You are on page 1of 59

Industrial Organization:

Price discrimination

Università di Roma “Tor Vergata”

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Overview

Context: Frequently, firms charge different prices to different


market segments
Concepts: market segmentation, elasticity rule
Economic principle: If you charge different prices for the same
product, expect arbitrage

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Motivation

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Example: laptop pricing

Production cost is $1,200


Three types of buyers:

Type W.T.P. ($) No. (K) Cum. No.


1 3000 10 10
2 2000 20 30
3 1000 30 60

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Example: laptop pricing (cont.)

Strategy 1: Price at $3000


Profit = ($3000−$1200) × 10K = $18m
Strategy 2: Price at $2000
Profit = ($2000−$1200) × 30K = $24m
Strategy 3: Price at $3000 for Type 1
Price at $2000 for Type 2
Profit = ($3000−$1200) × 10K +
+ ($2000−$1200) × 20K = $34m
Bottom line: If price discrimination is possible, it pays

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Perfect price discrimination

Personalized prices: each customer is charged a different price


— exactly his/her willingness to pay
Discriminatory pricing policies are common in some markets,
where the number of customers is relatively small and the
seller has considerable information about buyers
Examples are the so-called ‘customer markets’: ready-mixed
concrete, large commercial aircraft, enterprise software,
landing fees at airports, tug boat push services in ports, but
also, plumber, lawyer, piano teacher.
Although there is a list price (rack rate), each customer
receives a discount (often negotiated)
Final price depends on
customer’s willingness to pay
bargaining power of the two parties (ignored here)

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Perfect price discrimination (ctd.)

With respect to normal pricing,


The seller gains: revenue and profits go up
The low-price buyer often gains
The high-price buyer often loses
Net welfare effect:
overall welfare (consumers + producers’ surplus) increases
distribution of welfare changes dramatically: zero consumers
surplus,
if distribution matters, not clear whether PPD is good or bad
for society as a whole
Next: Different forms of price discrimination, which are all
approximations to perfect price discrimination

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Types of price discrimination
Perfect price discrimination often hard to use
Many approximations to PPD
By indicators (a.k.a. group pricing): market segment can
be directly identified
Trick: apply elasticity rule to each market segment
By self-selection (a.k.a. menu pricing): market segment
cannot be directly identified
Trick: offer options such that each consumer will pay what
they are willing to pay
Practical difficulties
Market research: sometimes difficult to know the willingness
to pay of different customers
Arbitrage: resale, gray markets
Legal limits, US: injury to competition
Legal limits, EU: single market
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Discrimination by indicators

Different segments can be identified directly


(i.e., it’s easy to know who’s who)
Examples?
Rule: different demand fcts ⇒ different elasticities ⇒
different prices.
Specifically, higher prices in less elastic markets (elasticity
rule):

pi − MC 1
=−
pi i
where
dqi pi
i ≡
dpi qi

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Graphical illustration

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Graphical illustration

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Graphical illustration

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Markups on European cars

Model Belgium France Germany Italy UK


Fiat Uno 7.6 8.7 9.8 21.7 8.7
Nissan Micra 8.1 23.1 8.9 36.1 12.5
Ford Escort 8.5 9.5 8.9 8.9 11.5
Peugeot 405 9.9 13.4 10.2 9.9 11.6
Mercedes 190 14.3 14.4 17.2 15.6 12.3

What’s going on here?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Markups on European cars

Model Belgium France Germany Italy UK


Fiat Uno 7.6 8.7 9.8 21.7 8.7
Nissan Micra 8.1 23.1 8.9 36.1 12.5
Ford Escort 8.5 9.5 8.9 8.9 11.5
Peugeot 405 9.9 13.4 10.2 9.9 11.6
Mercedes 190 14.3 14.4 17.2 15.6 12.3

What’s going on here?


home bias
import restrictions
costs
taxes

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Takeaways

Key issues for price discrimination are:


Identifying market segments
Avoiding arbitrage
With clear, separate segments: apply elasticity rule to each
separately

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Price discrimination by self-selection: Overview

