Motivation
Model
Estimation/Results
Summary
Definition of relevant market in beer industry:
Application of LA-AIDS model
T. Houška1 J. Bil2
1Chief Economist Department
Office for protection of competition ;
Department of Economics
Faculty of Economics and Administration, Masaryk University,
2Department of Economics
Faculty of Economics and Administration, Masaryk University
MUES, 11.10. 2012
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Motivation
Market definition - first step in market power assessment or
in investigating the intensity of market competition.
Why beer industry?
permanent growth in market concentration during last
decades, primarily due to mergers
the relevant market considered beer in nearly all market
decision in beer brewing industry
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
Procedure
2-stage demand estimation
Critical Loss Analysis
SSNIP test and on its basis definition of relevant markets
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
Data
Panel scanner data from Dominick’s Finer Food company
(36 stores in Chicago metropolitan area)
Weekly data from June 91 to November 95 (220
observation)
Detrended and seasonally adjusted (temperature,holidays)
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
Model overview
Multi-stage budgeting, 2 levels (+ more efficient estimates;
- weak separability assumption)
Beer brands divided into 6 segments: Premium, Light,
Craft, Imported, Dark, Non-alcoholic; each containing 3-5
beer brands
Linear approximation of Almost Ideal Demand System
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Data
Model specification
LA-AIDS specification
Bottom level (within segment):
wint = αin + βi ln(Ymnt /Pmnt ) +
I
j=1
γij ln pjnt + int , (1)
Lagged Stone index (+ avoid simultaneity bias)
ln Pmnt =
I
i=1
win(t−1) ln pint . (2)
Upper level (between segment):
wmnt = αmn + βm ln(Ybnt /Pbnt ) +
M
k=1
γmk ln Pknt + mnt , (3)
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Estimation
SUR-GLS fixed effect estimator (capture covariances of
error terms,between data variability very low)
Haussman test in most cases suggests endogeneity, but
we didn’t use IV estimator ("weak instruments",price
setting independent to demand shocks)
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
CLA
First version:
CLA calculates the level of sales, that a hypothetical
monopolist could lose due to a 5-10 % price increase in
order to preserve at least the level of profit before the price
increase
Break-even critical demand elasticity:Price elasticity of
demand, under which the original level of profit is equal to
the level of profit after the SSNIP
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
CLA cont.
Second version:
Profit maximizing CLA calculates the level of sales, that a
hypothetical monopolist could lose due to 5-10 % price
increase in order to achieve a profit-maximizing price
Critical demand elasticity is equal to the elasticity of
demand, under which the hypothetical monopolist would
increase its price by at least 5-10 % in order to set its price
on profit-maximizing level
Evaluation:
ε > εc,be, then the set of products does not constitute a
relevant market rendering the SSNIP not profitable.
ε < εc,be , the tested set of products is the relevant market.
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Outline
1 Motivation
2 Model
Data
Model specification
3 Estimation/Results
Estimation
Critical loss analysis
Results
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Bottom level results
Segment Brand Elasticity Standard Error Conditional elasticity
Premium Budweiser -4.285 0.128 -4.162
Old Style Regular -3.194 0.131 -3.145
Miller Genuine Draft -3.363 0.037 -2.920
Miller General Draft -3.467 0.036 -3.057
Light Amstel -2.998 0.197 -2.929
Michelob -2.377 0.102 -2.207
Coors Regular -2.434 0.132 -2.282
Miller -2.197 0.055 -1.417
Dark Augsburger -2.678 0.095 -2.438
Berghoff -2.221 0.073 -1.858
Becks -2.976 0.139 -2.706
Lowenbrau -1.472 0.136 -1.358
Killians Irish Red -3.153 0.121 -2.687
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Bottom level - results cont.
