F r ode St een and M or t en Sæt hr e
20t h June 2017
Norwegian School of Economics, Depart ment of Economics, Helleveien 30, NO-5045 Bergen. T his t hesis was writ t en as a part of t he Mast er of Science in Economics and Business Administ rat ion at t he Norwegian School of Economics. Please not e t hat neit her t he inst it ut ion nor t he examiners are responsible, t hr ough t he approval of t his t hesis, for t he t heories and met hods used, or result s and conclusions drawn in t his work.
†E-mail: damian.olsen@st udent .nhh.no.
‡E-mail: mart in.olsen@st udent .nhh.no.
T he aim of t his paper has been t o examine t he local compet it ion bet ween grocery ret ail st ores in t he Norwegian grocery ret ail market . Our at t ent ion has been on explaining t he e ect of market st ruct ure on prices. For t his purpose we assemble an original dat aset consist ing of a select ion of grocery ret ail st ores in t he cit y of Oslo, Norway. We const ruct several mar ket st ruct ure variables based on di erent st ruct ural measures and employ a random-e ect s est imat or t o det ermine t he relat ionship bet ween market st ruct ure and prices, cont rolling for cost and demand fact ors as well as st ore charact erist ics.
First , our ndings suggest t hat t he variat ion in prices is direct ly relat ed t o t he chain-concept a liat ion of each st ore. Nonet heless, under t he assumpt ion t hat market st ruct ure is exogenous in our model, we est imat e t hat (1) t he dist ance t o t he near est r ival does not a ect a st or e’s pricing behaviour, (2) t he number of r ival st or es have a negat ive e ect on a st ore’s price level, and (3) prices increase when t here is at least one rival for mat in t he local market .
However, if t he assumpt ion of exogenous market st ruct ure does not hold, which t here is reason t o believe, t hen our est imat ed relat ionship bet ween price and market st ruct ure only expresses t he correlat ion between t he two.
local pr ice compet it ion, t he norwegian grocery ret ail indust r y, spat ial compet it ion, st ruct ure-price relat ionship, indust rial economics
suggest ions and comment s, as well as for all t he inspiring discussions. T he paper has also bene t ed from comment s by Jarle Slørst ad at NorgesGruppen A SA and Simen A ardal Ulsaker from t he Depart ment of Economics, Norwegian School of Economics, as well as ot her part icipant s in t he NorgesGruppen proj ect . We would also like t o ext end our sincere grat it ude t o our loved ones for t heir uncondit ional support .
A part from t his, t he paper has bene t ed fr om unhealt hy sleep habit s, ca eine subst ances, and t he involunt ary family t ies of t he aut hors.
Here is a prayer for you. Got a pencil? ... ’Prot ect me from knowing what I don’t need t o know. Pr ot ect me fr om even knowing t hat t here ar e t hings t o know t hat I don’t know. Pr ot ect me fr om knowing t hat I decided not t o know about t he t hings I decided not t o know about . A men.’ ... T her e’s anot her pr ayer t hat goes wit h it . ’Lord, lord, lor d. Prot ect me from t he consequences of t he above prayer.’
—– Douglas Adams, Mostly Har mless (2009)
ii
to our loved ones, should we meet again post thesis
v i v i 1 3
2.1 M ar k et St r uct ur e . . . 4
2.2 M aj or Pl ayer s . . . 8
2.2.1 Nor g esGr uppen A SA . . . 8
2.2.2 Coop Nor ge SA . . . 9
2.2.3 Reit ang r uppen A S . . . 10
2.3 Compet it i on Par amet er s . . . 10
12 16 19 5.1 D ef ining t he L ocal M ar k et . . . 21
5.2 Pr oduct Cat egor ies and M ar k et B ask et of Goods . . . 23
5.3 T he Var iabl es . . . 27
5.3.1 Pr ice L ev el I ndex . . . 27
5.3.2 M ar k et St r uc t ur e and Compet it ion I nt ensit y Var iabl es. . . 28
5.3.3 Ot her Fac t or s. . . 31
5.4 Pr el i minar y Dat a A nal ysis . . . 33
37 6.1 M odel Spec if ic at ion. . . 37
6.1.1 Exog eneit y of M ar k et St r uct ur e. . . 38
6.2 St or e Pr i ce L ev el : M ar k et Bask et of Goods. . . 39 45 47 iv
51 A Pr oduct Cat egor ies and M ar k et B ask et of Goods . . . 51 B Ot her St r uct ur e-Pr i ce M odel s . . . 52
v
I Descript ive St at ist ics for t he Norwegian Grocery Ret ail Market . . . . 5 I I Number of St ores in t he Norwegian Grocery Ret ail Market , By Grouping
and Geographical Area . . . 7 I I I Select ed Market Dat a for 15 Local Grocery Market s in Oslo . . . 24 IV Descript ive St at ist ics for t he Price, Market St ruct ure, and Firm
St ruct ure Variables . . . 34 V Int ra-Concept St ore Level Price Index Di erences for the Market Basket
of Goods . . . 35 VI Relat ion Between Number of Chains in Local M arket and t he St ore
Level Price Index by Format . . . 36 VI I Paramet er Est imat es of t he Baseline Speci cat ions of t he Price Level
of Grocery Ret ailing St ores in Oslo . . . 41 VI I I Est imat ion Equat ions Explaining t he Price Level of Grocery Ret ailing
St ores in Oslo wit h Alt ernat ive St ruct ural Measures . . . 43 IX Product Cat egories in Market Basket of Goods . . . 51 X Est imat ion Equat ions Explaining t he Price Level of Grocery Ret ailing
St ores in Oslo wit h Addit ional St ruct ural Measures . . . 53
vi
1 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
T he Norwegian grocery ret ail market has over t he last decade been primarily charact erized by chain format ion and cent ralizat ion, wit h t hree groupings cont rolling nearly all of t he market . T his was furt her emphasized by ICA’s wit hdrawal from t he market in 2015, which saw t he dominant players furt her st rengt hen t heir posit ion.
However, following major st ruct ural changes, t he market has st abilized in 2016. T his has led t o an increased growt h in t he grocery ret ail market , which is closer t o t he level t hat t he market experienced half a decade ago.
In t he paper, we examine t he local compet it ion between grocery ret ail st ores in t he Norwegian grocery ret ail market —focusing at t ent ion on evaluat ing t he impact of market st ruct ure on grocery ret ail prices. We assemble an original dat a set wit h price informat ion from 49 st ores locat ed wit hin t he city of Oslo for t he period week 11, 2016, t o week 9, 2017, and observe a sample of discount , convenience and supermarket st ores operating in ft een dist inct local market s. More precisely, our primary object ive is t o st udy if di erences in grocery ret ail prices across st ores and local market s can be at t ribut ed solely t o chain-concept a liat ion aft er cont rolling for fact ors t hat may a ect t he cost and demand condit ions for the st ore, as well as for alternat ive measures of market st ruct ure. Under t he hypot hesis t hat each st ore is direct ly managed by t he chain-concept t hey belong t o, we would expect st ruct ural measures t o have no impact on prices. M oving forward, we will also provide a det ailed discussion on t he endogeneity of market st ruct ure in our approach, as a st ore’s price decision, as argued by Gullst rand and Jörgensen (2012), is a ect ed by t he local market , and t he local market st ructure may not be disengaged from t he pricing behavior of t he stores in t he part icular market .
