# 14.282 Midterm 1 - FINAL ANSWERS (Iteration 2)
**Student:** 思 중군 함대🐅
**Course:** 14.282 Organizational Economics, Fall 2025
---
[[iter(eval)]]
## Problem 1: Career Concerns and Incentive Contracts
### Part (a): Pure Career Concerns (p observable, not contractible)
#### I. Equilibrium Specification
**Wages:** $\{w_1, w_2\}$
**Efforts:** $\{a_1, a_2\}$ where $a_t = (a_{1t}, a_{2t})$
**Beliefs:** $E[\eta | p_1]$
#### II. Solution by Backward Induction
**Period 2:**
Since no future exists, the agent has no career concern motive. The agent chooses effort to maximize current payoff, but with no explicit incentive contract and no future benefit:
$
\max_{a_2} \quad w_2 - c(a_2) = w_2 - \frac{1}{2}(a_{12}^2 + a_{22}^2)
$
**Result:** $a_2^* = (0, 0)$
**Wage determination:** With competitive principals and rational expectations:
$
w_2^* = E[y_2 | p_1] = E[f_1 a_{12} + f_2 a_{22} + \eta + \varepsilon_2 | p_1] = E[\eta | p_1] = \varphi p_1
$
where in equilibrium, $\hat{a}_1 = a_1^*$, so $E[\eta | p_1] = \varphi(p_1 - g \cdot a_1^*) = \varphi p_1$ after substituting equilibrium belief.
Actually, more precisely: $E[\eta | p_1] = \varphi(p_1 - g_1 a_{11}^* - g_2 a_{21}^*)$. In equilibrium with correct beliefs, market knows effort, so:
$
w_2^* = \varphi p_1 \quad \text{(simplified, as market correctly anticipates } a_1^* \text{)}
$
**Period 1:**
The agent maximizes:
$
\max_{a_1} \quad w_1 - c(a_1) + \delta[w_2^* - c(a_2^*)]
$
Substituting $w_2^* = E[\eta | p_1]$ and $a_2^* = 0$:
$
\max_{a_1} \quad w_1 - \frac{1}{2}(a_{11}^2 + a_{21}^2) + \delta \varphi(p_1 - g \cdot \hat{a}_1)
$
where $p_1 = g_1 a_{11} + g_2 a_{21} + \eta + \phi_1$.
In equilibrium, $\hat{a}_1 = a_1$, so:
$
\max_{a_1} \quad w_1 - \frac{1}{2}(a_{11}^2 + a_{21}^2) + \delta \varphi(g_1 a_{11} + g_2 a_{21} + \eta + \phi_1 - g_1 a_{11} - g_2 a_{21})
$
$
= w_1 - \frac{1}{2}(a_{11}^2 + a_{21}^2) + \delta \varphi(\eta + \phi_1)
$
Wait, this doesn't look right. Let me reconsider.
Actually, the market updates belief about η based on observed p₁. The agent's marginal benefit from effort is the effect on market's belief. Let me redo this properly.
**Correct approach:**
Market observes $p_1 = g \cdot a_1 + \eta + \phi_1$.
Market's belief: $E[\eta | p_1] = \varphi(p_1 - g \cdot \hat{a}_1)$ where $\hat{a}_1$ is the market's conjecture of agent's effort.
Agent's payoff from period 1:
$
w_1 - c(a_1) + \delta E[\eta | p_1]
$
Agent chooses $a_1$ taking $\hat{a}_1$ as given (since agent is atomistic). Taking derivative:
$
\frac{\partial}{\partial a_{i1}}\left[- \frac{1}{2}(a_{11}^2 + a_{21}^2) + \delta \varphi(g_1 a_{11} + g_2 a_{21} + \eta + \phi_1 - g \cdot \hat{a}_1)\right] = 0
$
$
-a_{i1} + \delta \varphi g_i = 0
$
**First-order condition:**
$
a_{i1}^* = \delta \varphi g_i \quad \text{for } i = 1, 2
$
Or in vector form:
$
a_1^* = \delta \varphi g
$
where $g = (g_1, g_2)$.
