## _wp0791

## Source details

**Canonical URL:** [_wp0791](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0791.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0791.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0791.pdf.json)

---

### Introduction — Overview and setup
- Delegation implemented via long-term employment contracts insulates appointees from job insecurity and political interference.
- Two formalized rationales for delegation:
  - Hamilton’s learning-by-doing: long-term contracts lengthen horizon and increase incentives to invest effort, improving performance.
  - Blinder’s insulation: insulation from elections prevents politicians’ perverse incentive to follow public opinion when they possess superior private information.
- Key questions:
  - Why do politicians voluntarily delegate power to technocrats?
  - Do politicians delegate enough from society’s perspective, and under what conditions might delegation be socially optimal but privately suboptimal?
- Political economy assumptions:
  - Representative democracy with homogeneous agent preferences; two-period model, t = 1, 2.
  - Single officeholder can make policy decisions or delegate to an insulated technocrat.
  - Citizen utility and period payoff:
    - U_j^t = E_{t=1}^2 v_j^t
    - v_j^t = v_c^t(i_t, \bar r_t) + R_j − α/2 (e_j^t)^2  (equation (1))
    - R_j is an “ego rent”: R if citizen j is the officeholder (j = o), zero otherwise.
  - Policy choice binary: i_t ∈ {i_H^t, i_L^t}.
  - Shock \bar r_t ∈ { \bar r_H, \bar r_L } with Prob = 1/2 each.
  - Payoff normalization:
    - v_c(i_H, \bar r_H) = v_c(i_L, \bar r_L) = 0  (equation (2))
    - v_c(i_H, \bar r_L) = v_c(i_L, \bar r_H) = −1  (equation (3))

### Private signal, ability, and learning-by-doing
- Officeholder private signal σ_t ∈ {σ_H, σ_L}; signal accuracy depends on effort and ability.
- Period-1 signal accuracy for new officeholder:
  - Prob(Right) = 1/2 + θ + e_1
  - Prob(Wrong) = 1/2 − θ − e_1  (equation (4))
- Ability θ ∈ {θ_G, θ_B = 0} with Prob 1/2 each.
- Learning-by-doing for incumbent staying into period 2:
  - Prob(Right) = 1/2 + θ + β e_1 + e_2  (equations (5)-(6))
  - Prob(Wrong) = 1/2 − θ − β e_1 − e_2  (equations (7)-(8))
  - β > 0 interpretable as task complexity: first-period effort e_1 improves period-2 forecasting via β e_1.
- If replaced, period-2 forecasting reverts to period-1 technology (equation (4)).

### Costs and benefits of delegation (trade-offs)
- Costs:
  - Insulated technocrat cannot be dismissed even if incompetent; cost larger when candidate ability cannot be perfectly ascertained ex ante.
  - Delegation is more costly if technocrat is less responsive to changing electorate preferences.
  - Institutional examples cited (appointment terms for U.S. Federal Reserve governors, U.S. Supreme Court justices).
- Benefits:
  - Learning effect (Hamilton): longer horizons increase incentives to exert effort that yields better future performance via β e_1.
  - Insulation effect (Blinder): insulated technocrats avoid politicians’ incentives to follow public opinion that may conceal politicians’ low ability.
- Model origins/extensions:
  - Builds on Holmström (1982/1999) career concerns framework; replaces wage dynamics with electoral discipline; introduces learning-by-doing (β e_1) and a public signal.

### Key comparative insights (qualitative)
- Delegation desirable when task complexity (β) is larger.
- Larger private benefits of office (R) reduce politicians’ incentive to delegate.
- Greater uncertainty about candidate competency reduces private incentives to delegate.
- Delegation is always socially optimal when policymakers base decisions on public opinion, even if politicians may not privately prefer to delegate.
- Equilibrium delegations, when they occur, are socially optimal; but some tasks not delegated in equilibrium would improve social welfare if delegated.
- Strategic mechanism: politicians may mimic public opinion to avoid revealing low ability; insulated technocrats do not face this signaling constraint.

---

### Public signal — structure, institutions, and timing
- Public signal ρ_t ∈ {ρ_H, ρ_L} about shock \bar r_t.
- Public-signal accuracy:
  - Prob(Right public signal) = 1/2 + E_p θ = 1/2 + 1/2 θ_G
  - Prob(Wrong public signal) = 1/2 − E_p θ = 1/2 − 1/2 θ_G
  - E_p θ is expected ability of whole population (average cognitive abilities).
- Citizens do not exert effort to improve public signal; public signal tends to be worse than private signal of an officeholder who exerts positive effort.
- Institutional regimes:
  - Delegation: at t = 1 politician appoints a technocrat for both periods; politician’s utility becomes representative citizen utility plus small political rent R_D; delegation credible.
  - Political control: politician decides policy at t = 1, faces election at beginning of period 2; opponent randomly selected; citizens vote.
- Individual rationality: under both regimes officeholder must prefer being in office (net rent positive).
- Timing: two-period game with delegation decision at t = 1; election at beginning of t = 2 under political control.

---

### III. Hamilton’s Hypothesis: Learning — equilibrium and incentives

### Period-2 effort (common across regimes)
- Officeholder maximizes period-2 utility:
  - U_2 = [R − 1/2 α e_2^2] − {1/2 − E_a θ − e_2}
- If not previously in office then E_a θ = 1/2 θ_G; if incumbent then E_a θ = 1/2 θ_G + β e_1.
- First-order condition: α e_2 − 1 = 0
- Equilibrium effort:
  - e^* = α^(−1) > 0  (equation (10))
- e^* independent of E_a θ or tenure.