In t he special case when compet it ion is e ect ive, grocery ret ail prices would be det ermined solely on t he basis of marginal cost s and demand cont rols, and t here would not exist a syst emat ic st ruct ure-price relat ionship (Lamm, 1981, p. 69).
However, models of oligopolist ic behavior generally agree t hat t he compet it ion increases wit h, int er alia, t he number of st ores in a local market , as argued by A splund and Friberg (2002), and t hat t he equilibrium prices should fall as compet it ion increases. A s demonst rat ed in Dobson and Wat erson (2005, 2008), pricing according t o local condit ions should in fact be t he pro t -maximizing st rat egy in local market s. Chains may, nonet heless, nd it pro t able t o set a uniform price
when subject to different intensities of competition across various markets. In sum, exactly how equilibrium prices are related to market structure hinges crucially on the nature of the short-run interaction, and the potential for implicit collusion as national pricing can only be sustained with a credible, visible commitment to uniform pricing (Asplund and Friberg, 2002; Dobson and Waterson, 2008).
Our empirical strategy builds upon previous empirical research methods and insights from modern market theorists. However, where most of the previous empirical studies of market structure have focused on the analysis of cross-sectional data, we employ a random effects estimator to our panel data structure. That being said, to our knowledge, no papers on structure-price relationships have previously examined the Norwegian grocery retail market. The paper also extensively combines market structure measures employed in previous empirical research as we examine a vast amount of structural measures with the purpose of relating price variation to market structure. In addition, where the majority of the previous studies on price competition in grocery retailing have examined only the effect of market structure on supermarket prices, we also incorporate discount retailers and convenience stores in our sample.1
Our results seem to support the findings in Asensio (2014) in the sense that most of the variation in prices is directly related to the chain-concept each store belong to. Roughly 93 percent of the variations in prices can solely be attributed to the chain-concept affiliation of the store. However, unlike in Asensio (2014), we observe nonetheless a behavior of market structure measures and local socioeconomic attributes affecting prices. When assuming that market structure is exogenous in our model, our main findings is that (1) the distance to the nearest rival does not affect a store’s pricing behaviour, (2) the number of rival stores have a negative effect on a store’s price level, and (3) prices increase when there is at least one rival format in the local market. However, the causal interpretation of the estimated effects are only valid if the assumption of exogeneity holds. If this is not the case, which there is reason to believe, then the estimated relationships can only be interpreted as correlations.
The organization of the paper is as follows. In Section 2 we present an overview of the Norwegian grocery retail market. Section 3 includes a review of previous research on competition in local retail markets, whereas Section 4 introduces the concept
1Cleeren et al. (2010) use an empirical entry model to study the degree of intra- and interformat competition between discounters and supermarkets in Germany, while Zhu et al. (2009) examine competition between the three major firms in the retail discount industry.
3 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
of cust omized and uniform pricing of ret ail chains t o provide a just i cat ion for t he decision of grocery retail chains t o x t heir prices nationally as opposed t o following a policy of local pricing. In Sect ion 5 we present t he dat a t hat we use in our empirical analysis, explain how t he local market s and market basket of goods are de ned, and present t he variables we employ in our analysis. Sect ion 6 speci es the model of price compet it ion and report s t he empirical result s. Finally, t he paper summarizes major conclusions and discusses t he corresponding limit at ions of our select ed approach.
Over t he last decade t he development in t he Norwegian grocery ret ail market has been charact erized by chain format ion and cent ralizat ion, t he development of privat e labels, and increased vert ical cooperat ion (NI LF, 2013). I n t he most recent years I CA ’s wit hdrawal from t he Norwegian market has been t he key driver of change, wit h most of ICA’s st ores being acquired by Coop and t he rest being most ly divided between NorgesGruppen and Bunnpris (Solem, 2017). However, following t hese major st ruct ural changes, t he market has st abilized in 2016. T his has led t o an increased growt h in t he grocery ret ail market , which is closer t o t he level t hat t he market experienced half a decade ago.
M oreover, t he growt h comes despit e t he fact t hat bot h food boxes and online delivery have a wider range of users now than one year ago (Nielsen, 2017). In addit ion, grocery st ores are experiencing increased compet it ion from ot her market channels such as rest aurant s and kiosks (NorgesGruppen ASA, 2017). T herefore, de ning t he relevant grocery ret ail market is far from unambiguous. Because even t hough t he grocery ret ail market is dominat ed by grocery ret ail groupings such as NorgesGruppen and Coop, groceries are also sold t hrough ot her st ores and channels which are not uniquely ident i ed within t he grocery retail market . For t he purpose of our t hesis, we narrow down t he grocery ret ail market t o include only t he market players t hat have t heir main emphasis on groceries sold t hrough physical st ores, t hus excluding online delivery, kiosks, and gas st at ions as well as rest aurant s and fast -food chains.
The primary charact erist ic of t he Norwegian grocery ret ail market is it s high market concent rat ion of ret ailers, wit h t hree groupings cont rolling 96.1 percent of t he market as of 2016 (Nielsen, 2017). When including t he fourt h largest grouping, Bunnpris2, t he t ot al market share rises t o over 99.9 percent (Nielsen, 2017). The high market concent rat ion is furt her emphasized by the dominant posit ion of t he largest grouping, NorgesGruppen, whose revenues const it ut e about 40 percent of t he t ot al grocery ret ail market (Nielsen, 2017).
T he market shares in t he period from 2014 t o 2016 are present ed in Figure I.
We see t hat t he t hree largest groupings – NorgesGruppen, Coop, and Rema 1000 – have increased t heir market shares over t he period. Coop has experienced t he largest increase in market share, wit h an increase of over 30 percent . Coop’s success may part ly be explained by it s acquisit ion of I CA Norway in mid-2015. T his may have increased Coop’s compet it iveness through greater economies of scale and scope as well as increased purchasing power.
Furt her, in Table I we present select ed descript ive st at ist ics for t he Norwegian grocery ret ail market for t he period 2011-2016. T he t able includes t he t ot al revenues for t he market , as well as t he market shares by concept and by format . We see from Table I that t he t ot al revenues in t he grocery ret ail market has seen an increase of over 50 percent during t he last decade, t ot alling almost 170 billion NOK ex VAT as of 2016 (Nielsen, 2017). T his implies a revenue growt h t hat is almost one percent age point
higher t han t he growt h in t he combined ret ail market , which grows approximat ely t hree percent yearly (St at ist ics Norway, 2017).
