Market Structure and the Dynamics ofRetail Food Prices

Robert D. Weaver, Peter Chattin, and Aniruddha Banerjee

The effect of retail grocery market structure on the speed of adjustment of retail foodprices to changes in producer prices, real wages, and the cost of energy wasexamined for SMSAS. Evidence failed to support the implication of the Mason-Bainparadigm that increased concentration reduces market efficiency as reflected in speedof retail price adjustment. Evidence of strong intertemporal relationships betweenchange in producer prices and retail prices found for the categories meat, poultry,fish, eggs and cereal and baker products provide support to the hypothesis ofcost-push inflation.

Introduction

Whether coordination occurs explicitly, orthrough a gaming process, information flowand communication have been recognized ascrucial determinants of the success of outputand price control. The presence of imperfectinformation, high degrees of product differen-tiation, and search costs observed in contem-porary markets has motivated the so-calledmodern market theories of Schmalensee(1987), Salop and Stiglitz among others, whichargue that concentration may enhanceefficiency of information collection and pro-cessing leading to lower prices. These theoriesdirectly contradict the predictions of theMason and Bain paradigm that concentrationleads to higher profits and prices. In the ab-sence of resolution at a theoretical level, theeffect of market structure on market perfor-mance is clearly an empirical issue that de-serves consideration.

The objective of this paper is to analyze theimpact of market structure on a dimension ofeconomic performance not typically consid-ered: the speed of adjustment of prices tochanges in cost structure as reflected bychanges in input prices. By traditional percep-tion, changes in input prices would lead acompetitive market price to adjust through a

Au(hom are Associate Prot’e\so]- of Agricultural I+onomic>, Re-search A\s(stant in Agricultural Economic\, and AssisGurt Pro-fessor of Economics, The Pennsylvmria S[:itc University, respec-tively.

process of entry-exit forced by changes inprofits. In contmst, the presence of output andprice control in a concentrated market of im-perfectly competitive firms implies coordina-tion and this may suggest efficiency in adjust-ment of prices to maintain desired profitlevels, As Domberger noted:

“The implications for dynamic pricing behavior arenot difficult to rationalize in concentrated industrieswhere costs of search and communication amongsellers are relatively low, price adjustments can beeffectively coordinated and equilibrium in the in-dustry restored t%irly rapidly. The predictionwhich emerges (is) that the speed of price adjustmentwill tend to rise with the level of industrial concentra-tion, .“

This relation between speed of adjustment andmarket structure also emerges from the mod-ern market theorists cited above. As Salopand Stiglitz have argued, in the presence ofimperfect information, stability in sales mayrequire rapid adjustment in prices to preventconsumers from searching for lower pricesand switching stores. Price dispersion and,therefore, sales uncertainty are argued to in-crease with the cost of information. It followsfrom this modern market theory that when in-put prices vary over time, firms have a strongincentive to adjust prices quickly, and in har-mony, to minimize resulting price dispersion,high search costs for consumers, and possibleloss in sales.

Effects of market structure on speed ofprice adjustment are implied by a variety of

Weaver, Chattin, and Banerjee Market Structure and Food Prices 161

theories, and their existence is suggested byevidence of a relation between market struc-ture and prices and profits. Nonetheless, thenature of this relation is unresolved by eitherthe traditional theories following Mason andBain, or more recent modern market theories,leaving it an empirical question. The followingsections will present evidence concerning therelationship between the market structure ofretail grocery markets and the speed of adjust-ment of retail grocery prices to changes in inutprices as reflected by raw product prices andwages. The conceptual analysis has broaderapplication to the question of dynamics ofprice adjustment and the role of market struc-ture. Studies of the retail grocery industryhave generally established a relationship be-tween structure and either profit level (seeMarion et al. or Hall et al.) or price level (seeLamm (1981) or Cotterill) however, the effectof structure on speed of adjustment has beenleft largely unexplored.

Model Specification

The relationship between market structureand price level or dynamics requires economicmotivation. Consider a market equilibriumprice function derived from a micro-economictheory of firm behavior for the imperfectlycompetitive case. For example, for the casewhere outpu~ pricing power exists, the first-order condition for the profit-maximizing ofthe jth output choice by the hth firm in the ithmarket implies the following market specificprice function for each output j:

(1) P; = [dC$/dQ~][Ni j/(Nij + 6;)]

where

dC~/13Q~is the marginal cost for the hth firm,ith market, jth output,

Nti = (tlQti/dPti) (Pti/Qij), and

O; = (dQij/~Q!j) (Q~/Qu)

For particular specifications for the costfunction, market demand function, and reac-tion functions, (1) implies output price can bewritten as a function of the determinants ofthose functions. Since 6; = O for the case ofperfect competition, O! = 1 for monopoly, and0<81<1 for other cases of pricing power, O;reflects market power. Expression (1) can,therefore, be interpreted as establishing therelationship between market specific price

levels and the structure of market power with-in that market. Appelbaum (1979, 1982) andSchroeter have recognized this in approachesto parametrically estimate 0~ as a measure ofmarket power. Oligopoly theories of Cournot,Chamberlain, and Stigler as well as modernmarket theories suggest market power is de-termined by a vector Zij of market structurecharacteristics suggesting that 61 may be writ-ten:

(2) Oi = 6h(Zij)

Both Lamm (1981) and Cotterill recognizedthe implied relationship between market spe-cific price level and market power in studies ofretail grocery pricing in SMSAS and local mar-kets, respectively. In each case, they foundaggregate food price index levels to be sig-nificantly and positively affected by marketpower as measured by concentration ratios,market shares, or critical levels of marketshare. Solution of reaction functions (l),together with (2) imply the existence of an in-dustry level relationship between marketspecific prices and market structure:

(3) P; = Pij(Rj$ @j, Zij)

where

Rj and @jare arguments of C! reflecting in-put prices and fixed factors.

