Title: Industrial Organization or Imperfect Competition Consumer Search II
1Industrial Organization or Imperfect
Competition Consumer Search II
- Univ. Prof. dr. Maarten Janssen
- University of Vienna
- Summer semester 2008
- Week 7 (April 28, 29)
2First, search is exogenous I JR 2001
- Consumer wants to have house painted
- May ask one or several (N) firms to do job
- Each firm is asked (active) with prob a
- An active firm thinks that with probability
(1-a)N-1 it is a monopolist - Firms have cost c
- Consumer has willingness to pay v (suppose this
is known). - Which price will a firm charge?
3First, search is exogenous II JR 2001
- No price equilibrium in pure strategies
- How to construct equilibrium in mixed strategies?
- F(p) cum. symmetric equilibrium distribution
function - Write down expected profits ind. firm given F(p)
- Equate with certain profits of charging upper
bound what is upper bound? - Derive mixed strategy distribution
- Mixed strategies interpreted as price dispersion
identical goods are sold at different prices
4Comparative statics wrt N
- Individual profits declining in N
- As ind. profits are a(1-a)N-1 v
- Industry profits Na(1-a)N-1 v are also declining
in N - By choosing a can be made to mimic empirical
relation of industry profits to number of firms
5Endogenous Fixed Sample Search BJ 1983, JM
2004
- Consumers can decide how many firms to search 0,
1, 2, - each search has cost s
- willingness to pay v
- N Firms choose prices as before
- Symmetric Nash equilibrium where
- Consumer search behaviour is optimal given
strategy of firms - Firm pricing behaviour is optimal given strategy
of other firms and consumers
6Endogenous Fixed sample Search Ruling out
equilibria
- Can there be a sym. equilibrium in which firms
charge pure strategies? - No. Suppose they did. Consumers will only search
once. But if they do, firms have incentive to set
pv and then consumers wont search - Due to unit demand assumption cf., Diamond
result - Can there be a sym. equilibrium where all
consumers search at least two times? - No. Suppose they did. No firms would like to
charge highest price in F(p). All price equal
marginal cost, but then consumers would like to
search only once - Same argument if consumers choose either at least
two searches or not to search at all - Mixed strategy eq where some consumers search
only once
7Endogenous Fixed sample Search Ruling out
equilibria
- If consumers search one time, their exp pay-off
is v - E(p) s - If they dont search exp pay-off is 0
- If they search k times v- E(minp1,.., pk) ks
- Consumers cant randomize between one search and
no search at all (firms would price at v) - Consumers cant randomize between 1 and 3 or more
firms v - E(p) s v- E(minp1,.., pk) ks.
But E(minp1,.., pk) is decreasing in k and at a
decreasing rate. Searching more than once and
less than three times would be better - Consumers have to randomize between 1 and 2
times.
8Endogenous Fixed sample Search mixed strategy
equilibria
- If consumers search once or twice (with prod. a,
resp 1- a - ?(p) 2(1- a)(1-F(p))/N a /N p
- ?(v) av /N
- F(p) 1- a(v-p) / 2(1- a)p
- E(p) ? pdF(p) (av / 2(1- a)) ln (2- a)/a
- How to determine a?
- v - E(p) s v- E(minp1,p2) 2s
- E(minp1,p2) ? pdF(minp1,p2) 2 ?
p(1-F(p))f(p) dp - No explicit solution for a (only implicitly
defined, or numerically) - Equilibrium solution independent of N
9Endogenous Sequential Search
- Two types of consumers fraction ? fully
informed, fraction 1-? bears search cost sfor
each additional search Max. willingness to pay v
for both groups - After each search, consumers can decide whether
or not to continue searching - Perfect recall of prices
- How to decide whether to start searching?
- First search is for free or not
- N Firms choose prices as before
- Symmetric Nash equilibrium where
- Static game, despite sequential search
- Consumer search behaviour is optimal given
strategy of firms - Firm pricing behaviour is optimal given strategy
of other firms and consumers
10Optimal search rule I
- Suppose F(p) is firms pricing strategy and p is
lowest price consumers have observed so far. - Buy now yields v-p
- Continue searching yields ??? (at least v Ep
s) but take into account optimal behaviour after
search - Start at possible end when consumer has observed
N-1 prices. - Continue search v s (1F(p))p - F(p)E(pp lt
p) - Price ? that makes consumer indifferent between
two options is ? s F(?)E(pp lt ?) - Claim largest price in support of F(p) cannot be
above min (?, v) - Suppose it were, consumers will continue to
search will find lower price with probability 1 - Thus, F(?) 1 and ? s E(p)
- In last period, consumer buys iff price is at or
below min (?, v)
11Optimal search rule II
- So, in last period, consumer buys iff price is at
or below ? - Consider penultimate period
- Buying yields v p
- Continue searching yields v s Ep (given that
all firms charge below ? - Price ? that makes consumer indifferent between
two options is ? s E(p) - Stationary process optimal search is
characterized by reservation price ? buy iff p
min (?, v) - Due to perfect recall
- This reservation price is equal to maximum price
in support of F(p)
12Characterization of F(p) and ? when ? v
- Write down profit function for p lt ? v
- ?(p) ?(1-F(p))N-1 (1- ?)/N p
- ?(?) (1- ?)?/N
- F(p) 1 (1- ?)(?-p)/?Np 1/(N-1)
- E(p) ? pf(p) dp ? p dy (by using the change
of variables y 1 - F(p)) - Ep ? ? dy/1bNyN-1, where b ? / (1- ?)
- Reservation price ? s/ 1 - ? dy/1bNyN-1
- Can be larger than v if s is large enough.
13First Search (and last Search)
- When do consumers want to start searching?
- When first search is for free (Stahl 1989),
dominant strategy to search at least once. - When first search costs s, pay-off of first
search is v Ep s v - ?. - Thus, if ? v, uninformed consumers want to
search - Otherwise, they prefer not to search, but this
cannot be an equilibrium (as with only active
informed consumers prices would be equal to 0) - In both cases, as no firm charges above min (?,
v), consumers buy immediately
14Characterization of F(p) and ? when ? gt v
- When first price quotation is for free, write
down profit function for p v - ?(p) ?(1-F(p))N-1 (1- ?)/N p
- ?(v) (1- ?)v/N
- F(p) 1 (1- ?)(v-p)/?Np 1/(N-1)
- When first price quotation is not for free, only
part µ of uninformed consumers are active - ?(p) ?(1-F(p))N-1 µ(1- ?)/N p
- ?(v) µ(1- ?)v/N
- F(p) 1 µ(1- ?)(v-p)/?Np 1/(N-1)
- Ep v ? dy/1bNyN-1/µ and µ such that vEps
0
15When which equilibrium?
- When sis relatively large
- Partial consumer participation equilibrium
- When many fully informed consumers (? large)
- Full consumer participation equilibrium
16Comparative statics
- What is impact of increase in son Ep?
- For small c, ? v and Ep increases in c
- When sis close to 0, then model close to Bertrand
competition and Ep is almost 0 - For larger s, ? gt v and Ep decreases in s(as v -
Ep s 0) - Non-monotonic
- What is impact of increase in N on Ep?
- In partial participation equilibrium none
- In full participation equilibrium increasing
- When N increases transition from full to partial
participation equilibrium
17Conclusions
- With consumer search, prices above marginal cost
- When s becomes small, convergence to Bertrand
model - Consumer search can explain price dispersion in
homogeneous goods markets - Involves calculation of mixed strategy
distributions - Mathematical complications
- Interesting comparative statics