Moreover, in Table I I we present t he number of st ores in t he Norwegian grocery ret ail market by grouping and by geographical area. We observe t hat t he t ot al number of st ores have declined over t he period, which implies t hat t he average revenue per store has increased substant ially. According to (NILF, 2013), t his revenue growt h can mainly be at t ribut ed t o t he e ciency improvement s and rest ruct uring measures t hat have been carried out by t he major players in t he market .
2Bunnpris is an independent chain but has had a pr ocurement and dist ribut ion cooperat ion wit h REM A 1000 since 2012, where Rema 1000 has been responsible for t he procur ement negot iat ions and cont ract s as well as t he deliveries (NIL F, 2013). However, as of 2017, NorgesGruppen have overt aken t hese t asks on behalf of Bunnpris (Norwegian Compet it ion A ut hority, 2016).
5 LOCAL PRICE COMPETITION IN GROCERY RETAILING
Table I—Descriptive Statistics for the Norwegian Grocery Retail Market
2011 2012 2013 2014 2015 2016
Revenues, in MNOK 143,717 148,119 153,506 160,145 164,310 169,413
Market Shares by Chain (in Percent)
Rema 1000 21.3 22.2 23.1 23.7 24.2 24.4
Kiwi 15.2 16.0 16.9 17.7 18.9 19.9
Coop Extra 1.8 2.3 3.2 6.1 7.9 11.5
Meny 9.0 9.7 10.3 10.2 10.7 10.9
Spar/Eurospar 6.8 6.8 6.8 6.7 6.9 7.0
Coop Obs 5.5 5.6 5.5 5.4 5.5 5.4
Coop Prix 6.6 6.6 6.1 4.4 4.2 5.3
Coop Mega 5.7 5.2 4.5 3.7 3.7 4.3
Bunnpris 3.8 3.7 3.6 3.4 3.9 3.9
Joker 3.3 3.4 3.4 3.5 3.6 3.7
Market Shares by Format (in Percent)
Discount 54.9 57.4 59.7 61.8 63.4 65.1
Supermarket 25.6 25.4 25.0 23.7 23.0 22.2
Convenience 10.3 9.2 8.5 8.3 7.8 7.3
Hypermarket 9.3 8.1 6.7 6.1 5.8 5.4
Table I.Notes to Table I. The table includes the following descriptive statistics for the Norwegian grocery retail market for the period 2011-2016: total revenues in million NOK, the market shares by concept in percent, and the market shares by format in percent. The convenience format does not include sales from kiosks and gas stations. There are missing observations for Coop with regards to number of stores in 2011 and 2012. (Nielsen, 2015, 2016a, 2017; NorgesGruppen ASA, 2017; Coop Norge SA, 2016; Reitangruppen AS, 2016)
T he development in t he Norwegian grocery ret ail market has over t he last decade been charact erized by fewer but larger st ores and longer opening hours (NILF, 2013). Over t he last years, however, t he three largest groupings have all increased t heir number of stores, in direct cont rast t o t he market in general. This increase in number of st ores can most likely be att ributed t o ICA’s wit hdrawal from t he Norwegian market, wit h most of ICA’s st ores being acquired by Coop, NorgesGruppen, and Rema 10003 (Solem, 2017).
Furt hermore, when we decompose t he reduct ion in t he t ot al number of st ores by geographical area, we observe t hat t he decline in number of st ores has not been uniformly dist ribut ed across areas. According t o Gullst rand and Jörgensen (2012), t his patt ern may be explained by dist ribut ion costs and scale economies, which enable only t he largest chains t o be successful in remot e areas wit h low populat ion density;
anot her possible explanat ion could be changes in populat ion pat t erns across areas.
Eit her way, Oslo is t he only area t hat has seen an increase in t he number of st ores during the period, whereas the number of st ores in Nort hern Norway has seen a decline of almost t en percent .
Figur e I — M ar k et Shar es in t he Nor w egian Gr ocer y Ret ail M ar k et , 2014-2016
F igur e I . M arket shares in t he Norwegian grocery ret ail market in t he period 2014-2016, in percent ages.
M arket shares are based on Nielsen Norway’s Grocery Regist er and include all co-operat ive and privat e grocery st ores in Norway, excluding Svalbard. Grocery sales from gas st at ions and kiosks as well as food box es and online groceries, are not included in t he gure. T he market shares of I CA Norway are included under ’Ot her st ores’ in 2014 and 2015, following ICA ’s wit hdrawal from t he Norwegian market during 2015. (Nielsen, 2015, 2016a, 2017)
3Rema 1000 acquired several leasing cont ract s from I CA (Solem, 2017).
7 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
Tabl e I I — Number of St or es i n t he Nor w egian Gr ocer y Ret ail M ar k et , B y Gr ouping and Geogr aphical A r ea
2012 2013 2014 2015 2016 %
I n T ot al 3,917 3,899 3,899 3,806 3,814 -2.6
By Gr ouping
– NorgesGruppen 1,681 1,714 1,768 1,806 1,850 10.0
– Coop — 793 804 1,259a 1,250 57.6
– Rema 1000 505 528 541 565 594 17.6
By Geogr aphical A r ea
– West ern Norway 843 841 839 825 831 -1.4
– West -East ern Norway 721 723 720 698 698 -3.2
– Oslo 365 376 369 371 377 3.3
– East -East ern Norway 806 799 808 785 792 -1.7
– Cent ral Norway 636 632 634 625 622 -2.2
– Nort hern Norway 546 528 529 502 494 -9.5
Tabl e I I . Not es t o Table I I . T he t able includes t he number of st ores in t he Norwegian grocery ret ail market for t he period 2012-2016 by geographical area, as well as t he percent age change over t he period. No informat ion on t he number of Coop st ores in 2012 was found. a For 2015, t he I CA st ores are included under Coop, since Coop’s acquisit ion of ICA Norway went t hr ough in mid-2015. (Nielsen, 2015, 2016a, 2017; NorgesGruppen A SA , 2017; Coop Norge SA , 2016; Reit angruppen A S, 2016)
T he geographical di erences in number of st ores also ext end t o t he chain level.
NorgesGruppen is part icularly present in East ern and West ern Norway, but has a relat ively low number of st ores in Nort her Norway (NOU 2011:4, 2011b). Coop, on t he ot her hand, is well represent ed in Nort hern Norway, as well as in Central Norway, whereas Rema 1000 is more or less equally represent ed across all areas (NOU 2011:4, 2011b).
T here are t en main grocery ret ail concept s in t he Norwegian grocery ret ail market , which we present in Table I. T he t hree largest concept s in t erms of market share are Rema 1000, K iwi, and Ext ra, const it ut ing over half of t he market ’s revenues (Nielsen, 2017). T hese t hree concept s all operat e wit hin t he discount format of t he market , and t oget her wit h Coop Prix and Bunnpris t hey make up almost two t hirds of t he grocery ret ail market , which is almost an increase of
20 percent for t he discount format in t he last ve years. A lt hough t he discount format has seen a st eep increase in t he recent years, t his format no longer only consist s of ret ail concept s wit h limit ed product ranges and low prices; t here is a clear t rend t hat ret ail concept s wit hin t he discount format have increased t heir product ranges, focusing also on fresh produce, int er alia (Virke, 2015).