**Wage:** Competition among principals drives profits to zero:
$
w_1^* = \bar{u}
$
The agent's participation constraint binds in period 1.
#### III. Why is $(a_{11}, a_{21}) \neq (0, 0)$?
**Intuition:**
Even without an explicit performance contract, the agent exerts positive effort in period 1 due to **career concerns**. The mechanism:
1. Higher effort → higher observed performance $p_1$
2. Market interprets high $p_1$ as signal of high ability $\eta$
3. Higher inferred $\eta$ → higher period 2 wage $w_2$
The implicit incentive is $\delta \varphi$, which acts like an endogenous bonus rate. This is the "career concern" effect identified by Holmström (1982/1999).
**Mathematical insight:**
$
\frac{\partial w_2}{\partial a_{i1}} = \varphi g_i \delta
$
This marginal return to effort equals marginal cost at equilibrium:
$
a_{i1}^* = \delta \varphi g_i
$
#### IV. How does $\{a_t\}_{t=1}^2$ depend on $\cos(\theta)$?
**Define alignment:** $\cos(\theta) = \frac{f \cdot g}{\|f\| \|g\|}$
**Key insight:** In part (a), where $p$ is not contractible, $\cos(\theta)$ does NOT affect effort levels directly.
**Why?**
- Agent always exerts effort along direction $g$ (the observable metric)
- $\cos(\theta)$ measures alignment between firm value ($f$) and performance metric ($g$)
- Effort level: $\|a_1^*\| = \delta \varphi \|g\|$ (independent of $\theta$)
**What does $\cos(\theta)$ affect?**
The **social value** of effort:
$
\text{Value created} = f \cdot a_1^* = f \cdot (\delta \varphi g) = \delta \varphi \|f\| \|g\| \cos(\theta)
$
**Interpretation:**
- High $\cos(\theta)$ → effort well-aligned with firm value → efficient
- Low $\cos(\theta)$ → effort creates "window dressing" → inefficient
- $\cos(\theta) = 0$ → effort orthogonal to value → complete waste
**Summary:** $\cos(\theta)$ affects EFFICIENCY, not effort level, in pure career concerns.
#### V. How does $\{a_t\}_{t=1}^2$ depend on $\varphi = \frac{h_\phi}{h + h_\phi}$?
**Definition:** $\varphi$ is the **precision weight** in Bayesian updating, measuring how accurately performance $p_1$ reflects ability $\eta$.
**Bayesian interpretation:**
$
\varphi = \frac{h_\phi}{h + h_\phi} = \frac{\text{signal precision}}{\text{signal precision} + \text{prior precision}}
$
**Effect on effort:**
**Period 1:**
$
\frac{\partial a_1^*}{\partial \varphi} = \delta g > 0
$
Higher signal precision → stronger career concerns → more effort.