### Period-1 effort under delegation (technocrat)
- Technocrat maximizes:
  - U^T_1 = [2R − 1/2 α e_1^2 − 1/2 α e_2^2] − {1/2 − 1/2 θ_G − e_1} − {1/2 − 1/2 θ_G − β e_1 − e_2}  (equation (11))
- First-order condition: α e^T_1 − (1 + β) = 0
- Technocrat first-period equilibrium effort:
  - e^T_1 = 1/α (1 + β) > 0  (equation (12))
- Relation:
  - e^T_1 = (1 + β)/α > e^* = 1/α > 0  (labelled (e1A))

### Period-1 effort without delegation (political control)
- Solve Perfect Bayesian Equilibria (PBE).
- Voter re-election condition (incumbent thrown out if):
  - q(s_1) θ_G + β e_1 < 1/2 θ_G  (equation (13))
- Incumbency advantage arises from β e_1.
- Using Bayes’ rule for wrong forecast:
  - q_W = 1/2 − θ_G − e_1 / 2(1/2 − 1/2 θ_G − e_1)
- Incumbency survival condition yields:
  - β e_1 < 1/2 θ_G^2 / (1 − θ_G − 2 e_1)  (equation (14))
- For right-signal case:
  - q_R(e_1, θ_G) = 1/2 + θ_G + e_1 / 2(1/2 + 1/2 θ_G + e_1)  (equation (15))
- Incumbent politician’s period-1 expected utility:
  - U^P_1 = [R − 1/2 α e_1^2] + {1/2 + 1/2 θ_G + e_1}(R − 1/2 α e_2^2) − {1/2 − 1/2 θ_G − e_1}  − {1/2 + 1/2 θ_G + e_1}{1/2 − q_R(e_1, θ_G) θ_G − β e_1 − e_2} − {1/2 − 1/2 θ_G − e_1}{1/2 − 1/2 θ_G − e^*}  (equation (16))
- First-order condition (with cancellation using (15)):
  - α e_1 = (R − 1/2 α) + 1 + β(1/2 + 1/2 θ_G + 2 e_1)  (equation (17) simplified)
- Solve for politician’s effort (assuming α ≠ 2β):
  - e^P_1 = {1 + β(1/2 + 1/2 θ_G)} + {R − 1/2 α} / (α − 2β) > 0  (equation (18))
- Interpretation: politician’s e_1 influenced by learning-by-doing and career concerns (tournament effect).

---

### B. Delegation of Political Power — welfare and trade-offs
- Politician delegates iff:
  - U_D > U_P  (equation (19))
- Politician utility when delegating (substituting equilibrium effort):
  - U_D = 2 R_D − 1 + θ_G + (1 + β)/2 + 1/α  (equation (20))
  - U_D is increasing in θ_G with slope 1.
- Politician utility when not delegating:
  - U_P = γ_1 + γ_2 θ_G + γ_3 θ_G^2  (equation (21)), with γ_3 > 0 for permissible parameters.
- U_P and U_D can intersect up to two points; only right-most intersection admissible. If ability spread between points A and B delegation dominates; right of B political control dominates.
- Example 1 (numerical illustration):
  - Parameters (θ_G, R, R_D, α, β) = (0.2, 0.2, 0.15, 100, 1) imply:
    - (e^T_1, e^P_1, e^*, U_D, U_P) = (0.02, 0.018, 0.01, −0.45, −0.46) ⇒ U_D > U_P (delegation dominates).
  - If θ_G = 0.4 (others unchanged):
    - (e^T_1, e^P_1, e^*, U_D, U_P) = (0.02, 0.0193, 0.01, −0.25, −0.2402) ⇒ U_D < U_P (politician retains control).
- Candidate solution must satisfy:
  - (i) implied probabilities ∈ [0,1];
  - (ii) effort positive;
  - (iii) individual rationality constraints satisfied.
- Drivers of the welfare trade-off:
  - U_D and U_P both increase in θ_G, but U_P increases more.
  - For small θ_G (easy identification), delegation tends to dominate.
  - As θ_G increases (greater uncertainty about ability), politician more likely to retain control because risk of irreversible incompetent technocrat increases.
  - Task complexity (affecting β) and rents from office (R, R_D) also matter.

---

### 3. Comparative Statics — marginal effects and propositions

### Effects of learning-by-doing (β)
- Marginal effect on U_D:
  - ∂U_D/∂β = e_1 = 2(1+β)/α
- Marginal effect on U_P (envelope theorem):
  - ∂U_P/∂β = e_1(1/2 + 1/2 θ_G + e_1)
- At intersection (e_P1 < e_T1) deduced inequality:
  - ∂U_P/∂β < ∂U_D/∂β
- Comparative-static insight:
  - Increasing β shifts U_D and U_P up; U_D increases more than U_P.
  - Higher β (task complexity) makes delegation more desirable.

### Effects of private information (α^−1)
- ∂U_D/∂α^−1 = (1+β)^2 + 1
- ∂U_P/∂α^−1 = 1/2 α^2 (e_P1)^2 + {1/2 + 1/2 θ_G + e_P1} 1/2
- At intersection (e_P1 < e_T1): ∂U_P/∂α^−1 < ∂U_D/∂α^−1
- Insight:
  - Increasing α^−1 (more private information) analogous to increasing β; raises incentive to delegate.

### Effects of rents (R)
- Increasing R shifts politician’s utility U_P up, making delegation less likely.
- Intuition: larger private rent reduces politician’s concern for public good and raises cost of forfeiting managerial rents.

### Proposition 1 (endogenous delegation)
- Equilibrium variables (e_T1, e_P1, e^*, U_D, U_P) depend on (θ_G, R, R_D, α, β).
- Delegation occurs if U_D > U_P.
- Both delegation and non-delegation equilibria can exist depending on parameters.
- Marginal results:
  i) Increase in task complexity (α^−1 or β) increases incentive to delegate.
  ii) Increase in rent from office (R) reduces incentive to delegate.
  iii) Increase in uncertainty of officeholder ability (θ_G − θ_B) reduces incentive to delegate.

### Social welfare and delegation
- Politician’s utility = private rent + citizen utility (U_D^c or U_P^c).
- Delegation socially optimal if:
  - 2R_D < 2R − c(e_P1) − c(e_P2)  (equation (22))
  - Interpretation: delegation socially optimal when rent politician obtains by delegating is smaller than net rent extracted by performing the task himself.
- Proposition 2:
  - If (22) holds, when delegation occurs it is socially optimal.
- Caveat:
  - Possible U_D^c > U_P^c but U_D < U_P — society prefers delegation, politician prefers to retain power due to private rents; constitutional forcing of delegation could be optimal.