T he second largest grocery ret ail format is t he supermarket , const it ut ing almost a quart er of t he market . T he supermarket format o ers a wider range of product s t han t he discount format and compet es for cust omers not only t hrough pricing but more import ant ly t hrough it s assort ment (Virke, 2015). Meny has long held a leading posit ion wit hin t he supermarket format , which also includes Spar/ Eurospar and Coop Mega. However, t he supermarket format has seen a st eady decline in t he recent years, losing market shares t o t he discount format .
T he t hird largest grocery ret ail format is convenience. T he convenience format consist s of Joker and Coop Marked as well as some independent ret ailers. This format o ers t he smallest range of product s of all format s in t he market ; st ores belonging t o t his format are usually locat ed in t he dist rict s and oft en provide t he only opt ion for purchasing groceries and adjoining services t o resident s in t he area (V irke, 2015).
The last grocery ret ail format is t he hypermarket . T he hypermarket format o ers t he widest range of product s of all t he format s, wit h a variety of non-food product s and fresh produce encouraging one-st op shopping behavior in consumers (Virke, 2015).
Nonet heless, t his format has seen t he st eepest decline in market share of all t he format s, leaving only one ret ail concept , Coop OBS!, as of 2016. Hence, t he t rend in t he Norwegian grocery ret ail market seems t o adduce t hat t he format s t hat do not clearly emphasize low prices are losing t ract ion.
NorgesGruppen is t he market leader in t he Norwegian grocery ret ail market , wit h 1,850 grocery st ores dist ribut ed t hroughout Norway, of which 812 are wholly owned (NorgesGruppen ASA, 2017). Besides its grocery ret ail activit ies, NorgesGruppen also has operat ions wit hin wholesale, real est at e, and convenience and is one of Norways’s largest purchasing organizat ions, wit h large purchases annually for grocery, service
9 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
and large-scale households (NorgesGruppen ASA, 2017).
Wit hin t he grocery ret ail market, NorgesGruppen operates wit h ve main concept s:
Joker, Spar/ Eurospar, M eny, K iwi, and Nærbut ikken. K iwi operat es wit hin t he discount format and is t he largest of NorgesGruppen’s concept s wit h a market share of about 20 percent as of 2016, const it ut ing almost half of NorgesGruppen’s revenues from it s grocery ret ail act ivit ies (Nielsen, 2017). NorgesGruppen’s second largest concept is Meny, const it uting over a fourt h of NorgesGruppen’s grocery ret ail revenues (Nielsen, 2017). Unlike K iwi, M eny operat es wit hin t he supermarket format , as do Spar/ Eurospar. Joker and Nærbut ikken are smaller concept s, operat ing wit hin t he convenience format .
Furt hermore, NorgesGruppen has an ext ensive range of privat e labels. First Price is t he chain’s range of low-cost goods, which are available in all t he chain’s st ores, whereas Jacobs Ut valgt e is t he chain’s premium label (NILF, 2013). In addit ion, NorgesGruppen has ot her privat e labels wit hin foods, ingredient s and food st orage product s (NILF, 2013).
Coop is t he second largest grouping in t he Norwegian grocery ret ail market , wit h 1,250 st ores as of 2016. Over t he last few years Coop has st rengt hened it s posit ion t hrough t he acquisit ion of ICA Norway in 2015. Unlike t he ot her groupings, Coop is owned by t he consumers t hrough regional cooperat ives (NI LF, 2013). A lt hough Coop has no ownership in t he st ores, it owns t he right s t o t he concept s and is responsible for procurement , supply chain, market ing, and chain management (NILF, 2013).
Coop is also t he only grouping in t he market t hat has concept s wit hin all format s, with Coop Obs being t he only concept wit hin t he hypermarket format as of 2016. Coop Obs mainly focuses on grocery goods, but also o ers product s wit hin most branches of specialist ret ailing (NILF, 2013). Wit hin t he discount format Coop operat es wit h t wo concept s, Prix and Ext ra. Coop Ext ra is Coop’s largest concept as of 2016, experiencing a st eep increase in market share over t he last years. Coop Mega operat es within t he supermarket format , whereas Coop Market operat es within the convenience format and is t he smallest of t he concept s in t he Coop grouping.
Furthermore, Coop has several privat e labels in t heir product range. X-t ra is Coop’s
range of low-cost goods, which covers an ext ensive range of di erent product cat egories (Norwegian Consumer Council, 2015). Smak-forskjellen is Coop’s premium label, di erent iat ing t he goods along quality and origin paramet ers (Norwegian Consumer Council, 2015). In addit ion, Coop Änglamark consist s of organic and environment ally friendly goods, whereas Coop K a e is one of Norway’s largest co ee producers (NILF, 2013).
Reit angruppen AS is t he only grouping in t he Norwegian grocery ret ail market t hat operat es wit h a single concept , Rema 1000. Moreover, as opposed t o NorgesGruppen and Coop, all Rema 1000 stores are operat ed as franchises, where each st ore is operat ed independent ly under condit ions set by t he chain management (Norwegian Consumer Council, 2015). As of 2016, Reitangruppen is t he t hird largest grouping in t he market, with 594 st ores and a market share of over 24 percent . In addit ion t o grocery ret ailing, Reit angruppen also operat es kiosks and gas st at ions as well as having a separat e dist ribut ion subsidiary (NILF, 2013).
Furt hermore, Rema 1000 operat es wit hin t he discount format and focuses on dist ricts wit h high population density, which has cont ribut ed t o Rema’s growth in t he recent years (NILF, 2013). In addt ion, Rema’s product range wit hin privat e label has also cont ribut ed t o t he chain’s growt h, wit h product s wit hin bot h food and non-food cat egories. Wit hin t he food cat egory Rema 1000 has labels such as Nordfjord, Solvinge and Godehav, covering meat , chicken and sh product s.
In t he recent weeks, Rema 1000 has been t he subject of crit icism aft er a series of long-t erm and exclusive agreements were signed in t he beginning of 2017 (Valvik and Lynum, 2017). T he exclusive agreement s not only reduced t he number of brands in Rema’s product range, but also forced several of Rema’s former suppliers to undert ake ext ensive rest ruct uring and downsizing measures (Valvik and Lynum, 2017).
Price, product range, and locat ion are t he t hree most import ant fact ors when consumers in t he Norwegian grocery ret ail market decide between st ores (NorgesGruppen ASA, 2016). Ot her compet it ion paramet ers include, int er alia, opening hours, service levels,
11 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
and quality of goods, as well as brand image (NOU 2011:4, 2011a).