**Period 2:**
$
a_2^* = 0 \quad \text{(independent of } \varphi \text{)}
$
**Intuition:**
| $\varphi$ Level | Market Interpretation | Period 1 Effort | Logic |
|-----------------|----------------------|-----------------|-------|
| $\varphi \to 1$ | "Performance = Ability" | $a_1^* \to \delta g$ | Effort strongly affects reputation |
| $\varphi \to 0$ | "Performance = Noise" | $a_1^* \to 0$ | "It's all luck anyway" |
| $\varphi = 0$ | No learning | $a_1^* = 0$ | Career concerns vanish completely |
**Extreme cases:**
- $\varphi = 0$: No information in $p_1$ about $\eta$ → no career concerns → $a_1^* = 0$
- $\varphi = 1$: Perfect signal → maximum career concerns → $a_1^* = \delta g$
---
### Part (b): Contractible Performance ($w_t = s_t + b_t p_t$)
#### I. Equilibrium Contracts
**Period 2: Pure Moral Hazard**
Agent's problem:
$
\max_{a_2} \quad b_2 p_2 - c(a_2) = b_2(g \cdot a_2 + \eta + \phi_2) - \frac{1}{2}\|a_2\|^2
$
**FOC:**
$
b_2 g = a_2
$
Thus:
$
a_2^* = b_2 g
$
Principal wants to maximize firm value. If principal could choose any $a_2$, optimal would be:
$
a_2^{FB} = f
$
Since agent responds along $g$, principal chooses $b_2$ to induce effort in that direction:
$
b_2^* g = \text{projection of } f \text{ onto } g = \frac{f \cdot g}{\|g\|^2} g
$
Therefore:
$
b_2^* = \frac{f \cdot g}{\|g\|^2} = \frac{\|f\|}{\|g\|} \cos(\theta)
$
**Resulting effort:**
$
a_2^* = \frac{f \cdot g}{\|g\|^2} g
$
**Salary:** Competition implies zero profit:
$
s_2^* = E[y_2 | p_1] - b_2^* E[p_2 | p_1] = E[\eta | p_1] - b_2^* (g \cdot a_2^* + E[\eta | p_1])
$
Simplifying (since $E[\phi_2] = 0$):
$
s_2^* = -b_2^* g \cdot a_2^*
$
**Period 1: Explicit + Implicit Incentives**
Agent's problem:
$
\max_{a_1} \quad b_1 p_1 - c(a_1) + \delta[E[w_2 | p_1] - c(a_2^*)]
$
With $w_2 = s_2 + b_2 p_2$ and $E[w_2 | p_1] = s_2^* + b_2^* E[p_2 | p_1]$, but the key is:
$
E[w_2 | p_1] = E[\eta | p_1] + \text{contract terms}
$
The career concern component is:
$
\frac{\partial E[w_2 | p_1]}{\partial a_1} = \frac{\partial E[\eta | p_1]}{\partial p_1} \cdot \frac{\partial p_1}{\partial a_1}
$
With public contracts, market knows $b_1$, so:
$
E[\eta | p_1, b_1] = \varphi(p_1 - g \cdot a_1^{exp}(b_1))
$
where $a_1^{exp}(b_1) = b_1 g$ (expected effort given $b_1$).
The implicit incentive from updating is:
$
\frac{\partial E[\eta | p_1, b_1]}{\partial a_1} = \varphi g
$
**Total marginal return to effort:**
$
(b_1 + \delta \varphi) g
$
**FOC:**
$
a_1^* = (b_1 + \delta \varphi) g
$
**Optimal contract:**
To induce same effort as period 2 (if interior solution):
$
b_1 + \delta \varphi = b_2^*
$
Thus:
$
\boxed{b_1^* = b_2^* - \delta \varphi}
$
**Constraint:** $b_1^* \geq 0$. If $\delta \varphi > b_2^*$, then:
$
b_1^* = 0 \quad \text{(corner solution)}
$
This is **Holmström's "Wisdom of Weak Incentives"**: when implicit incentives are strong enough, no explicit incentive needed.
**Salary:**
$
s_1^* = \bar{u} - b_1^* E[p_1]
$
#### II. What makes $b_1^*$ different from $b_2^*$?
**Key difference:** Presence of implicit career-concern incentive $\delta \varphi$ in period 1.
**Substitutability:**
The relationship $b_1^* = b_2^* - \delta \varphi$ shows that:
- Explicit incentive ($b_1^*$) and implicit incentive ($\delta \varphi$) are **perfect substitutes**
- One unit increase in implicit → exactly one unit decrease in explicit
- Total incentive remains constant: $b_1^* + \delta \varphi = b_2^*$
**Economic interpretation:**
| Component | Period 1 | Period 2 |
|-----------|----------|----------|
| Explicit ($b_t$) | $b_2^* - \delta \varphi$ | $b_2^*$ |
| Implicit (career) | $\delta \varphi$ | $0$ |
| **Total** | $b_2^*$ | $b_2^*$ |
The firm "gets free incentives" from career concerns, so reduces explicit bonus accordingly.
