To maximize joint profits a cartel must determine the level of output at which

3 The durability of cartels

The cartel problem involves surmounting both external challenges (production by nonmembers) and internal problems (calculating the optimal cartel production, allocating production, detecting cheating, and deterring cheating). In this section we examine factors which seem to influence the ability of producers to meet these challenges. In short, we examine conditions that favor cartel formation and factors which contribute to cartel durability.

We begin by noting that the answers to these questions are complex and not well understood in any systematic way. Nevertheless, received theory does provide some valuable insights. In a strictly rational world where cartel agreements were enforceable a cartel would be undertaken if and only if the present value of the cartel's joint profits from monopoly pricing exceeded the present value of the expected cost of operation and enforcement. In such a world, the formation of potentially profitable cartels is blocked only by its lack of inventiveness with respect to mechanisms for detection and deterrence.

In reality, the existence of net gains from cartelization will generally not suffice to enable a cartel to be formed. Cartel formation is often hampered by the inability of the potential cartel members to strike a deal, and to manage a deal once it is struck. For an efficient cartel, it is not enough to divide markets or agree on a common price. In order to minimize the aggregate cost of producing the joint profit-maximizing output, it will ordinarily be necessary to devise some revenue-sharing scheme involving side payments. Without side payments, the feasible locus of efficient profit outcomes will contain points inconsistent with joint profit maximization. Thus, an important potential limit on the ability of producers to reach an agreement is posed by contractual difficulties and uncertainty. Differences in sellers' objectives (assuming they are not strict profit maximizers) and their forecasts of market demand contribute to these costs. Where a depletable natural resource is involved in which user costs are substantial, then differences in discount rates and forecasts of future reserve additions compound the problem. Oligopoly theory in the Fellner (1965) and Williamson (1975, ch. 12) tradition recognizes how such inconsistencies prevent cartels from ever simulating pure monopoly outcomes.

In an interesting paper, Cramton and Palfrey (1990) highlight the difficulties posed by asymmetric information about cost and demand for cartel members bargaining about production and revenue-sharing rules. They show that, for at least some common environments attaining in the absence of collusion, the excess payments necessary to induce truth-telling in the revelation mechanism with incomplete cost information are larger than the gains from collusion when the number of industry participants is sufficiently large. The opposite result holds for the common value situation of asymmetric demand information.

Even though theory indicates that cartels fall short of complete joint profit maximization, they can extract monopoly rents of nontrivial magnitudes, and the industrial organization literature has endeavored to identify the structural conditions and sellers' strategies capable of sustaining a noncooperative market bargain. It emphasizes the importance of market structure and the fewness and similarity of producers. Few sellers are a necessary but not sufficient condition for collusion; many sellers are a sufficient though not necessary condition for competition. It also emphasizes the importance of inelastic demand, not so much because it implies that the monopoly price premium is larger the more inelastic the demand, but because it suggests the absence of competition from outside the industry that might tend to upset and undo the collusive deal worked out among the industry participants.

Economic theory emphasizes the fragility of cartels in the face of the temptation which every member has to cheat by surreptitiously providing the market with a little extra production. The size of the temptation swings on at least three classes of factors, which we now examine.

The first is the behavior of short-run marginal cost in the neighborhood of the individual firms post cartel level of output. If the gap between marginal cost and price is large and if marginal cost continues to fall for the individual producer, the per unit profit is likely to be substantial and the incentive to cheat enormous.

The second factor is the elasticity of the individual producer's demand curve. This determines the responsiveness of sales to whatever discounts the cheater offers to move his output. Obviously, the more elastic the demand, the less the discount that must be offered and the greater the incremental profit. The less the product differentiation, the higher the elasticity. The ability of the cheater to price discriminate also affects the profitability of this behavior. If clandestine price cuts can be offered to lure new buyers while preserving the price structure in place to existing buyers, then cheating is especially seductive to the producer.

A third influence on the incentive to cheat is the probability of detection and the costliness of the punishment which the other cartel members are able to impose. Orr and MacAvoy (1965) have developed a model in which the price information is transmitted only with a lag so that the price cutter enjoys some increased profits before discovery, although reduced profits afterward. If enforcement takes the form of matching the cheater's price cuts, the potential cheater can calculate the optimal price cut. The present value of the profits expected from cheating will exceed those of remaining loyal if the lag before detection is long enough. Besides the lag, the price cutter's expected return depends on the likelihood of detection, which depends upon the form in which information passes through the market. The ‘trigger price' literature [most notably Stigler (1964), Green and Porter (1984) and Abreu, Pearce and Stacchetti (1985)] reinforces the difficulty of collusion when prices are imperfectly known.