By its derivation from (l), (3) implies in-stantaneous adjustment in output price tochanges in its determinants. Modern markettheories cited above suggest that the presenceof imperfect information, costs of adjustment,and costs of coordination may imply the exis-tence of inertia in adjustment, and variation ininertia across markets with different marketstructures. These differences imply the func-tion Pij(.) would be expected to vary acrossmarkets.

Specific dynamic adjustment models couldbe derived from specific theories of firm be-havior, e.g. a general first-order differentialmight be implied (dropping subscripts):

(4) dP, = (1 – B)P, = f(P; – Pt-l)

where

P; is the equilibrium price given firm behav-ior.

The partial adjustment model consistent withminimization of costs of adjustment used byGriliches and adopted by Domberger is a

162 October 1989 NJARE

special case of (4). The implication of (3) and(4) is that the observed market price P, can bewritten in a dynamic reduced form determinedby a vector of current and lagged determinantsof P:. To proceed, for each market area (i) andjth retail product category, a generalized re-duced form of (3) and (4) may be written:

(.5) Pij = Cijpfi(L) + etifori=, . ..nandj= . . ..m

where

eij is drawn from a covariance stationaryunivariate stochastic process withplim eij = O, asymptotic variance-covariance u~ IT, and plim eij cij =o,

Pti is a T x 1 vector of retail priceobservations,

Cij is a T x K matrix of the determinantsof market specific price level, e.g.,Rj, @j, Zij in (3), and

Bij(L) is a K x 1 vector of polynomials inthe backward shift opertor L.

Equation (5) can be interpreted as a dynamicreduced form of a multiple product priceadjustment process.

While the parameters of (3ti(L) are consid-ered fixed over time, their variation acrossmarkets provides the basis for investigation ofthe relation between market structure and thespeed of adjustments of price as reflected bythe form of ~ij(L). In principle, the followingapproach will be taken. Dropping the productsubscript j, define S((3i) as a K x 1 vectorfunction of the elements of pi to be interpretedas a statistic summarizing the speed of priceadjustment in market i to each respective de-terminant. In order to investigate the effects ofmarket structure on S((3i), the following linearmodel is proposed:

(6) S(PJ = Zi~ + Vi

where

Zi is a K x P vector of measures of marketstructure in market

y is a P x 1 vector of parameters, andvi is identical y, and independently distrib-

uted with plim Vi = O, and plim ~Vi =o.

Definition of S(~i) results in characteriza-tion of the temporal distribution of adjust-ment. Four alternatives are considered. Forthe kth determinant of retail price define:

Total Magnitude of Response:

S~i = ~ b~-~~=o

Mean Lag of Response:

~=o

Percentage of Response:

Level of Response:

where

@’_Tis the coefficient of the t – ~ order inthe kth polynomial in pi.

Consistent estimates of ~~ _, follow fromordinary least squares (OLS) estimation of (5)and the assumed asymptotic properties of eij.Using Slutsky’s Theorem, it follows that S~foreach definition are consistent estimates of S:based on (3{_,. The following equation is im-plied for each definition of S!:

(7) $’~(~i) = S~(~i) + Wi,

where

wi is identically and independently distrib-uted with plim (wi) = plim(WieiJJ = plim (WiVi) = O.

Finally, combining (6) and (7) single equa-tion OLS provides consistent estimates of y.Asymptotic efficiency of estimates of y and(3ij follows directly from the assumed sto-chastic properties. In the simplest case whereSk(~J = ~~-~, (5) and (6) can be inter-preted as a variation of the random coefficientmodel of Swamy. Results presented by Dun-can for M-estimators for sequential estimationmay also be used to establish consistency ofthese estimates.

Empirical Implementation

Empirical implementation of(5) and (6) for theretail food industry requires specification ofsample characteristics such as definition ofproducts, store type, and geographic market,and identification of appropriate and availabledata. The retail food industry is composed ofsupermarkets with highly diversified food andhome products, grocery stores which are spe-cialized in food products, and small specialty

Weaver, Chattin, and Banerjee Mariie[ Structure und Food Price.~ 163

or convenience stores. Within any marketarea the composition of stores is relativelystable over time, while across markets greaterdifferences would be expected to reflect differ-ing consumer preferences. The focus of thisstudy will be on the retail food industry as awhole rather than on a particular store type.