**Wisdom of Weak Incentives:**
When $\delta \varphi \geq b_2^*$:
- Career concerns alone sufficient
- $b_1^* = 0$ (corner solution)
- Effort: $a_1^* = \delta \varphi g < a_2^* = b_2^* g$
#### III. How does $a_1^*$ differ from $a_2^*$? Why?
**Case 1: Interior Solution** ($\delta \varphi < b_2^*$)
$
a_1^* = (b_1^* + \delta \varphi) g = b_2^* g = a_2^*
$
**Result:** $a_1^* = a_2^*$ (same effort level)
**Why?** Firm adjusts explicit bonus to offset career concerns, maintaining total incentive at optimal level $b_2^*$.
**Case 2: Corner Solution** ($\delta \varphi \geq b_2^*$)
$
b_1^* = 0 \implies a_1^* = \delta \varphi g
$
While:
$
a_2^* = b_2^* g
$
**Result:** $a_1^* \neq a_2^*$
Specifically:
- If $\delta \varphi = b_2^*$: $a_1^* = a_2^*$
- If $\delta \varphi > b_2^*$: $a_1^* > a_2^*$ (!)
**Why different?** When career concerns are very strong ($\delta \varphi > b_2^*$), the firm cannot reduce explicit bonus below zero. The agent "over-exerts" effort in period 1 relative to first-best, driven by excessive career concerns.
**Summary table:**
| Condition | $b_1^*$ | $a_1^*$ | Relationship |
|-----------|---------|---------|--------------|
| $\delta \varphi < b_2^*$ | $b_2^* - \delta \varphi > 0$ | $(b_1^* + \delta \varphi)g = b_2^* g$ | $a_1^* = a_2^*$ |
| $\delta \varphi = b_2^*$ | $0$ | $\delta \varphi g = b_2^* g$ | $a_1^* = a_2^*$ |
| $\delta \varphi > b_2^*$ | $0$ | $\delta \varphi g > b_2^* g$ | $a_1^* > a_2^*$ |
#### IV. Why is public observability of contracts important?
**Mechanism:**
Public observability affects the **strength of implicit incentives** through its impact on market inference.
**Private Contracts (not observable):**
Market observes only $p_1$, not $b_1$.
Belief updating:
$
E[\eta | p_1] = \varphi(p_1 - g \cdot \hat{a}_1^{private})
$
where $\hat{a}_1^{private}$ is market's conjecture without knowing $b_1$. Market must guess effort level, leading to:
- **Higher implicit incentive** (market attributes more of high $p_1$ to ability)
- **Lower optimal $b_1^*$** (firm exploits this)
**Public Contracts (observable):**
Market observes both $p_1$ and $b_1$.
Belief updating:
$
E[\eta | p_1, b_1] = \varphi(p_1 - g \cdot a_1^{exp}(b_1))
$
where $a_1^{exp}(b_1) = (b_1 + \delta \varphi)g$ is the expected effort given known contract.
Market can **decompose** performance:
$
p_1 = \underbrace{g \cdot a_1^{exp}(b_1)}_{\text{effort component}} + \underbrace{\eta}_{\text{ability}} + \phi_1
$
Result:
- **More accurate inference** about ability
- **Lower implicit incentive** (effort component factored out)
- **Higher optimal $b_1^*$** (firm must compensate)
**Formal comparison:**
Let $\delta \varphi^{private}$ and $\delta \varphi^{public}$ denote implicit incentives under each regime.
- Private: $\varphi^{private}$ reflects uncertainty about effort → larger
- Public: $\varphi^{public}$ reflects only ability uncertainty → smaller
Thus:
$
b_1^{private*} = b_2^* - \delta \varphi^{private} < b_2^* - \delta \varphi^{public} = b_1^{public*}
$
**Economic insight:**
Public contracts act as **hard evidence** (à la Grossman-Milgrom), preventing the firm from "free-riding" on excessive career concerns. This:
1. Reduces career distortions
2. Makes reputation system more efficient
3. Increases explicit incentive pay
**Connection to Problem 2:** This parallels the hard evidence vs. cheap talk distinction. Public contracts = verifiable information → better information transmission → more efficient outcomes.
