Matching with Transfers

1 From Matching to Markets

• We have studied a problem of allocating items to people (or people to people)

without money.

• We now study efficient allocation with prices.

− Need to define efficiency.

− Study auction designs to achieve efficient allocation.

• Assignment model

− Different items and individuals with different preferences

− But each person wants just one item (will be relaxed later).
Example 1. Room Assignment

• Problem: assign dorm rooms to students (each room with one space).

• Efficient assignment without prices or money.

− Assign random numbers, choose in order of numbers.

− Assignment will be Pareto efficient for any random order of students ⇒ may be
many Pareto efficient assignments.

• Suppose that students can trade rooms for money.

− Will the outcome of a room draw still be Pareto efficient?

− What does a Pareto efficient outcome look like?

− What sort of approach might lead to an efficient allocation?

2 Assignment Model

• Assignment problem: (N, K, {vij })

− Individuals: N = {1, . . . , n}, plus a seller indexed by 0. Each individual wants

at most one item.

− Items (or objects): K = {1, . . . , k}, owned initially by the seller.

− Preferences: Individual i ∈ N has value vij ≥ 0 for item j. If i gets item j and
pays pj , utility is vij − pj .

• An assignment is a matching of items to individuals, so that each individual gets

at most one item. Formally, a mapping

µ : N → K ∪ {∅}

such that µ(i) ̸= µ(i′) if i ̸= i′ and µ(i), µ(i′) ∈ K.

• Compared with the matching model in which each individual can rank-order items,
each individual now has a clear monetary value for each item, and cares about value
minus price.

• So, the preferences are now described as cardinal values, instead of ordinal rank-
ings. Of course, the former induces the latter.

• An assignment is Pareto dominated if it is possible to move to a new assignment

and make cash payments between individuals so that everyone is better off.

− An assignment is Pareto efficient if it is not Pareto dominated.

• The total value (or surplus) of an assignment that gives item µ(i) to each i is

v1µ(1) + v2µ(2) + · · · + vN µ(N ).

• What’s the relationship between efficiency & total value?

Theorem 1.
An assignment is Pareto efficient if and only if it maximizes total value.

Proof. Suppose an assignment does not achieve maximum value. Then there is
another assignment of items that will lead to strictly greater total value, and it
will be possible to move to this assignment and find transfer payments between
individuals so that everyone is strictly better off.

Suppose an assignment does achieve maximum bidder value. Then any change in
the assignment reduces the total value, so someone must lose from this change.
• Suppose each item is owned by a separate supplier, who values it at zero.

• Stable Payoffs: A vector of payoffs ({si}i∈N , {pj }j∈K ) that are feasible in the
sense that there exists an assignment µ(·) such that

− si + pµ(i) = viµ(i) for all i,

− unassigned buyer or seller gets zero payoff,

− si, pj ≥ 0 for each i and j (Individual rationality)

− si + pj ≥ vij for each i and j (No blocking by a pair).

• Connection with efficiency?

• A competitive equilibrium is a price, or a price vector, satisfying the con-
dition that, when individuals maximize their utility, or net gain, by optimally
choosing their preferred item at that price, the market clears.

− Assumption: Individuals cannot change prices.

− Market clears (supply=demand).

− Prices of all items are nonnegative, and unsold items must be priced at zero
(sellers’ reservation utilities), or else market clearing cannot happen.
Example 2. Three buyers and two items. Assume that the valuations are

X Y
Buyer 1 30 60
Buyer 2 20 40
Buyer 3 10 20

• What is the efficient (value-maximizing) allocation?

• What are the competitive equilibrium prices?

• What are stable payoffs?

• Market clearing prices and stable payoffs

− One market clearing prices: pX = 10, pY = 30.

− The payoffs are stable: 1 enjoys 30 which is no less than if she try to block with
X, i.e., 30 + 10 ≥ 30. Likewise for 2, 10 + 30 ≥ 40.

− Other market clearing prices: pX = 10, pY = 40, or pX = 20, pY = 40, or ...

• µ(1) = Y, µ(2) = X: unique assignment consistent with market clearing.

− 3 demands nothing: if 3 demands an item, then 1 and 2 with higher values will
demand items and there is excess demand.

− 1 and 2 demand an item: or else, demand < supply.