---

### IV. Blinder’s Hypothesis: Insulation — public signal equilibria and welfare
- Introduce public forecast ρ_t; citizens form forecasts with average expected ability and zero effort.
- Politician can reveal private forecast or hide it by following public signal.
- Three possible equilibria:
  1) Delegation to insulating technocrat.
  2) Politician retains power and uses private information (P).
  3) Politician retains power and mimics public opinion (public opinion equilibrium, PO).
- Public opinion equilibrium (PO):
  - Politician maximizes (period 1):
    - max_{e1} U_PO1 = {2R − 1/2 α e1^2 − 1/2 α e2^2} − {1/2 − 1/2 θ_G} − {1/2 − 1/2 θ_G − β e1 − e2}  (equation (23))
  - Equilibrium effort:
    - e_PO1 = β/α > 0  (equation (24))
  - Because e_PO1 > 0, incumbency advantage ensures re-election.
  - Proposition 3: If politician follows public opinion he is always re-elected.
  - Politician utility when following public opinion:
    - U_PO = 2R − 1 + θ_G + (1+β)^2 / (2α)  (equation (25))
- Comparative statics for choice PO vs P (Proposition 4):
  - Equilibrium variables depend on (θ_G, R, α, β); politician follows public opinion if U_PO > U_P.
  - Marginal results:
    i) Increase in β reduces incentive to follow public opinion.
    ii) Increase in R increases incentive to follow public opinion.
    iii) Increase in θ_G − θ_B reduces incentive to follow public opinion.

### Insulation vs public-opinion politician — trade-offs and welfare
- For parameters where politician prefers public opinion (U_D > U_PO), trade-off governed by α:
  - For low α, technocrat invests more effort and attains superior information relative to public; politician following public opinion invests little.
- Partial derivatives:
  - ∂U_D/∂α^−1 = 1+β^2 + 1+2β > ∂U_PO/∂α^−1 = (1+β)^2 / 2
  - ∂U_D/∂β = e_T1 = (1+β)/α > ∂U_PO/∂β = e_PO1 = β/α
- Proposition 5:
  - The more complex the task (increase in β and α^−1), the more likely a politician who is a slave of public opinion is to delegate to an insulated technocrat.
- Proposition 6:
  - If politician ignores private information and follows public opinion in equilibrium, delegation to an insulated technocrat is always socially optimal; however, it may not be privately optimal for the politician.
- Public signal effect:
  - Public signal generally makes delegation less likely because it can only make a politician who retains power better off.
  - To make public signal increase delegation requires altruistic politician motives or belief in commitment value from current delegation.

---

### V. Example: Central Bank Independence (Phelps problem embedded)
- Representative citizen period utility (Kydland-Prescott preferences):
  - v_ct = −1/2 π_t^2 − 1/2 λ x_t^2  (equation (26))
- Expectation-augmented Phillips curve:
  - π_t = κ x_t + π_t^e, t = 1,2  (equation (27))
- Demand side:
  - x_t = −ε(i_t − r̄_t), t = 1,2  (equation (28))
- Central bank sets i_t before shock realization; r̄_t ∈ {r_H, r_L} with Prob 1/2 each.
- Second-period officeholder problem:
  - max_{e,i,π} E[v_o]  (equation (29))
- Derived equilibrium policy values (for S_σ ∈ {HR, LR, HW, LW}):
  - i_H = μ[1/2 + E_2[θ] e] r̄_H + μ[1/2 − E_2[θ] − e] r̄_L  (equation (34))
  - i_L = μ[1/2 − E_2[θ] e] r̄_H + μ[1/2 + E_2[θ] + e] r̄_L  (equation (35))
  - π_HR = ε κ [1/2 − E_2[θ] − e] (r̄_H − r̄_L)  (equation (36))
  - π_HW = −ε κ [1/2 + E_2[θ] + e] (r̄_H − r̄_L)  (equation (37))
  - π_LR = −ε κ [1/2 − E_2[θ] − e] (r̄_H − r̄_L)  (equation (38))
  - π_LW = ε κ [1/2 + E_2[θ] + e] (r̄_H − r̄_L)  (equation (39))
  - x_s = k^{-1} π_s; s ∈ {HR, LR, HW, LW}  (equation (40))
  - E_2[θ] = q θ_G + (1−q) θ_B + e_1
- Effort solution with endogenous loss function:
  - e_s1/2 = φ/(α − φ) E_s1/2 θ, s_1 ∈ {HR, LR, HW, LW}
  - Cost c(e) = 1/2 α e^2; φ = 4(1 + λ/κ^2) ε^2 k^2 (r_H − r_L)^2
  - Equilibrium requires α > φ for positive effort.
- Implications:
  - Higher effort reduces output and inflation variability.
  - Delegation implies higher effort; central bank independence reduces both output and inflation variability.
  - With social loss v_ct = −μ_t[1/2 π_t^2 + 1/2 λ x_t^2] (equation (41)), higher effort also decreases average inflation when society weights recessions more than booms (μ_HR < μ_LR and μ_HW < μ_LW).
  - Delegation complements standard time-inconsistency motives for central bank independence; does not rely on inflation bias to justify independence.
  - Conclusion: instrument independence is socially optimal in this framework; goal independence is not.

---

### VI. Conclusion — summary of determinants and policy implications
- Delegation trades off the cost of potentially incompetent technocrats against benefits of:
  - (i) technocrats investing more effort in specialized tasks (learning);
  - (ii) technocrats being insulated from public-opinion pressures (insulation).
- Main determinants of delegation:
  - Task complexity (β, α^−1) — higher complexity increases incentive to delegate.
  - Rents from office (R) — higher rents reduce incentive to delegate.
  - Uncertainty about ability (θ_G − θ_B) — greater uncertainty reduces incentive to delegate.
- When condition (22) holds delegation is socially optimal, yet politicians may resist due to private rents.
- Formalization of Blinder’s insulation and Hamilton’s tenure rationales: secure tenure raises effort and learning incentives.
- Central bank application: delegation (instrument independence) reduces inflation and output variability and can lower average inflation under asymmetric social loss.
- Extensions and limitations:
  - Paper abstracts from dynamic inconsistency; focuses on election pressures.
  - Assumes homogeneous preferences; heterogeneous preferences and hidden agendas could further increase delegation incentives.
  - Multi-period extension likely preserves mechanisms and allows analysis of optimal election frequency and term length.