---
## Problem 2: Cheap Talk with Reputation
### Setup Summary
**Players:** Agent (advisor), Principal (decision-maker)
**State:** $s \in \{0, 1\}$ with $P(s=1) = p$
**Types:**
- Unbiased (u): $u^u = -y(s-d)^2 + \lambda \phi(m)$
- Biased (b): $u^b = xd + \lambda \phi(m)$
- Prior: $P(\text{unbiased}) = q$
**Message:** $m \in \{0, 1\}$ (cheap talk)
**Decision:** Principal chooses $d$ after observing $m$
**Reputation:** $\phi(m) = P(\text{unbiased} | m)$
---
### Part (a): No Full Separation
**Claim:** There exists no equilibrium where $m(s) = s$ for both types.
**Proof by contradiction:**
Suppose separating equilibrium exists:
- Unbiased: $m^u(0) = 0$, $m^u(1) = 1$
- Biased: $m^b(0) = 0$, $m^b(1) = 1$
**Principal's response:**
Using Bayes' rule on path:
- $m=0$ could come from either type when $s=0$
- $m=1$ could come from either type when $s=1$
- Both messages equally informative about type: $\phi(0) = \phi(1) = q$
Principal's optimal decisions:
- $m=0$: $d_0 = E[s | m=0] = 0$ (both types report truthfully)
- $m=1$: $d_1 = E[s | m=1] = 1$
**Check biased type's incentive when $s=0$:**
**Truth-telling** ($m=0$):
$
u^b(d_0, m=0, s=0) = x \cdot 0 + \lambda \cdot q = \lambda q
$
**Lying** ($m=1$):
$
u^b(d_1, m=1, s=0) = x \cdot 1 + \lambda \cdot q = x + \lambda q
$
**Comparison:**
Since $x > 0$:
$
x + \lambda q > \lambda q
$
The biased type strictly prefers to lie. The reputation term $\lambda q$ is identical in both cases (both on equilibrium path), so the biased type optimizes only the decision payoff $xd$.
**Conclusion:** Biased type deviates when $s=0$. Equilibrium breaks. $\square$
**Crawford-Sobel (1982) insight:** With cheap talk (costless messages) and preference misalignment, full separation impossible. The biased type can always mimic unbiased to get preferred decision.
---
### Part (b): No Equilibrium where $m^u = 0, m^b = 1$
**Claim:** No equilibrium exists where unbiased type always sends $m=0$ and biased type always sends $m=1$.
**Proposed equilibrium:**
- Unbiased: $m^u(0) = 0$, $m^u(1) = 0$
- Biased: $m^b(0) = 1$, $m^b(1) = 1$
**Principal's beliefs:**
- $m=0 \implies$ unbiased type $\implies \phi(0) = 1$
- $m=1 \implies$ biased type $\implies \phi(1) = 0$
**Principal's responses:**
- $m=0$: $d_0 = E[s | \text{unbiased}] = P(s=1) \cdot 1 + P(s=0) \cdot 0 = p$
- $m=1$: $d_1 = \arg\max_d E[u_p | \text{biased}] = 1$ (since biased always wants high $d$, principal infers they're advocating for it)
Actually, let me reconsider. If principal knows sender is biased, principal should optimize own payoff:
- $u_p = -(s-d)^2$
- With $P(s=1) = p$: $d_1^* = p$
Hmm, but the problem says "biased always prefers higher decisions no matter what is s", so principal knowing sender is biased doesn't help much. Let me use the off-equilibrium belief that supports this: principal believes biased would send $m=1$ regardless of $s$, so $m=1$ is uninformative.
**Better approach:**
After $m=1$, principal believes sender is biased. Since biased type's message is independent of $s$, the message is uninformative about state. Principal uses prior:
$
d_1 = p
$
After $m=0$, principal believes sender is unbiased. Unbiased sends $m=0$ regardless of $s$ (both when $s=0$ and $s=1$), so again uninformative:
$
d_0 = p
$
Wait, this doesn't work either. Let me reconsider the structure.
