1 Modelng Produce Procurement And Merchandsng Strateges Of Tradtonal Retalers In The ace Of Competton Wth Supermarkets In The São Paulo Regon. Elzabeth M. M. Q. arna, Rubens Nunes, Gulherme. de A Montero Unversdade de São Paulo - aculdade de Economa, Admnstração e Contabldade (EA/USP) epartamento de Economa São Paulo, Brazl Abstract Ths artcle proposes an explanaton for evdence that challenges the vew receved n relaton to the advance of the large supermarket chans n the food market. It was thought that the large, more effcent chans would expel the tradtonal forms of retal. The evoluton of the food retal structure n Brazl shows, however, growth n the number of tradton frms and ndependent supermarkets, n the last 9 years, wth lttle evdence of loss of mportance n food sales, especally as to the ndependent supermarkets whose share n food sales has grown. It was further beleved that the concentraton n retal would mply the power to rase prces, whch would occur after the expulson of the smaller frms. What was verfed, however, were fallng real prces n the food retal. The text s dvded n fve sectons. The frst descrbes the food retal structure n Brazl. ollowng, a model of olgopoly wth a compettve frnge s presented, n whch the products of the nucleus and the frnge are dfferentated by the convenence offered to the consumer. Accordng to the model, the concentraton n the nucleus s not suffcent to rase prces above the margnal costs. The thrd secton analyzes the behavor of food prces n retal. nally, some observatons are made on extra-prce competton. The conclusons follow. Key Words: retal, supermarkets, competton, strateges, São Paulo 1. ood Retal Structure n Brazl 1.1. menson of food retal n Brazl n In 2002, the consumpton of food n Brazl was approxmately US$ 41,9 bllon, and the exportaton of ndustralzed food reached US$ 10,7 bllon, whch corresponds to 11,6% of the GNP of Of the amount of food consumed n the country, 84% (US$ 35,1 bllon) corresponds to nhome consumpton and 16% (US$ 6,7 bllon) to away-from-home meals. Supermarkets 3 are the man dstrbuton channel of foods for n-home consumpton (fgure 1). We estmate, based on the data from AC Nelsen and ABRAS (Brazlan Supermarket Assocaton), the percentage of food sales (for n-home consumpton) at 49% 4. The retal self-servce equpment 1 The amounts expressed n US dollars (US$) are the result of the applcaton of the average exchange rate for the year 2002 (2,9207 R$/US$) to the amounts orgnally expressed n reals (R$). 2 In 2002 the Brazlan GNP totaled US$ 450,8 bllon. 3 Self Servce Stores, accordng to ACNelsen's defnton, are stores that n addton to beng classfed as ood Stores essentally have a checkout counter, a counter at the ext, wth ether a casher, or a pont-of-sale termnal, or a calculator, or whatever equpment s used to add and check purchases. In addton to that, they should offer shoppng baskets or strollers to customers. Most products, n such stores, are merchandsed n a manner so that shoppers may serve themselves. 4 The ABRAS defnes supermarket as a self-servce store wth at least 2 checkouts. Therefore ABRAS adopts a narrower defnton than Nelsen.