Context: Frequently, firms cannot directly identify the


different segments
Concepts: versioning and self-selection, two-part tariffs,
bundling; incentive and participation constraints
Economic principle: If you charge different prices for the same
product, expect arbitrage — unless you make the
products/offers different

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Self-selection schemes

In most cases, seller cannot directly identify consumer type,


but can still induce consumers to distinguish themselves
Versioning: design product lines that appeal to different
consumers
Examples?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Versioning 1.0

WTP WTP
Type # Not Restricted ticket Restricted ticket Cost
Tourist 10 350 300 0
Business 10 800 200 0
Strategy 1:
Strategy 2:
Strategy 3:

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Versioning 1.0

WTP WTP
Type # Not Restricted ticket Restricted ticket Cost
Tourist 10 350 300 0
Business 10 800 200 0
Strategy 1: Price single ticket (NR) at 350
Revenue = 350 × 20 = 7,000
Strategy 2: Price single ticket (R) at 800
Revenue = 800 × 10 = 8,000
Strategy 3: Price (R,NR) at (300,800)
Revenue = 300 × 10 + 800 × 10 = 11,000

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Versioning 1.1
Let’s now change WTPs slightly
WTP WTP
Type # Not Restricted ticket Restricted ticket Cost
Tourist 10 350 300 0
Business 10 800 400 0
Strategy 3: Price (R,NR) at (300,800)
Revenue = 300 × 10 + 800 × 10 = 11,000
This strategy won’t work anymore!! Business traveller will buy
restricted fare.
The utility a business travelers derives from a Non Restricted
ticket is now larger than the utility she derives from a
Restricted ticket.
UB (R) − pR > UB (NR) − pNR
400 − 300 > 800 − 800
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Versioning 1.1
WTP WTP
Type # Not Restricted ticket Restricted ticket Cost
Tourist 10 350 300 0
Business 10 800 400 0
Since Strategy 3 won’t work any more, the trick is to make
less attractive for a Business traveller to buy a ticket aimed a
Tourist traveler
How: we can reduce the price of the Non Restricted ticket
Strategy 4: Price (R,NR) at (300,700)
Revenue = 300 × 10 + 700 × 10 = 10,000

UB (NR) − pNR ≥ UB (R) − pR

800 − 700 > 400 − 300

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Versioning 1.1 (ctd)

If we want to make sure that every traveller buys the ticket


aimed to his/her type, so that
R → Tourist
NR → Business
then
UB (NR) − pNR ≥ UB (R) − pR
UT (R) − pR ≥ UT (NR) − pNR
⇒ It must hold for both types of customers (but it binds for
only one of them...)
The constraints are called INCENTIVE COMPATIBILITY
CONSTRAINTS

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Versioning summary

A scheme to induce customers to select themselves into high


and low prices
Key constraint (incentive compatibility constraint): you can’t
make the inexpensive version too attractive to those willing to
pay more
→ more attractive consumers get a positive surplus,
just enough to make the IC constraint to hold
Additional constraint (participation constraint): cheap version
must be sufficiently cheap that low types are willing to
purchase
→ less attractive consumers get utility equal to 0

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Practice: baby iMac

Market segment H (1 million) willing to pay $1,500 for iMac,


$800 for stripped-down version
Market segment L (2 million) willing to pay $600 for iMac,
$500 for stripped-down version
Production cost: $300 (either version)
What is optimal pricing policy?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Practice: baby iMac

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Practice: baby iMac

Candidate strategy 1: sell full version, charge $1,500


Profit: (1500 − 300) × 1 m = $1.2 bn
Candidate strategy 2: sell full version, charge $600
Profit: (600 − 300) × 3 m = $.9 bn
Candidate strategy 3: sell full version for $1,200,
stripped-down version for $500
Profit: (500 − 300) × 2 m + (1200 − 300) × 1 m = $1.3 bn
Note: $1,200 = 1, 500 − (800 − 500)

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Non-linear pricing

Definition: unit price varies with quantity purchased


Typical examples:
two-part tariff: fixed entry fee (F ), per-unit price (p)
quantity discounts
What is the optimal structure? What are the main obstacles
to implementation?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 1.0

