Thursday, April 10, 2008

Dunning on "endogenous oil rents"

At the comparative politics workshop yesterday, Thad Dunning presented some work in progress on "Endogenous Oil Rents." The puzzle that he is addressing is as follows: oil income is often treated as "manna from heaven" in the political economy literature, but examination of some cases (e.g. Venezuela and Mexico) reveals that the share of oil revenues that governments take for themselves varies over time. Why would leaders choose to limit the share of revenues that they take for themselves? Dunning acknowledges that factors independent of domestic politics account for some of this variation; these include conditions on loans from IFIs, ideology of leaders toward size of government, etc.

But he is convinced that domestic politics also play a role. Here is a simplified version of his logic (he has a much richer model, but the following captures the basics). Suppose that an elected incumbent, call him/her "A", can choose whether the amount of oil revenues that the government can take is "high" or "low", and that this choice cannot be overturned by future leaders. This oil revenue can be used by whomever is in power to buy votes in any forthcoming election. Now consider the possibility that there is some exogenous bias in the electorate that either works in favor or against the electoral chances of A (or A's party) versus some other politician, B. The claim is that when the bias against A is sufficiently strong (i.e. when A is "weak"), A may have incentive to limit the government's take of oil revenues.

Why? Well, the game starts with A choosing whether oil rents should be high or low, weighing consequences over eight possible future states of the world. The first four states are associated with bias being against A's electoral chances. The other four states are associated with bias in favor of A. Consider the first four states, holding the bias against A as fixed. In the future, A is either in power or not, and oil rents can be either high or low. So we have 4 states: (1) A in power, high rents; (2) B in power, high rents; (3) A in power, low rents; and (4) B in power, low rents. For states (3) and (4), rents are low, and so electoral outcomes are determined by the bias in the electorate. So A's chances of retaining office or successfully challenging B are low, but pretty much the same in (3) and (4). For state (2), the electorate's bias combined with B's access to high oil rents means that B will surely retain office. In (1), the oil rents don't do much to compensate for the bias against A in the electorate. Now, note that A's choice at the start of the game determines whether the future alternates between states (1) and (2) or whether it alternates between (3) and (4). In this case, A prefers alternation between (3) and (4), and so A chooses to lock in low oil rents. Therein lies the answer to the puzzle. For the other four states, it is easy to show that A's dominant strategy is to choose high rents. (Try it.)

One comment was that these sorts of counterintuitive outcomes probably occur only rarely, so this probably won't serve as a model of "normal" politics of oil. Another comment suggested that this work echoes work by Terry Moe from the mid-1980s explaining why some leaders seem to create government agencies that went against their interests. Take Nixon and the EPA. Environmental fervor was at an all time high in the early 1970s, so the political payoff of establishing the agency compensated for the offense to supporters from big business; softening the blow even more was that by taking control of the agenda, Nixon could lock in an EPA that would be minimally harmful to the interests of big business in the future. There was some discussion as well about how one could test this kind of logic. Some at the workshop were not convinced by the type of "small N" comparative case study evidence that was presented. Perhaps more convincing would be to collect quotes from people that structure government oil revenue allocation deals, along the lines of the quotes from letters between traders that Avner Greif uses as evidence in his work on medieval traders.