On Market Power And Cartel Detection: The South African Flour Cartel

Abstract

For several different reasons, competition authorities are expected to identify the source of market power, which may not be the result of anti-competitive conduct. Empirical techniques could help to gauge the degree of market power, and firm conduct. We ask the question, do observed prices and quantities in the flour industry reflect switching from collusive to non-collusive behavior and test for this empirically using the monthly time series data from September 2003 to December 2008. Within the framework of a structural model of equilibrium pricing, we specify a simultaneous equation switching regression model in which the parameters of the demand and cost functions are estimated. Our estimated conduct parameters suggest that the level of market power exercised by the flour millers is quite low in both periods. The perfect collusion hypothesis is rejected by the data. In addition, the estimated level of the conduct parameter diverges from those implied the models of competition and therefore we cannot define precisely the firm behaviour.

Notes

¹ The NEIO approach addresses concerns with the structure-conduct-performance paradigm which treats market structure as exogenous. The base hypothesis of structure-conduct-performance paradigm is that more concentrated industries imply less competitive conduct, and hence better performance. The NEIO approach stresses that: (1) one should be careful with the use of accounting cost data; (2) focus on industry-level analysis, and account for institutional idiosyncrasies of the industry; and (3) identify causal relationships rather than correlations. For example, profits and concentration are jointly determined. A correlation between profits and concentration do not necessarily reflect collusive conduct, but rather significant differences in efficiencies among firms. Econometrically, there is an endogeneity problem.

² The intuition is simply that consumer behaviour governs the ability of firms to raise price above marginal cost. If consumers are very price sensitive, (for example a large number of consumers stop purchasing the product in response to a price increase), firms in the industry are unable to exercise market power. As a result, the unit price-cost margin is small.

³ Within the framework of a structural model of equilibrium pricing, for a homogenous good market, we specify a simultaneous equation switching regression model in which the parameters of the demand and cost functions are estimated.

⁴ See for instance Verboven (1997).

⁵ The main attraction this model is that it results in a two-equation linear system that explain equilibrium industry price and quantity data. The two reduced-form equations show explicitly how the two endogenous variables (and) are jointly determined, and how changes in the exogenous variables affect the endogenous ones. Fortunately, we have industry data on prices and quantities. In competition law practice and some academic writings, however, it is easier to obtain pricing data than quantity data, and therefore common to see only the reduced-form price equation. That is, data on prices and the independent variables are used to estimate values for the reduced-form regression parameters. Then, by using the actual values of the independent variables and the reduced-form regression coefficients, predicted values for prices, as well as but-for prices, can be generated for the calculation of overcharges.

⁶ This approach, while reducing the number of estimating equations, is not without limitations. In aggregating the first-order conditions, one cannot estimate separate conjectural and cost parameters for each firm and time period. To reduce the dimensionality of the parameters in the industry supply function, Porter (1983) assumes that the firm-level values of θ_t times the associated market shares are the same constant. This assumption has the important computational advantage of reducing the number of conjectural and cost parameters to two. It makes it easy to calculate equilibrium prices and quantities in perfectly competitive and monopoly (collusive) markets. However, this simplifying assumption has some disadvantages. For example, it is unclear why conjectural parameters should vary inversely with market shares.

⁷ See Baum, C.F., M.E. Schaffer, and S. Stillman, ivreg2: Stata module for extended instrumental variables/2SLS, GMM and AC/HAC, LIML and k-class regression. See http://ideas.repec.org/c/boc/bocode/s425401.html.

⁸ A valid instrument should meet three requirements; (1) the instrument must be correlated with the endogenous variable, conditional to the other covariates; (2) the instrument cannot be correlated with the error terms in the explanatory equation and (3) Exclusion should also be applied when instrumental variables are selected to differentiate the supply and the demand equation. An exogenous supply shifter does not affect demand except through its effect on price and can be used as a valid instrument for price in the demand equation. Similarly, an exogenous demand shifter does not affect supply except through its effect on price and can be used as a valid instrument for price in the supply equation.

⁹ Food and Agriculture Organisation (FAO) (2013). FAOSTAT-Agriculture database. Food and Agriculture Organization. [online] Available at: http://www.fao.org

¹⁰ Assume a linear inverse demand function as below, where Q_t is demand in period t, P_t is price in period t and X_t is an exogenous demand shifter.

The supply function can be written as:

where Wt is the exogenous cost shifter and Q_t is the quantity in period t. We solve the identification problem by assuming θ_t = {θ^c, θⁿ} takes two values, θ^c indicates the value under a collusion regime and θⁿ indicates the value under a non cooperative regime.

¹¹ To illustrate, suppose the industry level conduct parameter estimate is 0,5 whereas the Cournot is 0,25 (in a four-firm market), the difference may simply be due to misspecification

¹² For competition authorities, the objective of detecting a cartel is not to identify industries with high price-cost margins but rather to uncover prosecutoriable cases of collusion.

¹³ Other reasons for a high price-cost margin are greatly differentiated products, production technologies protected by patents and trade secrets, and high search costs for consumers; and one suspects these are much more ubiquitous than collusion.

¹⁴ Applications include Porter (1983) and Ellison (1994).