*Source: _wp0791 (PDF chapter/section)_

### Introduction............................................................................................................

### Introduction

### Overview
- Delegation of critical public policy tasks (e.g., interpreting the constitution, conducting monetary policy) is often implemented via long-term employment contracts that insulate appointees from job insecurity and political interference.
- Two classical rationales for delegation are formalized:
  - Hamilton’s learning-by-doing argument: long-term contracts give appointees a longer horizon and increase incentives to invest effort, improving performance.
  - Blinder’s insulation argument: insulation from elections prevents politicians’ perverse incentive to follow public opinion when they possess superior private information.
- Key questions addressed:
  - Why do politicians voluntarily delegate power to technocrats?
  - Do politicians delegate enough from society’s perspective, and under what conditions might delegation be socially optimal but privately suboptimal?

### Political economy setting and assumptions
- Representative democracy with homogeneous agent preferences.
- Two-period model, t = 1, 2.
- One political office occupied by a single officeholder (the “politician” if elected); the politician can either:
  - Make policy decisions himself, or
  - Delegate to a technocrat who is not subject to re-election (i.e., insulated/longer term).
- Citizen utility and period payoff (notation retained exactly):
  - U_j^t = E_{t=1}^2 v_j^t
  - v_j^t = v_c^t(i_t, \bar r_t) + R_j − α/2 (e_j^t)^2  (equation (1))
  - R_j is an “ego rent”: R if citizen j is the officeholder (j = o), zero otherwise.
- Policy choice is binary: i_t ∈ {i_H^t, i_L^t}.
- Shock \bar r_t takes two values \bar r_H or \bar r_L with equal probability: Prob(\bar r_t = \bar r_H) = Prob(\bar r_t = \bar r_L) = 1/2.
- Payoff normalization in baseline:
  - If correct forecast: v_c(i_H, \bar r_H) = v_c(i_L, \bar r_L) = 0  (equation (2))
  - If incorrect forecast: v_c(i_H, \bar r_L) = v_c(i_L, \bar r_H) = −1  (equation (3))

### Private signal, ability, and learning-by-doing
- Officeholder receives a private signal σ_t ∈ {σ_H, σ_L} about \bar r_t.
- Signal accuracy depends on effort and ability.
- Probability of a right signal for a new officeholder (period 1):
  - Prob(Right signal for the officeholder) = Prob(σ_H | \bar r_1 = \bar r_H, e_1, θ) = Prob(σ_L | \bar r_1 = \bar r_L, e_1, θ) = 1/2 + θ + e_1  (equation (4) numerator)
  - Prob(Wrong signal for the officeholder) = 1/2 − θ − e_1  (equation (4))
- Ability θ is a random draw taking two values with equal probabilities 1/2: θ_G > θ_B = 0; types a ∈ {G, B}.
- Learning-by-doing for an incumbent staying in period 2:
  - Prob(Right signal for officeholder) = Prob(σ_H | \bar r_2 = \bar r_H, e_2, θ) = Prob(σ_L | \bar r_2 = \bar r_L, e_2, θ) = 1/2 + θ + β e_1 + e_2  (equations (5)-(6))
  - Prob(Wrong signal for officeholder) = 1/2 − θ − β e_1 − e_2  (equations (7)-(8))
  - β > 0 interpreted as complexity of the task: first-period effort e_1 translates into better period-2 forecasting via β e_1.
- If a new officeholder is in period 2 (i.e., the period-1 officeholder is replaced), the forecasting technology reverts to the period-1 expressions (equation (4)); period-1 incumbents cannot pass on individual job-specific knowledge if removed.

### Costs and benefits of delegation (trade-offs formalized)
- Cost of delegation:
  - Insulated technocrat cannot be dismissed even if incompetent; this cost is larger when candidate ability cannot be perfectly ascertained ex ante.
  - Delegation is more costly if technocrat is less responsive to changing electorate preferences.
  - Example institutional fact: U.S. Federal Reserve governors are appointed by the president for a 14-year term (with the exception of the chairman); U.S. Supreme Court justices are appointed for life.
- Benefits of delegation:
  - Learning effect (Hamilton): longer employment horizons increase incentives to exert effort that yields better future performance (learning-by-doing captured via β e_1).
  - Insulation effect (Blinder): technocrats insulated from elections avoid politicians’ incentives to follow public opinion when doing so would conceal signals about politicians’ judgment or ability.
- Model origins and extensions:
  - Builds on Holmström (1982/1999) career concerns framework; extends by:
    - Replacing future wage dynamics with electoral discipline.
    - Introducing learning-by-doing (β e_1).
    - Introducing a public signal about the policy problem to formalize insulation.
  - Persson and Tabellini (2000) introduced electoral discipline earlier; the learning and insulation channels are newly emphasized here.

### Key comparative insights and qualitative results (as stated)
- The model underlines a basic cost–benefit trade-off for delegation driven by:
  - Task complexity: the more complex the policy task, the more desirable delegation becomes (via larger role for β).
  - Private benefits of office (e.g., corruption): the larger these are, the lower the politician’s incentive to delegate.
  - Uncertainty about candidate competency: greater uncertainty reduces private incentives to delegate.
- Welfare and equilibrium observations:
  - Delegation is always socially optimal when policymakers base decisions on public opinion, even if politicians may not privately prefer to delegate.
  - If delegation occurs in equilibrium, it is always socially optimal; conversely, some tasks are not delegated in equilibrium even though delegation would improve social welfare.
- Strategic mechanism emphasized:
  - Politicians may follow public opinion to avoid revealing low ability via adverse outcomes; technocrats insulated from elections do not face this signaling constraint and hence may choose better policies when public and private signals conflict.

