Examples Introductiontoempiricalindustrial...

Introductionto empiricalindustrial

organization

Paul Schrimpf

QuestionsExamples

Methodology

Structuralempiricalmodels in IOEconomic Model

Econometricspecification

Example: Bresnahanand Reiss (1991)

References

Introduction to empirical industrialorganization

Paul Schrimpf

UBCEconomics 565

January 3, 2017


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

1 QuestionsExamples

2 Methodology

3 Structural empirical models in IOEconomic ModelEconometric specificationExample: Bresnahan and Reiss (1991)


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

References

• Aguirregabiria (2012) chapter 1

• Reiss and Wolak (2007) through section 4

• Einav and Levin (2010)


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Section 1

Questions


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Industrial organization

Industrial organization is about the structure of industriesin the economy and the behavior of firms and individuals inthese industries

• Departures from perfect competition• Strategic behavior• Scale economies• Transaction costs• Information frictions

• Impact on firms’ profits and consumers’ welfare


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

General Approach• Goal: model how profits and welfare are influenced by“exogenous” factors such as:

• Demand• Technology• Institutional features and regulation

• Also interested in:• Market structure: number of firms and their respectivemarket shares

• Market power: ability of firms to earn extraordinaryprofits

• Useful for:• Firm managers for e.g. choosing prices, evaluating amerger, predict the effect of introducing a new product,etc.

• Governments for e.g. choosing how to regulate naturalmonopolies, identifying and punishing anti-competitivebehavior, predicting the effects of taxes andenvironmental policy, etc


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Example 1

• These three examples come from Aguirregabiria (2012)• New product: A company is considering launching anew product, e.g., a new smartphone.

• Goal: choose price and estimate profits• Needs to predict demand and response of other firms• Data on sales, prices, and product attributes along withmethods from this course can be used


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Example 2

• Environmental policy: A government imposes newrestrictions on the emissions of pollutants fromfactories in an industry.

• New policy encourages adoption of a new cleanertechnology

• Changes cost structure, which will affect competition• E.g. if the new technology reduces variable costs butincreases fixed costs, then expect a decline in thenumber of firms and an increase in the average size(output) of a firm in the industry

• Data on prices, quantities, and number of firms in theindustry, together with a model of oligopolycompetition, we can evaluate the effects of this policychange in the industry on both firms and consumers


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Example 3

• Intel-AMD duopoly: in the CPU market has lasted manyyears with clear leadership by Intel with more thantwo-thirds market share.

• Questions: why has market structure and market powerbeen so persistent?

• Possibilities: large sunk entry costs and economies ofscale, learning-by-doing, consumer brand loyalty, orpredatory conduct and entry deterrence

• Data on prices, quantities, product characteristics, andfirms’ investment in capacity allow us to measure thecontribution of these factors


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Section 2

Methodology


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Methodology

Empirical industrial organization has a distinctmethodology focusing on structural economic models

• Complete economic model tailored to the question andindustry being studied

• Econometric model closely tied to economic model

• Trade off between breadth of questions you can answerand strength of assumptions


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Historical approaches toempirical IO

• 1940s and earlier: case studies• Careful descriptions of specific industries, firms, orevents

• Little quantification or formal tie to theory• 1950s-1970s: structure-conduct-performance

• Cross-industry regressions relating market structure tomarket outcomes

• E.g. regress Lerner index, (P − MC)/P, onHerfindahl-Hirschman index,

∑Ni=1 share2

i• Drawbacks: (1) ignores industry heterogeneity, (2) doesnot identify causal effect

• Late 1980s-present: new empirical industrialorganization

• Analyses of individual industries• Empirical analysis framed in terms of an economictheory of the relevant industry or a set of competingtheories


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Section 3

Structural empirical models in IO


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Ingredients of a structuraleconomic model in IO

1 Question2 Economic model

• Key features of the industry that are important toanswer our empirical question

• Should not be needlessly complicated

3 Data4 Econometric specification of model

• Economic models are deterministic and will nevermatch data, so need to add heterogeneity and/or shocks

5 Estimation

6 Reporting of results


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Model ingredients I

1 Description of the economic environment1 the extent of the market and its institutions;2 the economic actors; and3 the information available to each actor.

2 List of primitives1 technologies (e.g., production sets);2 preferences (e.g., utility functions); and3 endowments (e.g., assets).

3 Variables exogenous to agents and the economicenvironment

1 constraints on agents’ behavior; and2 variables outside the model that alter the behavior of

economic agents

4 Decision variables, time horizons and objectivefunctions of agents, such as:


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Model ingredients II

1 utility maximization by consumers and quantitydemanded; and

2 profit maximization by firms and quantity supplied.

5 An equilibrium solution concept, such as:1 Walrasian equilibrium with price-taking behavior by

consumers; and2 Nash equilibrium with strategic quantity or price

selection by firms.


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Econometric specification I

• Economic models are deterministic and will nevermatch data, so need to add heterogeneity and/orshocks

• Unobserved heterogeneity• E.g. firms vary in their productivity• Must clearly specify to whom what is

observed/unobserved — e.g. all firms’ productivities areunobserved by the econometrician, and firms observetheir own productivity but not others

• Optimization errors• Agents fail to exactly maximize their payoffs

• Measurement errors

• Functional forms and distributional assumptions• Economic models involve utility, profit, etc. functions ofunknown form. For estimation we often restrictfunctions and distributions to be of a known parametricform.


