Market Concentration and

1
Market Concentration and Farm-Retail Price
Asymmetry in the U.S. Fluid Milk Market Brian
W. Gould Joseph Schoen Kyle Stiegert Send
comments to gould_at_aae.wisc.edu Applied Economics
Workshop October 19, 2005
2
Overview

• Introduction
• US Fluid Milk Prices Farm and Retail
• Overview of Food Retailer Concentration
• Previous Evidence of Milk Price Asymmetry
• Overview of the Econometric Model of Price
  Transmission
• Use of Non-Reversible Functions
• Panel Data with AR errors
• Overview of Market Level Data
• Preliminary Results (Farm Price Change)
• Conclusions/Next Steps

3
Asymmetric Price Transmission

• Why has the pricing of fluid milk become a major
  concern to the point that a Government
  Accountability Office (2004) inquiry was
  undertaken in 2004?

4

• Fluid milk price volatility has dramatically
  increased in recent history

5
(No Transcript)
6

• Retail-farm price spread continues its upward
  trend

7
Asymmetric Price Transmission

• Increased food retailing concentration over the
  last 3 decades
• A measure of concentration Herfindahl-Hirschman
  Index (HHI)
• where si is the value of sales by the ith firm
  and S is the total value of market sales in a
  particular market
• U.S. Dept. of Justice Guidelines wrt HHI
• 1000-1800 Moderately Concentrated
• gt1800 Concentrated
• Mergers adding 100 to HHI in concentrated
  markets raise antitrust concerns

8
Retail Grocery Sector HHI Values Selected Cities
9
Asymmetric Price Transmission

• Evidence of asymmetric cost pass-through for
  fluid milk retailers
• Kinnucan and Forker (1987) Annual U.S.
  Aggregate Data for Major Dairy Products
• Hansen et. al. (1994) Annual U.S. Aggregate
  Fluid Milk Data
• Romain et. al. (2002) Aggregate Monthly New
  York State Fluid Milk Data
• GAO (2004) Meta analysis of fluid milk pricing
• Estimates of price transmission of farm price
  increases were greater than for decreases
• Capps and Sherwell (2005) Monthly Sales Data
  for 2 and Whole Milk, 7 US Cities

10
Asymmetric Price Transmission

• Why should we be concerned with the
  characteristic of the price transmission process
  for fluid milk?
• Relationship between farm and retail prices
  provide insights into marketing efficiency
• Asymmetric price transmission may indicate
  possible market failure or abuse of market power
• Higher market power can increase asymmetry
  (Carman and Sexton, 2005 Sexton, Xia and
  Carman, 2004)
• Implications as to consumer and producer welfare
• Higher than normal retail prices
• Cotteril(2003) 1.00 more due to market power
  in NE
• Continued trend of variability of farm milk
  prices
• Continuation of concentration in food retailing

11
Asymmetric Price Transmission w/ Monopoly and
Concave Demand
P
MC?-MC MC-MC' but P? P gt P P' ? Cost
increases are passed on to retail prices more
completely than cost decreases
P
P
MC
P
MC
MC
MR
D
Q

• Under oligopoly, firms price between MR and D
• More competitive ? less asymmetric

12
Research Questions

• Is there evidence of asymmetric farm-retail
  price transmission in the US fluid milk sector?
• What is the role of imperfect competition?
• To what extent does spatial as opposed to
  market concentration explain asymmetric price
  transmission (Azzam, 1999)?
• Does the magnitude of asymmetry vary with market
  and spatial concentration across cities?
• What is the role of price variability in the
  degree of asymmetry?
• Implications of increased farm price variability

13
Econometric Model

• Based on the model originally developed by
  Wolframm(1971) and extended by Houck (1977)
• Approach has been used extensively in analyses of
  agricultural commodity price transmission
• Pork Boyd and Brorsen (1988)
• Fed Cattle Bailey and Brorsen (1989)
• Peanuts Zhang, Fletcher and Carley (1995)
• Wheat Mohanty, Peterson and Kruse (1995)
• Apples Willett, Hansmire and Bernard (1997)
• Rice, Beans, Tomatoes Aguiar and Santana
  (2002)

14
Econometric Model

• Houck (1977) approaches used in the analysis of
  fluid milk price asymmetry
• Kinucan and Forker (1987)
• Romain, Doyon and Frigon (2002)
• Capps and Sherwell (2005)
• Carman and Sexton (2005)

15
Econometric Model

• We improve upon previous research
• More current data encompassing periods of
  dramatic farm (Coop Class I) price increases and
  decreases
• April 2003 13.74
• June 2004 25.16
• Incorporate explicit measures of concentration as
  explanatory variables
• CR2, CR4 ( of total retail sales accounted for
  2 and 4 largest firms, respectively)
• Herfindahl-Hirschman Index (HHI)
• Spatial Concentration (Supermarkets/1,000 pop.)