**Actually, the equilibrium specifies:**
- Unbiased always sends 0 (both states)
- Biased always sends 1 (both states)
So:
- $m=0 \implies$ unbiased $\implies d_0 = p$ (uninformative about $s$)
- $m=1 \implies$ biased $\implies d_1 = p$ (uninformative about $s$)
This gives same decision both ways, which seems odd. Let me re-examine.
Perhaps the specification is:
- $d_0$ should be chosen to maximize principal's payoff given belief sender is unbiased
- Since unbiased has same preferences as principal, principal would want to follow unbiased's implicit advice
- But unbiased sends $m=0$ for both states, so no advice conveyed
- Thus $d_0 = p$
Similarly, $d_1 = p$.
**Check unbiased type when $s=1$:**
**Stay** ($m=0$):
$
u^u = -y(p-1)^2 + \lambda \cdot 1 = -y(1-p)^2 + \lambda
$
**Deviate** ($m=1$):
$
u^u = -y(p-1)^2 + \lambda \cdot 0 = -y(1-p)^2
$
**Comparison:**
Stay payoff: $-y(1-p)^2 + \lambda$
Deviate payoff: $-y(1-p)^2$
Since $\lambda > 0$, staying is better. No profitable deviation.
Hmm, this suggests equilibrium might exist. Let me check other types.
**Check unbiased type when $s=0$:**
**Stay** ($m=0$):
$
u^u = -y(p-0)^2 + \lambda \cdot 1 = -yp^2 + \lambda
$
**Deviate** ($m=1$):
$
u^u = -y(p-0)^2 + \lambda \cdot 0 = -yp^2
$
Again, staying better due to reputation.
**Check biased type:**
**Stay** ($m=1$):
$
u^b = xp + \lambda \cdot 0 = xp
$
**Deviate** ($m=0$):
$
u^b = xp + \lambda \cdot 1 = xp + \lambda
$
**Profitable deviation!** Biased type prefers to deviate to $m=0$ to gain reputation, since decision is same either way.
**Conclusion:** Equilibrium fails because biased type deviates. $\square$
---
### Part (c): No Equilibrium where $m^u = 1, m^b = 0$
**Claim:** No equilibrium exists where unbiased always sends $m=1$ and biased always sends $m=0$.
**Proposed equilibrium:**
- Unbiased: $m^u(0) = 1$, $m^u(1) = 1$
- Biased: $m^b(0) = 0$, $m^b(1) = 0$
**Principal's beliefs:**
- $m=0 \implies \phi(0) = 0$ (biased)
- $m=1 \implies \phi(1) = 1$ (unbiased)
**Principal's responses:**
- $m=0$: $d_0 = p$ (biased uninformative)
- $m=1$: $d_1 = p$ (unbiased uninformative)
**Check biased type:**
**Stay** ($m=0$):
$
u^b = xp + \lambda \cdot 0 = xp
$
**Deviate** ($m=1$):
$
u^b = xp + \lambda \cdot 1 = xp + \lambda
$
**Profitable deviation!** Since decisions are identical ($d_0 = d_1 = p$), biased type strictly prefers to deviate to $m=1$ to gain reputation $\lambda$.
**Conclusion:** Equilibrium fails. $\square$
**General insight:** When both types pool on different messages but decisions are same, the type assigned worse reputation will deviate to improve reputation costlessly.
---
### Part (d): Pooling on $m=1$ Equilibrium
**Equilibrium candidate:**
$
m^u(0) = m^u(1) = m^b(0) = m^b(1) = 1
$
**Principal's response:**
Message $m=1$ is uninformative (all types send it). Using prior:
$
d^* = E[s] = p
$
**Off-equilibrium belief:**
For $m=0$ (off path), use pessimistic belief:
$
\phi(0) = 0
$
This implies principal believes only biased would deviate, so:
$
d_0 = p \quad \text{or} \quad d_0 = 1
$
(Biased type wants high decision regardless of state; if principal knows sender is biased and uninformative, best response is still $d_0 = p$. But for strongest deterrent, use $d_0 = 1$.)
Let's use $d_0 = 1$ for clearer analysis.
**Incentive compatibility: Unbiased type, $s=0$**
**Stay** ($m=1$):
$
u^u = -y(p-0)^2 + \lambda q = -yp^2 + \lambda q
$
**Deviate** ($m=0$):
$
u^u = -y(1-0)^2 + \lambda \cdot 0 = -y
$
**IC condition:**
$
-yp^2 + \lambda q \geq -y
$
$
\lambda q \geq y(1-p^2) = y(1-p)(1+p)
$
**Equilibrium condition:**
$
\boxed{\lambda \geq \frac{y(1-p^2)}{q}}
$
**Interpretation:**
For equilibrium to exist, reputation value $\lambda$ must be large enough that unbiased type (who sees $s=0$) prefers to:
- Accept wrong decision ($d=p$ instead of $d=0$)
- Maintain reputation ($\phi = q$)
Rather than:
- Reveal truth (send $m=0$)
- Get correct decision ($d=0$) but lose reputation ($\phi=0$)
**Key insight:** High reputation concerns ($\lambda$) **destroy information transmission**. Even honest types lie to protect reputation.
---
### Part (e): Pooling on $m=0$ Equilibrium
**Equilibrium candidate:**
$
m^u(0) = m^u(1) = m^b(0) = m^b(1) = 0
$
**Principal's response:**
Message $m=0$ uninformative:
$
d^* = p
$
**Off-equilibrium belief:**
For $m=1$, use:
$
\phi(1) = 0 \implies d_1 = 1
$
**IC: Unbiased type, $s=1$**
**Stay** ($m=0$):
$
u^u = -y(p-1)^2 + \lambda q = -y(1-p)^2 + \lambda q
$
**Deviate** ($m=1$):
$
u^u = -y(1-1)^2 + \lambda \cdot 0 = 0
$
**IC condition:**
$
-y(1-p)^2 + \lambda q \geq 0
$
$
\lambda q \geq y(1-p)^2
$
**Equilibrium condition:**
$
\boxed{\lambda \geq \frac{y(1-p)^2}{q}}
$
**Relationship to part (d):**
Condition (d): $\lambda \geq \frac{y(1-p^2)}{q} = \frac{y(1-p)(1+p)}{q}$
Condition (e): $\lambda \geq \frac{y(1-p)^2}{q}$
Since $(1-p)^2 < (1-p)(1+p)$ for $p > 0$:
$
\frac{y(1-p)^2}{q} < \frac{y(1-p)(1+p)}{q}
$
**Result:**
- High $\lambda$: Both equilibria exist, but (d) has tighter condition
- Intermediate $\lambda$: Only (e) exists
- Low $\lambda$: Neither exists (no pooling equilibrium)
**Summary:**
| $\lambda$ Range | Equilibrium |
|-----------------|-------------|
| $\lambda \geq \frac{y(1-p)(1+p)}{q}$ | Both pooling equilibria exist |
| $\frac{y(1-p)^2}{q} \leq \lambda < \frac{y(1-p)(1+p)}{q}$ | Only pool on $m=0$ |
| $\lambda < \frac{y(1-p)^2}{q}$ | Neither pooling equilibrium |
**Economic lesson:**
Reputation concerns create **babbling equilibria** where all types send same message → principal learns nothing → decisions suboptimal.
The paradox: caring about reputation (wanting to appear unbiased) makes agents unable to transmit information, even when they have valuable information to share.
This connects to **Crawford & Sobel (1982)**: cheap talk + preference misalignment + reputation concerns → information transmission fails.
**Connection to Problem 1:** Just as public contracts act as hard evidence to restore information transmission, here we'd need verifiable messages (hard evidence) rather than cheap talk to enable truthful communication.