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Econometric specification II

• Reasons: (i) computational tractibility, (ii) limited datasize, (iii) identification (often questionable)

• E.g. utility CRRA, Cobb-Douglas production function,prodictivity log-normal, etc

• Identification: given the distribution of the observeddata, is there a unique value of model parameters thatmatch that distribution?


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Example: Bresnahan and Reiss(1991)

• Can learn a lot from market entry with very limited data• Cross-section of isolated markets where we observe

• Number of firms• Some market characteristics (prices and quantities notneeded)

• Identify:• Fixed costs• Degree of competition: payoffs = f(number of firms)


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Setting

• Questions:• Degree of competition: how fast profits decline with nm• How many entrants needed to achieve competitiveequilibrium (contestable markets)

• Data:• Retail and professional industries (doctors, dentists,pharmacies, car dealers, etc.), treat each industryseparately

• M markets• nm firms per market• Sm market size• xm market characteristics


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Model I

• N potential entrants• Profit of each firm when n active = Πm(n)

• Πm decreasing in n

• Equilibrium:

Πm(nm) ≥ 0 and Pm(nm + 1) < 0

• Profit function:

Πm(n) = Vm(n)︸︷︷︸variable

− Fm(n)︸︷︷︸fixed

=Smvm(n) − Fm(n)=Sm

(xDmβ − α(n)

)−

(xcmγ + δ(n) + εm

)

where• α(1) ≤ α(2) ≤ · · · ≤ α(N)


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Model II

• δ(1) ≤ δ(2) ≤ · · · ≤ δ(N)• Entry deterrence, firm heterogeneity, real estate prices

• Key difference between variable and fixed profits is thatvariable depend on Sm, fixed do not


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Estimation I

• Parameters θ = (β, γ, α, δ)• MLE

θ̂ = argmaxθ

M∑

m=1

logP(nm|xm, Sm; θ)

• Assume εm ∼ N(0, 1), independent of xm, SmP(n|xm, Sm; θ) =P (Πm(n) ≥ 0 > Πm(n + 1))

=P(

SmxDmβ − xCmγ − Smα(n) − δ(n) ≥ εε > SmxDmβ − xCmγ − Smα(n + 1) − δ(n + 1)

)

=

cutoff(n+1) cutoff(n)

0.0

0.1

0.2

0.3

0.4

-4 -2 0 2 4epsilon

dnorm

(epsilo

n)

=Φ(SmxDmβ − xCmγ − Smα(n) − δ(n)

)−

− Φ(SmxDmβ − xCmγ − Smα(n + 1) − δ(n + 1)

)


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Data

• 202 isolated local markets• Population 500-75,000• ≥ 20 miles from nearest town of 1,000+• ≥ 100 miles from city of 100,000+

• 16 industries: retail and professions, each estimatedseparately


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Results

• For most industries, α(n) and δ(n) increase with n

• Define S(n) = minimal S such that n firms enter

S(n) = xCmγ + δ(n)xDmβ − α(n)

• Varies across industries

• S(n)n ≈ constant for n ≥ 5• Contestable markets (Baumol, Panzar, and Willig, 1982) :an industry can be competitive even with few firms ifthere is easy entry


organization

Paul Schrimpf

QuestionsExamples

Methodology




References


organization

Paul Schrimpf

QuestionsExamples

Methodology




References

Aguirregabiria, Victor. 2012. “Empirical IndustrialOrganization: Models, Methods, and Applications.”

Baumol, WJ, JC Panzar, and RD Willig. 1982. “Contestablemarkets and the theory of industry structure.” .

Bresnahan, Timothy F. and Peter C. Reiss. 1991. “Entry andCompetition in Concentrated Markets.” Journal of PoliticalEconomy 99 (5):pp. 977–1009. URLhttp://www.jstor.org/stable/2937655.

Einav, L. and J. Levin. 2010. “Empirical IndustrialOrganization: A Progress Report.” Journal of EconomicPerspectives 24 (2):145–162. URL http://www.aeaweb.org/articles.php?doi=10.1257/jep.24.2.145.

Reiss, P.C. and F.A. Wolak. 2007. “Structural econometricmodeling: Rationales and examples from industrialorganization.” Handbook of econometrics 6:4277–4415. URLhttp://www.sciencedirect.com.ezproxy.library.ubc.ca/science/article/pii/S1573441207060643.

http://www.jstor.org/stable/2937655

http://www.aeaweb.org/articles.php?doi=10.1257/jep.24.2.145

http://www.aeaweb.org/articles.php?doi=10.1257/jep.24.2.145

http://www.sciencedirect.com.ezproxy.library.ubc.ca/science/article/pii/S1573441207060643

http://www.sciencedirect.com.ezproxy.library.ubc.ca/science/article/pii/S1573441207060643

Examples Introductiontoempiricalindustrial...

Documents

Transcript of Examples Introductiontoempiricalindustrial...