EC231: Industrial Economics 1

Jonathan Cave
Search
Law of One Price
The law of one price (LOOP) states that in the absence of trade
frictions (such as transport costs and tariffs), and under conditions
of free competition and price flexibility (where no individual sellers
or buyers have power to manipulate prices and prices can freely
adjust), identical goods sold in different locations must sell for the
same price when prices are expressed in a common currency.
(Wikipedia)
Why does this hold, and where doesn’t it hold?
Why?

EC231 Industrial Economics I: Strategic Behaviour 2

Reasons for differences in prices across similar products
Product differentiation (e.g. Hotelling model)
– Different prices due to physically different products- either vertically or horizontally differentiated
(Covered in EC208) or firm actions
Price discrimination (or bundling) – covered in first section
Firms taking advantage of differences in consumer requirements/ characteristics (e.g. airline seats;
essentially similar but minor diffs)
Search and/ or switching costs
– Differences in consumer characteristics due to their own behaviour
– Another reason why Bertrand model may be inappropriate
– May be linked with price discrimination sometimes.
– Notice also that advertising is a competing means of consumers getting information.
Different market organisational structure:
– e.g. segmented markets/ vertical structures (relevant for tyres)
But prices are not the only thing that differs, of course

EC231 Industrial Economics I: Strategic Behaviour 3

Information and pricing
Consumers don’t have full information because they don’t seek it or because it
is difficult to obtain
– Information may concern product quality, service or price
– We mainly consider pricing (simplest) – but price can signal other thinsg
What are the implications, in particular for competition and pricing patterns?
– Depends on who, how and how much price search occurs
For specific tyres, there is minimal differentiation between products.
Additionally, no obvious price discrimination, so must be information (?).
Electricity (gas, car insurance): slightly more complex cases because more
information is required to get an accurate quote. See later

EC231 Industrial Economics I: Strategic Behaviour 5

Broad Explanations: Search cost and switching cost.
Search cost of making the purchase
– “Distress” or unconfident? (tyre example)
Switching cost of changing to a new supplier (Dual fuel example) - once you have
searched, you need to click through, contact the switching company by phone or
sign up online to arrange the transfer, but no one (usually) needs to visit home.
Some goods have search cost, but essentially no switching cost; switching cost can
be high or low; can analyse these cases independently
Notice both tyres and dual fuel contracts are essentially homogenous (not quite
true in terms of elec+ gas, from some consumers’ perspectives).
Don’t assume that internet search is typical; many people don’t, for most products.
For some, may be difficult. Maybe more interesting things to do or maybe range of
options undermines confidence!

EC231 Industrial Economics I: Strategic Behaviour 6

Modelling search costs (1)- Consumer side
How much search should a consumer (logically) do?
Search Technologies:
– Cost of search is per search (e.g. phoning dealers); call and time cost, or
– A lump sum payment for a set of searches (e.g. comparison site); cost of
completing form, finding info (may also include hidden commission)
Search Approaches:
– Sequential: Make a decision after every search
– Non-sequential: Make a decision to carry out a given number of searches or
until a trigger is reached.
– Sequential search is only likely to be relevant for the first technology
– (Note car insurance example, though; may be mixed cases. May also be
‘anchored’ by renewal offer)

EC231 Industrial Economics I: Strategic Behaviour 7

Modelling search costs (2)
If search costs an amount every time, only limited search.
Assume prices differ across firms.
On average, the expected lowest price falls as more firms searched - next two slides.
Continue until the expected gain from another search is less than the expected cost-
i.e. stop just before this (possibly a target being reached).
– Whether search is sequential or non-sequential
[Of course, if search costs a fixed amount , then either do it or not, dependent on the
fall-back position- insurance example; “go compare”]
Note- not every search can be carried out on the internet (e.g. haircut)
Internet search costs include expected ‘hassle’ cost and delay of returning
unsatisfactory items.
Some costs may be psychic, can be a big part.
In other realistic cases, beliefs/preferences are affected by search (bait & switch, etc.)