Food products are widely recognized to bemarketed in unique local markets, the di-mensions of which are defined by travel costsand, as implied by modern market theories,search costs. In this study the market area willbe defined as an SMSA. This specification hasalso been adopted as an approximately accu-rate market area for retail foods by Lamm(1981), and Geithman and Marion.

The time. period of 1969–81 was selectedas a period in which input prices for foodretailing were effected by significant generalinflation processes. Monthly Bureau of LaborStatistics (BLS) data are available fortwenty-three SMSAS for food product catego-ry CPIS and individual product prices. Prod-ucts sampled by BLS vary across markets ac-cording to an observed base period salesvolume. This variation led Geithman and Mar-ion to conclude that use of BLS CPI indexesfor cross-sectional study of structure-price re-lationships may result in a negative bias in es-timates of any concentration-price relation-ship. For the current study, the BLS data isused to determine speed of adjustment pricesover time, a use for which BLS prices weredesigned (Rothwell).

The extent that cross-sectional variation inbase period product category composition isreflected in changes in prices over time is dif-ficult to determine. Geithman and Marionargued that BLS sampling methods are biasedtoward high volume store brands leading theselow-priced products to be included more fre-quently in concentrated than unconcentratedmarkets. By implication, as recognized byLamm (1979, 1981), price index levels arenegatively biased in concentrated markets.Whether this sampling bias would bias priceadjustment measures depends on whetherprice adjustment varies by product volume.To the extent that high volume products arehighly visible to the consumer, modern markettheories would suggest price adjustment forthese products would be faster. The conclu-sion could be drawn that BLS sampling pro-cedures may bias price adjustment measuresto reflect faster adjustment in concentratedmarkets than unconcentrated markets. To the

extent that this bias occurs, results reportedhere would overstate actual relationships be-tween speed of adjustment and market con-centration,

A wide range of specifications of producttypes and levels of aggregation (both withinand across type) would be both appropriateand of interest in the study of market structureeffects on the speed of price adjustment.Lamm (1981) used a product category annualcost, while Cotterill and others have usedstore level market baskets. The present studywill limit its focus on the three categories ofproducts for which BLS category CPI indexesare available: (1) cereal and bakery products,(2) meats, poultry, and fish, and (3) fruits andvegetables. These categories account for im-portant portions of U .S. farm production, yetthey present variation in the extent of process-ing. By use of category indexes rather than theaggregate food price index, prices reflect vari-ation of product categories across the sampledSMSAS. Respectively, in 1982 these catego-ries accounted for 1470, 327o, and 16’%0of aver-age weekly per person at home food ex-penditures for U. S, urban households.

The composition of the BLS categories is asfollows: cereal and bakery products (CB) in-cludes basic grains, milled and unmilled (e.g.wheat flour and rice), and bakery productssuch as bread, crackers, and cookies; meats,poultry, fish, and eggs (MPFE) focuses onwhole and cuts of fresh meat with the excep-tion of canned tuna and sardines and fresheggs; fruits and vegetables (FV) includesfresh, frozen, and canned fruits and vege-tables. Producer prices for CB and MPFE forthe same categories were employed as avail-able on a monthly basis from BLS. BLSchanges the category definition between theCPI for FV and the PPI for FV where onlyfresh fruits and vegetables are sampled. Be-cause wholesale or producer prices are deter-mined in nation-wide markets rather than localSMSA markets, national aggregate producerprices were employed.

In addition to wholesale and raw productprices, prices of other inputs involved in re-tailing food products must be considered asdeterminants of the retail price level. Al-though omitted by Cotterill, German andHawkes find wages and energy account for thepredominant proportion of the cost of provid-ing retail services. The roles of these inputprices will be explored below. Labor costswere collected from county level unemploy -

164 October 1989 NJARE

ment insurance data available for ten SMSASfrom BLS monthly employment and quarterlypayroll reports for the retail food industry.Use of county level data allows a closeapproximation to SMSA labor markets. Em-ployment and Earnings data was unavailablefor the retail food industry on a monthly basisexcept at a national aggregate level, Themonthly SMSA level CPI for energy based onthe cost of gas and electricity was used for theprice of energy.

Market Structure

Measurement of market structure by the ele-ments of Zi in (6) follow from theories of oli-gopoly (Cournot, Chamberlain, and Stigler)which suggest that the number of firms partici-pating, disparity in size, and achievement ofthresholds in output control affect the feasibil-ity and effectiveness of price setting power.The implication of modern market theories isthat imperfect information leads to coordina-tion costs which increase with (1) the structureof sales distribution as measured by marketconcentration, (2) market growth, or thestability of market size, and (3) the existenceof barriers to entry.

