Expert advising under checks and balances (Q2247961)

It is well known that dividing agenda-setting authority and veto authority between different principals is a standard interpretation of so-called ``checks and balances''. By \textit{E. Spolaore} [``Adjustments in different government systems'', Econ. Polit. 16, No. 2, 117--146 (2004; \url{doi:10.1111/j.1468-0343.2004.00134.x})] the conventional wisdom was formalized by showing that the dictator system may adjust too often, while the checks and balances system may fail to adjust when adjustment is optimal. The contribution to this classical topic is done by adding an informational dimension based on cheap talk games. The author's novel prediction that checks and balances may do a better job adjusting to pressing big policy change because of informational advantage is quite counterintuitive. The model is as follows: an organization of two principals (\(P_{1}\) and \(P_{2}\)) has to screen among a range of policies that differ from each other by a one-dimensional parameter \(\omega\) with uniform prior on \([0, 1]\). There is an interested expert (E) who is privately informed about \(\omega\). Under \(P_{1}\) (or \(P_{2}\)) dominance, the expert sends a cheap talk message to \(P_{1}\) (or \(P_{2}\)) before the policy is chosen by the principal who has power. Under checks and balances the advising and policy making proceeds as follows. The expert sends a private cheap talk message to \(P_{1}\) (\textit{agenda-setting stage advising}). \(P_{1}\) proposes a policy recommendation \(y\) to \(P_{2}\) (\textit{agenda setting}). \(P_{2}\) chooses to accept or reject the agenda. In case of rejection, an exogenous status quo policy \(y_0\) will be chosen (\textit{veto-stage policy making}). Let player \(i\)'s utility function be \(U^{i}\), with \(i = E, P_{1}, P_{2}\). The authors admit the following assumptions about the players' preferences: \[ \begin{aligned} U^{E}(y,\omega,b)&=-(y-(\omega+b))^2, \quad b>0\\ U^{P_{1}}(y,\omega,b_{1})&=-(y-(\omega+b_1))^2, \quad 0\leq b_{1}\leq b\\ U^{P_{2}}(y,\omega)&=-(y-\omega)^2.\end{aligned} \] The optimal policy for the expert (and, respectively, \(P_1\) and \(P_2\)) is \(\omega + b\) (and, respectively, \(\omega + b_{1}\) and \(\omega\)), because they all want a higher policy for a larger state of the world. The perfect Bayesian equilibrium of the game under both authority allocations is chosen. The exact equilibrium partition structure is defined in Propositions 4.1, 4.2 which approve the existence of a perfect Bayesian equilibrium under some specified conditions. In case of different equilibrium information structures, the distribution of information must be considered. The tradeoff between these two forces, considering lowest informational losses, is given in Proposition 5.1. The essence of this theorem is consistent with the perception that checks and balances are more efficient in case the conflict of interest is relatively small. This result holds true for more general specifications of the model parameters. The authors conclude that ``dividing agenda-setting authority and veto authority between two different principals has the benefit of assuring a partisan expert to reveal more credible information (measured by the number of signals) to the principals as a whole since neither principal can unilaterally exploit expert advice. There is a central tradeoff between signal quantity and distributive quality under checks and balances''.