Price is perhaps t he most import ant compet it ion paramet er in t he Norwegian grocery ret ail market . T he increased at t ent ion given t o prices by consumers and t he frequent price comparisons in t he media, have among ot hers t hings cont ribut ed t o increased awareness t o prices and margins in t he indust ry (NOU 2011:4, 2011a).
T his is, int er alia, illust rat ed in NorgesGruppen’s annual report , where it is st at ed t hat NorgesGruppen st ores should always be compet it ive on price (NorgesGruppen ASA , 2017). T he focus on price in t he market is perhaps furt her emphasized by t he considerable growt h experienced by t he discount format in t he recent years, alt hough increases in product range may also have played a part in t his growt h. Furt hermore, t he price focus in t he grocery ret ail market has also led t o t he emergence of loyalty programs4, which at t empt t o at t ract and ret ain cust omers by o ering personalized discount s, int er alia. M oreover, t he price compet it ion has not only increased t he use of personalized discount s, it has also sparked t he development of privat e labels5 in t he market (V irke, 2015). Part of t he development in privat e label is driven by t he increase in t he discount format , and t herefore it has become increasingly import ant for chains t o have product s in t heir range t hat can drive t he price compet it ion (NOU 2011:4, 2011a). It has proven more advant ageous t o t he chains t o sell cheaper privat e labels t han t o reduce t he price and margins of ot her brands, alt hough result s suggest t hat t he int roduct ion of privat e labels in Sweden has cont ribut ed t o lower prices on nat ional brands as well (NOU 2011:4, 2011a; Asplund and Friberg, 2002).
In addit ion t o prices, bot h product range and locat ion are import ant compet it ion paramet ers in t he grocery ret ail market . Using product range as a compet it ion paramet er has primarily been reserved t o t he supermarket and hypermarket format s. However, t he increase in t he number of privat e labels in t he market has cont ribut ed t o product range becoming an increasingly import ant compet it ive fact or between di erent grocery chains as well as wit hin product cat egories, t hrough t he means of product exclusivity (NILF, 2013).
4A s of 2016, 66 percent of all cost umers in t he grocery r et ail market part icipat e in a loyalt y program (Nielsen, 2016b). However, since t hen Rema 1000’s has int roduced a new loyalt y program, Æ, which led t o K iwi, Coop, and M eny releasing t heir own loyalty pr ograms soon aft er. We t herefore expect t hat t he part icipat ion in loyalt y programs is even higher as of 2017.
5Privat e labels include product s t hat are sold exclusively wit hin t he grocery chain under a br and name t hat t he chain owns and cont rols (V irke, 2015).
According t o NILF (2013), privat e labels help build and st rengt hen cust omer loyalty t o t he chain as opposed t o brands t hat are not exclusive t o t he chain.
Furt hermore, according t o NOU 2011:4 (2011a), consumer choices suggest t hat t he relevant grocery market s are local and t hat consumers oft en decide between stores based on locat ion. In addit ion, t he Norwegian Consumer Council (2015) argues t hat consumers are oft en reluct ant t o change between st ores if t he st ores are far apart , even when t here is money t o be saved by doing so.
In summary, alt hough t he paramet ers we have discussed all in uence t he compet it ion bet ween rms in t he Norwegian grocery ret ail market , t he focus of t his paper is on price as t he compet it ion paramet er.
T here has been a growing empirical lit erat ure dealing wit h t he relat ionship between prices and compet it ion in t he grocery ret ail market , where t he lit erat ure is most developed in t he case of horizont al compet it ion (Connor, 1999; A sensio, 2014). T he st udy on ret ail food prices and compet it ion focuses largely on whet her increased compet it ion in a geographical de ned area, measured by concent rat ion, market st ruct ure or new ent ries, has any disciplinary e ect on prices or not (Gullst rand and Jörgensen, 2012).
Dat a for t he st udies on t he ret ail food prices and compet it ion is usually obt ained by sampling from di erent geographical market s on municipality or met ropolitan level, usually de ned as regions or urban areas, from which market st ruct ure, prices and variables driving demand and cost s are observed (A sensio, 2014, p. 4). In short , as Asensio (2014) highlight s most aut hors have found t hat higher concent rat ion is associat ed wit h higher prices. Nevert heless, most of t he variat ion in prices is explained by fact ors speci c to t he st ore, such as chain a liat ion or store size, implying t hat t he magnit ude of local compet it ion is relat ively small.
Pricing practices in t he grocery retail market have long been of int erest bot h from a posit ive and prescript ive standpoint (Connor, 1999, p. 121). Posit ive studies, which almost exclusively are concerned wit h pricing under di erent degrees of compet it ion, have primarily been wit hin t he scope of indust rial-organizat ion (IO) economics. T he
13 LOCAL PRICE COMPETITION IN GROCERY RETAILING
studies vary considerably in price indexes, time period, concentration indexes, control variables, sample size, and level of aggregation (Yu and Connor, 2002). According to Connor (1999), there are several noteworthy cross-sectional empirical studies of grocery retail price indexes in the IO tradition. Marion et al. (1979) uses extensive price-check data for grocery retailers operating in 36 cities, and find, by the means of a market-basket price index of 94 branded food items, that markets shares and concentration are positively related to the market-basket price index. The results in Marion et al. (1979) were verified by Cotterill (1986) using a cross section of subpoenaed price data of a product basket from 35 supermarkets in eighteen mostly small, isolated Vermont towns and cities, finding that prices are higher in markets where supermarket concentration is high. In addition, Lamm (1981) also finds, for eighteen major Standard Metropolitan Statistical Areas, that concentration is positively related to food prices, drawing on the price of a homogeneous market basket of food for a family of four published by the Bureau of Labor Statistics.
On the other hand, with emphasis on the geographic variations in prices among proximate rivals firms, Fik (1988) examined spatial competition in the retail food market in the metropolitan area of Tucson in the U.S. Fik (1988) models the price competition as price-reaction functions and by using individual store prices together with the distance to the nearest competitor, the study finds that there is statistical evidence that the intensity of price reaction is a decreasing function of the distance between rival chains. The study conducted by Zhu et al. (2009) on the competition among Wal-Mart, Kmart and Target in the U.S. food market stresses in addition the importance of store characteristics for understanding the spatial competition.
They suggest that the competitive pressure from a store is prominent on other stores located within a few kilometres. Moreover, the paper finds that the impact rapidly declines with additional distance, with the Wal-Mart supercenters being the only ones competing beyond 15 kilometres.
One of the few studies that fails to find a positive relationship between local market concentration and grocery prices is one authored by Newmark (1990). However, Yu and Connor (2002) examines the sensitivity of Newmark’s analysis to a number of methodological and measurement factors. Yu and Connor substitutes, inter alia, the absolute purchase cost employed by Newmark for a true index of food prices. The initial retesting was highly successful in the sense that the correction of the flaws led to a strongly positive and highly significant concentration estimate. On the other hand,
according to Asplund and Friberg (2002), the lack of sufficient geographical variation in the data, which is necessary to trace the relatively small effect of the price-concentration relationship, is the main explanation for the absence of statistically significant results in Newmark (1990). Nevertheless, the retesting by Yu and Connor (2002) shows the importance of careful statistical craftsmanship and good data, especially for the independent variables (Connor, 1999).