_Italicized source attribution: _wp0791 - Introduction (PDF chapter/section)_

### 2.  Public signal

### 2. Public signal

### Public signal: structure and properties
- There is a public signal ρ_t about the realization of the shock ̄r_t that can take the values ρ_H and ρ_L.
- The public signal is created through a process similar to the private signal:
  - Prob(Right public signal) = Prob(ρ_H | ̄r_1 = ̄r_H) = Prob(ρ_L | ̄r_1 = ̄r_L) = 1/2 + E_p θ = 1/2 + 1/2 θ_G
  - Prob(Wrong public signal) = Prob(ρ_H | ̄r_1 = ̄r_L) = Prob(ρ_L | ̄r_1 = ̄r_H) = 1/2 − E_p θ = 1/2 − 1/2 θ_G
- E_p θ is the expected ability of the whole population; public opinion is formed by average cognitive abilities.
- Citizens do not expect to be pivotal and thus do not exert effort to improve the public signal; consequently the public signal tends, on average, to be of worse quality than the private signal of an officeholder who exerts positive effort.

### Institutions: delegation versus political control
- Politician selection and choice:
  - A politician is randomly selected from citizens and, once “elected” at t = 1, decides the institutional regime for producing the public good.
- Delegation:
  - At t = 1 the politician appoints an agent (technocrat) randomly selected from citizens; the technocrat is (contractually) in office for both periods.
  - The politician’s utility becomes that of a representative citizen (he is not in charge of policy, so j ≠ o in (1)) except he receives a small political rent from office R_D.
  - Delegation is assumed credible: the politician can make a binding promise not to fire the delegate for the duration of the game.
- Political control:
  - At t = 1 the elected politician decides to make the policy decision himself.
  - The politician is in office during t = 1 but faces an election at the beginning of period 2; an opponent is randomly selected and citizens vote; the winner is officeholder in t = 2.
- Individual rationality: under both regimes the officeholder must prefer to be in office than not (net rent from office is positive).

### Timing
- The timing of events follows the game described (periods t = 1 and t = 2; delegation decision at t = 1; election at beginning of period 2 under political control).

---

### III. Hamilton’s Hypothesis: Learning

### Overview
- The section formalizes the learning effect from delegation (Hamilton’s hypothesis) abstracting from the public signal; Section 4 (not included here) adds the public signal to generate an insulation incentive.
- The politician compares utilities under delegation and political control; these depend on equilibrium effort and ability.
- Solve by backward induction: derive period 2 effort (same under both regimes), then period 1 effort which depends on regime.

### Equilibrium — 1. Period 2 effort
- Officeholder maximizes period-2 utility:
  - U_2 = [R − 1/2 α e_2^2] − {1/2 − E_a θ − e_2}
- If officeholder was not in office in previous period then E_a θ = 1/2 θ_G.
- If officeholder was in office in previous period then E_a θ = 1/2 θ_G + β e_1 due to learning-by-doing.
- First-order condition: α e_2 − 1 = 0
- Equilibrium effort:
  - e^* = α^(−1) > 0  (equation (10))
- e^* does not depend on E_a θ or tenure.

### Equilibrium — 2. Period 1 effort under delegation
- Technocrat (superscript “T”) maximizes period-1 expected utility:
  - U^T_1 = [2R − 1/2 α e_1^2 − 1/2 α e_2^2] − {1/2 − 1/2 θ_G − e_1} − {1/2 − 1/2 θ_G − β e_1 − e_2}  (equation (11))
- First-order condition: α e^T_1 − (1 + β) = 0
- Technocrat’s first-period equilibrium effort:
  - e^T_1 = 1/α (1 + β) > 0  (equation (12))
- Relation between efforts:
  - e^T_1 = (1 + β)/α > e^* = 1/α > 0  (labelled (e1A))
- Interpretation: first-period effort includes myopic gain (1) and learning-by-doing marginal gain (β).

### Equilibrium — 3. Period 1 effort without delegation (political control)
- Game of incomplete information; solve for Perfect Bayesian Equilibria (PBE).
- Period-2 expected utility from electing a new politician:
  - U^NP_2 = −{1/2 − 1/2 θ_G − e^*}
- Period-2 expected utility from re-electing incumbent:
  - U^IP_2 = −{1/2 − q(s_1) θ_G − β e_1 − e^*}
- Voters throw out officeholder if U^IP_2 < U^NP_2, i.e., if:
  - q(s_1) θ_G + β e_1 < 1/2 θ_G  (equation (13))
- Incumbency advantage arises from β e_1.
- Compute q(s_1) by Bayes’ rule:
  - For wrong forecast:
    - q(s_1 ∈ Wrong signal) = q_W(θ_G, e_1) = Prob(θ_G | ̄r_1 = ̄r_L, σ_H, e_1)
    - q_W = 1/2 Prob(σ_H | ̄r_1 = ̄r_L, e_1, θ_G) / [1/2 Prob(σ_H | ̄r_1 = ̄r_L, e_1, θ_G) + 1/2 Prob(σ_H | ̄r_1 = ̄r_L, e_1, θ_B)]
    - q_W = 1/2 − θ_G − e_1 / 2(1/2 − 1/2 θ_G − e_1)
- Substituting yields incumbency survival condition:
  - β e_1 < 1/2 θ_G^2 / (1 − θ_G − 2 e_1)  (equation (14))
- Model focuses on election versus appointment by assuming condition (14) is satisfied (so incumbent can be thrown out).
- For right-signal case:
  - q(s_1 ∈ Right signal) = q_R(e_1, θ_G) = 1/2 + θ_G + e_1 / 2(1/2 + 1/2 θ_G + e_1)  (equation (15))
- Incumbent politician’s period-1 expected utility (taking e_2 as given):
  - U^P_1 = [R − 1/2 α e_1^2] + {1/2 + 1/2 θ_G + e_1}(R − 1/2 α e_2^2) − {1/2 − 1/2 θ_G − e_1}  − {1/2 + 1/2 θ_G + e_1}{1/2 − q_R(e_1, θ_G) θ_G − β e_1 − e_2} − {1/2 − 1/2 θ_G − e_1}{1/2 − 1/2 θ_G − e^*}  (equation (16))
- First-order condition with respect to e_1:
  - α e_1 = (R − 1/2 α) + 1 + β(1/2 + 1/2 θ_G + 2 e_1) + θ_G {q_R − 1/2} + ∂q_R/∂e_1 θ_G {1/2 + 1/2 θ_G + e_1}  (equation (17))
- Using (15) the last two terms cancel; assuming α ≠ 2β to ensure solution existence, solve for e^P_1:
  - e^P_1 = {1 + β(1/2 + 1/2 θ_G)} + {R − 1/2 α} / (α − 2β) > 0  (equation (18))
- Interpretation: two incentive effects influence politician’s effort — learning-by-doing and career concerns (tournament effect).

