Transcript Berry-Levensohn-Pakes EMA (1995) Paper Notes

Berry-Levensohn-Pakes EMA
(1995) Paper Notes
ECON 721
Goals of paper
• To estimate demand and cost parameters for a
class of oligopolistic differentiated products
markets
– Demand side – recover own and cross price
elasticities and elasticities with respect to product
attributes
– Supply side – recover cost as a function of product
characteristics
• Method uses product-level and aggregate
consumer-level data
• Estimation is consistent with a structural model
of equilibrium in an oligopolistic industry
Key features of the modeling framework
• Joint distribution of consumer characteristics and
product attributes that determines preferences
over the products marketed
• Price taking on the part of the consumers
• Nash equilibrium assumptions on the behavior
of producers in an oligopolistic, differentiated
product market (new car market)
Notation
• ζ : characteristics of the individual making the
choice
• (x,ξ,p): product characteristics (observed,
unobserved and price)
– ξ represents for example style, prestige, reputation
• U(ζi,pj,xj, ξj ;θ ) = utility if person i makes choice j
• Choice 0 is the outside option of not purchasing
any of the goods (cars)
• If M is the number of consumers in the
market, then the vector of market
demands is given by
iid case, with no unobserved product heterogeneity
• Model not that appealing
– subject to IIA criticism
– Does not allow for interaction between
consumer and product-level characteristics (no
preference heterogeneity)
– Substitution patterns only depend on the δ’s.
Conditional on market shares, substitution
patterns do not depend on attributes of the
products
• Therefore, generalize the model to allow consumers to
have different preferences for different observable
characteristics (random coefficients model)
• Have information on aggregate distribution of
consumer’s income, that they want to incorporate
– Income distribution will come from CPS, not car data
• Final specification (Cobb-Douglas form):
• Price is expected to be correlated with
unobserved product attributes ξ – need
instruments
• Will estimate demand system by Method of
Moments, jointly with supply side
• Quantity demanded
• Use an inversion procedure of Berry (1994) to
transform the observed quantity, price and
characteristics data into a linear function of ξ
Cost side
• Multiproduct firms 1…F
• Each produces a subset of J possible products (car manufacturers
produce multiple types of cars that may compete with each other)
• Cost of producing a product is assumed to be independent of output
levels and log linear in vector of cost characteristics (wj,ωj)
Nash assumption
• Firm chooses prices that maximize profit-taking,
given the attributes of its products and the prices
and attributes of competitor’s products
• Only price is endogenous to the model,
attributes are not endogenous (do not decide
what types of cars to bring to market)
FOC
instruments
• Assume
– E(ωj |z)=0
– E(ξj |z)=0
• Instrument set consists of functions of the
characteristics and cost shifters of all other
products
Empirical Results
• Use data from the Automotive News Market Data Book
– Number of cylinders, number of doors, weight, engine
displacement, horsepower, length, width, wheelbase, EPA MPG,
front wheel drive, automatic transmission, power steering, air
conditioning