2 wth only one checkout sold the equvalent of 9%. The tradtonal retal 5 was responsble for 15%. Complementarly, away-from-home meals 6 were responsble for 19%. The Brazlan agrfood system allows for dfferent retal machnes, whch at the same tme compete among themselves for consumer preference and complement each other, by presentng specfc advantages n relaton to the varety of products and costs of purchase. IGURE 1: CONSUMPTION AN ISTRIBUTION O OOS IN BRAZIL Source: prepared by the authors wth data from: Brazlan ood Industry Assocaton (ABIA), Brazlan Assocaton of ood Wholesalers and strbutors (ABA), Brazlan Assocaton of Collectve Meal rms (ABERC), AC Nelsen, The evoluton of the food retal structure The openng of the economy, ntated at the end of the 1980s, and stablzaton of the nflaton rate at low levels (Real Plan, of 1994) allowed for the tghtenng of competton n food retal, whch led to the reducton of margns and, as a response from the frms, growth n productvty. There was an ncrease n the degree of concentraton n the supermarket chans, n parallel to the entry of foregn companes. Added to the local sources of supply was the nternatonal market. As to the consumer, an ncrease was verfed n the dversty of the profles and ther capacty to express ther preferences not only n relaton to products, but also n relaton to the servces offered by the dfferent retal machnes. BNES 7 (2000b) emphaszes the contnuty of the consoldaton and modernzaton process of the large frms of super and hypermarkets observed snce the md 1990s. The ratonalty of the acqustons s the reducton of costs and the ncrease of effcency and productvty, by means of the 5 Tradtonal stores, accordng to ACNelsen s defnton, are stores where a salesperson or clerk must be present to serve shoppers. 6 Comprehends the food servces, such as bars and restaurants, and collectve meals (cafeteras and restaurants wthn frms) 7 BNES stands for The Brazlan Natonal evelopment Bank.
3 exploraton of economes of scale and scope. The largest frms would have compettve advantages such that the best alternatve for the smaller frms would be the transfer of ponts of sale and other assets to the better postoned compettors. In addton to the acquston of other chans, the growth of the leadng frms also occurred by means of the constructon of new ponts of sale. Wal-Mart ntally opted to nstall ts own stores. However, n 2004 t expanded ts number of establshments by means of the purchase of the thrd largest Brazlan chan of supermarkets. The man lmtaton of the strategy of nternal growth (greenfeld nvestment) s the scarcty of avalable areas n well-postoned ponts and the hgh prce of rents, as a result of sad scarcty of adequate urban areas. Most of the chans, however, opted for the purchase of chans that were already nstalled, as was the case of the Portuguese groups Sonae and Jerônmo Martns. The strategy of growth by acquston should be exhausted nsofar as the leadng frms are acqurng the smaller chans. BNES (2000b) verfed, however, that, n 2000, there were stll supermarket chans to be acqured. The movement of acqustons s not restrcted to the large frms: acqustons occur among the smaller chans, whch n turn have been ncorporatng ndependent unts. Ths movement would tend to elevate the degree of concentraton n food retal, specfcally n the large urban centers, such as São Paulo and Ro de Janero, where hypermarkets domnate the market (BNES, 2000b). The effects of concentraton of the sector have permtted consderable ncrements n the operatonal result of supermarket chans, by means of the reducton of expenses and, manly, the reducton of costs wth the acquston of merchandse, obtaned prncpally due to prce negotaton, condtons of payment, and added servces from supplers. CB, for example, constructed a mega dstrbuton center n Greater São Paulo, to attend super and hypermarkets of dfferent chans ( Extra (hyper) Pão de Açúcar (super) and Baratero (dscount store). The growng returns of the large chans would have as a consequence the possblty of prce reductons to the consumer and the elmnaton of less effcent formats of retal. The advantages of sze obtaned n the negotatons wth supplers allow a sngle store to operate wth varous formats, dentfed by dstnct brand names or flags, thus beneftng from the segmentaton of the consumer market. Varables such as sze of the stores, number of tems offered, added servces, and prce polcy can be combned, amng to explore opportuntes opened by dfferent groups of consumers. The elevated costs of the ponts of sale n areas of ntense traffc or of hgh populaton densty are overcome wth smaller stores, supermarkets whose areas vary from 300 m² to 800 m². The format of the small store lnked to a chan would be the entry door for the dspute for the consumer of a lower ncome level, spread out around the perpheres of the metropoltan regons. The ndependent supermarkets that operate n these areas would have dffcultes n competng wth the prces of supermarket chans. Supermarket chans 8 held, n 2000, 1,1% of the stores and 42,8% of the volume of food sales. Wth 16,6% of the stores, ndependent supermarkets were responsble for 44,0% of food sales. The crushng majorty of establshments, 82,3% of them, are made up of tradtonal retal, whch s responsble for 13,2% of the volume of sales. (Table 1) 8 Chan Stores are made up of 5 or more stores under the same name and CNPJ.
