The Determinants of Cartel Duration

Abstract

In this paper we model cartel duration as a mixed proportional hazard model and condition on cartel characteristics such as the agency first detecting the cartel, industry, if it is a bid rigging or price fixing cartel, the number of countries affected, the affected sales, and measures of the economic cycle and trend. Results are intuitive and fairly consistent across models, and conform well with theory and prior empirical work. We also found that the model results are sensitive to the presence of unobserved heterogeneity.

Among other results, we find that cartels first detected by United States or European Union agencies tend to be longer-lived, likely because those detected by other jurisdictions are primarily follow-ups of related larger and older cartels first uncovered in the United States or Europe. Bid rigging cartels tend to be longer-lived than others, while cartels distributed across geographies tend to be shorter-lived. Cartel durations are increasing in the size of a cartel’s affected sales and sanctions. Industries such as Petroleum & Coal, Finance & Insurance, and Food, Feed, Tobacco & Transportation have shorter-lived cartels, while industries such as Electronic Products have longer-lived cartels. The state of the economy can impact the duration of a cartel as well. Cartels where the leading firm has a market share of at least 40% have longer durations. However, the wide variation in the unobserved frailty factor suggests that additional, significant covariates remain unaccounted for in our information set.

Keywords: Cartel Duration, Screening, Mized Proportional Hazard Model, Collusion, Detection, Screening

JEL Classification: C41, K14, L14