A system of open price quotas often confounds cheating as the same price will generally have to be offered to all buyers and will become known to all sellers simultaneously. If price quotations are made on a customer-by-customer basis, several outcomes appear possible. The buyer with the special deal may wish to cooperate in keeping it secret, fearing that if other buyers hear of it they will demand equally favorable terms. This is likely if the cartel commodity is an intermediate product, the acquisition price of which affects the buyers' competitive position in downstream markets. It is possible, however, that the buyer may judge the best course of action to be playing one cartel member off against another in the hope of getting a better price. This runs the risk, however, of providing other cartel members with the information they need to share in order to discipline the cheater. Obviously, the most likely outcome is difficult to predict except upon various assumptions about the internal structure and operating mechanisms of the cartel.

The fundamental tension in cartel arrangements is that there are incentives for firms both to form and to undermine cooperative institutions. There has been a great deal of attention paid in the last few years to models that exploit these incentives to predict when in the business cycle cartels are most likely to occur. Stigler (1964) and Green and Porter (1984) note that when prices are not directly observable and demand is subject to random fluctuations, undercutting an agreement on prices is observationally equivalent to an inward shift of the demand curve. When cartel members see market price falling below a certain level (the trigger price), they rationally respond by expanding their own outputs for a finite period of time, even if there has been no cheating and members know there has been no cheating! Despite the eventual occurrence of this unwarranted punishment, a trigger price scheme is attractive to the cartel since a sufficiently long reversion phase will deter cheating. The empirical implication of the trigger price mechanism is that cartel agreements are more likely to break down when demand is falling, as during business cycle downturns.

Recently, Rotemberg and Saloner (1986) have argued that they expect exactly the opposite result: cartels are more likely to fail when demand is rising since this increases the benefit from cheating. Thus, while theory has identified economic fluctuation as an important determinant of cartel durability, it is unable to predict the direction of incentives in this case; we must rely on empirical evidence to sort things out.

In an important test of these competing theories and other predictions about cartel behavior, Suslow (1992) examined the durability of international cartels in 45 industries between 1920 and 1939. During this period cartels were often overt, with European firms taking the lead due to relatively lax antitrust laws. Most of the cartels considered were governed by formal contracts between members, with United States participants being an occasional exception.

Suslow finds that the average non-censored cartel episode lasted 2.8 years, and the longest episode was more than 13 years. Censoring in this case means that the cartel agreement was cut short by an exogenous factor. Table 1 lists the reasons, endogenous and exogenous, for the demise of cartels studied by Suslow.

Table 1. Reasons for termination of cartel contracta

The number of firms comprising the cartel appears to influence durability. Information on firm membership was available for 41 episodes; of these 64% had five or fewer members, and 83% had 10 or fewer members. Further, “formal cooperation, rather than tacit coordination, is chosen in markets with relatively few firms.” [Suslow (1992) p. 12]

Another significant finding of Suslow's study is that longer-lived cartels tended to employ more complex and specialized governance structures. In particular, there seems to be a requirement that successful cartels put production quotas and punishments into the contract.

Finally, economic activity appears to be an important factor governing cartel durability. Using indices of industrial production for the United States, the United Kingdom, and France, Suslow concludes that economic volatility, or positive and negative ‘surprises', contributes to the collapse of cartels. Her results also indicate that sub-trend economic activity tends to decrease cartel survival. Suslow's investigation thus lends empirical support to Green and Porter's characterization of cartel behavior.