Measurement of market concentration re-quires choice of a statistic that reflects varia-tion in the extent of market dominance orprice setting power, as well as the distributionof that power among firms. Alternativesadopted in empirical studies have ranged fromindividual firm market shares (Kwoka 1979,1981), and 4-, 8-, and 20-firm concentrationratios, to Stigler and Schmalensee’s use ofHerfindahl indexes to describe the distributionof firm level sales. Lamm (1981) found 2-, 3-,and 4-firm concentration ratios to be positive-ly related to retail food price levels though useof firm market shares improved model fit mar-ginally, More recently, Cotterill found theHerfindahl index, market shares, and marketconcentration ratios to effectively representmarket structure as a determinant of retailfood price levels in a cross-section of Vermontfirms. However, Cotterill finds the Herfmdahlindex to outperform market share measuresand market share measures to outperformmarket concentration ratios, a result alsofound by Shephard, Ravenscraft and Marionet al, In this study, data availability limits theenquiry to the use of firm concentration ratios.The 4, 8, and 20 firm concentration ratios(CR4, CR8, and CR20) are employed as re-

ported for 1972 in Grocery Retailing Concen-tration (BLS). These ratios reflect differentperspectives on the distribution of marketpower among firms. While the 4-firm measurefocuses on the extent of highly concentratedmarket power, the 8-firm measure indicatesthe extent to which a larger group of firms mayexist which could coordinate prices, In con-trast to the 4-firm ratio, the 20-firm ratio in-dicates the extent to which market sales areclaimed by a group of firms which is largeenough that coordination costs may precludeexpression of market power.

In addition to the sales structure of an in-dustry, traditional oligopoly theory as well asmore recent theories of entry deterrent pricing(e.g. Gaskins) suggest that the extent of barri-ers to entry influences the firm’s pricingstrategy and, thereby, an effect on the speedof adjustment would be expected. Kwoka(1981) has found a variety of measures of scaleof plant size to be positively related to thelevel of the price-cost margin. In studies of re-tail food prices, Lamm (1981) has found aver-age store size (measured by the number ofmarket baskets sold per store) to be significantand negatively related to price level, This re-sult suggests scale economies resulted in costreductions rather than barriers to entry whichled to increased prices, Cotterill also in-troduced store size as a determinant of firmlevel prices and found it insignificant in linearmodels, though significant when introduced inquadratic form. These results were interpretedas indicating an initial negative effect on pricesas scale reduced costs; however, as scale con-tinues to increase Cotterill argued store differ-entiation is achieved through expanded prod-uct line which establishes market power toset higher prices. In contrast, Salop and Stig-litz’s modern market theory would predict theeffect of scale on price level and speed ofadjustment to be negative, and positive, re-spectively. In this study, firm size is measuredby firm average square feet of retail sales areafor 1972 as reported in Grocery Retailing.

Market growth has been argued to result inincreased prices and profit levels as oligopo-lists attempts to ration available capacity. Al-though Kwoda (1979) among others havefound significant positive relations betweenmarket growth and prices, others have foundnegative relations. Scherer, Schmalensee, andParker and Connor found a significant andpositive relationship in a price-cost marginmodel as well as one explaining the difference

Weaver, Chattin, and Baneijee

between store and manufacturer brand prices.Cotterill found an insignificant and negativerelationship with firm level prices. Modernmarket theories would suggest that as marketgrowth implies instabilityy in past results ofcoordination, firms may be expected to reduceprices, and adjust them more quickly in re-sponse to exogenous factors. Market salesgrowth were employed as a measure of marketgrowth as available from Bureau of Census,Retail Trade Division, Subject Statistics for1972.

Empirical Results

Dynamic reduced forms such as (5) were es-timated for the three product categories (CB,MPFE, and FV) for the twenty-three SMSASfor which CPI indexes are available, Prior toestimation, sufficient conditions for covari-ance stationarity were established for all pricevariables by first differencing in logarithms.For the MPFE CPI series, annual seasonalitywas also removed by differencing. 1After pre-

1Independentvariablesin time series are definitionallystochas-tic regressors. Covariance stationarity is required to ensure that

Market Structure and Food Prices 165

filtering, the dynamic reduced forms were es-timated as linear forms implying variables canbe interpreted as rates of change and es-timated parameters as coefficients of rates ofchange, or elasticities of the CPI to a change inthe PPI.

The most general model including PPI,wages, and the price of energy was estimatedfor ten SMSAS. Results indicated that wageswere consistently insignificant (out to sixteenlags) across products and SMSAS. The dy-namic reduced forms were re-estimated with-out wages for the complete set of twenty-threeSMSAS for which data are available. For allSMSAS for MPFE and CB, the lag structurefor the PPI indicated statistically insignificanteffects beyond the sixth period. Table 1summarizes the estimated lag structures forthe PPI. For the FV category, no statisticallysignificant relation was found between the PPIand the CPI. In part, this may have been due

the sampleinformationmatrixconvergesasymptoticallytotheun-derlyingcovariancematrixwhichcharacterizesthestochasticpro-cesswhich generated the sample. Sample specificprefilteringtoachieve sufficientconditions for covariance stationarity is a pre-requisite for establishing an asymptotic relation between es-timated sample and populationstatistics, see Goldbergeror Fom-by, Hill, and Johnson.

Table la. Response of Percent Change in CPI to Percent Change in PPI by SMSA: Meat,Poultry, Fish, and Eggs Category

Lag (months)

City o 1 2 3 4 5 6 R2 D.W.