Imported Heineken -4.046 0.075 -3.773
Becks Regular -4.907 0.075 -4.672
Molson Golden -3.054 0.058 -2.947
Moosehead -3.237 0.059 -3.106
Craft Augsburger Regular -2.849 0.054 -2.523
Berghoff Regular -2.409 0.040 -2.009
Leinenkugel Premium Regular -2.038 0.074 -1.833
Samuel Adams Lager -5.069 0.135 -4.840
Non-alcoholic Coors Cutter -0.750 0.112 -0.895
Miller Sharps -1.784 0.035 -1.057
Odouls -2.524 0.066 -2.025
Table: Overall own price elasticities
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Bottom level - result evaluation
All own-price elasticity falls in (-4,-2), only non-alcoholic
beers inelastic demand
The most elastic ones: premium, imported
The least elastic ones: non-alcoholic, dark
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Upper (segment) level
Premium Light Dark Imported Non-alc. Craft
ln(Y/P) -0.006* 0.004* 0.006* -0.004* 0.009* 0.010*
ln(PPremium) -0.102* 0.050* 0.007* 0.010* 0.002 0.033*
εPremium -1.218 0.110 0.017 0.024 0.005 0.074
ln(PLight) 0.050* -0.115* 0.016* 0.028* -0.013* 0.034*
εLight 0.374 -1.902 0.125 0.213 -0.105 0.266
ln(PDark) 0.007* 0.016* 0.017* -0.007* -0.002 -0.032*
εDark 0.062 0.222 -0.760 -0.113 -0.039 -0.465
ln(PImported) 0.010* 0.028* -0.007* -0.004 -0.012* -0.015*
εImported 0.097 0.222 -0.053 -1.028 -0.091 -0.112
ln(PNon−alc.) 0.002 -0.013* -0.002 -0.012* 0.028* -0.003
εNon−alc. -0.024 -0.134 -0.025 -0.125 -0.747 -0.034
ln(PCraft) 0.033* 0.034* -0.032* -0.015* -0.003 -0.019*
εCraft 0.254 0.295 -0.288 -0.142 -0.032 -1.176
Constant 0.167* 0.167* 0.167* 0.167* 0.167* 0.167*
s.d. (0.001) ( 0.001) ( 0.000 ) ( 0.001) (0.000) (0.001)
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Estimation
Critical loss analysis
Results
Upper level - result evaluation
Size of elasticities in line with bottom level: higher for
premium light segment, lower for non-alc. and dark
segment
Interesting result: negative cross-price elasticities with
non-alc. beer segment
Results of SSNIP
SSNIP = 0.05 SSNIP = 0.1
m ε εc,be εc,pm εc,be εc,pm
Premium 0.175 1.218 4.45 3.64 3.64 2.67
Light 0.161 1.902 4.74 3.83 3.83 2.77
Dark 0.108 0.760 6.31 4.80 4.80 3.24
Imported 0.084 1.028 7.47 5.44 5.44 3.52
Non-alcoholic 0.185 0.747 4.26 3.51 3.51 2.60
Craft 0.150 1.176 5.00 4.00 4.00 2.86
Relevant market isn’t larger than segment
Jaroslav Bil
Motivation
Model
Estimation/Results
Summary
Summary
Hypothetical monopolist over any one of our proposed
segments would have enough market power to implement
the SSNIP, making each of the analysed segments a
separate relevant market.
Competition concern might be focused on the certain
segment not on all of beer.
Limitation:
credibility/reliability of segment classification
we analyse only demand side substitution of the market
(what about supply-side substitution ?)
Jaroslav Bil
Appendix For Further Reading
For Further Reading I
Baltagi, B. H.: Econometric Analysis of Panel Data. John
Wiley & Sons, 2008.
Davis, P., and Garces, E.: Quantitative techniques for
competition and antitrust analysis. Princeton University
Press, 2010.
Bronnenberg, B., Mela, C., and Boulding, W.: The
Periodicity of Pricing, Journal of Marketing Research 43 3
(2006), 477–493.
Deaton, A., and Muellbauer, J.: An Almost Ideal Demand
System, The American Economic Review 70 3 (1980),
312–326.
Jaroslav Bil
Appendix For Further Reading
For Further Reading II
European Commission: Notice on the Definition of Relevant
Market For The Purposes of Community Competition Law,
Official Journal of the European Communities 1209 97
(1997), 1–12.
Gokhale, J., and Tremblay, V.: Competition and Price Wars
in the U.S. Brewing Industry, forthcoming (2011).
Green, R., and Alston, J.: Elasticities in AIDS Models,
American Journal of Agricultural Economics 72 2 (1990),
442–445.
Jaroslav Bil
Appendix For Further Reading
For Further Reading III
Hausman, J., Leonard, G., and Zona, D.: Competitive
Analysis with Differenciated Products, Annals of Economics
and Statistics / Annales d’Économie et de Statistique 34
(1994), 159–180.
Johnson, F. I.: Market Definition Under the Merger
Guidelines: Critical Demand Elasticities , mimeo (1986),
1–13.
Rojas, Ch., and Peterson, E. B.: Demand for differentiated
products: Price and advertising evidence from the U.S.
beer market 26 1 (2008), 288–307.
Jaroslav Bil
Appendix For Further Reading
For Further Reading IV
Subhash, J.: Global Competitivness in the Beer Industry: A
Case Study , University of Connecticut, Food Marketing
Policy Center Working paper (1994), 1–36.
Jaroslav Bil