Suppose each consumer demands several units (minutes of


calls, hours at the gym, etc)
Let D(p) be each consumer’s demand curve
How can a two-part tariff extract more surplus from these
consumers?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Non-linear pricing

Price per unit set to maximise the consumer’s surplus


⇒ p equal to MC ⇒ p = MC
Fixed fee set to extract consumer’s surplus ⇒ equal to
CS(p = MC )
⇒ F equal to CS(p = MC ) ⇒ F = CS(p = MC )
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Practice: gym
Monthly individual demand for hours: q = 1 − p
Marginal cost: zero
Optimal (uniform) price per hour:

Optimal two-part tariff:

Effect on profit ?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Practice: gym
Monthly individual demand for hours: q = 1 − p
Marginal cost: zero
Optimal (uniform) price per hour: p = 12 (from q = 12 )
Profit per customer: 12 × 12 = 14
Optimal two-part tariff: usage fee = marginal cost = 0
Fixed fee: 12 (1 × 1) = 12 (consumer surplus)
Profit per customer: 21
Huge increase in profit (why?)

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0

Suppose that different consumers have different demand


curves Di (p) for each unit they consume
For simplicity, assume 2 types only, H and L
How can a menu of two-part tariffs allow seller to implement
a versioning strategy?
How are types defined?
What are the participation and incentive constraints?
Focus on INFORMATION ON CONSUMERS
Define FULL INFORMATION (FI) the case when the firm has
an observable signal to distinguish between consumers (as in
what ???)
Define ASYMMETRIC INFORMATION (AI) the case when
the firm has NO observable signal to distinguish between
consumers. Types are consumers’ private information

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under FI)

Under FI, the firm offers a menu of


contracts, but any consumer is
constrained to buy the package
which has been tailored for his/her
type
Using a NON LINEAR price,
price equal to marginal cost
an individually tailored fixed part,
equal to the consumer’s surplus
full extraction of the consumers’
surplus

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under AI)
Under AI, firms offer a menu of contracts.
Any consumer can buy the package (s)he prefers, without
being constrained to buy the package which has been tailored
for his/her type.

Different price per unit and different fixed fee.


Examples ??
Why this structure ??
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Two-part tariffs 2.0 (under AI) (ctd)

Because of ASYMMETRIC INFORMATION (the inability of


the firm to observe the private information of the consumer’s
type), the offers that fully extract the consumers’ surplus are
not available to the firm
If the firm offers in the market the same two combinations
as in FI
consumers of type H → TH = {pH , FH } = {c 0 , CSH (c 0 )}
consumers of type L → TL = {pL , FL } = {c 0 , CSL (c 0 )}
⇒ Both types of consumers will prefer to buy at the offer
aimed at consumers L

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under AI) (ctd)

Under AI, the firm cannot use the


“optimal” two-part offers.
H-type consumers would prefer to
buy at the offer aimed at L-type
consumers, because

CSH (c 0 ) > CSL (c 0 )

The gain each H-type consumer


make by “cheatin” is the yellow
area
This is also a measure of the firm’s
lost profits for each “cheating”
consumer

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under AI) (ctd)
What is then the nature of the offers that, given the
asymmetry of information, the firm can offer in order to
maximise its profits?
These profit-maximising offers must statisfy these two
constraints:
INCENTIVE COMPATIBILITY CONSTRAINT (IC): Each
consumer must prefer to buy at the offer aimed at his/her own
type
UL (pL , FL ) ≥ UL (pH , FH )
UH (pH , FH ) ≥ UH (pL , FL )
INDIVIDUAL RATIONALITY (PARTICIPATION)
CONSTRAINT (IR): Each consumer must be willing to buy

UL (pL , FL ) ≥ 0

UH (pH , FH ) ≥ 0
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Two-part tariffs 2.0 (under AI) (ctd)

An example of (silly. . . ) offers


which satisfy the two constraints
consumers of type L
→ TL = {pL , FL } = {c 0 , CSL (c 0 )}
consumers of type H