AtlantaBaltimoreBostonBuffaloChicagoCincinnatiClevelandDallasDetroitHonoluluHoustonKansas CityLos AngelesMilwaukeeMinneapolisNew YorkPhiladelphiaPittsburghSt. LouisSan DiegoSan FranciscoSeattleWashington, DC

0.39830.31400.26440.40790.41550.42760.43810.41930.49880.06540.42360.38050.37300.34010.34670.34290.38210.37620.43070.43170.39030.31970.3826

0.21950.23710.20850.20580.20950.28250.34150.34590.26170.21750.27930.28850.29260.25000.21260.20400.23760.26070.24750.30190.28780.26070.2811

0.19970.16750.15620.20290.12930.10700.08920.0763

0.;780.13440.16340.07830.15810.14880.14670.11120.14880.1445

—0.07440.14380.1526

—————

0.0753———

0.0567————

0.0701——————.—

—

0.;55—

0.0881—

0.;210.1181

——

0.08740.0857

0.;98—

0.1325—

0.09410.08220.11990.1395

—

————

0,0857——————————————————

— .63— .64— .57— .60— .58— .63— .64

.690.;45 .650.0566 .51

— .63— .61— ,65— .61— .62— .63— .60— .55— .64— .67— .64— .54— .56

2,372.292.342.092.642.482.422.122.311.982.192.242.232.122.362.342.412.382.331.992.312.382.41

Only parameter estimates significant at the 0.1 level (using two-tailed tests) are reported.

166 October 1989

to differences in category definiticm. The CPIincludes frozen and processed, while the PPIincludes only fresh. Further analysis of the FVcategory was not pursued.

Because covariance stationarity was es-tablished before estimation it is not surprisingthat estimated residuals were found to bewhite noise in all cases. The estimated lagstructure suggests that substantial adjustmentof the MPFE product CPI occurs within twomonths, while for the CB product categorythe lag structure varies significantly acrossSMSAS. These results are consistent withthose found by Lamm (1981) and Lamm andWestcott for national aggregate data.

The estimated lag structures for CPI adjust-ment to changes in PPI were employed to es-timate five alternative SMSA specific speed ofadjustment statistics. Substantial variation ineach of these statistics was apparent acrossSMSAS. Interpretation of the speed of adjust-ment coefficients follows directly from theinterpretation of estimated coefficients aselasticities. The magnitude, mean, and per-centage response statistics represent the sum,mean, and period specific percentage of theelasticity of the CPI for change in the PPI overthe lag structure identified and estimated. Es-timates of equation (6) are reported in Table 2

NJARE

and 3 for each of the speed of adjustment coef-ficients, and for alternative concentrationratios,

For the cereal and bakery product category,Table 2 indicates variation in total magnitude,and percentage responses in t, t-1, and t-3appears related to market structure. In allcases, results are robust across different con-centration ratios indicating that concentrationis a significant positive determinant of totalmagnitude and percentage of response duringthe current period t and period t-3. A negativerelation is indicated for period t-1. These re-sults provide empirical evidence in support ofmodern market theories which hypothesizethat concentration increases speed of adjust-ment through reduction in costs of and in-crease in feasibility of coordination amongfirms. As Stiglitz has noted, this implies anefficiency gained due to concentration overpurely competitive structures where coordina-tion is impossible, and information and searchcosts are high.

Barriers to entry as measured by firm aver-age floor space was found a significant andnegative determinant of total magnitude andpercentage response in period t, a result con-sistent with that found by Lamm and Cotterillin models of price levels. A significant positive

‘I’able lb. Response of Percent Change in CPI to Percent Change in PPI by SMSA: Cerealand Bakery Products Category

Lag (months)

Citv o 1 2 3 4 5 6 R2 D.W.

AtlantaBaltimoreBostonBuffaloChicagoCincinnatiClevelandDallasDetroitHonoluluHoustonKansas CityLos AngelesMilwaukeeMinneapolisNew YorkPhiladelphiaPittsburghSt. LouisSan DiegoSan FranciscoSeattleWashington, DC

—0.3631

0.3227.

0.37950.22140.3956

—0.34340.23470.5634

—0.39000.48690.2421

—0.45000.28470,36200.37450.36470.39420.49680.1771

—

0.20280.35820.28530.22850.2185

————

0.55870.3292

0.3398——

0.1402

——

0.19300.21170.2429

—

0.2607—

.55 2.74

.48 2.45

.56 2.46

.50 2.61

.58 2.32,47 2.48.37 2.29.53 2.50.18 2.88.56 2.10.57 2.29.59 2.28.62 2.09.19 2.82.59 2.14.69 2.38.56 2.31.57 2.43.47 I .99.52 2.45.68 2.13.57 2.38.39 2.21

0.24110.2381

—–0.2332

——

—0.2434

— ——

———

——

0.;71—

—0.19620.22320.2709

—

—0.56750.25820.2483

———

———

—0.20520.17300.1683

—0.4153

—

—0.;73

—

————

—0.1726

.0.24770.1891

———

——

0,2385——

0.2195——

—–0.2665–0.1873

—–0,2761

——

——

0.2334—

0.2879—

———

0.1838— —

———

0.20730,3207

——

–0.3099

——

0.2658

—0.2525

——

0.2674

Only parameter estimates significant at the 0.1 level (using two-tailed tests) are reported.