4 TABLE 1 - BRAZIL Number of stores and partcpaton n food retal by type of retal machne Number of Stores Year Tradtonal Chans Independent Total Brazl Volume of Sales Year Tradtonal 14,90% 15,30% 15,60% 15,40% 15,60% 13,70% 13,20% 21,00% 20,40% Chans 45,10% 44,40% 44,60% 44,90% 46,60% 44,70% 42,80% 44,70% 43,85% Independent 40,00% 40,30% 39,80% 39,70% 37,80% 41,60% 44,00% 34,52% 35,74% Obs: There was a change n the methodology utlzed n calculatng the volume of Sales as of 2001, so that the values cannot be compared wth the prevous years. Untl 2000, the volume of sales collected n the channels was focused on the categores audted by the varous ndces of ACNelsen. The new methodology consders the total volume of Sales of each channel, ncludng all the categores that each channel sells. Source: AC Nelsen Census, apud ARINA (2002) The number of stores belongng to supermarket chans presented a declne of 5,3% between 1994 and 2000 (Table 2). In ths perod, the number of tradtonal retal establshments grew 27%, or 4,1% per year, whle the number of ndependent supermarkets grew 60%, or 8,2% per year. In the same perod, the GNP grew 16,4%, equvalent to 2,6% per year. Even though the demand for foods has grown at rates greater than the rates of GNP growth, due to the dstrbutve effects of the nflaton control, t s very lkely that food retal has grown more than demand, especally as of 1997 when the GNP growth rates were very reduced. In consequence, the competton n food retal has become more and more ntense. The values on volume of sales ndcated n Table 1 for 2001 and 2002 are not comparable wth the data referrng to prevous years, because there was an mportant change n the methodology of the Nelsen Retal Census. Untl 2000, the total volume of sales (on whch the shares are calculated) was the sum of revenues wth the products audted by the AC Nelsen Insttute; as of 2001, the volume of sales began also to nclude the amount of sales wth non-audted products, that s, the total revenue of the food retal establshments, and not only the amount of sales of a set of pre-selected products. The change of methodology revealed that the presence of the tradtonal retal was more mportant than was prevously judged. espte the ncrease of the degree of concentraton n the segment of supermarket chans, ARINA and NUNES (2002) noted a dscrete trend of deconcentraton n food retal at a natonal level. Ths trend s assocated to the expanson of the ndependent supermarkets (n number of establshments and market share) and of tradtonal retal (only n number of establshments). The regonal realty, however, shows mportant dstnctons. The more densely populated metropoltan regons wth hgher ncome presented greater advances of supermarket chans. These were responsble for more than 70% of food sales n the large Metropoltan Regons of São Paulo and Ro de Janero. In the other areas, ths number was between 20% and 50%. The smlar dstrbuton n
5 the large metropoltan areas suggests that the urban densfcaton favors the presence of supermarket chans, whose proftablty would depend on large consumer markets. In Greater São Paulo, the share of chans grew 14,01% between 1994 and In Greater Ro de Janero, the market share of chans presented a small drop of 2,22% n the same perod (table 2). In all the areas researched, wth excepton of area II and Greater São Paulo, the relatve weght of the chans n food sales was reduced. The reducton of the market share of chans s assocated to the expanson of ndependent supermarkets. In the large metropoltan regons the share of ths form of retal was stable, around 12% to 13% n Greater Ro and 23% to 25% n Greater São Paulo. In the other areas, the slce of ndependent supermarkets has grown between 1998 and In Area VII, belongng to the Center West, whch s undergong a process of urbanzaton related to the expanson of the agrcultural fronter, the growth trend of ndependent supermarkets has been clear snce 1994 (growth rate). Tradtonal retal was always below 30% of food sales. The lowest share s n Greater São Paulo, whch regstered a relatve fall, from 7,7% n 1994 to 5,5% of the food sales n In areas V (São Paulo countrysde) and VI (South regon) a drop was observed n the share of tradtonal retal between 1998 and 2000 n the volume of food