[SOUND] Now, let's discuss an alternative approach to the external agency problem which is accountability ex ante. As I explained earlier, accountability ex ante is about society's approval of what governments intend to do prior to these actions are taken. Let's talk about how strict and rigid should be a society's control over government's intended actions. On the one hand, if this control is excessively strict and excessively rigid then governments are denied much of their executive authority. They are unable to act swiftly, decisively, expeditiously in their reaction to changing circumstances could be slow, delayed, ineffective, and also such government can be easily kept captive by different relatively small interest groups which emphasize effective veto power over government action simply because it requires almost universal, unanimous society's of approval of anything that governments intend to do. And in reality such strict control is implemented first by strong majority requirements and also by various types of checks and balances which constrained government's activities. On the other hand, if society's control over government is weak. In other words, if it takes just relatively small group to support what the government wants to do to have these actions sanctioned, in that case, there could be multiple problems because even nearly disastrous policies could benefit a relatively small group of economic agents, and this group will provide the required support to this policy. Or you can imagine a predatory government that keeps safeguarded its political base from its predation or, perhaps, even redistributes some of its illicit gains from the rest of the economy to this group to make sure that it keeps enjoying this group's support. And, in that case, this government will have complete freedom of hands to expropriate wealth from the rest of society, and we certainly don't want that to happen. So, the question is what is the optimal level of society's ex ante control over what governments intend to do. And in discussing this question I would like to use a relatively recent article by Aghion, Alesina, and Trebbi which appeared ten years ago. Here is a model that describes this choice and sheds light on what the solutions could be. Suppose that there is status quo before any action of government is implemented and the status quo outcome equals 1 across the society. And I should mention that the society is assumed to be a unit continuum of agents. So, in the status quo every agent has utility equals 1. Government's policies are denoted by a. And a drawn randomly from this range: minus A capital all the way to A capital. Now, policy outcome, if policy is a, is a plus mu. And the second component mu describes variation of individual policy outcomes across a unit continuum of agents. And I assume that this variation is distributed according to cumulative distribution function F of mu, furthermore, I will assume that the average value of mu is equal to the median value and this is quite an essential assumption. It is met, for example, if the distribution of mu is symmetric, but quite obviously, it could also take place for some asymmetric distributions. As far as policy itself is concerned, as I said, we assume it is drawn randomly from this range, and its distribution is G of a. Now, we want to describe the strength, the rigidity of societies' ex ante control over government policies by a majority rule. And let M be the majority rule which is in place. M is a portion of the agents whose support is required to endorse government's policy, to endorse it ex ante. When M equals one half, then we have a conventional 50% majority rule. We can conceivably imagine situations when M is less than one half, and, in this case, we have essentially minority rule. Or, alternatively, M can be more than one half, in that case we have a supermajority. So, the question is what is the optimal majority rule which makes the society's ex ante control over government most effective. Let's find out what this solution is. To do so, first of all, let's note that there is a break-even agent who is indifferent between the status quo, where his utility is 1, and this policy, in which case his utility is a plus mu. So, for this break-even agent then the individual variation mu should be equal to 1 minus a. And now we can easily figure out how many agents would support policy a. And this will be the agents whose individual evaluation of this policy mu will be above this threshold. And quite certainly, the number of these agents will be 1-F(1-a). Now, we want this to be greater than or equal to the majority threshold M. In which case this policy will be approved by a required majority. And that means that our majority rule allows policies for which a, which is the overall policy quality, is greater than or equal to this quantity - 1 minus the inverse of F(1-M). This majority rule, as I said, would allow policies with quality a above this particular threshold. Now, I want to find out what is the optimal value of this threshold. But let me do it in two steps. Let me first find out what is the optimal value of the policy threshold a, and then I will realize what is the optimal majority. So, let's see how we can choose optimally the policy quality threshold a-star. We want to do so to maximize expected social gains from this particular threshold. because, of course, the threshold has to be set in advance for all policy alternatives proposed by the government. If the quality of policy is low, if a is below this threshold a-star, then this policy will not be approved by the society, and then the status quo outcome equals 1 will obtain. And, as a result, the society's gains in a policy being disapproved will be also equal 1. Recall there is a unit continuum of economic agents. Now, if a policy is above this threshold, a-star, it's different. This policy