Weaver, Chattin, and Banerjee Market Structure and Food Prices 167

Table 2. Market Power and the Speed of Adjustment of Retail Food Prices: Cereal andBakery Products

Speed ofAdjustment Concentra- Concentra- Barriers Market De~.St&istic tion Ratio Intercept tion Ratio to Entry Growth R2 Me&r

Total Magnitude

Mean Lag

PercentageResponse (t)

PercentageResponse (t – 1)

PercentageResponse (t – 2)

PercentageResponse (t – 3)

4-firm

8-firm

20-firm

4-firm

8-firm

20-firm

4-firm

8-firm

20-tirm

4-firm

8-firm

20-tirm

4-firm

t+-firm

20-firm

4-firm

8-firm

20-tirm

t-statistics are presented in parentheses.

.69(2.49)

.76(2.56)

.71(2.02)

1.09(.92).85

(.71).39

(.27)

.41(1.12)

.46(1.20)

.45(.99)

.86(1.93)

1.16(2.16)

.91(1.97)

– .070(::;:)

(.34)– .29(.56)

– .21(.70)

–.14(.42)

–.14(.34)

.0044(1.35)

.0029(.93).0037

(.85).015

(1.11).017

(1.34).02.5

(1.41)

.0033(.78).0022

(.54).0024

(.43)

– .0098(1.89)– .013(2.03)– .0094(1.92)

– .0014(-.28)

– .000099(.022).0021

(.33)

.00094- (2,67)

.0072(2.04)

.0074(1.43)

– .000038(1.80)– .000039(1.78)– .000044(1.89)

.000051(.56).000044

(.49).000012

(.13)

– .00005(-1.79)

– .00005(1.79)– .00005(1.80)

.000042(1.21)

.000062(1.76)

.000045(1.31)

.000017(.55).000017

(.54).000014

(.41)

– .000019(.85)

– .000022(.89)

– .000030

.11(.49).079

(.33).066

(.27).16

(.17).14

(.15).084

(.089)

– .020(.06)

-.061(. 16)

– ,061(.20)

– .38(1.02)–.31(.86)

– .34(.93)

.44(1.26)

.45(1,34)

.48(1.41)

.031(.12)

– .032(.12)

– .079

.26

.22

.21

.10

,12

.14

.20

,18

.18

.27

.29

.28

.12

.11

.12

.34

.24

.15

.74

2.26

.26

.41

,21

.15

(1.11) (.28)

relationship with percentage response in pe-riods t-1 was found. Results were, in general,robust across models with different concentra-tion measures. The results suggest that slowerimmediate and less total adjustment as scaleincreases. After one period, in t-1 larger scalefirms adjust faster in the period. These resultsare consistent with increased scale introduc-ing diseconomies into the speed of firm reac-tion to changes in cost. While increased con-centration appears to enhance coordinationand speed of adjustment among firms, in-creased size of firm appears to decrease the

firm’s speed of adjustment. Market growthappears to be a consistently insignificant de-terminant of speed of adjustment, a result alsofound by Cotterill in a model of price levels.Overall, the results appear to support themodern market theories, not the Bain andMason paradigm.

Results for Meat, Fish, Poultry, and Eggsare reported in Table 3. For this product cate-gory, results vary substantially across speedof adjustment statistics. Results fail to indicateany relationship between percentage responsein t, t-2, and t-3 and measures of market struc-

168 October 1989 NJARE

Table 3. Market Power and the Speed of Adjustment of Retail Food Prices: Meat, Fish,Poultry, and Eggs

Speed ofAdjustment Concentra- Concentra- Barriers Market Dep.Statistic tion Ratio Intercept tion Ratio to Entry Growth R2 Mean

Total Magnitude

Mean Lag

PercentageResponse (t)

PercentageResponse (t – 1)

PercentageResponse (t - 2)

PercentageResponse (t - 3)

4-firm

8-firm

20-firm

4-firm

8-tirm

20-firm

4-firm

8-firm

20-firm

4-firm

8-firm

20-firm

4-firm

8-firm

20-firm

4-firm

8-tirm

20-firm

.86(5.09)

.87(5.01)

.92(4.46)2.70

(6.40)2.75

(6.36)2.89

(5.62)

– .072(.36)

– .077(.36)

–.13(.51)

.38(4.27)

.37(4.03)

.36(3.29)

.25(1.56)

.21(1.29)

.17(.89)

.14(1.03)

.19(1.43)

.26

– .0016(.84)

– .0016(.89)

– .00024(.95)

– .0064(1.33)– .0065(1.42)– .0088(1.39)

.00032(.139).00036

(. 16).0011

(.37)

.0007(.76).00086

(.88).0010

(.75)

.00089(.48).0013

(.78).0020

(.84)

.00098(.64)

– .000071(.048)

– .0010(1.58)

t-statistics are presented in parentheses.

