know whether firms in a given market (or industry) are in a cartel or not. Every once in a while a cartel is exposed, and investigations and court proceedings may eventually establish the years in which it existed.

We show how the model can be tailored to fit a period when cartels were legal, and take the model to data on Finnish legal cartels. Our aim is to spot how many cartels were in Finland in the manufacturing industries between 1950 and 1990.

According to our analysis, Finnish manufacturing was essentially not cartelised after the Second World War. Slowly, the proportion of industries that had a cartel increased, and jumped up significantly in the early 1970s, reaching over 90%. This jump coincides with a positive GDP shock that the Finnish economy experienced in the early 1970s. Ignoring small fluctuations, the degree of cartelisation did not decrease after the jump and may have increased slightly during the last years of the 1980s. Our analysis suggests that by the end of our observation period, nearly all Finnish manufacturing industries had a cartel.

We study the jump of the early 1970s in some detail and are able to show that the relentless process that generates new cartels is even more important than the jump. This proves important because legal Finnish manufacturing cartels turn out to be very long-lasting. A cartel had a life-expectancy of 20 years in post-war Finland. This means that there was very little outflow from cartels and therefore even a modest inflow would have eventually lead to the same situation – a cartel in nearly all industries.

Our approach is not, without extra effort, able to say much about the welfare damage wrought by a cartel. But our results do strongly suggest that competition policy deserves the high priority it has lately been given.


Chang, Myong-Hun and Joseph E Harrington Jr (2009), “The Impact of a Corporate Leniency Program on Antitrust Enforcement and Cartelization”, mimeo, John Hopkins University.
Harrington, Joseph E Jr and Myong-Hun Chang (2009), “Modelling the Birth and Death of Cartels with an Application to Evaluating Antitrust Policy”, Journal of the European Economic Association, 7(6):1400-1435.
Hyytinen, Ari, Frode Steen, and Otto Toivanen (2010), “Cartels Uncovered”, CEPR Discussion Paper 61.
Miller, Nathan (2009), “Strategic Leniency and Cartel Enforcement”, American Economic Review, 99(3):750–768.

105 Reads