E. Maskin & J. Tirole

 A Summary

In a Nutshell

A study of when decision making powers should be exercised under direct democracy (DD), representative democracy (RD) (accountable officials), or judicial rule (JR) (non-accountable officials). In order to do so they posit a two-period game in two different systems. In the first system the voters do not discover whether the optimal policy was chosen before moving on to the second period (“no feedback”). In the second system, they do discover if the optimal policy was chosen (“feedback”).

 In the no feedback system DD results in the policies preferred by the electorate i.e. the decision is determined by the electorate’s prior beliefs about the best policy, so the popular action is selected in each period. JR results in the policies preferred by the judge as the official need not worry about reelection. The actions of the representative depend upon how much he values being in office. If he values it a lot he will pander to public opinion (choose the popular action) so the best form of government is DD of JR depending on how much the electorate know ex ante about the optimality of each action. When he values office holding little, he will act in order to protect his legacy and this provides the electorate with the possibility of weeding out noncongruent politicians. In this case RD is better than JR but may still be inferior to DD (depending on how much the electorate knows ex ante about the optimality of each action). This implies that decisions of great importance (where the legacy motive will dominate) should be taken by representatives not judges.

 In the feedback system there is some probability that the electorate find out if the period 1 decision was optimal before the second period. DD and JR result in the same outcomes as under the no feedback system. If the probability of feedback is sufficiently high the politician always chooses the optimal policy in the first period. If the feedback probability is only moderate noncongruent officials may pander, but otherwise officials act on their legacy motive.


  • Power is delegated to representatives as they are expected to make better decisions as they have experience, knowledge etc. and they have better incentives to decide well than the average citizen due to propensity to free-ride on effort of other citizens. Such delegation also can help to prevent the tyranny of the majority.

What motivates representatives to act in the public interest?

  1. Legacy – they want to be remembered for doing good things
  2. Power – they like office for its own sake 
  • The public can harness these motivations by making them accountable (i.e. by holding elections and reelections). Elections induce noncongruent officials (one’s with opinions different to the public) to act in the public interest. This is the “moral hazard” correcting element of accountability. Elections also allow the electorate to weed out noncongruent officials altogether. This is the “adverse selection” correcting element of accountability.
  • Yet there are problems. In order to get reelected the official may pander i.e. choose a policy that is popular rather than good for society. This conflicts with the rationale of representation being that representatives make better decisions. Also, if minority rights are a concern, such a system would give the majority too much power to shape the government.

 The Model

  • 2 periods and a pair of possible actions {a,b}
  • The optimal choice is not known by the electorate, but the probability of a=optimal has a probability p(>0.5). This is known by the electorate and hence a is the “popular” choice (i.e. it would be chosen if the electorate did the deciding). The parameter p is a measure of how much the electorate knows about the issue, and this is affected by its technicality and its familiarity.
  • Unless DD is specified, the decision is allocated to an official. With probability π(>0.5) the official is congruent (preference ranking is the same as voters would have if fully informed). With probability 1-π they are non-congruent. The official knows what decision is best for himself, and for society (reflecting the incentive to be better informed).
  • δ is a discount factor which measures eagerness to hold on to power. It is the ratio of the official’s payoff from remaining in office for period 2, to that from choosing her preferred action in period 1.
  • Officials will always choose their preferred policy in period 2, and this will be the optimal policy with probability π.
  • q is the probability that the electorate find out if the optimal decision was taken in period 1 before period 2 occurs.
  • Payoff to electorate is 1 if optimal policy chosen in 1st period, 2 if chosen in both, 0 if chosen in neither.


q = 0 

  1. DD – decision determined by prior beliefs so popular action selected at each period. Welfare: WDD = 2p
  2. JR – official selects his preferred policy as no need to worry about reelection. Welfare: WJR = 2π
  3. RD
    • Strong office holding motive (δ>1). The representative prefers to remain in office relative to selecting his preferred policy. He selects popular action a regardless of his preferences or what is optimal. Voters will only reelect if he chooses a. This is “full pandering”. Welfare: WRD = p + π. Note that WRD < max{WDD, WJR}. In other words RD is the same as DD in period one, and JR in period 2. RD is dominated by either DD or JP. Whether DD or JP is better will depend upon whether p > π or π > p.
    • Weak office holding motive (δ<1). The representative chooses his preferred policy in period 1. There is no pandering. The electorate can now draw inferences about the politician from this choice. He will only be reelected if he chose the popular action. Thus RD produces the same outcome as JR in the first period, but RD now dominates JR as the electorate can replace an official whose behavior suggests non-congruence. [The algebra on page 12 is too much for me].
  • The only benefit of accountability in the no-feedback case is the opportunity to screen out noncongruent officials. Accountability does not induce the official to act on behalf of the electorate (when δ> 1 the official panders; when δ<1 he acts in his own interests).
  • If the decision is particularly important for the legacy of the official (going to war etc.) then the parameter δ will be much smaller and thus less likely to generate pandering. Thus such decisions should be allocated to RD not JR [note that decisions in period 1 are not better under RD, only that RD provides an opportunity to screen officials.


q > 0


  1. DD – as above: WDD = 2p
  2. JR – as above: WJR = 2π
  3. RD – [it gets somewhat complex here].
    • Full Pandering. The official selects the popular action in period 1, and is reelected only if he adheres to this equilibrium. This can only occur when the feedback is sufficiently slow, and there is a large desire to stay in office. As with No-Feedback, RD is dominated by either JR or DD depending on the same conditions as above.
    • Forward Looking Pandering (δq > 1). The official selects the optimal action in the first period regardless of his own preference. He is pandering not to current voters but to future voters who may have received feedback about his first period performance. This is the moral-hazard correcting effect of accountability (there is no adverse selection effect). This happens with rapid feedback. Welfare: WRD = 1 + π. This is better than JR (WJR = 2π)
    • Partial Pandering (δq < 1, q > 0). Either the official chooses his preferred policy (as in JP), or, if noncongruent, he possibly panders (improving welfare compared to JP). If there is no feedback then the period two outcome is the same is JP (chooses preferred policy), but if there is feedback then the likelihood of a congruent official is improved as noncongruents will be thrown out of office. This equilibrium thus incorporates aspects of both moral-hazard and adverse-selection corrections.


  • When there is no feedback, elections help in the adverse selection correction but does nothing to solve the moral hazard problem. Either δ>1 in which case politicians pander and RD is dominated or δ<1 in which case behaviour is the same without accountability and elections offer merely the opportunity to remove noncongruent officials.
  • With rapid enough feedback (δq > 1) the forward looking pandering equilibrium exists and solves the moral hazard problem in the first period but does not help with adverse selection.
  • With moderately fast feedback (δq > 1, q > 0) the partial pandering equilibrium exists and to some extent counteracts both the moral hazard and the adverse selection problems.
  • Non-accountability is desirable when a) the electorate is poorly informed about the optimal action(p < π) and b) when feedback is slow. So technical decisions may best be taken by JR
  • The most important decisions should be taken by elected officials. 


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: