Take-It-or-Leave-It Pricing with Adverse Selection

Game Theory · Medium · Free problem

A seller holds a box whose value $V$ is drawn uniformly from $[0, 100]$. You are the buyer. You post a single take-it-or-leave-it price $x$ -- the seller either accepts or walks away. If a sale occurs, the seller also receives an exogenous rebate of $\$10$ on top of the price.

Two scenarios:

  1. (Uninformed seller) The seller does not observe $V$ before deciding. They accept if their expected gain from selling is non-negative, i.e., $x + 10 \geq E[V]$.
  1. (Informed seller) The seller observes $V$ before deciding. They accept if $V \leq x + 10$ (their value is covered by price plus rebate).

For each scenario: (a) derive your expected profit $\pi(x)$ as a function of $x$, (b) find the price $x^{*}$ that maximizes $\pi(x)$, and (c) compare the two optima -- what does the difference tell you about the effect of seller information on buyer profitability?

Hints

  1. Think about who accepts your offer in each scenario. In the uninformed case, the seller's decision does not depend on $V$ -- so what does the distribution of $V$ look like conditional on a sale?
  2. In the informed case, the seller accepts iff $V \leq x + 10$. Write expected profit as $P(\text{sale}) \times E[V \mid \text{sale}] - x \times P(\text{sale})$, substituting the conditional expectation for a truncated uniform.
  3. For Scenario B, after simplifying $\pi_B(x) = (100 - x^2)/200$, the maximum over $x \geq 0$ is at $x = 0$. For Scenario A, the seller only participates if $x + 10 \geq 50$, so the binding constraint pins $x_A^{*} = 40$.

Worked Solution

How to Think About It: This is an adverse selection problem in the style of Akerlof's lemons market. When the seller is uninformed, they accept or reject without conditioning on $V$, so you buy a random box with $E[V] = 50$. The rebate shifts the seller's participation constraint -- they need $x + 10 \geq 50$, so you must pay at least $40$. Your profit is $50 - x$, maximized at $x = 40$. When the seller is informed, they only sell when $V$ is low -- specifically when $V \leq x + 10$. You face the classic lemons problem: the average box you actually buy is worth far less than 50, because the seller screens out the high-$V$ boxes. This kills your profit margin dramatically.

Quick Estimate: Before any algebra: in the informed case, if you post $x = 20$, the seller sells whenever $V \leq 30$. You buy a box worth $E[V | V \leq 30] = 15$ on average, and you pay 20 for it -- that is a loss of 5 per trade. Clearly a high price is bad. At $x = 0$, seller sells if $V \leq 10$: you get a box worth $E[V | V \leq 10] = 5$ and pay nothing, netting 5 per trade, but only with probability $10/100 = 0.1$. Expected profit $\approx 0.5$. In the uninformed case at $x = 40$: seller always sells (since $40 + 10 = 50 = E[V]$), you buy a box worth $50$ and pay $40$ -- profit of 10 per trade, with certainty. The informed case is much worse for you.

Approach: Compute the probability of trade and the conditional expectation of $V$ given a sale, then optimize over $x$.

Formal Solution:

*Scenario A -- Uninformed Seller:*

The uninformed seller does not know $V$, so their expected value of the box is $E[V] = 50$. They accept if the price plus rebate covers this expected value: $$x + 10 \geq 50 \implies x \geq 40.$$

If $x \geq 40$: the seller always accepts. A sale occurs with probability 1, and $V$ is drawn unconditionally from $\text{Uniform}[0, 100]$. Your expected profit is: $$\pi_A(x) = E[V] - x = 50 - x.$$

This is decreasing in $x$, so the optimal price is $x_A^{*} = 40$, giving: $$\pi_A(40) = 50 - 40 = 10.$$

If $x < 40$: the seller rejects (since $x + 10 < 50$, selling is a net loss for them). No trade occurs, $\pi_A(x) = 0$. So $x_A^{*} = 40$ is confirmed.

*Scenario B -- Informed Seller:*

The informed seller observes $V$ and accepts iff their net gain is non-negative: $$x + 10 \geq V \implies V \leq x + 10.$$

Assume $0 \leq x \leq 90$ (so $x + 10 \in [10, 100]$ and some trade occurs). The probability of a sale is: $$P(\text{sale}) = P(V \leq x + 10) = \frac{x + 10}{100}.$$

Given a sale, $V$ is uniform on $[0, x + 10]$, so: $$E[V \mid \text{sale}] = \frac{x + 10}{2}.$$

Your expected profit (per round, not per trade) is: $$\pi_B(x) = P(\text{sale}) \cdot \left(E[V \mid \text{sale}] - x\right) = \frac{x + 10}{100} \cdot \left(\frac{x + 10}{2} - x\right).$$

Simplify the per-trade profit: $$\frac{x + 10}{2} - x = \frac{10 - x}{2}.$$

So: $$\pi_B(x) = \frac{(x + 10)(10 - x)}{200} = \frac{100 - x^2}{200}.$$

This is a downward-opening parabola in $x$, maximized at $x = 0$: $$x_B^{*} = 0, \quad \pi_B(0) = \frac{100 - 0}{200} = 0.5.$$

Sanity check: at $x = 0$, the seller sells only when $V \leq 10$. Trade probability is $10\%$, average box value given sale is $5$, you pay $0$ -- profit of $5$ per trade, $0.5$ expected. Correct.

*Comparison:*

| Scenario | $x^{*}$ | $\pi(x^{*})$ | |---|---|---| | A (Uninformed seller) | $40$ | $10$ | | B (Informed seller) | $0$ | $0.5$ |

Answer: Seller information reduces buyer profit by a factor of 20 -- from $\pi = 10$ to $\pi = 0.5$. The optimal price also collapses from $40$ to $0$. In the informed case, the buyer can only profit by offering a very low price and accepting that trades are rare; any higher price attracts sellers with low-$V$ boxes and the adverse selection wipes out the margin.

Intuition

The stark contrast between the two scenarios -- profit of 10 versus 0.5 -- is a clean illustration of Akerlof's lemons problem. When the seller is uninformed, they cannot condition their accept/reject decision on $V$, so you effectively buy a random box. The rebate just shifts the participation threshold, and you can extract most of the surplus by setting price equal to the threshold. But when the seller observes $V$, they act as a gatekeeper: they only sell you the low-quality boxes ($V$ small) and hold the good ones. The more you raise the price to attract high-$V$ sellers, the more you also attract sellers with $V$ just below the new threshold -- but you overpay relative to the average quality you receive. The optimal response is to post a very low price and accept that you trade infrequently, skimming only the very cheapest boxes.

In practice, this is the economics behind bid-ask spreads and why market makers widen quotes when they suspect informed flow. The rebate in this problem plays the role of a liquidity premium or make-rebate on an exchange: it shifts who participates but does not change the fundamental adverse selection logic. The key lesson is that information asymmetry is not a second-order friction -- it can reduce buyer surplus by an order of magnitude, and the optimal pricing response is counterintuitive: post a lower price, not a higher one.

Open the full interactive solver →