EC231 Industrial Economics I: Strategic Behaviour 8

Dice illustration for sequential search case
Obviously just an illustration- assume prices distributed according to the numbers on the die
(i.e. 1/6 of firms charge 1, 1/6 charge 2, etc.). Consumer searches randomly.
Average value if shake once
Suppose search cost per throw is
Assume sequential search
So search again as long as current lowest price is or a , because on average you will gain more
than the cost
Once lowest sampled price drops to (or below), stop searching.
So is reservation price.
Example assumes prices are given (in this case uniform between upper and lower value)
Also assumes that you can return to a previous offer “free” – sampling with replacement.

EC231 Industrial Economics I: Strategic Behaviour 9

Dice illustration for non-sequential search case
Maintain price assumption prices distributed according to a die (i.e. of firms charge each price )
Average value from one sample is 3.5
Consider average lowest value from 2 samples
– Each combination of prices has probability ;
– 1 way to get =6: ; 3 ways to get =5: ; 5 ways to get =4: , etc.
– So expectation of

Collect k samples
– unless all ; this has probability
– if all but not all ; this has probability
– Similarly, the probabilities that
– The expectation of is

k 1 2 3 4 5 6 7 8 9 10
v* 3.500 2.528 2.042 1.755 1.569 1.440 1.346 1.276 1.222 1.180
E(gain) 0.972 0.486 0.286 0.186 0.129 0.094 0.070 0.054 0.042

EC231 Industrial Economics I: Strategic Behaviour 10

Dice illustration for non-sequential search case
So if , non-sequential strategy is to commit to doing 2 searches then stop
If cost drops to , commit to doing 3 searches
Example assumes distribution of prices is given and free return to pick earlier price (makes
more difference here)
But what will the distribution of prices look like?
Need to look at the firm side of the equation
Why? Because firms can react to consumer behaviour!

EC231 Industrial Economics I: Strategic Behaviour 11

“Tourist trap” (Diamond) model - Firm side
A first look at the firms’ decisions, given consumer search costs
Assume that firms trade currency. Each consumer wants to buy units.
Consumer has a general idea about the distribution of prices, searches
randomly.
Cost to check price is per store.
So consumer, in effect, pays: at first store.
Go to second store if
i.e. would not engage in more search if
So, first firm can charge and not lose custom.
But any firm could think like that

EC231 Industrial Economics I: Strategic Behaviour 12

 Tourist trap (cont)
In other words, however many stores there are, price will be above competitive price at some stores.
If all firms think alike, they may all charge .
But then a firm could raise its price to and still make sales
Argument continues until all firms charge
Result: If a single-price equilibrium exists, it will be at and there would be no search!
The Diamond paradox
– Analogous, but opposite, to the Bertrand paradox

If there is no single price equilibrium, and firms charge different prices, there may be no equilibrium or a
mixed price equilibrium.
But this basic model doesn’t give a very satisfactory model of search!

EC231 Industrial Economics I: Strategic Behaviour 13

Recap and preview
Consumers search
Firms react to consumer search
If all consumers have some search cost
– and firms know this
– then if firms’ prices are close enough consumers will not search
– and there will be an equilibrium at monopoly price
– Diamond (or Tourist trap) equilibrium
But not a very good theory of search!
Now, two other approaches to search modelling, plus some empirical analysis.

EC231 Industrial Economics I: Strategic Behaviour 14

“Tourists and Natives” model (Salop/Stiglitz)

Assume two types of firms, charging high and low prices

Half (say) of consumers (natives) are well-informed, half (tourists) ill-informed.
All natives go to low price stores. Half the tourists do (by chance) and the rest
go to high price stores.
On average, high price stores make of sales, low price stores get .
Can an equilibrium exist, with high price and low price stores?
Yes, given enough tourists: Monopolistically competitive equilibrium
Profits are same for both types; high price stores charge larger mark-up but
spread fixed costs over fewer customers.