.000021(1.60)

.000022(1.65)

.000025(1.81)

.000038(1.18)

.000041(1.27)

.000052(1.53)

.000010(.67).000010

(466).0000087

(.53)

– .000012(1.75)– .000012(1.81)– .000013(1.88)

– .0000063(.51)

– .0000069(.56)

– .0000095(.74)

– .0000084(.82)

– .0000083(.79)

– .0000067(.61)

– .096(.67)

– .09(.65)

– .085(.62)

– 1.98(3.38)

–1.17(3.39)

–1.15(3.33)

.18(1.078)

.18(1.088)

.18(1.12)

.18(2.43)

.18(2.45)

.17(2.39)

.072(.53).075

(.57).072

(.55)

.039(.35).020

(.18).014

(.13)

ture. However, a strong significant relation isfound for mean lag and percentage response int-1 (the period where mean percentage re-sponse is largest). Results for total magnitudeof response suggest a significant positive roleof average firm size, opposite of that found forcereal and bakery products. Results are not in-variant across speed of adjustment statistics.For the mean lag, market concentration isnegatively related to speed of adjustment, aresult that is robust across models withdifferent concentration ratios. Barriers to en-try as measured by average firm size is posi-

.20

.20

.21

.49

.49

,49

.079

.079

.086

.41

.42

.47

.047

.068

.074

.071

.047

.061

.85

1.93

.11

.45

.297

.16

tively related to mean lag. In both cases, theconcentration ratio and the barriers to entrymeasure are only marginally significant. Mar-ket growth, in contrast to findings for cerealand bakery products, is a significant, negativedeterminant. This result is consistent with themodern market theories which suggest that inexpanding markets, firms adjust prices morerapidly as a strategy for maintenance of mar-ket share,

The distribution of speed of adjustmentacross periods is reflected in the percentage ofresponse statistics. The model of this statistic

Weaver, Chattin, and Banerjee Market Structure and Food Prices 169

for t-1 indicates that barriers to entry have aconsistently negative significant effect on thewithin period adjustment. As firm size in-creases speed of adjustment slows. This sameresult was found for cereal and bakery prod-ucts. Market growth has a positive and signif-icant effect with period response in t-1, a re-sult consistent with the finding that marketgrowth reduces mean lag of response, One re-sult found consistently across speed of adjust-ment coefficients was the absence of a role ofconcentration. Across all speed of adjustmentcoefficients and concentration ratios, the con-centration ratio was insignificant except forthe mean lag where it is only marginally sig-nificant.

Conclusions

Specific conclusions can be drawn from the re-sults presented above, Variation in the rate ofchange of the retail food prices over the timeperiod studied can be explained by variation inraw material prices measured by the PPI andenergy prices. This confirms that raw materialprices play an important role in retail foodprice inflation processes. This result suggeststhat effective agricultural price stabilizationprograms may lead to benefits realizedthrough impacts on the retail food price infla-tion process, Wages did not appear to play asignificant role; however, difficult y of measur-ing this variable may have led to this weakrelationship. The speed of adjustment of retailprices was found by vary substantially acrossSMSAS and this variation was explained bymarket structure. Results appear invariant tothe definition of the concentration ratio. Dataavailability precluded the use of other mea-sures of market structure such as marketshares or the Herfindahl index.

The absence of a role of concentration mea-sures for the meat category (MPFE), and agenerally positive relationship found for cerealand bakery products provide support for themodern market theory that concentration mayenhance efficiency in price adjustment and, forthe case of price adjustment, fails to supportthe implication that may be drawn from theBain-Mason paradigm that concentration re-duces the efficiency of market performance.Importantly, results support the conclusionthat consumers may be benefited by concen-tration which leads to more rapid price adjust-ment.

Overall the results for these product catego-ries provide evidence that is consistent withthe hypothesis that market structure affectsthe speed of price adjustment in the retailgrocery industry, The evidence does not rejectthis hypothesis; however, it can not be judgedas strongly supporting the hypothesis. Im-portantly, the results fail to support the im-plications of the Mason-Bain paradigm and, in-stead, provide support for the modern markettheories which suggest concentration may en-hance some market performance characteris-tics.

References

Applebaum, E. “Testing Price Taking Behavior. ” Jour-nal of Econometrics, 9(1979):283–294.

Applebaum, E. “The Estimation of the Degree of Oli-gopoly Power. ” Journal of Econometrics,10(1982):313-331.

Bain, Joe S. “Relation of Profit Rate to Industrial Concen-tration: American Manufacturing, 1936-40. ” Quar-terly Journal of Economics, 65(1951):293-324.

Bureau of Labor Statistics, U.S. Department of Labor.Consumer Price Index, Detailed Report, variousyears.

Bureau of Labor Statistics, U .S, Department of Labor.Employment and Earnings, various years.

Bureau of Labor Statistics, U.S. Department of Labor.Estimated Retail Food Prices by City, various years.

Bureau of Labor Statistics, U.S. Department of Labor.Producer Prices and Price Indexes, various years.

Bureau of the Census, Retail Trade Division, U.S. De-partment of Commerce, Grocery Retailing Concen-tration in Metropolitan Areas, Economic CensusYears, 1954-1972.

Chamberlain, Edward. The Theory of Monopolistic Com-petition, sixth edition (London: Oxford UniversityPress), 1949,

Cotterill, Ronald W. “Market Power in the Retail FoodIndustry: Evidence from Vermont. ” Review of Eco-nomics and Statistics, 68(3),(1986):379-386.