− 2 demands X: if 2 wants Y , then pY − pX < 20, and 1 also wants Y .

− 1 demands Y : if 1 wants X, then pY − pX > 30, and 2 also wants X.

• A complete set of market clearing prices

− Set pX ≥ 10 and pY ≤ 20. (2 wants X but 3 doesn’t)

− Set pY ≥ pX + 20 and pY ≤ pX + 30 (1 prefers Y to X, but not 2)

pY

50

40

30

20

10

10 20 pX
• Generality of the observations

Theorem 2.

(i) A competitive equilibrium yields stable payoffs.

(ii) Conversely, stable payoffs are implementable by competitive equilib-

rium prices.

(iii) Competitive equilibrium admits an efficient assignment that maxi-

mizes total value.

(iv) The set of market clearing prices have the lowest element (buyer-
optimal equilibrium) and the highest element (seller-optimal stable equi-
librium).
Proof of (i): Suppose ({pj }j∈J ) is a market clearing price vector, and µ(·) is the
associated equilibrium assignment. Then, at the equilibrium, individual i enjoys
the payoff of si = viµ(i) − pµ(i). [If µ(i) = ∅, then si = 0.] We argue that
({si}i∈N , {pj }j∈K ) are stable. Check individual rationality and no blocking pair.

Proof of (ii): Suppose ({si}i∈N , {pj }j∈K ) is stable. Then, there exists a feasible
assignment µ(·). Show that µ(·) maximizes utility for the individuals given the
prices ({pj }j∈K ).

Proof of (iii): Suppose ({pj }j∈K ) is a market clearing price vector, and µ(·) is
the associated equilibrium assignment. Pick some alternative assignment in which
i gets item ν(i). We know that for each bidder i, viµ(i) − pµ(i) ≥ viν(i) − pν(i) since
individuals maximize their utility. Sum these inequalities:
X X X X
viµ(i) − pµ(i) ≥ viν(i) − pν(i).
i∈N i∈N i∈N i∈N
 We show that µ(·) is Pareto efficient. Suppose not. Then, there exist a coalition
N ′ ⊂ N and another allocation z(·) such that viz(i) − pz(i) ≥ viµ(i) − pµ(i) for all
i ∈ N ′ with at least one strict inequality. Thus,
X X X X
viz(i) − pz(i) > viµ(i) − pµ(i)
i∈N ′ i∈N ′ i∈N ′ i∈N ′

Therefore,
X X X X
viµ(i) − pµ(i) + viz(i) − pz(i)
i∈N \N ′ i∈N \N ′ i∈N ′ i∈N ′
X X X X X X
≥ viµ(i) − pµ(i) + viµ(i) − pµ(i) = viµ(i) − pµ(i),
i∈N \N ′ i∈N \N ′ i∈N ′ i∈N ′ i∈N i∈N

which is a contradiction.

Proof of (iv): Omitted.

2.1 Market Desgin 1: Vickery-Clarke-Groves(VCG) mech-
anism

• Suppose items are initially owned by a “seller” that has zero value for all of the
items.

• The seller wants to allocate the items efficiently.

• VCG Mechanism:

− Each buyer reports her values of all items.

− The seller picks the assignment that maximizes the total (reported) value, and
charges “appropriate” prices.
• W (N ′, K ′) = total value that are maximized when assigning a set N ′ ⊆ N of
individuals to a set K ′ ⊆ K of items.

• Suppose at the efficient assignment, an individual i gets item µ(i). Then, she is
charged the payment

pVCG(i) = W (N − i, K) − W (N − i, K − µ(i))

• The payment equals the maximal surplus that would be generated without her
minus the surplus that accrue to the other buyers with her around.

• Effectively, each agent is charged the “net externality” she is causing to the re-
maining buyers.
• Recall Example 2.

X Y
Buyer 1 30 60
Buyer 2 20 40
Buyer 3 10 20

− For buyer 1:

pY = pVCG(1) = W (N − 1, K) − W (N − 1, K − µ(1)) = 40 + 10 − 20 = 30.

− For buyer 2:
pX = pVCG(2) = 60 + 10 − 60 = 10.

− This is precisely the buyer-optimal equilibrium prices.

 Theorem 3.
VCG mechanism is strategy-proof and individually rational, yields an effi-
cient assignment, and admits buyer-optimal stable payoff.