TH = {pH , FH } = {c 0 , CSL (c 0 )}
Even if these offers satisfy the two
constraints, the amount of lost
profits is identical to the previous
case (yellow area)
Is there a way to reduce this
amount of lost profits?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under AI) (ctd)

The loss of profits comes from the


gain that H-type consumers make
pretending to be L-type.
Possible to reduce this gain by making
the offer aimed at L-type consumers
less attractive to H-type consumers
In order to do so, RAISE the PRICE
L-type consumers pay. This
L-type buys less (-)
L-type gets a lower CS and pay a
lower fixed fee (-, green area,
with a positive price-cost margin)
H-type gain less when cheating
(smaller yellow area) and pay a
larger fixed fee (+, darker pink
area)
Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION
Two-part tariffs 2.0 (under AI) (ctd)

p ◦ is larger
the smaller is the loss from L-type consumers, due to the
smaller quantity they consume
the greater is the gain (in terms of a smaller loss) from H-type
consumers, due to the larger fixed fee the firm can charge
An optimal p ◦ will trade-off these two effects and depend on
Size of the two groups
Shape of the demand functions

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Two-part tariffs 2.0 (under AI) (ctd)

Very specific example, but some very general results


Less attractive consumers pay a higher per unit price
More attractive consumers pay a lower price
. . . but also
Less attractive consumers pay a lower fixed fee but are left
with zero surplus
More attractive consumers pay a higher fixed fee but are
left with a positive surplus

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bundling

When more products are on sale in a “bundle”


Pure bundling and mixed bundling
pure bundling: products are NOT on sale separately
mixed bundling: products are on sale in a bundle AND
separately
Examples
A form of versioning (why?)

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bundling: an example

Utility Utility
Type # Mozart Queen
Classical 40 50 0
Rock 40 0 50
Eclectic 20 30 30

Strategy 1: Price at 50 per ticket


Revenue = 50 × 40 × 2 = 4,000
Strategy 2: Price at 30 per ticket
Revenue = 30 × (40+20) × 2 = 3,600
Strategy 3:
.

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bundling: an example

Utility Utility
Type # Mozart Queen
Classical 40 50 0
Rock 40 0 50
Eclectic 20 30 30

Strategy 1: Price at 50 per ticket


Revenue = 50 × 40 × 2 = 4,000
Strategy 2: Price at 30 per ticket
Revenue = 30 × (40+20) × 2 = 3,600
Strategy 3: Price at 50 per ticket or 60 for series
Revenue = 50 × 40 × 2 + 60 × 20 = 5,200

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Other examples

General lesson: need an additional (non-price) dimension to


screen customers
Many applications
Damaged goods
Low value version has equal or even higher production cost
than high value version
Clearly motivated by market segmentation
Intertemporal discrimination
The durable goods monopoly curse
Does e-commerce make price discrimination easier or more
difficult?

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Auctions

Context: You have an object to sell; what’s the best way to


do it?
Concepts: fixed-price, auction
Economic principle: auctions may be the best form of price
discrimination by self-selection

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Alternative selling mechanisms

Who sets the price or prices?


Firm: pricing
Buyer: auctions
Both: negotiations
Pros and cons of auctions vis-a-vis pricing (and negotiations).

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Fixed price vs auctions

Seller owns a widget, no value for it


Two interested buyers; identical valuations either $100 or $150
(equal probability)
Perfect correlation: buyers have same valuation
Buyers know their valuation, seller does not
Fixed price:

Ascending price auction:


Idea:

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Fixed price vs auctions

Seller owns a widget, no value for it


Two interested buyers; identical valuations either $100 or $150
(equal probability)
Perfect correlation: buyers have same valuation
Buyers know their valuation, seller does not
Fixed price:
if p = 150, then for E(π) = 75
if p = 100, then for E(π) = π = 100 → best fixed price
Ascending price auction:
each bidder willing to pay up to her valuation (bi = vi ), so
E(π) = 125
Idea: auction as ultimate strategy for price discrimination by
self-selection

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Auction formats

Ascending auction (a.k.a. English): The good is sold to the


last bidder for the highest bid. Common to sell art and
antiques.
First-price descending (a.k.a. Dutch): The seller reduces the
price until someone accepts the offered price and buys at that
price.
First-price sealed-bid auction: Bidders submit a bid
simultaneously without seeing anyone else’s bid and the
highest bidder wins. The winner pays its own, highest bid.
Second-price sealed-bid auction (a.k.a. Vickrey): Bidders
submit a bid simultaneously without seeing anyone else’s bid
and the highest bidder wins. The winner pays the amount bid
by the second-highest bidder.