83.1 ?
16
Econometric Model

• We improve upon previous research
• Incorporate a variety of definitions of our
  dependent variable to examine asymmetry to
  examine robustness of our conclusions
• Change in retail price (reported here)
• Change in farm-retail margin
• Percent change in farm-retail margin
• Given our use of city/retail firm data can
  examine differences across city and/or firms
• Incorporate alternative methods for accounting
  for panel nature of our dataset (Fixed Effects
  used here)

17
Econometric Model

• Basic Houck model
• RPt retail price in the tth period
• WPt wholesale price in the tth period (COOP
  price)
• ?WPIt change in wholesale price if there is an
  increase from t-1 to t
• ?WPDt change in wholesale price if there is a
  decrease from t-1 to t
• WPIt sum of period to period increases in
  wholesale price
• WPDt sum of period to period decreases in
  wholesale price
• Trend trend variable

18
Econometric Model

• Basic Houck model
• Nonreversibility (asymmetry) occurs if ß1 ? ß2
• The value of RP at any point t is
• Substituting this into the basic model

19
Econometric Model

• Basic Houck model

gt 0
lt 0
20
Example Data Construction
Time WP ?WPI ?WPD WPI WPD RP ?RP0
0 2 --- --- --- --- 5 ----
1 6 4 0 4 0 7 2
2 9 3 0 7 0 11 6
3 5 0 -4 7 -4 9 4
4 7 2 0 9 -4 9 4
5 6 0 -1 9 -5 10 5
6 3 0 -3 9 -8 8 3
7 5 2 0 11 -8 8 3
8 9 4 0 15 -8 12 7
9 14 5 0 20 -8 16 11
WP Wholesale (Farm) Price RP Retail Price
21
Econometric Model

• Our empirical model
• Analysis of Private Label Whole Milk eliminates
• Brand effect differences across firm/city
• Milk quality effects (fat content) across
  firm/city
• Test for asymmetry wrt both farm price and
  marketing cost changes (only farm price
  discussed)
• Incorporate Almon lag structure for marketing
  cost and coop price variables (displayed ßs
  derived from estimated Almon lag coefficients)
• AR(1) error structure (city/firm-specific ?s)
• Test for long-run and short-run asymmetry

22
Econometric Model
Number of Lags

• Our empirical model
• t time period, j city/firm, T total no. of
  months, J total no. city/firms, N P refer to
  pos/neg changes, MC marketing cost index, COOP
  cooperative Class I price, MONTH monthly
  DVs variables, CITY_FIRM city/firm DVs,
  NE_COMP Northeast Dairy Compact DV, Q volume
  of milk sales by firm

23
Hypothesis Tests

• Long-run farm price transmission asymmetry (H.1)
• Short-run farm price transmission asymmetry
  requires at least one of the following to occur
  (H.2)
• Incomplete transmission of farm price changes
  (H.3)
• Speed of Adjustment joint test (H.4)

24
Description of Our Data

• Retail Price Information
• Information Resources Incorporated (IRI)
• UPC Level
• For this analysis private label, whole milk
• Sales Values and gallons sold by firm (not
  outlet)
• 10 cities included in this analysis
• 1 city in each Federal Order
• Atlanta, Boston, Charlotte, Chicago,
• Cleveland, Dallas, Kansas City, Phoenix,
• Salt Lake City, Seattle, Tampa
• Monthly (weekly) March 1997-Dec. 2004 (94
  months)
• Data for Key Players (firms) in each city

25
Description of Our Data

• Farm Level Milk Price
• Announced Cooperative Class I price
• Prices charged fluid milk distributors by dairy
  cooperatives and bargaining federations
• Assume uniform processor costs
• Reasonable proxy for the cost of raw product to
  fluid milk bottlers
• Monthly
• Prices specific to each city