However, fewer structure-price studies have been performed outside the U.S.
According to Connor (1999), a probable reason is, inter alia, that reliable food-price surveys are not available or do not cover enough cities for cross-sectional statistical testing. For the Swedish grocery retail market, Gullstrand and Jörgensen (2012) examine the competitive situation by using a detailed dataset covering all Swedish food retailers. The results are unambiguous and suggest that price competition is substantial but that the effect wears off quickly, implicating that a variation in competition may be an important explanation for price variations within Sweden.
More precise, Gullstrand and Jörgensen find that the price competition is substantial among neighbouring stores within a kilometre, with no significant effect between stores separated by a distance of more than one kilometre. Consequently, they conclude that the competition among Swedish food stores is indeed local, and that the area of a municipality should be considered as many small local markets for food retailing.
Their definition of local competition is hence more narrowly defined than in most previous studies, supporting the notion found in studies of Swedish consumer behavior stating that the consumers’ main food store is close in terms of distance. The results also support the notion that the size of a store substantially lowers prices, and that prices are positively associated with population and wealth.
Asensio (2014), on the other hand, conducts an empirical structure-price analysis of supermarkets located in the city of Barcelona, Spain. He estimates the extent to which variation in supermarket prices depend on neighbourhood and store characteristics, the degree of local competition, as well as chain policies The degree of local competition is measured by different market structure variables. However, only the prices corresponding to the second quarter of 2011 for stores with a selling area above 400 square metres are used in the study. Asensio finds that the supermarkets do not respond to local competitive conditions, and that the only variable that seems to have an impact on prices beyond the chain affiliation of the supermarket is the size of the store, in the sense that economies of scale leads to lower prices. As Asensio
15 LOCAL PRICE COMPETITION IN GROCERY RETAILING
acknowledges, the results contradict the conclusion reached by Asplund and Friberg (2002) on the competition between Swedish grocery retail stores, which is found not to depend on chain affiliation at the local market level. However, to depict whether each store is operated independently, Asplund and Friberg (2002) uses HHI on store, chain and region level as structural measures. Asensio (2014), on the other hand, includes instead the number of supermarkets located at a given distance from the store whose prices are observed, to measure the degree of competition.
In Chile, Lira et al. (2012) investigates empirically the relationship between market structure and consumer prices in the supermarket industry. They use a panel of monthly data from 16 cities and find a positive relationship between local competition and prices as well as evidence of lower prices in the presence of major national chain in the cities, underscoring the importance of formats. Cleeren et al. (2010) further emphasizes the importance of formats as the results suggests that intra-format competition is significantly stronger than inter-format competition among supermarkets.
A more relevant problem with the empirical structure-price literature is however, according to Asensio (2014), that it often does not take into account the potential endogeneity of market structure, as ’observed market structures are not randomly assigned (e.g. levels of concentration result from strategic decisions by firms when deciding whether to enter or exit a given market)’ (p. . Not correcting for the endogeneity of the variables used to measure intensity of local competition may bias the results Singh and Zhu (2008). As reported by Cotterill (2006), the majority of previous literature estimating price-competition relationships in supermarkets, does not seem to have addressed the potential endogeneity bias. Two exceptions are the previously mentioned studies by Gullstrand and Jörgensen (2012) and Asensio (2014) on the grocery retail market in Sweden and Barcelona, respectively. To instrument different structural measures, Asensio (2014) argues that ’the obvious [instruments]
are the socioeconomic attributes of the neighbourhoods that have been shown not to be related to prices, but which would influence the presence of a supermarket’ (p. 11).
More precisely, he uses population density, income and land values as instruments.
Gullstrand and Jörgensen (2012), on the other hand, uses the economic structure in the broad neighbourhood of each store as instruments (i.e. regional dummies, HHI (based on sales) within 50 kilometres, and the store size of the nearest competitor based on the Euclidean distance). In Section 6.1.1, we supply a discussion on the endogeneity of market structure.
I n t he following sect ion we will int roduce t he concept of cust omized and uniform pricing of ret ail chains t o provide t he rat ionale behind t he decision of grocery ret ail chains t o x t heir prices nat ionally as opposed t o following a policy of local pricing.
A s just i ed and illust rat ed by Dobson and Wat erson (2005, 2008), ret ail chains essent ially eit her set a chain- or count ry-wide price, or t hey cust omize prices t o t he st ore level according t o local demand and compet it ive condit ions.6 By commit t ing not t o cust omize prices at t he st ore level and inst ead adopt uniform pricing across all st ores in t he chain, raising overall pro t s t hereby, t he ret ail chains could under cert ain circumst ances have a st rategic incent ive to soft en competit ion in compet it ive market s.
If t he chains do not modify t heir pricing policy according t o local circumst ances, we would not observe any relat ionship between market st ruct ure measures at t he local level, and prices. T he reasoning, as ment ioned in A sensio (2014), is t hat alt hough pricing according t o local condit ions should be t he pro t -maximizing st rat egy in local market s, chains may nd it pro t able t o set a uniform price when subject t o di erent int ensit ies of compet it ion across various market s. Hence, if we do not nd any signi cant impact of market st ruct ure on prices, we may have reason t o believe t he chains set a uniform price across all local market s.
Dobson and Waterson (2005, 2008) argue t hat di erent ret ail locat ions have, int er alia, di erent degrees of compet it ion. Hence, we might expect prices to be cust omized across locat ions built on t he not ion t hat rms are bet t er o pract icing t hird degree price discriminat ion bet ween locat ions of di ering compet it ive int ensity. However, under t hese circumstances rms may nevert heless pract ice uniform pricing rat her than varying prices across locat ions. As highlighted by Særvoll and T jøm (2013, p. 13), t he Norwegian grocery ret ail chains operat e bot h on a nat ional and local level, where e.g.
K iwi follows a nat ional pricing policy, while Rema 1000 set s prices according t o local condit ions. Drawing heavily on Dobson and Wat erson (2005, 2008), we will in t he cont inuat ion provide insight int o t he nat ure and ext ent of t he circumst ances where a uniform pricing st rat egy o ers t he st ores operat ed by a mult i-market ret ail chain great er pro t t han a local pricing st rat egy.
6K eep however in mind t hat in our case any given chain could have several di erent concept s who seemingly act independent of one anot her. T hus, chains could in t his case be viewed as chain-concept s.