• Intuition for the strategyproofness:

− Each buyer i becomes a residual claimant and collects her marginal social con-
tribution

viµ(i) − pVCG(i) = viµ(i) − W (N − i, K) + W (N − i, K − µ(i))

= W (N, K) − W (N − i, K)

− Note that W (N − i, K) does not depend on i.

− So, she has an incentive to maximize it by reporting truthfully.

2.2 Market Design 2: Ascending (clock) auction

• Ascending (clock) auction (informal version)

− Seller has a price clock for each item.

− Price of each item starts at $0.

− At each point, buyers demand at most a single item.

− Prices advance for goods in “excess demand.”

− Auction ends when no items are in excess demand.

• Assumptions

− At any given moment, each buyer bids for the item it wants most at the current
clock prices (“truthful bidding”).

− Prices rise continuously rather than “jumping” discretely.

Example 3. Consider Example 2

pX pY 1 2 3
0 0 Y Y Y − When pX = 0 and pY = 20, 2 is indif-
0 5 Y Y Y ferent between X and Y.
0 10 Y Y X
− If pX increases, 2 chooses Y .
0 15 Y Y X
− If pY increases, 2 chooses X.
0 20 Y X X
1 20 Y Y X − Hence, px, pY increase together with
1 21 Y X X pY − pX = 20, until auction ends at
2 21 Y Y X px = 10, pY = 30.
.. .. .. .. ..

10 29 Y Y
10 30 Y X
• Ascending (clock) auction (formal version)

− The informal version may be susceptible to miscoordination among buyers when

they are indifferent among several items.

− Demange, Gale, and Sotomayer (1986) provide a formal version:

− At each point, each buyer demands “all” items that she prefers the most.

− If all individuals can be assigned to some items they are demanding, then stop.

− Otherwise, prices are increased by minimal currency increment for those items
that are “minimally overdemanded.”

− The rest is the same.

• Math for checking overdemanded set: Hall’s marriage theorem.

− Suppose there are a finite set M of women and finite set W of men.

− Each man is compatible with some subset of women.

− What is the condition for all men to be matched with compatible women?

Theorem 4 (Hall).
All men M can be matched with compatible women if and only if for each
subset M ′ of men, the number of women who are compatible with at least
one man in M ′ is no less than the number of M ′.
• Illustration of the condition

Boys Girls

B1 G1

B2 G2

B3 G3

B4 G4

− The condition fails since for set {B1, B2, B3} the set of compatible girls {G1, G2}
has a fewer cardinality.
• Back to Ascending (clock) auction

− At each price vector, for each set of “active” bidders B, consider the set of items
D(B) demanded by at least one bidder in B.

− If #D(B) ≥ #B for each subset B of active bidders, then market clears at the
current price.

− Otherwise, there exists B such that #D(B) < #B. We say that D(B) is
over-demanded set.

− A minimal over-demanded set of items is an over-demanded set which do not

contain a proper subset of over-demanded set.

− Prices are raised by minimal currency increment for minimal over-demanded set,
and move to the next round.
 Theorem 5.
In the assignment market setting, the ascending auction with truthful bid-
ding will
(i) Finish at the lowest market clearing prices.
(ii) Result in an efficient (value-maximizing) assignment.
(iii) In short, the ascending auction implements the VCG outcome.
3 Assignment Auction

• Ask bidders to submit a value for each item

• Using the reported values

− Find the assignment that maximizes total value.

− Find the lowest market clearing prices.

− Assign the goods accordingly at those prices.

• Claim: assignment auction has the following properties

− Bidders have no incentive to misreport their values.

− Outcome is efficient, and prices are competitive.

• Recall Example 2:

X Y
Buyer 1 30 60
Buyer 2 20 40
Buyer 3 10 20

• Assignment incentives (for buyer 3)

− Can bidder C gain by misreporting? NO!

− To win X, would have to report a value of 20 or more.

∗ That would mean paying pX = 20, not worth it.

− To win Y, would have to report a value of 50 or more.

∗ Reporting 50 means (1, X), (3, Y ) generates 80.

∗ But reporting 50 or more means pX = 20 and pY = 50.
• Assignment incentives (for buyer 2)

− Consider a bid that still results in 2 winning X

∗ To clear market, 3 must demand nothing, so pX ≥ 10.

