Industry Empirical Studies NEIO and Industry Models of Market Power Based on the lectures of Dr Christos Genakos (University of Cambridge)
Dec 16, 2015
IndustryEmpirical Studies
NEIO and Industry Models of Market Power
Based on the lectures of Dr Christos Genakos (University of Cambridge)
1. NEIO and the Structural Approach
2. Identification
3. Estimation and Hypothesis Testing
4. Examples: Porter (1983); Genesove and Mullin (1998)
5. Reduced form and Non-Parametric approaches
OUTLINE
New Empirical Industrial Organization (NEIO)
Most important weakness of the SCP paradigm was the lack of feedback mechanisms emphasized by game theory
Structure, Conduct and Performance are jointly determined by underlying primitives, institutional details and equilibrium assumptions
Two important lessons during the 70-80’s: every industry has many potentially important idiosyncrasies and these details matter a lot for the predicted conduct and performance
Perhaps we should abandon the hope of finding common patterns across industries and instead look at each industry more carefully
New Empirical Industrial Organization (NEIO)
Key features of NEIO:
No use of accounting data for costs and price-cost margins
Estimate market power fore each industry separately
Behavior of firms is estimated based on theoretical oligopoly models. This allows for explicit hypothesis testing on the degree of market power.
The degree of market power is identified and estimated. The inference of market power is based on the conduct of firms.
The Structural Approach
Suppose you had data on the following homogeneous goods market:
•P industry price
•qi output for each firm and Q the whole industry
•Y variables that shift the demand curve (income, weather, price of substitutes)
•W variables that shift the supply curve (price of inputs, weather, technology)
Could you uncover the extent of market power?
YES! Use the data to simultaneously estimate the elasticity of demand, marginal costs and firm conduct!
The Structural Approach
The key aspect of this approach is that it uses theory to specify the structure of demand and supply and in the process firm conduct is identified (pure magic!)
Let’s see how:
Demand function
Supply function
Profit function
),,( YQPP
),,( WqCC i
),,(),,( WqCqYQP iii
The Structural Approach
Marginal cost
Marginal Revenue
λi is a parameter which measures conduct; λi=0 price taker, λi=1 Cournot, λi=1/si Monopoly.
Optimality Condition gives us the supply relationship:
i
ii dq
WqdCWqMC
),,(),,(
iii qdQ
dPPYMR ),,(
i
iii
iii
s
P
WqMCP
qdQ
dPWqMCP
),,(
),,(
The Structural Approach
Two interpretations of λi parameter: (i) measures the gap between price and marginal cost, and (ii) an “aggregate conjectural variation”
Problem with interpretation (i): can justify only few values, not a continuous index
Problem with interpretation (ii): Corts (1999) critique that estimation of λi only unbiased if underlying method is the result of a conjectural variations eq.; underestimate if firms collude
Mkt Structure λ L
Competition/Bertrand
0 0
Cournot 1 -si/ε
Monopoly (collusion) 1/si -1/ε
1. NEIO and the Structural Approach
2. Identification
3. Estimation and Hypothesis Testing
4. Examples
5. Reduced form and Non-Parametric approaches
OUTLINE
Identification
Can we identify the market power parameter λi given only market level data on P, Q, Y and W?
Remember our supply function is:
Identification Problem is that Q and P are equilibrium values, simultaneously determined by the interaction of consumers and firms
QdQ
YQdPWQMCP
),,(),,(
Identification
To trace the supply equation we need variables that shift the demand curve (like income) but not the supply relationship
P1
Q1
P2
Q2
P3
Q3
D(Y2)D(Y3)
D(Y1) S
P
Q
Identification
Similarly, to trace out the demand curve we need variables that shift the supply (like wages) but not the demand relationship
P1
Q1
P2
Q2
P3
Q3
D
S(W1)
S(W2)
S(W3)
P
Q
Market Power Identification
Hence to identify demand (supply) function, we need at least one exogenous variable in the supply (demand) relationship that does not enter the demand (supply) function.
What about the market power?
Assume demand is given by
(1)
Assume also that marginal cost (not observable) is given by
(2)
Hence, supply relationship is
(3)
24131210 YQYYQP
WQMC 210
WQYQP 213110 )(
Market Power is NOT Identified
Shifting only the intercept of the demand curve does not identify market power
P1
Q1 Q
PMCC
MCMP2
Q2
D(Y2)
MR(Y2)D(Y1)
MR(Y1)
Market Power IS Identified
Shifting ΒΟΤΗ the intercept and the slope of the demand curve identifies market power
P1
Q1 Q
PMCC
MCM
D(Y2)
MR(Y2)D(Y1)
MR(Y1)
P2c
Q2c Q2
m
Market Power Identification
Hence, using econometric estimates of the demand and supply parameters (equations 1 and 3) we can obtain an estimate of the degree of market power, in our example here:
Note: identification is based on (arbitrary?) assumptions on the functional form of both the demand and marginal cost functions.