sales. Ths trend, however, was not verfed n the other areas, where tradtonal retal presents a modest, but postve, growth. All the Nelsen areas presented an ncrease n the number of tradtonal retal stores. The annual growth rates vared between 0,8% per year, n the South regon (area VI) and 5,6% per year n the Center West. In the South we see the hghest growth rate of ndependent supermarkets, 16% per year, between 1994 and Ths concdence suggests that, at least n the South regon, part of tradtonal retal may have been transformed nto ndependent supermarkets, adaptng ther equpment to selfservce 9. The number of ndependent supermarkets fell n the two largest metropoltan regons, Greater São Paulo and Greater Ro, where falls of 3,74% and 4,1% per year were regstered between 1994 and In the other areas, the growth of the ndependent supermarkets was vgorous, wth annual rates between 4,7% (area II) and 16% (area VI). In Greater Ro, the reducton of the number of ndependent supermarkets was not accompaned by an mportant reducton n the fracton of ths type of establshment n the volume of food sales, whch suggests that there was an ncrease n the average sze of the ndependent supermarket store, perhaps due to the ext of smaller unts. The slce of supermarket chans dd not grow, despte the reducton of the number of ndependent stores. Only n Greater São Paulo (area IV) was an ncrease n concentraton verfed n food retal, wth the market share of supermarket chans growng nasmuch as the number of ndependent supermarkets fell. 9 The Nelsen Census overestmates the number of supermarkets, when compared to ABRAS, snce they only requre one checkout for the store to be consdered an ndependent supermarket.
6 AC Nelsen Regons Varaton of the number of stores ( )-% Varaton n food sales ( )-% Varaton n total sales ( )- % Total Average rate of annual growth Total Average rate of annual growth Tradtonal 47,33 5 4,39 0,78-0,017 Regon I Chans -40,48-4,57-17,6-2,87-0,011 Independent 106,38 10,01 20,39 3,64 0,024 Tradtonal 32,65 3,79-2,2 0,047-0,022 Regon II Chans 8,22 1,5 8,22 1,51-0,075 Independent 43,57 4,71-2,64-0,34 0,038 Tradtonal 55,57 5,83 12,6 2,05 0,343 Regon III Chans 58,52 6,5-2,22-0,35-0,05 Independent -53,02-4,16 1,72 0,7-0,166 Tradtonal 31,02 3,58-28,57-5,27-0,06 Regon IV Chans 15,86 2,07 14,01 2,26 0,028 Independent -35,94-3,74-22,37-3,84-0,044 Tradtonal 31,52 3,58-23,12-3,93-0,086 Regon V Chans -21,07-2,27-14,07-2,39 0,031 Independent 58,46 6,14 15,6 2,55 0,014 Tradtonal 6,23 0,87-25,51-4,1 0,04 Regon VI Chans -15,73-0,95-13,72-2,3-0,08 Independent 188,8 15,82 22,58 3,52 0,05 Tradtonal 52,97 5,68 4,02 0,68 0,041 Regon VII Chans 0 1,18-27, ,08 Independent 65,21 7,07 15,88 2,5 0,032 Tradtonal 33,5 3,72-11,41-1,86-2,86 BRAZIL Chans -20,69-2,47-5,1-0,83-1,9 Independent 7,44 7, ,72 3,53 Source: Prepared by authors based on the ACNelsen Retal Censuses ( )
7 2. The survval of small retal The emprcal evdence does not support the theory of the dsappearance of small retal caused by the expanson of the supposedly more effcent large supermarket chans. Ths secton presents an explanaton for the survval of small retal, based on the fact that large supermarkets and small retal offer ther clents dfferent combnatons of prces and convenence, or purchasng costs. The consumer has avalable two perfect substtute goods, one of them sold exclusvely at the stores of the domnant nucleus and the other good avalable only at the stores of the compettve frnge. To buy n each type of retal machne, the Nucleus or the rnge, the consumer ncurs dfferent purchasng costs, and, whch nvolve transport costs, dstances moved nsde the store, tme lost n choosng the products and n the checkout lne, etc. We stpulate, for smplcty s sake, that the purchasng costs n each type of retal machne are, for each consumer, fxed, n the sense of not dependng on the quantty purchased. The purchasng costs dffer among consumers, whether for the dfferent dstances between the resdence and the stores, or the dfferences n preferences. The -th consumer faces the problem of maxmzaton of the followng utlty functon, subject to the budgetary restrcton: q τ, se compra em U = q τ, se compra em t p q + p q = w where q (q ) s the quantty purchased n the store of the domnant nucleus (frnge) and ( ) s the respectve purchasng cost for the -th consumer; w s hs ncome or wealth. The consumer wll be ndfferent to shoppng n stores of the domnant nucleus or of the frnge f the net utltes n the two channels are equal or f the dfference n the quantty that can be purchased n the two channels compensates exactly the addtonal dsutlty correspondng to the dfference n the purchasng costs ( τ ) = U ( q τ ) U q U w p w p = U ( τ τ ) The purchasng costs n the establshments of the domnant nucleus are, by assumpton, hgher. In general, the stores of the large supermarket chans are less numerous and less spread out than those of the compettve frnge. urthermore, as the stores of the domnant nucleus are larger, t s necessary to mover greater dstances wthn the store, whch mples more tme spent shoppng. There are also long checkout lnes at peak tmes. Part of the purchasng costs are transacton costs, because the sectons of the supermarkets that compete wth the specalzed retal treat the clent mpersonally and have lttle flexblty to change the products (meat cuts, for example) n accordance wth the clent s taste. rom the condton of ndfference, a relatonshp s establshed between the prces of supermarkets and of tradtonal retal. Ths relatonshp s restrcted by the purchasng costs n the two dstrbuton channels. If the relatonshp between the prces that makes the consumer ndfferent to the type of store s not respected, the consumer wll buy exclusvely n the channel that wll provde greater net surplus. If the relatonshp s respected, the choce of the consumer remans undetermned. We stpulate, n ths case, that the expected quantty of purchase s 0,5 w/p, because the consumer wll choose the supermarket and the tradtonal retal wth 50% of probablty each. The demands of the domnant nucleus and the frnge are:
8 q q = q q w w w =, se U p p p w w = 0, se U p p w w w =, se U 2 p p p t > U ( τ τ ) < U ( τ τ ) = U ( τ τ ) q q = q q w w w =, se U p p p w w = 0, se U p p w w w =, se U 2 p p p t < U ( τ τ ) > U ( τ τ ) = U ( τ τ ) At the market, there are n consumers wth dfferent purchasng costs and ncomes. The aggregaton of the ndvdual demands generates two surfaces of demand, of supermarkets and of tradtonal retal. These demands are contnuous, decreasng n prces themselves and ncreasng n relaton to the prces of the other channel. n Q Q Q = q( p, p, w ) ; < 0 ; > 0 = 1 p p n Q Q Q = q( p, p, w ) ; < 0 ; > 0 p p = 1 The food retal structure s descrbed as an olgopoly wth a compettve frnge. The frms of the domnant nucleus (large supermarket chans) compete va prces, accordng to the Bertrand Model. We stpulate that the supermarket chans dfferentate lttle among themselves; for the effect of analyss, the vast dfferentaton occurs between the supermarket chans and tradtonal retal and ndependent supermarkets stuated on the compettve frnge. or each frm of the domnant nucleus, the demand s hghly (nfntely) elastc to prces. ollowng the example of the avaro and Spller model (1984) each supermarket chan, by changng ts decson as to prces hopes that the reacton of other chans wll be retalaton and of the frnge wll be to settle nto a model of domnant frm. As of a gven volume of sales, the margnal costs of the supermarkets located n the domnant nucleus are constant. The compettve frnge (tradtonal retal and ndependent supermarkets) has hgher margnal costs than those of the domnant nucleus and they are, furthermore, ncreasng. The competton n the domnant nucleus has as a drect result the determnaton of prces equal to the margnal cost (constant) of ths segment (rght sde of gure 2, p * = MC ). Wth prce p defned, a curve of demand s determned for the compettve frnge (on the left sde of gure 2, n mrrored mage, Q = f(p p ). The frms of the compettve frnge are prce takers and defne the quanttes offered prce to margnal cost. The curve of margnal cost of the frnge (MC ) s the aggregate supply of ths segment of retal. The ntersecton of the demand (Q ) and supply (MC ) curves defne the prce (p *) and the quantty (Q *) of equlbrum of the frnge. Wth the frnge prce defned, a demand curve s determned for the domnant nucleus (Q = g(p p )). On ths aggregate demand curve of the domnant nucleus s defned, to prce p *, the quantty demanded of the domnant nucleus (Q *).