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Other elements of auctions

Value
Private value: Individual bidders know how much the good is
worth to them but not how much other bidders value it.
Common value: The good has the same value to everyone, but
no bidder knows exactly what that value is.
Number of Units
Auctions can be used to sell one or many units of a good:
uniform price or discriminatory
Some examples:
art, wine
eBay
government procurement
flowers, fish
T-bills, IPOs

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bidder strategy

Trade-off similar to monopoly pricing


Seller: margin vs quantity sold
Bidder: margin vs winning probability
Optimal bid results from MR = MC -type of rule

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bidding Strategies in Private Value Auctions

Second-Price sealed bid Auction


The amount that you bid affects whether you win, but it does
not affect how much you pay if you win, which equals the
second-highest bid.
Second-Price Auction Best Strategy
Bidding your highest value is your best strategy (weakly
dominates all others).
Suppose that you value an object at $100. If you bid $100 and
win, your CS = 100 - 2nd price. If you bid less than $100, you
decrease your probability of winning. If you bid more than
$100, you risk ending up with a negative CS.
So, bidding $100 leaves you as well off as, or better off than,
bidding any other value.

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bidding Strategies in Private Value Auctions

Ascending Auction
The amount that you bid affects whether you win and pay.
English Auction Best Strategy
Your best strategy is to raise the current highest bid as long as
your bid is less or equal than the value you place on the good.
Suppose that you value an object at $100. If you bid an
amount b and win, your surplus is $100 – b. Your surplus is
positive or zero for b ≤ 100. But, negative if b > 100. So, it is
best to raise bids up to $100 and stop there.
If all participants bid up to their value, the winner will pay
slightly more than the value of the second-highest bidder.

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Bidding Strategies in Private Value Auctions

Dutch and First-Price Sealed Bid Auction


The amount that you bid affects whether you win and pay.
Best Strategies
The best strategy for both games is to bid an amount that is
equal to or slightly greater than what you expect will be the
second-highest value, given that your value is the highest.
Bidders shave their bids to less than their value to balance the
effect of decreasing the probability of winning and increasing
CS. The bid depends on the beliefs about the strategies of
rivals.
Since everybody is bidding their expectation of the highest
losing value conditional on them winning, the mean winning
bid will equal the second highest value (the highest losing
value).

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


The revenue equivalence theorem

Thus, the expected outcome is the same under each format


for private value auctions:
The winner is the person with the highest value, and their
expected payment equals the expectation of the
second-highest value.

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Winner’s curse occurs in common-value auctions

The winner’s curse occurs in common-value auctions: the


winner’s bid exceeds the common-value item’s value. So, the
winner ends up paying too much.
The overbidding occurs when there is uncertainty about the
true value of the good,
Suppose valuations are common (an extreme case of positively
correlated valuations)
One informed bidder, one uninformed
If the uniformed bidder wins, it is because it values the good
too much
Uninformed bidder must beware of adverse selection problem
Uninformed bidder bids more conservatively, wins less often
than informed bidder
Evidence from offshore drilling auctions

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Multi-unit auctions

Sometimes, seller has multiple, identical units


(e.g., T-bills, electricity)
Alternative auctions
Uniform price
everyone pays the lowest bid
Discriminatory
everyone pays his/her bid
Multiple, similar objects (e.g., spectrum licenses)
Problem: valuations are interdependent
Common solution: simultaneously-ascending auction

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION


Takeaways on auctions

Different selling mechanisms trade-off


transaction costs
ability to extract consumer surplus
Auctions allow for better discrimination (by self-selection)
than fixed price
Comparison between auctions and negotiations less obvious

Università di Roma “Tor Vergata” INDUSTRIAL ORGANIZATION

You might also like