---

### B. Delegation of Political Power

### 1. Welfare and existence of trade-offs between regimes
- Politician delegates iff U_D > U_P  (equation (19)).
- Politician utility when delegating (substituting equilibrium effort):
  - U_D = 2 R_D − 1 + θ_G + (1 + β)/2 + 1/α  (equation (20))
  - U_D is a function of R_D, θ_G, β, and α; it is increasing in θ_G with slope 1.
- Politician utility when not delegating can be written as:
  - U_P = γ_1 + γ_2 θ_G + γ_3 θ_G^2  (equation (21))
  - γ_3 = [1/4 β^2 α − 2β / (α − 2β) + 1/4 β^3 (α − 2β)^2 − 1/8 α β^2 (α − 2β)^2 + 1/4] > 0  (textual expression showing γ_3 > 0)
  - It is shown (appendix) that γ_3 must be positive for permissible parameters.
- Generic form: U_P (politician) and U_D (technocrat) as functions of θ_G (or θ_G − θ_B) can intersect in up to two points; only the right-most intersection (point B) is an admissible equilibrium.
- If ability spread is between points A and B, delegation dominates; to the right of B political control dominates.
- Existence of trade-off established by example parameters (Example 1).

Example 1 (numerical illustration):
- Suppose (θ_G, R, R_D, α, β) = (0.2, 0.2, 0.15, 100, 1).
  - This implies (e^T_1, e^P_1, e^*, U_D, U_P) = (0.02, 0.018, 0.01, −0.45, −0.46), i.e., U_D > U_P and delegation dominates.
- If instead θ_G = 0.4 (other parameters unchanged):
  - (e^T_1, e^P_1, e^*, U_D, U_P) = (0.02, 0.0193, 0.01, −0.25, −0.2402), i.e., U_D < U_P and delegation is dominated, so politician retains control.

- Candidate solution must satisfy:
  - (i) implied probabilities of every event are between 0 and 1;
  - (ii) effort is positive;
  - (iii) individual rationality constraints of politician and technocrat are satisfied.

### 2. The welfare trade-off: interpretation and drivers
- Endogenous delegation can occur; whether it happens depends on model parameters — notably the nature/complexity of the policy task, ability spread, learning-by-doing (β), rents from office (R, R_D), and technology parameter α.
- Key comparative statics and intuition:
  - U_D and U_P both increase in θ_G, but U_P increases more (slope of U_P greater than slope of U_D).
  - For small θ_G (low ability dispersion / easy identification of qualified candidates), delegation to technocrats tends to dominate.
  - As θ_G increases (greater spread or greater uncertainty about ability), the utility of keeping power rises and crosses U_D — politician more likely to retain control.
  - Political explanation: appointing an independent technocrat creates risk that society is stuck with an incompetent technocrat who cannot be fired; higher uncertainty about candidate suitability (larger θ_G − θ_B) argues for politician retaining power.
  - Practical examples: Governor of the Federal Reserve or Supreme Court justice: qualifications and objectives are relatively identifiable/clear — arguments favor delegation; tasks like fiscal policy where electorate wishes shift and objectives are less clear — argument for politician to retain power.
- Other parameters also affect delegation: complexity of task (affects β and learning-by-doing) and rents from office influence the politician’s incentives.

*Source: _wp0791 - 2.  Public signal*

### 3.  Comparative Statics

### 3.  Comparative Statics

### Learning-by-doing (β)
- Marginal effect on utility under delegation:
  - ∂U_D/∂β = e_1 = 2(1+β)/α
- Envelope-theorem derivative for politician:
  - ∂U_P/∂β = e_1(1/2 + 1/2 θ_G + e_1)
- At the intersection of the two utility curves:
  - e_P1 < e_T1
  - ∂U_P/∂β = e_P1(1/2 + 1/2 θ_G + e_P1) < (1+β)/α(1/2 + 1/2 θ_G + e_P1) < ∂U_D/∂β
  - Inequality uses (1/2 + 1/2 θ_G + e_1) < 1
- Comparative-static insight:
  - Increasing β shifts both U_D and U_P up and increases welfare in both regimes.
  - Utility under delegation (U_D) increases more than utility under politician control (U_P).
  - Mechanism: technocrat’s effort in period 1 is more sensitive to β because the technocrat cannot be fired and thus has a larger stake in office.
- Interpretation:
  - β can be viewed as task “complexity”; higher β makes delegation to a technocrat more desirable, particularly when learning on the job matters.

### Private information (1/α)
- Higher α^−1 increases effort by both technocrat and politician; as α^−1 approaches its feasible limit, policymakers become perfectly informed.
- Marginal derivatives:
  - ∂U_D/∂α^−1 = (1+β)^2 + 1
  - ∂U_P/∂α^−1 = 1/2 α^2 (e_P1)^2 + {1/2 + 1/2 θ_G + e_P1} 1/2
- At the intersection (e_P1 < e_T1):
  - ∂U_P/∂α^−1 < ∂U_D/∂α^−1
- Comparative-static insight:
  - The effect of increasing α^−1 is analogous to increasing β: higher α^−1 raises the incentive to delegate.
  - α^−1 can be interpreted as the extent of “private information” the policymaker has relative to the public — another dimension of task complexity.