9 gure 2 Market Equlbrum n the domnant nucleus (rght) and the compettve frnge (left) MC Q =f(p P * ) P, P Q =g(p P ) P * P * MC Q Q * Q * Q It s mportant to observe that the demands of the nucleus and the frnge are not ndependent. gure 3 llustrates the effect of a rse n prces of the domnant nucleus. Suppose that the supermarket chans decde to coordnate ther prce polces, callng a truce n the prce war and fxng prces above the margnal costs (MC ). The prces of the nucleus would go from p 0 to p 1 ; p 1 > p 0. In response to the rse n prces, some consumers mgrate to the frnge. The demand curve of the frnge moves up and to the left (n the mrrored part of gure #1, the broken lne), determnng a new equlbrum n the frnge (p 1, Q 1 ), wth greater prces and quanttes. The demand curve of the nucleus moves up and to the rght, n a poston consstent wth the new prce of the frnge (p 1 ). The new demand curve of the nucleus s represented by the broken lne n the rght part of the graph. The domnant nucleus loses market share (Q 1 < Q 0 ), but less than n the hypothess of the frnge not respondng to the ncrease n nucleus prce. The loss of market wll be greater the more elastc the supply curve of the frnge. gure 3 Comparatve statstc: the effect of a rse n prces of the domnant nucleus Q 1 Q 0 Q 0 =g(p 0 P 0 ) Q 1 =g(p 1 P 1 ) P 1 P 1 P 0 P 0 Q 1 Q Q Q Q Q
10 Supposng that the orgnal stuaton (varables ndexed wth 0) were of long-term equlbrum, the ncrease n prces on the frnge (nduced by the ncrease of prces n the nucleus) would begn to provde economc proft on the frnge and to attract the entry of new frms. Wth the movement of the supply curve, the prces of the frnge would fall and the demand curve of the nucleus would move down and to the left. In the long term, the loss of market share of the domnant nucleus would be more ntense, endng wth a quantty lower than Q 1. espte the concentraton of food retal, the frms of the domnant nucleus have dffculty rasng prces and extractng more surpluses from the consumer. In the short term, competton between frms of the nucleus, whch occurs prmarly va prces, causes the strategy to sustan hgh prces to be domnated by the strategy to lower prces, snce the demands of the ndvdual frms are hghly elastc to prces. In the long term, the growth of the compettve frnge puts n check the attempt of nucleus frms to coordnate ther prce polces (a result that could emerge from the repetton of the Bertrand model). Small food retal survves despte havng hgher costs than the large supermarket chans because t offers more convenence to the consumer, or, n the terms of the model presented, nvolves purchasng costs to the consumer that are lower than the large chans. There s a tradeoff between prces (lower n the supermarkets and the hypermarkets) and purchasng costs (lower n the tradtonal retal and ndependent supermarkets). As the consumers have dstnct preferences n relaton to prce and convenence- and a sngle consumer can even, n dfferent crcumstances, choose dfferent dstrbuton channels-, there s space n the market for tradtonal retal and for ndependent selfservce stores. As the alternatve of rasng prces results n loss of market for the domnant nucleus or for the compettve frnge, the large supermarket chans seek cost-reducng nnovatons, whether n the process of controllng merchandse flows, or n the negotatons wth supplers. As long as the other frms of the nucleus do not copy the nnovaton, the poneers can realze economc proft or reduce the prces to the consumer, wnnng parcels of the market from ther drect compettors. 3. Consequences on the prces to the consumer 3.1. Evoluton of the degree of concentraton and behavor of prces After the Real Plan, food real prces were drastcally reduced, precsely n the perod of greatest demand The data from INPC/IBGE show reductons of 20% on average for Brazl. Awayfrom-home meals, n turn, present a slow reducton n real prces, though over a longer perod of tme, reachng 2001 wth real prce ndces 15% lower than n That s, both n retal and n food servces, food prces present drops n relatve prces. GRAPH 1 ood Real Prce Index Brazl ( Natonal Consumer Prce Index INPC Brazl, deflated by the general INPC) ago/94 fev/95 ago/95 fev/96 ago/96 fev/97 ago/97 fev/98 ago/98 fev/99 ago/99 fev/00 ago/00 fev/01 ago/01 fev/02 ago/02 fev/03 ago/03 INPC Real _Almentação INPC Real_Almentação ora do lar Source: IBGE n
