Information exchange in cartels

• Author: Yu Awaya,Vijay Krishna
• Date: 01 June 2020
• Published date: 01 June 2020
• DOI: http://doi.org/10.1111/1756-2171.12320

RAND Journal of Economics

Vol.51, No. 2, Summer 2020

pp. 421–446

Information exchange in cartels

Yu Awa y a ∗

and

Vijay Krishna∗∗

Antitrust authorities view the exchange of information among firms regarding costs, prices, or

sales as anticompetitive. Such exchanges allowcompetitors to closely monitor each other, thereby

facilitating collusion. But the exchange of aggregate information, perhaps via a third party, is

legal. The logic is that collusion is difficult if the identity of a price-cutting firm cannot be as-

certained. Here, we examine this logic using Stigler’s model of secret price cuts. We firstidentify

circumstances such that when no information exchange is possible, collusion is difficult. We then

show that if firms’ aggregate sales are made public, nearly perfect collusion is possible.

1. Introduction

Antitrust authorities in most countries try to restrict the exchange of information among

competitors. In particular, the exchange of firm-specific data regarding costs, prices, and sales is

viewed, at the very least, with suspicion. But the sharing of aggregate data, perhaps via a trade

association or some other third party, is usually deemed to be legal. According to guidelines of

the European Commission:

“Exchanges of genuinely aggregated data, that is to say, where the recognition of indi-

vidualised company level information is sufficiently difficult, are much less likely to lead

to restrictive effects on competition than exchanges of company level data.” (European

Commission, 2011)

In the same vein, the US Federal Trade Commission suggests that competitors establish a

“safety zone” in which “the shared statistics are sufficiently aggregated that no participant can

discern the data of any other participant” (FTC, 2014). The Japan Fair Trade Commission too

∗University of Rochester; yuawaya@gmail.com.

∗∗Penn State University; vkrishna@psu.edu.

The research reported here was supported by a grant from the National Science Foundation (SES-1626783). Awaya

and Krishna thank Reiko Aoki, Kentaro Inomata, Jun Nakabayashi, Hiroyuki Odagiri, and Yosuke Okada for helpful

comments, especially concerning antitrust practice in Japan. We also thank Yossi Spiegel for his comments and the

referees and the editor for their suggestions.

© 2020, The RAND Corporation. 421

422 / THE RAND JOURNAL OF ECONOMICS

has ruled that the exchange of information “without clearly indicating the quantities, amounts,

etc. of the individual constituent enterprises … ” is permissible (JFTC, 2015).1

What is the logic of these guidelines? Individual data allow firms to monitor each others’

activities better than aggregates and so restrictions on the exchange of individual data limit the

possibility of coordination and collusion. Economic theory also seems to confirm this view. For

instance, Radner, Myerson, and Maskin (1986) show that in a repeated partnership game, imper-

fect public monitoring—where each participant observes only an aggregate—severely restricts

the prospects of cooperation relative to a situation in which individual choices are perfectly ob-

served. When only an aggregate is observed, individual transgressors cannot be identified. In

technical terms, the “pairwise identifiability” condition of Fudenberg, Levine, and Maskin (1994)

fails.2Leading antitrust scholars, perhaps influenced by the theory, have also claimed that

“[ … ] aggregating the data largely removes the value of information in facilitating collu-

sion.” (Carlton, Gertner, and Rosenfield, 1997)

This seemingly compelling logic notwithstanding, many cartels seem to operate quite suc-

cessfully on the basis of aggregate information alone. Members of the copper plumbing tubes

cartel

“[ … ] set up a new data exchange system—initially on a monthly, later on a quarterly,

basis through the World Bureau of Metal Statistics. WBMS statistics only contained ag-

gregated figures and no company specific information.” (European Commission, 2004)3

In this article, we study a repeated oligopoly with private monitoring in the style of Stigler

(1964). Firms produce differentiated products and compete in prices. Firms’ sales are determined

by all the prices but are subject to individual and market shocks. In the baseline model, firms

cannot observe the prices set by others and so secret price cuts are possible. Rather, firms observe

only their own sales and, as in Stigler (1964), when sales are rather noisy, collusion is hard

to sustain because of the difficulties firms face in monitoring each others’ activities. We then

consider an alternative situation in which a third party—say, a trade association—publicly reveals

aggregate data on sales to all the participants. We assume that the sales data are accurate, perhaps

having been verified by a sufficiently accurate audit.4Weidentify circumstances in which without

any information exchange, industry profits in any equilibrium are low, whereas with the public

availability of the aggregate, there exists an equilibrium in which firms’ profits are close to those

in a monopoly.This result suggests that the exchange of information among fir ms can be harmful

even when it is in aggregate form and the data of any individual participant cannot be discerned.5

Moreover, the potential increase in profits coming from the additional information available to

the firms can be quite large.

How can firms collude using only aggregate data? The key is that whereas individual sales

can be rather variable, aggregate sales are usually much less variable.Thus, if a firm cuts its price

from an agreed upon level, this will cause, with high probability, an increase in aggregate sales

1In a case involving the titanium sponge industry, the JFTC did not allowthe exchange of even aggregate infor-

mation. The reason was that there were only two firms in the industry and so aggregate information could be perfectly

disaggregated (JFTC, 1999).

2This article considers finite games. See Matsushima (1989) for a similar result that applies to games with contin-

uous actions (as in the current article).

3Prior to this, the cartel engaged in the direct exchange of sales figures. Thus the cartel chose to move to a system

where only aggregate information was exchanged. See Sugaya and Wolitzky (2018) for a rationalization of this change.

4In the well-studied lysine cartel case, AC-Treuhand, a Swiss consulting firm, actually audited sales figures (Mar-

shall and Marx, 2008). The sorbates cartel verified reported sales by cross-checking against export figures reported to

the government.

5Formally,as in Radner, Myerson, and Maskin (1986), the pairwise identifiability condition fails in our model.

C

The RAND Corporation 2020.