5.2 OPEC as a “Commodity Producer Club” But Not a Cartel

The cartel theory states that there are seven characteristics that must exist in a group of producers in order to be labeled a cartel: A cartel must assign quotas to its members, monitor members to avoid violations, punish violators, target a minimum price, take action to defend the price, have a large market share, and divide votes among members based on their market share. OPEC did not meet any of these characteristics during its peak of power between 1973 and 1981. OPEC assigned quotas in 1982 but failed to enforce them. OPEC did not set a price band until 2000. It might have been able to defend prices since 2001 with the help of technical, political, and natural factors that prevented many oil-producing countries from increasing production. OPEC never had a punishment mechanism or an effective monitoring system such as that of the Texas Railroad Commission, which used the National Guard and Texas Rangers to enforce its rules and cut production in Texas and Oklahoma. In addition, OPEC's market share is relatively small (approximately 40%). Even today, each OPEC member has one vote regardless of its production and reserves. Unlike the Seven Sisters, OPEC did not divide the world market and did not control investment. The market share of the Seven Sisters was more than double that of OPEC.

By U.S. standards, OPEC does not fit the description of a cartel. Monopolization has two elements: the acquisition of monopoly position and the intent to monopolize and exclude rivals. Official U.S. judicial records indicate that a monopoly position has been associated with companies whose market share approached or exceeded 80%. This is double the current OPEC market share and much larger than the 54% market share achieved in 1974. OPEC has thus failed, by a wide margin, to acquire a monopoly position. As for intent to acquire a monopoly and to exclude rivals, OPEC has never taken the steps mentioned previously to be a successful cartel. OPEC was labeled a cartel because it supposedly increased prices, but how could OPEC exclude rivals by increasing oil prices? Figure 1 shows that higher oil prices bring higher production, more producers, and more competition. OPEC did not act as a monopolist. OPEC has increased production many times in recent years to stabilize prices. It has become an active participant in the world community through negotiations with consuming countries and active aid programs to developing countries. Many researchers today, including lawyers and policy makers, view OPEC as an agent of stabilization in a volatile world. In fact, the Bush administration praised OPEC for its efforts to prevent prices from increasing to record levels during the invasion of Iraq.

None of the statistical tests in the literature support the cartel model for OPEC. Researchers who conducted these tests concluded that OPEC is a “partial,” “weak,” “clumsy,” or “loose” cartel. Such conclusions are mere journalistic expressions. No known theory in the economic literature supports them. The results of so-called statistical tests are not acceptable. For example, applying the same tests to non-OPEC countries shows that non-OPEC producers fit the cartel model better than OPEC members. In addition, these tests focus on parallel behavior, not on cartelization. Parallel behavior may exist in any market, including competitive ones.

The cartel model requires OPEC to operate on the elastic portion of its demand curve. Studies show that OPEC operated on the inelastic portion of its demand curve during its power peak in the 1970s.

Coordination among governments and commodity producer clubs is well-known. This coordination is not labeled as cartelization, despite periodic meetings and clear goals to curb production and maintain prices.

Researchers who attempt to prove that OPEC is a cartel by focusing on the difference between marginal cost and the price of oil ignore the basic principles of natural resource economics. The price of a nonrenewable resource in a competitive market reflects the marginal cost and the discount rate chosen by the producer. Therefore, in a perfectly competitive market the price of any natural resource is higher than its marginal cost, unlike renewable resources, the price of which equals the marginal cost. Hence, focusing on the difference between marginal cost and price to prove market power is conceptually wrong unless the researcher has perfect estimates of the user's cost.

Researchers who believe that OPEC is not a cartel argue that there are other explanations for OPEC wealth transfer that are not related to cartelization. Some researchers argue that it was both the market power of Saudi Arabia and U.S. price controls that led to wealth transfer to OPEC. Various studies show that U.S. oil price controls suppressed U.S. production, increased the world demand for oil, and raised Saudi Arabia's output. This unintended effect enabled OPEC members, mainly Saudi Arabia, which increased its production substantially in the late 1970s, to transfer wealth from the consuming countries to the producing countries.

Historical records indicate that OPEC was reacting to economic and political events rather than being a proactive cartel that consistently followed polices that would maximize its wealth. OPEC reacts pragmatically to compromise on the conflicting interests of its members. In addition, OPEC pricing has always followed the spot market.

