604 Interational Organization world oil prices is fraught with potential endogeneity.34 Low oil prices might cause OPEC to lower its production quotas,but if OPEC actually has market power,lower OPEC quotas would cause high oil prices.Thus on its own the (lack of)correlation between OPEC quotas and oil prices does not give one enough information to make valid inferences about its status as a cartel.35 Some sophisticated statistical techniques might be used to try to overcome this problem,but they are not satisfying.36 The second reason is perhaps even more important:production constraints are a necessary element of cartel behavior.If OPEC is not constraining its members'production,then it is not a cartel,by definition.37 Focusing on production allows one to directly inves- tigate the extent of collusion between OPEC members,rather than looking at its indir- ect effect on prices.Indeed,even if OPEC was somehow affecting market prices without constraining its members'production,it would not be doing so as a cartel. What evidence should one expect if OPEC is a cartel?Mankiw defines a cartel as a group of firms (or states,in this case)that creates agreements about quantities to produce or prices to charge,and further it"must not only agree on the total level of production but also on the amount produced by each member."38 This definition implies that a gap between market price and marginal cost of production is not by itself evidence of a cartel.39 Instead,one should see signs that the organization is cooperating to restrict production (to drive prices up).One should see the following kinds of evidence:new members of the cartel have a decreasing or decelerating pro- duction rate(first test);members should generally produce quantities at or below their assigned quota (second test);changes in quotas should lead to changes in production, creating a correlation(third test);and members of the cartel should generally produce lower quantities(that is,deplete their oil at a lower rate)on average than nonmembers of the cartel (fourth test).Failure to observe any of these phenomena would cast doubt about OPEC's status as a cartel,although none is totally determinative.The fourth test is perhaps the strongest because it is difficult to imagine how an organization that does not restrict output compared with nonmembers could be called a cartel how else could it increase average prices?40 To preview the results,OPEC fails all four of the tests. 34.Subsequent sections further discuss the link between oil prices and inferences about OPEC as a cartel. 35.A simple bivariate ordinary least squares (OLS)regression between world oil prices and OPEC's aggregate production target from 1982 through 2009 yields an R2 value of just 0.15. 36.To date,no one has identified a plausible instrumental variable or natural experiment.Other approaches exist but have not produced a widely accepted conclusion on the cartel question.See Dahl and Yuicel 1991;Gulen 1996;Alhajji and Huettner 2000:Reynolds and Pippenger 2010;and Bremond etal.2012. 37.ankiw2011,351. 38.bid.351. 39.Producers who stop producing before marginal costs equal market price (like some OPEC producers. possibly)are not behaving perfectly competitively,but that does not necessarily imply cartelization. 40.One of OPEC's stated goals is to stabilize prices.It is possible that an organization could seek to stabil- ize prices without affecting the long-run average price or production levels of its members.Yet such an organization could not be considered a classic cartel because it would not be profit maximizing.Itworld oil prices is fraught with potential endogeneity.34 Low oil prices might cause OPEC to lower its production quotas, but if OPEC actually has market power, lower OPEC quotas would cause high oil prices. Thus on its own the (lack of) correlation between OPEC quotas and oil prices does not give one enough information to make valid inferences about its status as a cartel.35 Some sophisticated statistical techniques might be used to try to overcome this problem, but they are not satisfying.36 The second reason is perhaps even more important: production constraints are a necessary element of cartel behavior. If OPEC is not constraining its members’ production, then it is not a cartel, by definition.37 Focusing on production allows one to directly inves￾tigate the extent of collusion between OPEC members, rather than looking at its indir￾ect effect on prices. Indeed, even if OPEC was somehow affecting market prices without constraining its members’ production, it would not be doing so as a cartel. What evidence should one expect if OPEC is a cartel? Mankiw defines a cartel as a group of firms (or states, in this case) that creates agreements about quantities to produce or prices to charge, and further it “must not only agree on the total level of production but also on the amount produced by each member.”38 This definition implies that a gap between market price and marginal cost of production is not by itself evidence of a cartel.39 Instead, one should see signs that the organization is cooperating to restrict production (to drive prices up). One should see the following kinds of evidence: new members of the cartel have a decreasing or decelerating pro￾duction rate (first test); members should generally produce quantities at or below their assigned quota (second test); changes in quotas should lead to changes in production, creating a correlation (third test); and members of the cartel should generally produce lower quantities (that is, deplete their oil at a lower rate) on average than nonmembers of the cartel (fourth test). Failure to observe any of these phenomena would cast doubt about OPEC’s status as a cartel, although none is totally determinative. The fourth test is perhaps the strongest because it is difficult to imagine how an organization that does not restrict output compared with nonmembers could be called a cartel— how else could it increase average prices?40 To preview the results, OPEC fails all four of the tests. 34. Subsequent sections further discuss the link between oil prices and inferences about OPEC as a cartel. 35. A simple bivariate ordinary least squares (OLS) regression between world oil prices and OPEC’s aggregate production target from 1982 through 2009 yields an R2 value of just 0.15. 36. To date, no one has identified a plausible instrumental variable or natural experiment. Other approaches exist but have not produced a widely accepted conclusion on the cartel question. See Dahl and Yücel 1991; Gülen 1996; Alhajji and Huettner 2000; Reynolds and Pippenger 2010; and Brémond et al. 2012. 37. Mankiw 2011, 351. 38. Ibid., 351. 39. Producers who stop producing before marginal costs equal market price (like some OPEC producers, possibly) are not behaving perfectly competitively, but that does not necessarily imply cartelization. 40. One of OPEC’s stated goals is to stabilize prices. It is possible that an organization could seek to stabil￾ize prices without affecting the long-run average price or production levels of its members. Yet such an organization could not be considered a classic cartel because it would not be profit maximizing. It 604 International Organization