### Rents and Corruption (R)
- Increasing R shifts the politician’s utility curve up, making delegation less likely.
- Intuition:
  - Higher private rent from office reduces politician’s concern for the public good and raises the cost (to the politician) of forfeiting managerial rents via delegation.

### Proposition 1 (Endogenous delegation of political power)
- Equilibrium variables (e_T1, e_P1, e^*, U_D, U_P) depend on (θ_G, R, R_D, α, β).
- Delegation occurs if U_D > U_P.
- Both delegation and non-delegation equilibria exist; which arises depends on (θ_G, R, R_D, α, β).
- Marginal results for the policy task:
  i) Increase in task complexity (α^−1 or β) increases incentive to delegate.
  ii) Increase in rent from office (R) reduces incentive to delegate.
  iii) Increase in uncertainty of officeholder ability (θ_G − θ_B) reduces incentive to delegate.

### Social Welfare and Delegation
- Politician’s utility = private rent from office + citizen’s utility (U_D^c or U_P^c).
- Delegation is socially optimal if:
  - 2R_D < 2R − c(e_P1) − c(e_P2)  (equation (22))
  - Interpretation: delegation is socially optimal when the rent politician obtains by delegating is smaller than the net rent he extracts by performing the task himself.
- Proposition 2:
  - Assume condition (22) holds. When delegation occurs, it is socially optimal.
- Caveat:
  - Possible outcome: U_D^c > U_P^c but U_D < U_P — society would prefer delegation but the politician prefers to retain power due to private rents; constitutionally forcing delegation could be optimal.

### IV. Blinder’s Hypothesis: Insulation — Public Signal and Public Opinion Equilibrium
- Introduce public forecast (ρ_t, t=1,2); citizens’ forecasts formed with average expected ability and zero citizen effort.
- Officeholder’s private forecast remains private information; politician can reveal it or hide it by following the public signal.
- Three possible equilibria:
  1) Delegation to an insulated technocrat.
  2) Politician retains power and uses private information (P).
  3) Politician retains power and mimics public opinion (public opinion equilibrium, PO).
- Public opinion equilibrium (politician bases period-1 decision on public signal):
  - Politician’s maximization (period 1):
    - max_{e1} U_PO1 = {2R − 1/2 α e1^2 − 1/2 α e2^2} − {1/2 − 1/2 θ_G} − {1/2 − 1/2 θ_G − β e1 − e2}  (equation (23))
  - Equilibrium effort:
    - e_PO1 = β/α > 0  (equation (24))
  - Because e_PO1 > 0, incumbency advantage: politician always re-elected.
  - Proposition 3: If the politician follows public opinion he is always re-elected.
  - Politician’s utility when following public opinion:
    - U_PO = 2R − 1 + θ_G + (1+β)^2 / (2α)  (equation (25))
  - Comparative statics for politician choosing public opinion vs private-information policy:
    - Proposition 4: Equilibrium variables (e_PO1, e_P1, e^*, U_PO, U_P) depend on (θ_G, R, α, β). Politician follows public opinion if U_PO > U_P. Both equilibria exist depending on parameters. Marginal results:
      i) Increase in β reduces incentive to follow public opinion.
      ii) Increase in R increases incentive to follow public opinion.
      iii) Increase in θ_G − θ_B reduces incentive to follow public opinion.

### Insulation vs Public-Opinion Politician — Trade-offs and Welfare
- For parameter ranges where politician prefers public opinion (U_D > U_PO):
  - Key trade-off works through α: for low α, technocrat invests more effort and attains superior information relative to the public; politician following public opinion invests little effort.
  - Partial derivatives:
    - ∂U_D/∂α^−1 = 1+β^2 + 1+2β > ∂U_PO/∂α^−1 = (1+β)^2 / 2
    - ∂U_D/∂β = e_T1 = (1+β)/α > ∂U_PO/∂β = e_PO1 = β/α
  - Proposition 5: The more complex the policy task (increase in β and α^−1), the more likely a politician who is a slave of public opinion is to delegate to an insulated technocrat.
  - Proposition 6: If the politician ignores private information and follows public opinion in equilibrium, delegation to an insulated technocrat is always socially optimal; however, it may not be in the politician’s private interest to delegate.
- Public signal effect:
  - The public signal makes delegation less likely because it can only make a politician who retains power better off.
  - To make a public signal increase delegation, one would need altruistic motives for politicians or beliefs about commitment value from current delegation.

### V. Example: Central Bank Independence (Phelps problem embedded)
- Representative citizen period utility (Kydland-Prescott preferences):
  - v_ct = −1/2 π_t^2 − 1/2 λ x_t^2  (equation (26))
- Expectation-augmented Phillips curve:
  - π_t = κ x_t + π_t^e, t = 1,2  (equation (27))
- Demand side:
  - x_t = −ε(i_t − r̄_t), t = 1,2  (equation (28))
- Central bank sets i_t before shock realization; r̄_t ∈ {r_H, r_L} with equal probability.
- Second-period officeholder problem:
  - max_{e,i,π} E[v_o]  (equation (29)); treat expectations as constants under discretion.
- Derived equilibrium policy values (for combinations S_σ ∈ {HR, LR, HW, LW}):
  - i_H = μ[1/2 + E_2[θ] e] r̄_H + μ[1/2 − E_2[θ] − e] r̄_L  (equation (34))
  - i_L = μ[1/2 − E_2[θ] e] r̄_H + μ[1/2 + E_2[θ] + e] r̄_L  (equation (35))
  - π_HR = ε κ [1/2 − E_2[θ] − e] (r̄_H − r̄_L)  (equation (36))
  - π_HW = −ε κ [1/2 + E_2[θ] + e] (r̄_H − r̄_L)  (equation (37))
  - π_LR = −ε κ [1/2 − E_2[θ] − e] (r̄_H − r̄_L)  (equation (38))
  - π_LW = ε κ [1/2 + E_2[θ] + e] (r̄_H − r̄_L)  (equation (39))
  - x_s = k^{-1} π_s; s ∈ {HR, LR, HW, LW}  (equation (40))
  - E_2[θ] = q θ_G + (1−q) θ_B + e_1
- Effort solution with endogenous loss function:
  - e_s1/2 = φ/(α − φ) E_s1/2 θ, s_1 ∈ {HR, LR, HW, LW}
  - Cost function c(e) = 1/2 α e^2; φ = 4(1 + λ/κ^2) ε^2 k^2 (r_H − r_L)^2
  - Equilibrium requires α > φ for positive effort.
- Implications:
  - For given ability, higher effort reduces output and inflation variability.
  - Since delegation implies higher effort, central bank independence (delegation) reduces both output and inflation variability.
  - With modified social loss v_ct = −μ_t[1/2 π_t^2 + 1/2 λ x_t^2] (equation (41)), higher effort also decreases average inflation when society weights recessions more than booms (μ_HR < μ_LR and μ_HW < μ_LW).
  - Delegation here complements standard time-inconsistency motives for central bank independence; it does not rely on inflation bias to justify independence.
  - Conclusion for monetary policy example:
    - Instrument independence is socially optimal in this framework; goal independence is not.