∗ So 2 cannot get X at a better price.

− Consider a bid that results in 2 winning Y .

∗ Again pX must deter 3, so pX ≥ 10. But also, pY must cause 1 to demand X

rather than Y , so pY − pX ≥ 30 ⇒ pY ≥ 40.

∗ So 2 cannot get Y at a price below pY = 40, not desirable.

• Assignment incentives (for buyer 1)

− Consider bid that still results in 1 winning Y

∗ At market clearing prices, still need 3 to demand nothing, so pX ≥ 10 and 2 to

demand X, so pY − pX ≥ 20 ⇒ pY ≥ 30.

∗ Cannot do better than truthful bidding.

− Consider bid that results in 1 winning X.

∗ At market clearing, need 3 to demand nothing, so pX ≥ 10.

∗ Better to get Y for pY = 30 with truthful bidding.

 Theorem 6.
In the assignment auction, bidders have incentives to bid their true values.
Assuming bidders do bid their true values,
(i) The items are assigned to maximize total value
(ii) Bidders pay the lowest market clearing prices.

• Notes:

− Incentives to bid truthfully depend on the bidders wanting only one item, oth-
erwise there may be incentive for demand reduction.

− There is a “second price” version of the auction, called the Vickrey auction, in
which incentives to bid truthfully are preserved.
4 Connection to Matching

• At the beginning we noted a possible connection to matching theory b/c of the

one-to-one assignment.

• Think of each bidder as forming a preference list that factors in both item and
money preferences (think of prices as being in discrete dollar increments)

− Example: first choice is to pay zero for item 1, second choice is to pay $1 for
item 1, third choice is to pay $0 for item 2, fourth choice is to pay $2 for item 1,
etc..

• Items prefer more money, but don’t care who offers it.
• Deferred acceptance?

− Each bidder submits a preference list

− Seller runs deferred acceptance algorithm:

∗ Bidders “propose” to the items.

∗ Items accept highest offer, reject others.

∗ Bidders continue down their preference list, “raising their bids” as the algorithm
proceeds.

∗ Algorithm will eventually terminate.

■ Auctions & Matching

• Matching algorithm (Gale and Shap- • Deferred acceptance auction (Kelso

ley, 1962) and Crawford, 1982)
Offers made by computer. “Bids” made by computer.

1. Men make offers to most preferred 1. Bidders offer most preferred re-
remaining acceptable woman. maining acceptable purchase.

2. Women hold best man, reject others. 2. Items hold best bid, reject others.

3. Rejected man strikes the woman 3. Rejected bidder strikes offer from
from his/her list. his/her list.

4. Process continues until no new offers 4. Process continues until no new offers
or rejections. or rejections.

5. Implement last held allocation. 5. Implement last held allocation.

■ Deferred acceptance “auctions”

• What we know from matching theory

− DA algorithm will converge to a “stable” allocation.

− Bidder-offering DA gives stable allocation preferred by bidders, and is strategy-

proof for the bidders.

• Stability: each bidder prefers the item they get, at the price they pay, to any other
item at the price it receives.

− So at the final item prices, demand = supply!

• Completing the auction/matching link

− A stable allocation is a competitive equilibrium

− Bidder-proposing DA gives the lowest market-clearing prices.

5 Summary

• Assignment model captures settings where bidders with diverse preferences must
be assigned to a diverse set of goods, and pricing is allowed.

• Competitive equilibrium is a natural candidate for a “good” outcome, esp. w/

bidder-optimal prices.

• A well-designed auction can elicit willingness-to-pay from bidders and identify

market clearing prices.

− VCG mechanism, ascending auction, deferred acceptance auction

• There is a close connection to matching theory, and a version of the DA can work
as an ascending auction.
References
Demange, Gabrielle, David Gale, and Marilda Sotomayer (1986), “Multi-item Auctions,” Journal
of Political Economy, 94, 863–872.

Gale, David and Lloyd Shapley (1962), “College Admissions and the Stability of Marriage,” Amer-
ican Mathematical Monthly, 69, 9–15.

Kelso, Alexander S., Jr., and Vincent P. Crawford (1982), “Job Matching, Coalition Formation, and

Gross Substitutes,” Econometrica, 6, 1483–1504.