11 The groups that led the rse n prces educaton, health, and housng present a large share of servces n ther composton and, n the case of housng, tarffs controlled by the publc sector (water, sewage, gas, telephone, and electrc power). The groups that most contrbute to the mantenance of low nflaton rates comprehend products n whch nputs have greater weght n the cost composton, wth lower partcpaton of labor (n comparson wth, for example, the groups of educaton and health). urthermore, these are products more drectly exposed to nternatonal competton, snce the commercal openng begun at the end of the 80s. The behavor of food prces n the nne metropoltan regons n whch prces are gathered for the calculaton of the INPC (Belém, Belo Horzonte, Cuabá, Curtba, ortaleza, Porto Alegre, Recfe, Salvador, and São Paulo) was very smlar, despte the dfferent retal structures dentfed n these regons. A possble nterpretaton of ths fact s that the macroeconomc factors that nfluenced food prces, such as exchange rate, nterest rate, and federal taxes, were more mportant than the dosyncratc factors of the metropoltan regons. Thus, for example, shocks such as the exchange devaluatons of 1999 and 2001 mpacted the costs of the food ndustry n a smlar manner, but not necessarly equal, n all the Brazlan regons. Changes n the degree of concentraton n food retal, on the other hand, were not sgnfcant to explan the varatons n the food prces to the consumer. Wth the objectve of testng the hypothess that the dfferent structures do not explan the prce dfference among the regons, ALMEIA (2003) utlzed the Spearman correlaton coeffcent (rs)10 between varatons of the prces and the degree of concentraton. or ths, two varables were created for each regon representng the varaton of the IPCA (VIPCA) and the partcpaton of supermarket chans n food retal (VCONC). Both represent the annual varaton of a gven regon n relaton to the average observed n Brazl n each year. That s, VIPCA j,t = (IPCAj t/ipcaj t-1) / (IPCA Brazl t/ipca Brazl t-1) VCONC j,t = (Concentraçãoj t/concentraçãoj t-1) / (Concentração Brasl t / Concentração Brasl t-1) The subscrpts n the expressons above represent, n the order that they appear, the regon and the date (year) of the observaton. The procedure to relate the value of the varables observed n each regon to the natonal average has the objectve of solatng the effects of factors common to the varatons of food prces n all the regons studed. The results are reproduced n Table 2. TABLE 2 Varaton of the IPCA-food and the concentraton n the food market (n %) and Spearman Correlaton Coeffcent r s I - NE II RJ III Gde RJ IV Gde SP VI Sul VIPCA VCONC VIPCA VCONC VIPCA VCONC VIPCA VCONC VIPCA VCONC 95/94 1,05 0,99 0,97 1 0,98 0,99 0,99 1,1 1,04 0,94 96/95 0,99 0,98 1,02 0,98 1 0,99 1 1,01 0,98 0,97 97/96 0,98 1,03 1 1,08 1,03 0,99 1,01 0,98 1, /97 5,88 1,03 1,07 0,99 3,49 0,96 2,81 1 1,92 0,99 99/98 1 0,88 1 1,14 1 1,02 1 1,02 1,05 1,08 00/99 0,98 0,96 1 0,96 0,98 1,08 1 1,08 0,98 0,94 10 One of the technques of nference that does not requre very strong hypotheses on the parameters, denomnated non-parametrc : the Spearman correlaton coeffcent depends only on the orderng of the objects accordng to the varables whose correlaton one wshes to gauge (Segel, 1981).