will be approved. And then every individual will earn the outcome a plus mu. And if I aggregate these outcomes across the society bearing in mind that this is a unit continuum, we'll have a plus the average of mu, which is also the total of mu. And this is what I have here. The optimal policy, a-star, will be chosen for maximizing this total. And if I want to identify this policy, one thing that I can do and those of you who are comfortable with simple calculus would probably prefer this option would be to differentiate this expression by a star and to write down the first order condition which says that this derivative should be equal to 0. This is probably not the best way to handle this problem because the first order condition is necessary but not sufficient for optimality as you know. And I would personally prefer to use a graph. On this graph the outcome that would obtain if the policy is not approved, in other words the status quo, is this straight line. And this is unity, one. If a policy is approved then the total payoff to the society is a plus the average of mu. And this is this straight line. And I think, from this graph it's pretty obvious that the optimal threshold a-star, is where these two lines intersect. Indeed, if I increase my a-star over this level then I would lose this triangle of social gains. If I diminish, decrease a-star below this level then I will miss this portion of social gains. So, if I do my best, then a-star has to be chosen here. And at this point, as we see, the default payoff which is 1, is equal exactly to a plus average of mu. And then indeed, my optimality condition would be 1 equals to a-star plus the average of mu. And, therefore, the inverse distribution function of one minus M equals to 1 minus a-star, as we know from our previous analysis. And it is equal to the average of mu. And, therefore, I am now ready to calculate my optimal majority rule by simply applying the cumulative distribution function F to the left- and right-hand side of this equality, F minus 1 of 1 minus M equals to the average of mu. And that gives me immediately M-star equals to 1 minus F of the average of mu and if I recall my assumption that the average of mu is the median level of mu, then of course, half of the median of mu equals one half by definition. And I'm back to the standard majority rule. And which basically says that in this level the optimal majority rule is the conventional simple majority rule. And, of course, it might not be the case if I change some assumptions of the model. And one thing that I would like to show you, and this is something which is very important for our discussion of the ex ante solution of the external agency problem, is to allow for public servants, for the executives that implement proposed policies might be of different types. We could think about a benevolent type executive, which is doing her best to serve the society but even in this case it's quite important for the society to exercise controls over this executive's actions simply because what this executive wants to do might be harmful for the society because the policy is an error but if the executive is of a different type, in other words, of a predatory type, and, in that case, that person would not try to serve the society to the best of her abilities but instead to expropriate resources from the society by abusing her or his office powers, then the society would need some defense from such an executive. And, of course, how likely is it that the chief executive will be of predatory type rather than of benevolent type depends on the expectations in the society of moral qualms of its politicians. Suppose that these expectations are such that the probability of the executive being of a predatory type is pi. And let's hope that pi is small. But small, but not 0. And the probability of the benevolent type is then 1 minus pi. Now, in both cases, the society's payoff as a function of a majority rule would be different. If you are confident that you're dealing with an executive of predatory type, then you might want to protect yourself as tightly as you can. And then an increase in majority rule will be always beneficial for the society, and this is reflected by this inequality. The society's welfare in the case it's governed by a predator type executive rises when this executive is under tighter societal control. And we've just studied what would happen if the executive is of benevolent type. In this case, the society's welfare will be a single-peaked curve. Let's see what happens here. And this is the society's welfare if the executive is of benevolent type. And it reaches maximum at one half, which is simple majority. And this is the society's welfare if the executive is of a predatory type. And it increases in M all along. And the third curve which is between the first and the second one is the society's welfare, expected welfare I should say, when there is a chance that an executive could be either of benevolent or of a predatory type. And for this linear combination of two welfare functions or rather of their derivatives, I see that simple majority rule is not optimal. And if there is a risk that an executive will be of predatory type, the society wants to have protection, which would be above the majority rule. In other words, which will be some supermajority. And this is clearly seen from this picture because this average, which is maximum not at one half but at some supermajority point. Which is to the right of one half. So, by and large, our conclusion is that the society has to carefully select its approval thresholds and majority rules being cognizant of various factors including among other things the quality of it's chief executives and, more generally, public servants. And if there are some doubts about moral qualms of such individuals, then a supermajority would be appropriate. [SOUND]