EC231 Industrial Economics I: Strategic Behaviour 15

Equilibrium in Salop/Stiglitz
High price stores and low price stores can each make normal profit, after allowing for fixed costs, i.e. fixed costs per
unit sold are higher for high price stores.
High price stores:
𝜋 h=(𝑝 h −𝑐)𝑞 h − 𝑓
Low price stores:
𝜋 𝐿 =(𝑝 𝐿 − 𝑐)𝑞 𝐿 − 𝑓
from previous assumptions
Carlton and Perloff discuss this in great detail in ch. 13.
Model relies on very specific assumptions (e.g. effectively very high search costs for “tourists” and focal point prices)
in order to get its equilibrium, and can only give rise to a two-price equilibrium, with 2 firm types, not one with
many prices
Again, no search (or rather search is exogenously given)
Next: Extensions to more complex models- mixed equilibria.
Why do more complex models?- to get nearer to observed prices

EC231 Industrial Economics I: Strategic Behaviour 16

Recap and preview
Three theoretical frameworks for search cost
– Diamond; equilibrium at monopoly price
– Salop/Stiglitz; two-price equilibrium, no search
– Varian/ Stahl; mixed strategy equilibrium with range of prices,
limited search, but can be extended
But: Do the predictions stack up empirically?
– One example of internet search and a generalisation

EC231 Industrial Economics I: Strategic Behaviour 28

Implications of search models generally
(Diamond) Non-zero search cost for everyone can mean very little search takes place. If so, all
prices may be very high; there need not be dispersion in prices. (Professions??)
It is rational to engage in only limited search.
Search costs can lead to multiple prices in the market, in particular to two-price or more likely
multi-price equilibrium. “Ignorant” consumers pay more on average.
Different levels of search/different search costs lead naturally to some price dispersion in the
market (complication of Stahl).
Individual firms selling in different ways may be able to take advantage of differential
searching behaviour (another topic)
As the share of consumers with complete information increases, average prices fall
Price distributions across firms may change over time.
Belleflamme and Peitz ch 7, Carlton and Perloff ch 13.

EC231 Industrial Economics I: Strategic Behaviour 29

Search on the internet
Typical search on the internet involves finding out about a number of suppliers
at the same time (e.g. car insurance; cameras). Use “search engines” or
comparison sites
Why don’t all consumers go for the lowest price product?
– “Frictionless commerce”
This would imply a Bertrand equilibrium, with no one making any money.
If so, why do the higher-price stores advertise there or allow their placement?
How can the comparison site persuade firms to list?
(also, why is there more than one comparison site??)

EC231 Industrial Economics I: Strategic Behaviour 30

Ellison-Ellison obfuscation model (Econometrica, 2009)
Consider a straightforward, basic product: Computer memory
Small firms selling computer parts at Pricewatch.com
Idea: Demand for these products is likely to be extremely price elastic, so firms
set low prices for basic product, but make money on add-ons.
Model: firms, versions, costs ; prices
Time cost of to visit the firm’s site
Basically, Diamond model adapted to upgrade pricing
Fraction who upgrade: (for firm )
E-E have price distribution for a year, plus one firm’s sales; estimated demands
based on price ranks, cross-effects

EC231 Industrial Economics I: Strategic Behaviour 31

Clearinghouse models and drip pricing
Clearinghouse models (e.g., Varian, Baye & Morgan) can handle the drip-
pricing of essential add-ons:
– "Dripped" shipping costs, taxes, airline baggage fees, etc.
Without drip pricing, consumers observe the complete list of prices and
purchase from the e-retailer charging the lowest price
Drip pricing transforms a visit to the "clearinghouse“ into a costly sequential
search environment
With drip pricing, consumer must click a given listed price (at cost ) to
determine the full price
Suggests "monopoly" pricing may result (due to Diamond Paradox)

EC231 Industrial Economics I: Strategic Behaviour 34

EC231 Industrial Economics I: Strategic Behaviour 35