17 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
T he analyt ical framework considers a chain-st ore (C) which serves all local market s in a count ry. T he count ry is made up of N ( 2) dist inct and economically separat e local ret ail market s. T he framework assumes t he exist ence of two type of market s, respect ively large, compet it ive market s and smaller, uncompet it ive market s. I n each of t he larger, compet it ive market s t he chain-st ore faces compet it ion from an independent local st ore (I ), such t hat t he compet it ive market s make up a local duopoly. T here exist s D ( N ) local duopolies where compet it ion is charact erized by Bert rand-Nash conduct . In t he smaller market s, labelled M (= N D ), t he chain-st ore enjoys a monopoly posit ion. We denot e each of t he M monopoly markets by k = 1 + D , ..., N , and each of t he D duopoly market s by h = 1, ..., D . T he bifurcat ion of t he local market s picks up t he fact t hat local market s di er in respect t o consumer demand, t he number of players operat ing, and t he int ensity of compet it ion.
Furt her, we assume t hat consumer demand is ident ical wit hin, but not between, each market type. The st ores have complet e informat ion about t he market. Moreover, t o ease t he exposition furt her, Dobson and Wat erson (2005, 2008) assume t hat wit hin each market type, consumer demand and operat ing cost s are ident ical. I n addit ion, t here is no consumer demand or cost connect ion between t he market s, such t hat pro t s are separable across market s. We also assume t hat t he operat ing cost s are ident ical for t he chain-st ore and t he independent st ore, and t hat t he st ores operat e under a const ant marginal and unit cost of zero.
Wit h a local pricing st rat egy (L ), the chain-st ores’ monopoly price will be dependent on t he consumer demand in t he monopoly market s ( ). In t he duopoly market s t he price will depend on t he int ensity of compet it ion ( ). T he local pricing equilibrium price in monopoly market s is given by
pm pLC k=
2 (0, 1)
while t he corresponding price in duopoly market s is provided by
pd pLC h = pLI h = 1
2 (0, 1)
When t he consumer demand funct ions are ident ical across all market s, i.e. = 1,
and t he compet ing st ores’ product s are viewed as being demand independent , i.e.
= 0, t hen t he chain-st ore will be indi erent between using local pricing and uniform pricing, and t he price in t he monopoly and duopoly market s will be equal t o 1/ 2.
Ot herwise (i.e. = 1) t he chain-st ore will st rict ly prefer t o use local pricing.
T he combined pro t s of t he chain-st ore across all markets under local pricing (L ) are
LC =
D
h = 1 C hL +
N
k = D + 1
C kL = D (1 )
(1 + ) + (2 )2 + M 2 4 .
W it h a uniform pricing st rat egy (U), on t he ot her hand, t he chain-st ore set s a single price across all market s t o maximise it s combined pro ts. The equilibrium prices when t he chain-st ore adopt s uniform pricing is given by
pUC = (1 )[D (2 + ) + 2 M (1 + )]
D (4 2) + 4M (1 2) ,
which depends, as we observe, on t he int ensity of compet it ion ( ), consumer demand in t he monopoly market s ( ), as well as t he number of monopoly (M ) and duopoly (D ) market s.
T he combined pro t s of t he chain-st ore for t he monopoly and duopoly market s under uniform pricing are t hus
U C =
D
h = 1 U C h +
N
k = 1 U
C k = (1 )(D + M (1 2))[(2 + )D + 2 M (1 + )]2 (1 )[D (4 2) + 4M (1 2)]2 .
To facilit at e the comparison of t he pro t s for t he chain-st ore under local pricing and uniform pricing, it is convenient t o subst it ut e D and M wit h t he paramet er µ = M / N (where µ (0.1)) which speci es t he proport ion of t he market s t hat are monopoly market s for t he chain-st ore. Equivalent ly, 1 µ is t he proport ion of duopoly market s. T herein we can add t hat as long as t he monopoly price is lower t han t he t he duopoly price, t hen t he chain-st ore, irrespect ive of t he value of µ, prefers local pricing.
In ot her words, if LC UC > 0 t hen a local pricing st rat egy will be t he most
19 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
pro t able for t he chain-st ore. Respect ively, uniform pricing across all market s will be t he most pro t able if LC UC < 0. Nat urally, t he chain-st ore is indi erent between local pricing and uniform pricing when LC UC = 0. However, t he default behavior of t he chain-st ore is always t o use local pricing, but the scope of uniform pricing increases subst ant ially if the chain-st ores (t acit ) coordinat e t heir pricing policy choices. Dobson and Wat erson (2005) argue t hat only wit h a visible, credible commit ment t o uniform pricing across all market s, a uniform pricing can be sust ained in equilibrium.
Dobson and Wat erson (2005, 2008) also show t hat t heir t heory of cust omized and uniform pricing st rat egies apply t o ot her market st ruct ure forms. We t herefore have reason to believe t heir results may help explain t he rat ionale behind pricing st rat egies in t he Norwegian grocery ret ail market . I n t he cont inuat ion we will draw on t heir basic idea when, int er alia, examining di erent market st ruct ure variables’ e ect on prices and t hus assess if t he grocery ret ail st ores in Oslo seem t o be direct ly managed by t he chain-concept t hey belong t o.
We assemble an original dat aset t hat includes weekly price dat a for 49 st ores wit hin t he city of Oslo for t he period week 11, 2016, to week 9, 2017. The dat aset consists of 99 unique goods wit hin nine grocery cat egories, tot alling 2,303 st ore-level observat ions across st ores and weeks.7 T he dat a has been made available t o us by NorgesGruppen ASA.
T he st ruct ure of t he dat aset is focused around t he local market s we ident ify in Sect ion 5.1, for which we collect an ext ensive set of measures to re ect market-speci c demand and cost condit ions. On t he local market level we have available demographic and socioeconomic informat ion, whereas on t he st ore-level t he dat aset includes informat ion about chain a liat ion, st ore concept and format , st ore size, and t he geographical coordinat es of t he st ores, as well as st ore revenues for 2015 for roughly half of t he st ores in t he dat aset .. T he dat aset includes price dat a for t hree di erent chains, seven di erent concept s, and t hree di erent format s.8 On t he art icle-level t he
7I n Sect ion 5.3.1 we elaborat e on t he pr ocess of const ruct ing st ore-week level observat ions.
8T he chains t hat are included in t he dat aset are N orgesGruppen, Coop, and Rema 1000, while concept s include K iwi, Rema 1000, Coop Ext ra, Joker, M eny Basis, M eny Pluss, and Spar M arked.
W it h respect t o format s, t he dat aset includes discount ers, supermarket s, and convenience st ores (see Table V ).
dataset includes information about price, campaigns and categorization.
The final dataset is constructed by merging and trimming various datasets that include price data for the NorgesGruppen stores, price data for the price-comparison stores, and information about store and market characteristics. First, we begin by defining the relevant local markets (see Section 5.1). Then we proceed to construct the market structure variables in Section 5.3.2 using the store characteristics we have available. Third, since the frequency of the price data for the NorgesGruppen stores and the price-comparison stores differs, we aggregate the price data for the price-comparison stores on a weekly level9to match the price data frequency for the NorgesGruppen stores. Furthermore, since we do not have price data available for all stores, we retain only the stores that we have price data available for. This reduces the dataset from 1,968,989 observations to 810,150 observations.
