This paper asks how an oversight institution should reveal information when elections both discipline incumbents and select future officeholders. In a two-period political agency model, an incumbent privately observes a policy shock, chooses costly effort, and is judged from the policy outcome and a public signal about the shock. Timing is central. If the signal is observed before effort, optimal revelation is polar: full revelation when discipline is valuable, and no revelation when selection is valuable. If the signal is observed after effort, partial revelation can strictly dominate both extremes. The designer sometimes withholds information, which weakens effort incentives, but clear signals remain likely enough to expose shirking. Delayed revelation therefore weakly dominates ex ante revelation and strictly improves welfare when selection concerns are sufficiently strong. The result identifies retrospective oversight as a distinct institutional tool, not merely delayed revelation.

Can a more competent government produce worse policy outcomes? In a dynamic two-party model, an office-motivated opposition can sabotage reform. Greater competence improves policy selection but also makes successful reform more electorally threatening, inducing more sabotage. Voter welfare therefore depends on effective competence: the incumbent's informational advantage multiplied by reform's survival probability. When sabotage responds sufficiently elastically near full competence, welfare is maximized at an intermediate level; this result ranks informational environments rather than prescribing lower competence. For sufficiently impatient parties, the competence--failure ordering survives jointly more precise signals for both parties and asymmetric priors; it also holds locally under weak partisan motives. Imperfect detection by voters can sustain a no-sabotage equilibrium when the detection probability reaches a competence-dependent threshold, but the threshold rises with competence. Voter ignorance can therefore both reward sabotage and make competent governments harder to protect from sabotage.

How do sellers compete when they face uncertainty about rivals’ costs and consumers have limited price information? We address this question in an oligopoly model that embeds cost uncertainty into the framework of Burdett and Judd (1983). Equilibrium prices respond heterogeneously to consumer information: when captive consumers become more informed, all sellers lower prices; when non-captive consumers become more informed, low-cost sellers cut prices while high-cost sellers raise prices. With endogenous information acquisition through consumer search, seller entry has no effect if search frictions are high, but induces price divergence when search costs are low. Greater information, whether exogenous or search-driven, improves efficiency overall, though it may harm captive consumers. Under third-degree price discrimination, aggregate consumer surplus remains unchanged relative to uniform pricing, with gains for informed consumers are exactly offset by losses for captive consumers.

We build a model of policymaking under the threat of unrest. A policymaker chooses how much effort to spend on a public good; effort is unobservable and the outcome conditional on effort is uncertain. A group of citizens protest if the outcome falls short of a reference point; the reference point is determined endogenously by expectations about the outcome and by the intensity of emotions. We show that the effects of stronger emotional reactions on policymaker’s effort and the probability of protest are nonmonotonic and depend on the group’s ability to inflict damage. Equilibrium may require the policymaker to randomize between providing some effort or no effort at all, in order to temper citizens’ expectations, in which case strong emotional reactions are counterproductive. Optimal emotional reactions are fine-tuned to minimize the probability of protest.

How frequently should voters review public officials when performance is noisy? We study a continuous-time Ferejohn model and test it with global data. At the beginning of each fresh term, the incumbent chooses a complete effort plan, and voters observe a discounted performance signal only at the scheduled review. Because the same kernel, $e^{-rt}$, weights effort in both utility and measured performance, a weighted-Jensen argument implies that every optimal plan is constant within the term. For every proposed term, the voter chooses the effort-maximizing performance cutoff. A constitutional designer anticipates this response and balances the precision gained from a longer observation window against the delay of the retention reward. Within the primitive class satisfying the global-concavity condition, the incentive-efficient normalized-term correspondence is invariant to office rent, the effort-cost function, and the Brownian noise scale; the numerical search finds $rT_0^*\approx0.182256$, or $T_0^*\approx0.182256/r$. Two separate extensions allow the electoral result to be disregarded at the review. In the autocracy extension, retention probability is $p+(1-p)\Phi$; in the weak-democracy extension, it is $(1-p)\Phi$. Each uses a single override probability $p$. Both mechanisms reduce effort for a fixed term, but they shift optimal term selections in opposite directions. Divergences between electoral results and office transitions can discipline $p$ within each model. We outline an office-level empirical implementation and a reform design for testing the model's term-length mechanism.