12 r s 0,54-0,6-0,37-0,71 0,43 Crtcal values: 0,94 e 0,82, for sgnfcance of 1% and 5%, respectvely. (SIEGEL, 1979) Source: ALMEIA (2003), based on ACNelsen Retal Censuses and IBGE. ALMEIA (2003) rejected n all the regons the hypothess that the ncrease of degree of concentraton n food retal s correlated to the ncrease of food prces to the consumer. Accordng to the author, t cannot be affrmed that the dfferent evolutons of IPCA verfed n each metropoltan regon can be explaned by the dfferent market structures. If the supermarket chans managed to conquer slces of market n some regons by nternal growth or acquston, ths dd not gve them power to rase the margns and the food prces to the consumer ood prces n supermarkets and n tradtonal retal 11 A hypothess to explan the coexstence of dfferent retal machnes was rased n BNES (2000c): One factor of great mportance resdes n the polcy and effectve practce of lower prces, on the part of smaller frms, snce one of the man reasons for consumers to shop n hypermarkets s the prce. Insofar as the neghborhood stores functon wth lower prces, there s a transfer of purchases from hypermarkets to these supermarkets. Ths argument contans some dffcultes: f the chans have cost advantages, for example, comng from economes of scale, they could practce lower prces than the ndependent supermarkets, ganng market share and ncreasng ther profts. The dversty of forms n retal could be understood f the purchasng costs to the consumer were taken nto account. In the large supermarkets, the prces are lower, but the purchasng costs (gong to the store, checkout lne, movement wthn the store) are hgh; n small retal, the opposte occurs (hgher prces, lower purchasng costs). If there s space for the two retal formats, t s because the advantages (pecunary and non-pecunary) of each format compensate each other n some way. Otherwse, the optmal retal form would domnate the entre market. BNES (2000c) went as far as to rase the hypothess that small retal would practce lower prces as a means of survvng the competton of hypermarket and supermarket chans. The evdence obtaned for the metropoltan regon of São Paulo n the data bank of IPC-pe suggests that ths hypothess s ncorrect for an mportant set of products. It could be sad n fact that the general rule s lower prces n supermarkets. To evaluate the prces practced n dfferent retal machnes, MONTEIRO (2004) utlzed the data bank of IPC-pe (Consumer Prce Index, prepared by the Insttute of Economc Research oundaton). Ths prce ndex refers only to the Metropoltan Regon of São Paulo. Hstorc seres of prces were gathered fro 12 products: whole chcken (graph 2), alcatra (beef cut) (graph 4), acém (beef cut), slced ham (graph 3), prato cheese, fresh rolls, specal mlk, long lfe mlk (UHT), tomato, carrots, shark, and fresh sardnes. GRAPH 2 PRICES O WHOLE CHICKEN IN SUPERMARKETS AN IN TRAITIONAL RETAIL- CURRENT R$ /KG 11 Ths tem s based on the development of the scentfc ntaton of Gulherme owler de Ávla Montero.