Note: If we assume constant marginal cost, we can estimate the degree of market power! Without shift in slope!!!
3
3
01 1
1
1. NEIO and the Structural Approach
2. Identification
3. Estimation and Hypothesis Testing
4. Examples: Joint Ex Committee; Genesove and Mullin
5. Reduced form and Non-Parametric approaches
OUTLINE
Estimation and Hypothesis Testing
Given a set of credible instruments, the econometrician estimates the demand and optimality condition either separately (2SLS) or as a system (3SLS, GMM) of equations
Two ways to estimate the market power parameter:
1) Estimate it as a “free” continuous variable. One then tests whether λ equals a value associated with a well-known model of competition (Bertrand, Cournot, collusion)
2) Estimate separate models corresponding to the various well-known models by imposing the particular value of λ and then use non-nested tests to choose among them
1. NEIO and the Structural Approach
2. Identification
3. Estimation and Hypothesis Testing
4. Examples: Joint Ex Committee; Genesove and Mullin (1998)
5. Reduced form and Non-Parametric approaches
OUTLINE
Genesove and Mullin (1998): conduct and cost in the sugar industry, 1890-1914
Genesove and Mullin’s aim is to test the validity of the NEIO methodology by comparing the estimated conduct parameter from a structural model to the calculated price-cost margins in the sugar industry
The simple production function together with its volatile history of high concentration, price wars and court cases at the beginning of the century make this industry the ideal test ground
Why should an industrial economist care about the answer?
The Sugar Industry and Production Technology
The industry during period of study is characterized by high levels of concentration, episodes of entry and price wars and later acquisition by or accommodation with ASRC
Refined sugar is a homogenous good with common technology:
RAWRAW PPkcc 075.126.0
Demand and Structural Model
The postulate a general demand formula
that encompass as special cases the quadratic, linear, log-linear and exponential
Optimality condition for a constant marginal cost, c, and conduct parameter, θ, is given by:
Instruments used: Cuban raw sugar imports, which are driven by harvest cycle, weather conditions, Cuban Revolution, Spanish-American War
)()( PPQ
c
cP )(
Supply Equation and Results
Substituting marginal cost function into pricing rule gives us:
Genesove and Mullin estimate different versions of their model depending on the demand function but also cost information availability
Results:
1) NEIO methodology does pretty good tracking calculated price-cost margins independent of the assumed demand function, although θ underestimated
2) Cost estimates sensitive to the model assumed, predictive power improved when add real info even if model misspecified
3) Estimating a “free” conduct parameter improves estimates
RAWkPc
P
0
1. NEIO and the Structural Approach
2. Identification
3. Estimation and Hypothesis Testing
4. Examples
5. Reduced form and Non-Parametric approaches
OUTLINE
Reduced form and Non-Parametric approaches
An alternative method to a full structural model is to use comparative statics and be able to distinguish firm behaviour
Good alternatives if important concerns on specification of structural model or data limitations
Basic idea: suppose that firms face a constant marginal cost; a shock causes the marginal cost to rise. If the market is competitive, the price will increase by the same amount as mc. If the market is oligopolistic, price will not change by the same amount.
Again we need to specify a demand function and functional form will matter for the results, but in principle we require less info than a full structural model
However, by imposing less structure we are able JUST to test whether the market is competitive or not, cannot measure the degree of market power
NEIO and Industry Models of Market Power:
References
*Bresnahan, T. (1982) “The Oligopoly Solution is Identified”, Economic Letters, 10: 87-92.
*Bresnahan, T. (1989) “Empirical Studies of Industries with Market Power”, Handbook of Industrial Organization, 1011-1057.
Corts, K. (1999) “Conduct Parameters and the Measurement of Market Power”, Journal of Econometrics, 88:227-250.
Genesove, D. and Mullin, W. (1998) “Testing Static Oligopoly Models: Conduct and Cost in the Sugar Industry, 1890-1914”, Rand Journal of Economics, 29:355-377.
Graddy K. (1995) “Testing for imperfect competition at the Fulton Fish Market”, Rand Journal of Economics, 26:75-92.
Next time: Differentiated Products Structural Models
*Berry, S (1994) “Estimating Discrete-Choice Models of Product Differentiation”, Rand Journal of Economics, 25:242-262.
*Hausman, J. (1997) “Valuation of New Goods Under Perfect and Imperfect Competition”, in Bresnahan and Gordon eds., The Economics of New Goods, NBER.
Nevo (2001) “Measuring Market Power in the Ready-to-Eat Cereal Industry”, Econometrica, 69:307-342.
*Nevo (2000) “A Practitioner’s Guide to Estimation of Random-Coefficients Logit Models of Demand”, Journal of Economics and Management Strategy, 9:513-548.
Berry, S., Levinsohn J. and Pakes, A. (1995) “Automobile Prices in Market Equilibrium”, Econometrica, 63:841-890.