### VI. Conclusion — Key takeaways and extensions
- Delegation trades off the cost of potentially incompetent technocrats against benefits of:
  - (i) technocrats investing more effort in specialized tasks (learning).
  - (ii) technocrats being insulated from public-opinion pressures.
- Main determinants of delegation: task complexity (β, α^−1), rents from office (R), uncertainty about ability (θ_G − θ_B).
- When condition (22) holds, delegation is socially optimal, but politicians may resist delegation due to private rents.
- Blinder’s (insulation) and Hamilton’s (tenure) rationales are formalized: longer, secure tenure raises effort and learning incentives.
- Example application to central bank independence shows delegation can reduce both inflation and output variability and average inflation under asymmetry in social loss.
- Limitations and extensions:
  - Dynamic inconsistency abstracted from here; paper focuses instead on election pressures.
  - Assumes homogeneous preferences; hidden agendas and heterogeneous preferences could further increase delegation incentives.
  - Extension to multiple periods (beyond two) is important and likely to preserve main mechanisms while allowing analysis of optimal election frequency and term length.

*Source: _wp0791 - 3.  Comparative Statics (IMF Working Paper excerpt).*

### References

### References

### Bibliographic entries

- Alesina, A., and R. Gatti, 1995, “Independent Central Banks: Low Inflation at No Cost?”American Economic Review, Vol. 82, pp. 196—200.
- –––, and L. Summers, 1993, “Central Bank Independence and Macroeconomic Performance”Journal of Money, Credit, and Banking, Vol. 25, pp. 151—62.
- –––, and G. Tabellini, 2006, “Bureaucrats or Politicians?,” mimeo Harvard and Bocconi Universities.
- Arrow, K., 1962, “The Economic Implications of Learning by Doing,”Review of Economic Studies, Vol. 29, pp. 155—73.
- Barro, R.J., and D.B. Gordon, 1983, “A Positive Theory of Monetary Policy in a Natural Rate Model,”Journal of Political Economy, Vol. 91, pp. 589—610.
- Besley, T., and S. Coate, 2003, “Elected Versus Appointed Regulators: Theory and Evidence,”Journal of the European Economics Association, Vol. 1, pp. 1176—1206.
- Drazen, Allan, 2001, “‘Laying Low’ During Elections: Political Pressure and Monetary Accommodation,” mimeo University of Maryland.
- –––, 2000,Political Economy in Macroeconomics, Princeton University Press, Princeton, NJ.
- Elster, J., 1995, “The Impact of Constitutions on Economic Performance,”Proceedings of the Annual World Bank Conference on Economic Development, Washington: The World Bank, pp. 209—226.
- Eggertsson, G., 2003, “How to Fight DeflationinaLiquidityTrap:CommittingtoBeing Irresponsible,”IMF Working Paper, No. 03/64, Washington DC.
- –––, and E. Le Borgne, 2003, “A Political Agency Theory of Central Bank Independence”IMF Working Paper, No. 03/144, Washington DC.
- Green, J., and N. Stokey, 1983, “Tournaments and Contracts,”Journal of Political Economy, Vol. 91, pp. 349—64.
- Holmström, B., 1999, “Managerial Incentive Problems: A Dynamic Perspective,”Review of Economic Studies, Vol. 66, pp. 169—82. Originally, 1982, inEssays in Economics and Management in Honor of Lars Wahlbeck(Stockholm: Swedish School of Economics).
- Kydland, F.E., and E.C. Prescott, 1977, “Rules Rather than Discretion: The Inconsistency of Optimal Plans,”Journal of Political Economy,.Vol. 85, pp. 473—91.
- Lazear, E.P., and S. Rosen, 1981, “Rank-Order Tournaments as Optimum Labor Contracts,”Journal of Political Economy, Vol. 89, pp. 841—64.
- -32-
- Le Borgne, E., and B. Lockwood, 2003, “Do Elections Always Motivate Incumbents? Experimentation vs. Career Concerns,”IMF Working PaperNo. 57/03, Washington DC, forthcoming inPublic Choice.
- Maskin, E., and J. Tirole, 2004, “The Politician and the Judge: Accountability in Government,”American Economic Review, Vol. 94, pp. 1034—54.
- Persson, T., and G. Tabellini, 2000,Political Economics: Explaining Economic Policy, MIT Press, Cambridge, MA.
- Prat, A., 2005, “The Wrong Kind of Transparency,”American Economic Review, Vol. 95, pp. 862—77.
- Rogoff, K., 1985, “The Optimal Degree of Commitment to an Intermediate Monetary Target,”Quarterly Journal of Economics, Vol. 100, pp. 1169—89.
- –––, and A. Sibert, 1988, “Elections and Macroeconomic Policy Cycles,”Review of Economic Studies, Vol. 55, pp. 1—16.
- Wilson, J.Q., 1989,Bureaucracy:  What Government Agencies Do and Why They Do It, Basic Books, New York.

*Source: _wp0791 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0791.pdf_