Conclusions

Although the cartel-versus-fringe model has been solved for the most relevant equilibrium concepts, a lot of work has still to be done. The assumptions underlying the analysis that is described are restrictive. Constant marginal extraction costs are less appealing and less realistic than stock-dependent extraction costs. A linear and stationary demand schedule for oil is also rather special. Moreover, one may wonder whether the Hotelling rule, underlying the entire analysis, is an adequate description of the behavior of oil suppliers. Adding complexity to the model is a scientific challenge. But there are also empirical challenges. For example, one would like to know whether the actual behavior on the oil market bears resemblance to the theory outlined above. An important reason for continued work on the cartel-versus-fringe model is also the strong relationship between climate change and emissions of CO2 as a consequence of burning fossil fuel. Most of the existing integrated assessment models do not take the exhaustibility of fossil fuels into account, nor do they recognize the imperfections of several nonrenewable resource markets. So, a further study of the cartel-versus-fringe model, with an integration of fuels such as coal, as well as the introduction of backstops, such as solar and wind power, would constitute a valuable and relevant extension.

6.6.2 The instability of cartels

The so-called ideal distribution of overall production among the members of a cartel rarely happens in practice. Very often the allocation is decided on the basis of past sales levels, or on the basis of the production capacity at the time the collusive agreement is signed. In other cases the market is divided on the basis of the geographic distribution of consumers. In such cases both the price and quantity of the product produced may well differ from firm to firm. In general, however, the distribution of profits is an essentially conflict-ridden problem, and laborious negotiations are inevitable to resolve it. This is why cartels are often unstable.

Another important source of instability in cartels arises from the incomplete information of its members. In particular, as we know from the Prisoner's Dilemma game, in many situations of oligopolistic interaction each player retains a unilateral incentive to act opportunistically. It is true that a cartel is a solution to a cooperative game – one that stipulates a binding agreement, with sanctions in the case of defection – but it is also true that in reality the actions of the partners are not precisely observable. That leaves room for opportunistic, self-interested actions that undermine the cartel's cohesion. Let's look at a simple example in Fig. 6.5 that illustrates this possibility.

Consider a group of producers that has set the cartel's equilibrium on the basis of the estimated product demand (curve D) and its production costs – that is, price pM, total quantity qM, and the individual production quotas. If all producers operating in the sector belong to the cartel, then the result coincides with a monopoly equilibrium.

Now imagine a cartel member that, in accordance with the agreement, tries to sell the quota assigned to it; it expects to be able to sell at the unit price pM. Suppose instead that the market is inclined to absorb that quota only at a price p < pM. What might happen? There are three possibilities:

one or more cartel members may have brought more to the market than they were assigned, such that the actual supply is equal to q∗, and the resulting price from the intersection of curve D and the inflated supply curve S″ is equal to p∗;

no cartel member violated the agreement, but demand fell (or the estimate erred on the high side), so the real demand curve D″ intersects the supply set by the cartel (curve S) at point A;

a member produced more than expected and demand decreased, so the new demand curve (between D and D″) and the new supply curve (between S and S″) intersect at one of the points on segment AB, all of which are of the order of p∗, the observed price. Now if one of the “loyal” members believes that the reason for the price drop is the first or third possibility, it will find it opportune to violate the agreement by also trying to produce and sell more than what is permitted by the agreement. In other words, the imperfect observability of the relevant economic variables and of the cartel partners' behavior can incentivize actions that lead to the cartel's dissolution.

6.2 Latin American Drug Cartels

Less tied currently to financial markets, the drug cartels involved many reputed institutions in money-laundering scandals. The cases of Wachovia and HSBC investigated recently are overwhelming due to the enormous size of illegal transactions.

In 2010, US authorities brought criminal proceedings13 against Wachovia, which settled for 110 million dollars in forfeiture, for allowing transactions later proved to be connected to drug smuggling, and incurred a 50 million dollar fine for failing to monitor cash used to ship 22 tons of cocaine. More shockingly, and more importantly, the bank was sanctioned for failing to apply the proper anti-laundering strictures to the transfer of 378.4 billion dollars, an amount equivalent to one-third of Mexico’s gross national product, into dollar accounts from currency exchange houses in Mexico with which the bank did business.

In December 2012 HSBC, Europe’s biggest bank,14 was accused of failing to monitor more than 670 billion dollars, almost twice the amount Wachovia laundered, in wire transfers and more than 9.4 billion dollars in purchases of US currency from HSBC Mexico. The bank settled for a historic top penalty of 1.25 billion dollars forfeiture and 665 million dollars in civil penalties under the settlement.