Cournot, Augustin. Researches into the MathematicalPrinciples of the Theory of Wealth, translated byNathanial Bacan, Macmillan, New York, 1897.

Domberger, S. “Price Adjustment and Market Struc-ture.” The Economic .lournal, 89(1979):96-108.

Duncan, Gregory M. “A Simplified Approach to M-Estimation with Application to Two-Stage Es-timators. ” Journal of Econometrics, 34(1987):373-389.

Fomby, T. B., R. C. Hill, and S. R. Johnson. AdvancedEconometric Methods, Springer-Verlag, NY, 1984.

Gaskins, D. W. “Dynamic Limit Pricing: Optimal PricingUnder Threat of Entry.” Journal of EconomicTheory, 3(1971):306-322.

Geithman, F. and Bruce Marion. “A Critique of the Useof BLS Data for Market Structure Analysis. ” Amer-

170 o~tober 1989 NJARE

ican Journal of Agricultural Economics,60( 1978):701-705.

German, G. and G. Hawkes. Operating Results of FoodChains, Cornell University, 1980.

Goldberger, A. S. Econometric Theory, John Wiley andSons, Inc., NY, 1964.

Griliches, Z. “Distributed Lags: A Survey. ” Economet-rics, 35(1967):16-49.

Hall, L., A. Schmitz, and J. Cothern. “Beef WholesaleRetail Marketing Margins and Concentration. ” Eco-nomics, 46(1979):295-300.

Kwoka, John. “The Effect of Market Share Distributionon Industry Performance. ” Review of Economicsand Statistics, 61(1981):101–109.

Kwoka, John. “Does the Choice of ConcentrationMeasure Really Matter?” Journal of Industrial Eco-nomics, 29(1981):445-453.

Lamm, R. M. “Dynamics of Food Price Inflation. ” West-ern Journal of Agricultural Economics 4(1979):119-132.

Lamm, R. M. “Prices and Concentration in the Food Re-tailing Industry. ” Journal of Industrial Economics,30(1981):67-78.

Lamm, R. M. and P. C. Westcott. “The Effects of Chang-ing Input Costs on Food Prices. ” American Jourmdof Agricultural Economics, 63(1981):187–196,

Marion, B. W., W. F. Mueller, R. W. Cotterill, F. E.Geithman, and J. R. Schmelzer. “The Price andProfit Performance of Leading Food Chains. ” Amer-ican Journal of Agricultural Economics, 61(1979):420-433.

Mason, E. S. “price Production Policies of Large ScaleEnterprises. ” American Economic Review, 29(1939):61-74.

Means, G. C. Industrial Prices and Their Relalive Irr-flexibility. United States Senate Document 13, 74thCongress, 1st Session, Washington, D. C., 1935.

Parker, R., and J. Connor. “Estimates of Consumer LossDue to Monopoly in the U.S. Food Manufacturing

Errata

Industries. ” Committee on Small Business, U.S.House of Representatives, 1980.

Ravenscraft, David. “Price-Raising and Cost-ReducingEffects in Profit-Concentration Studies: A MonteCarlo Simulation Analysis. ” Ph.D. thesis, North-western University, 1980.

Rothwell, Doris P. “Calculation of Average Retail FoodPrices. ” Monthly Labor Review, (1965):61-65.

Salop, Steven, and Joseph Stiglitz. “Bargains and Rip-offs: A Model of Monopolisticrdly Competitive PriceDispersion. ” Review of Economic Studies, 44(1977):493-510.

Scherer, F. M. Industrial Market Structure and EconomicPerformance, 2nd edition (Rand McNally, Chicago),1980.

Schmalensee, R. “Inter-Industry Studies of Structure andPerformance. ” To appear in Schmalensee, R. andWillig, R. (eds.), Handbook of Industrial Organiza-tion, Elsevier, Amsterdam, 1987.

Schroeter, J, R. “Estimating the Degree of Market Powerin the Beef Packing Industry. ” Review of Economicsand Statistics, 70(1),(1988): 158–162.

Shephard, William. “The Elements of Market Structure. ”Review of Economics and Statistics, 54(Feb.1972):100-107.

Stigler, George. “A Theory of Oligopoly. ” .70arna/ ofPolitical Economy, 92(1964):44-61,

Stigler, George. The Organization of Industry, Home-wood, Illinois: Irwin, 1968.

Stiglitz, Joseph. “Equilibrium in Product Markets withImperfect Information. ” American Economic Re-view, 69(Part 1; May, 1979):339–345.

Swamy, P. A. V. B. Statistical Inference in RandomCoef$cient Regression Models, Springer-Verlag,1971.

White, Lawrence J. “Searching for the Critical Concen-tration Ratio: An Application of the ‘Switching ofRegimes’ Technique. ” In S. Goldfield and R. Quandt(eds.) Studies in Non-Linear Estimation, Cambridge,Massachusetts, 1976.

Love, John, and William Lesser. “The Potential Im-pact of Ice-Minus Bacteria as a Frost Protestant inNew York Tree Fruit.” 18(1989):26-34.In Figure 2 on page 30, the cross-hatched area in-dicates where average date of last 28° freeze occursafter May 1.