26
Description of Our Data

• Marketing Cost Index
• Based on Romain, Doyon and Frigon (2002)
• 54 Processing
• 26 processing labor
• 24 packaging costs
• 4 energy
• 36 Retail
• 31 retail labor
• 5 energy
• 10 Transportation
• BLS data used for all components
• Aggregated using Tornquist-Theil Index with
  budget shares assigned as above

27
Description of Our Data

• Market Concentration Measures
• Obtained from bi-annual Market Scope
  publication
• Complete census of value of sales in major
  markets across the U.S. for all firms
• Markets defined to match IRI data
• Values are interpolated between publication
  periods

28
Description of Our Data

• Food retailers in each city included in our
  analysis

City Retailers Mean HHI
Atlanta 3 1934
Boston 4 1620
Charlotte 4 1879
Chicago 2 1972
Dallas 5 1352
Kansas City 1 1255
Phoenix 3 1754
Salt Lake City 2 1432
Seattle 2 1554
Tampa Bay 5 2320
Above DOJ Limit
29
Preliminary Econometric Model Results

• The above assumed a one-way fixed effects model
• Tested using F-test proposed by Baltiga (1995)
• H0 Pooled Data
• F-Stat of 71.8 ? reject H0 ? Fixed Effects
• Lag length for MC and COOP determined using AIC
  criterion
• ? All lags for MC removed
• ? Two lag periods for COOP price change

30
Preliminary Econometric Model Results

• Summary of AR(1) Results 31 ? values
• Estimated via Cochrane-Orcutt procedure

All ?s significantly different from 0
31
Preliminary Econometric Model Results

• Summary of Results Fixed Effect Coefficients
  (City/Firms)

City Retail Firm Number Retail Firm Number Retail Firm Number Retail Firm Number Retail Firm Number
Atlanta 11 13 22
Boston 3 16 18 19
Charlotte 2 5 7 22
Chicago 4 9
Dallas 1 11 12 14 20
Kansas City 8
Phoenix 1 6 16
Salt Lake 1 17
Seattle 1 16
Tampa Bay 1 10 13 21 22
Indicates stat. sign.
15 out of 31
32
Preliminary Econometric Model Results

• Summary of Results Misc. Coefficients

Variable Coefficient S.E.
Q -0.00992 0.00079
HHI 0.11827 0.04230
NE_COMP -0.01531 0.03360
significant at least the 0.05 level
33
Preliminary Econometric Model Results

• Summary of Results Cumulative Change
  Variables, MC and COOP

Variable Coeff. S.D. Variable Coeff. S.D.
MC0P -0.00187 0.00218 MC0N -0.00501 0.00213
COOP0P 0.73112 0.04670 COOP0N -0.20482 0.03995
COOP1P 0.04981 0.05853 COOP1N -0.26976 0.04099
COOP2P -0.02904 0.05236 COOP2N -0.17158 0.03587
significant at least at the 0.05 level
34
Preliminary Econometric Model Results

• Summary of Results Hypothesis Tests

Null Hypothesis Value S.D.
Long-run farm price transmission symmetry 0.1057 0.0317
Short run price transmission symmetry after current month 0.5263 0.0640
Short run price transmission symmetry with additional month 0.3063 0.0686
significant at the 0.001 level
35
Preliminary Econometric Model Results

• Summary of Results Hypothesis Tests

Price Transmission Characteristic Value S.D.
Positive change in coop price completely transmitted -0.2481 0.0394
Negative change in coop price completely transmitted 0.3538 0.0379
Price Transmission Elasticity (Postive, eP) 0.3820 0.0200
Price Transmission Elasticity (Negative, eN) 0.3283 0.0192
eP- eP 0.0537 0.0161
significant at the 0.001 level
36
Next Steps

• Extend method by which market concentration
  incorporated into the model
• Change functional form to allow market
  concentration measures to impact hypothesis
  tests wrt price asymmetry
• Current model does not allow for this
• Modify HHI to account for population differences
  across city
• More fully explore the time series aspects of the
  data
• Co-integration-type model (Error Correction
  Model, Capps and Sherwell, 2005)

37
The End!