Obviously these cases concern wire transfer and international Forex transactions, but with the pressure of controls on these types of operations, cartels might look in the future to launder through more complex methods involving securities or financial markets.

6 Outlook

The European Commission recognized the anti-trust and the cartel type behavior and started an ongoing investigation of the EURIBOR manipulation. However, a full-fledged picture of all the consequences has not yet been seen by any regulatory body, since in many cases of benchmark manipulation only the point in time impact of the trading book of the concerned banks are overseen by the regulators and investigators. The full effect on the real economy is often bigger and more important. In the case of the EURIBOR, the manipulation came at a time when the budgetary coordination of the Eurozone was weak and a small variation of the interest rate could have massive consequences in the way policies were designed and economic decision were made.

11.4 Heuristics, hedging, and bias in subjective estimation

The pressing economic and policy issues stemming from the OPEC cartel were manifest in two resource appraisals of national scope: NURE for uranium [US DOE (1980)] and the oil resource appraisal resulting in USGS Circular 725 [Miller et al. (1975)]. Both of these appraisals employed (1) subjective geological analysis as a means to estimating the magnitude of undiscovered deposits, and (2) the description of quantitative estimates by subjective probabilities. Because of the urgency of the policy issues that were examined with regard to estimates of the quantity and quality of undiscovered deposits, methodologies were scrutinized more carefully than ever before.

Initially, subjective probabilities of an expert were accepted as useful, first because of his expertise and second because of the seemingly reasonable assumption that man functions well as an intuitive statistician, meaning that the expert's perception of likelihood of an event corresponds generally to its relative frequency in nature. Moreover, Delphi methods had become accepted means for treating differences of opinions of experts, having taken for granted that convergence of opinions of members of the group to a single value provided a desirable and useful resolution of differing expert opinion. These perceptions were seriously challenged by the work of Tversky and Kahneman (1972, 1974) and Sachman (1974).

Tversky and Kahneman (1972, 1974) identified some of the heuristics used by man to estimate an uncertain event: anchoring and adjustment, representativeness, and availability. These heuristics were found to lead to predictable biases. Psychometricians found that man generally behaves as though he knows far more than he actually does, tending to provide subjective probability distributions that are narrower than they should be by approximately 50% [Slovic (1972)]. This tendency results from the inability of man to image mentally the combinations that result in extreme values of the random variable; heuristics used to approximate actual calculations understate probabilities for events in the tails of the distribution when stochastic events are conjunctive.

Sachman's (1974) investigation of traditional Delphi as a means of inquiry found that repeated cycles of estimate and statistical feedback violate blatantly the condition that estimates by individuals be independent in order for statistics such as mean and variance and for relative frequencies of estimates to be statistically meaningful. Moreover, he documented that convergence can be solely a psychological phenomenon and usually is not the result of exchange of factual information and logical deliberation.

These findings led to (1) the total elimination of Delphi or the replacement of traditional Delphi by modified Delphi, and (2) the structuring and support of subjective probability estimation so as to improve the use by the expert of his science and available geodata and to minimize the effects of heuristic biases.

4.1.2 Australian Focus

The match-fixing connection seems to be dominated by Asian cartels that have a well-established worldwide network with ties to various competitions and sports. A good example is the investigation of match rigging in the Victorian league.10 In 2013, the Australian police arrested 10 people as a result of a match-fixing investigation that was prompted by information provided by the Football Federation of Australia. Segaran Gsubramaniam, a 45-year-old from Malaysia known as “Gerry,” appeared to be the brain behind a match-rigging enterprise involving several British footballers and was allegedly involved in fixing games while playing for the Melbourne-based Stars in the second-tier Victorian Premier League. The match-fixing allegations related to the Stars’ last four games, in which they conceded 13 goals without scoring. They lost 16 of their 21 league matches that season, winning one and drawing four. The ring

leader Gsubramaniam acted as contact between Stars players and betting syndicates and placed bets that led to winnings of more than 1.17 million pounds. This fact proves that even in second-tier leagues the betting turnover can be significant.

8.1.2 Rule 394 and the third market

In 1970, NYSE Rule 394 prohibited member firms of the NYSE from trading outside the NYSE either as agent or as principal. Member firms, acting as agents, could not send customer orders to other markets (other than regional exchanges), nor could they trade with customers as principals outside the NYSE. This rule had the beneficial effect of forcing all orders to interact in one market – the NYSE, but it had the harmful effect of limiting competition from new markets. Over time, regulatory pressure weakened Rule 394 and caused it to be abolished in 2001. First, in 1976, the rule was changed to Rule 390, which permitted trades, where the NYSE member acted as agent, to be executed off the NYSE. This modification gave rise to a new third market as member firms sent customer orders to third market makers (such as Madoff and Co.) that promised to match NYSE prices. In addition the third-market-maker paid the broker for the order flow. The new third market specialized in the order flow of small, uninformed, customers in contrast to the third market of the 1970s, which was an institutional market to avoid high commissions.

Second, Rule 390 was weakened by SEC Rule 19c-3 that exempted any stocks listed after April 1979 from application of the rule. Under 19c-3, a NYSE member could trade with customers as a principal and could therefore make in-house market in eligible stocks, but, surprisingly, few members set up in-house markets in listed stocks. Finally, Rule 390 was abolished by the NYSE in 2001 because of SEC pressure and because the rule had become ineffective.

Thus by the year 2001, two of the key features of the NYSE cartel – fixed commissions and the restrictive Rule 394 – had been abolished. The one remaining feature of the cartel – limited direct access for the 1366 members – remains. The privilege of membership continues to have substantial value as NYSE seat prices in 2000 exceeded $2 million. Members are of three types:21 specialists (about 450), independent floor brokers (about 525) and floor brokers for retail firms (about 330). Specialists trade for their own accounts as market makers and keep the book of limit orders. Independent floor brokers receive commissions for executing customer orders. Floor brokers that work for retail firms execute the portion of the firms’ order flow that is not routed through the electronic DOT system.

2.7 Brokerage Operations

Until 1975, brokerage firms in the United States functioned as a rather exclusive cartel, fixing commissions and offering clients a wide array of services ranging from providing trade executions and margin opportunities to offering advice and management services. Brokerage firms were all said to be full service, and included firms such as Merrill Lynch, E.F. Hutton, Paine Webber, and Smith Barney. The industry was forced as of May 1, 1975, acting on SEC Rule 10b-3 to break its cartel and compete with respect to brokerage commission levels. Numerous discount brokers opened operations, competing against full service brokers by offering trade executions with lower brokerage commissions. Discount brokers initiating operations in the 1970s included Quick & Reilly and Charles Schwab. The Internet’s development in the 1990s and shortly afterwards created opportunities for Internet brokers such as Ameritrade, ScottTrade, Interactive Brokers, and E-Trade to further reduce commissions and execute transactions for clients without actually speaking live with brokerage account representatives.

Investors wishing to open brokerage accounts need to consider their trading needs and interests to properly select their brokers. Investors needing substantial advice and counsel along with other services may find their needs better fulfilled by full-service brokers such as Merrill Lynch, Raymond James, or UBS. In addition, there is some statistical evidence (discussed earlier) suggesting that larger full-service firms better provide for transaction price improvement (executing the client’s transaction at a price that is better than the best bid or offering at the time of the transaction). Investors who are comfortable with their trading experience and who expect to execute smaller trades, particularly more frequent smaller trades, might prefer the lower commissions offered by an Internet broker such as ScottTrade or T.D. Ameritrade. These on-line brokers will often offer low flat-rate commissions, sometimes even dropping to zero, particularly when the brokerage firms receive payment for order flow (rebates).

J.D. Power and Associates regularly provides broker quality information based on its customer satisfaction surveys. Important evaluation criteria include commission levels, percentage of orders showing price improvement, frequency of limit order executions, margin and cash interest rates, and customer satisfaction. Barrons conducts similar surveys. Such surveys are useful in making broker-selection decisions.

Most stock orders in the United States are placed through stockbrokers who are compensated with commissions. These commissions are typically charged as a fraction of the dollar volume of the transaction and as a function of the number of shares associated with the order, or in many instances, a simple flat commission. Sometimes, the broker’s commission will be a function of the bid-ask spread; that is, the difference between the lowest price a trader is offering for the security and the highest price an investor is willing to bid for that security. OTC markets and particularly exchanges will provide for trading priority rules, which designate which orders will be filled or the sequence in which orders will be filled. Frequently, priority rules are designated according to transaction price, size, location of origination of order timing, and so on.