Furthermore, information about grocery category is only available for the NorgesGruppen stores, and therefore we need to connect article IDs with category across all price-comparison stores. However, not all article IDs from the price-comparison stores were represented in the NorgesGruppen stores, resulting in a manual pairing process for these observations.10 Furthermore, due to the low store representation and few price data observations we remove all observations for the six first weeks in the sample, namely week 5 through week 10 in 2016, as well as for the last two weeks, namely week 10 and 11 in 2017. This reduces the dataset to with 737,065 observations. The reason that the sample in the first few weeks is inconsistent is that the data-gathering from the price-comparison stores was initiated in early 2016 and evidently it took some weeks before the data-gathering scheme came into full effect. Next we remove observations with unreasonably low or high prices, as these observations can most likely be attributed to data-collection errors.
Further, we proceed to construct the market basket of goods and the store level price index based on a number of criteria (see Section 5.2 and 5.3). During the process of constructing a market basket and a price index, we first remove goods which are not reported in all stores across the sample period, which reduces our dataset with 537,250 observations to 196,815. However, note that the selected goods are not necessarily
9Price data for the price-comparison stores is included on a semi-weekly basis; however, the number of price data observations for the price-comparison stores depends on the frequency of price gathering.
10As with all manual processes, there is a risk that the process is inconsistent. However, we believe that we have been sufficiently meticulous to avoid such problems in the process.
21 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
represent ed in every st ore, every week.11 By const ruct ing t he st ore level price index (on t he basis of t he market basket of goods) t o include only one price observat ion for
each st ore in each week, we reduce our sample t o 2,303 st ore-week observat ions.
Preferably we would want t o avoid explicit ly de ning local market s since this exercise may result in local market s t hat do not coincide wit h t he area of compet it ion for t he st ores. However, while previous research generally has used municipalit ies (e.g., Gullst rand and Jörgensen, 2012), municipal dist rict s (e.g., Asensio, 2014), cit ies (e.g., Lira et al., 2012), Met ropolit an St at ist ical Areas12(e.g., Lamm, 1981) or Labor Market A reas13 (e.g., Cohen and M azzeo, 2007) t o de ne market boundaries, we need t o explicit ly de ne local market areas, such as in e.g. Cot t erill (1986), as we only have demographics and socioeconomic at t ribut es available on t he local market level. T his will allow us t o cont rol for demographic and socioeconomic di erences across market s in Sect ion 6.
When de ning t he local market s t o represent meaningful economic dist inct ions, overlapping local market s should not exist wit hin t he de ned geographic markets, and consumers should not typically purchase groceries from st ores out side of t heir local market , as argued by Cohen and Mazzeo (2007, p. 66). To de ne the local market s we employ NorgesGruppen’s geolocation t ool, which is a dat abase of all grocery st ores in t he Norwegian grocery ret ail market and t heir locat ion. T he geolocat ion t ool makes available demographic informat ion on t he local market -level based on t he market parameters we de ne, as well as store characterist ics and t he geographical coordinat es of each st ore.
Since t he groupings in t he Norwegian grocery market have not made public t heir pricing regions, we begin by de ning our local market s around price-comparison st ores14. First ly, we use t he price-comparison st ores as a st art ing point because in
11We discuss t his pot ent ial pr oblem in Sect ion 5.2
12A geographical region wit h a relat ively high populat ion densit y at it s core and close economics t ies t hroughout t he area, de ned by t he O ce of M anagement and Budget in t he U.S.
13Provided and de ned by t he Bur eau of L abor St at ist ics t o represent int egrat ed economic areas in t he U.S.
14I n t his paper, price-comparison st ores are de ned as st ores which NorgeGruppen collect s price informat ion from wit hin t he limit at ions of t he common indust ry st andard. T he common indust ry st andard is developed by t he grocery chains in cooperat ion wit h t he indust ry organizat ion, V irke, and allows t he grocery chains t o collect price dat a from compet it ors for up t o 20 hours per week (NorgesGruppen A SA , 2016).
addition to price data on NorgesGruppen stores the price-comparison stores are the market competitors that we have price information available for. Secondly, there is reason to believe that NorgesGruppen’s choice of price-comparison stores has been motivated by the competition faced by these stores in particular, thus implying that the price-comparison stores are relevant competitors within their respective local markets. However, pricing data for the price-comparison stores is only available for certain areas within the city of Oslo. Therefore, our analysis is restricted to using local markets around the areas of Sagene and Lambertseter, which are located in the urban and suburban areas of Oslo, respectively.
Furthermore, the geolocation tool does not allow for market definitions other than those based on drive time. Nevertheless, both Norwegian Competition Authority (2015) and the UK Competition Commission, according to Dobson and Waterson (2008), employ drive time in determining the store choices consumers face at the local level, using different drive times depending on area characteristics. The relevant price-comparison stores are used as starting points for the drive time computation.
Since the areas of interest in our analysis differ in population density and settlement patterns, we decide on using a drive time of 2 minutes and 5 minutes for the urban and suburban area, respectively. The geolocation tool assumes that one minute of drive time equals a distance of roughly 750 meters. The difference in the choice of market definitions between the areas is motivated by the assumption that consumers are more likely to walk to grocery stores in urban areas, whereas they are more likely to drive in suburban areas, resulting in larger local markets. However, the drive time in the geolocation tool does not account for any traffic congestion patterns, resulting in local market definitions that we believe are too broad given our expectations to the actual drive time in the areas. Moreover, Cotterill (1986) argues that defining local markets that are too broad substantially reduces the ability of our models to explain pricing behavior. Hence, to take into account real drive time and the market definition concerns of Cotterill (1986), we discuss several different refinements to our local market definitions.
For the urban area we decide on narrowing down the area such that the maximum Euclidean distance between stores in the market is no more than 500 meters, which translates into an area of approximately 0.8 square kilometres. For the suburban area we decide to include stores within a maximum Euclidean distance of 1,500 meters,
23 LOCA L PRI CE COM PET I T I ON I N GROCERY RETA I LI NG
Fi gur e I I — L ocal M ar k et B oundar ies
F ig ur e I I . Physical locat ion of st ores wit hin t he local market boundaries; here represent ed wit h Ulsrud in t he suburban area of Oslo. (NorgesGruppen A SA , 2017)
which t ranslat es int o an area of approximat ely 7 square kilomet res. Using t he 2-t o-5 minut e rule, t he 500-t o-1,500 met er re nement , and t he nat ural boundaries of t he t ypography of t he respect ive areas, we ident ify 15 local market s. Table I I I report s select ed st at ist ics for t he 15 local market s, including number of st ores, demographics, and the number of observations for each local market . T he local markets range in size from Ryen wit h 11 grocery st ores and 11,781 resident s, t o Vossegat a, wit h 3 grocery st ores and 2,649 resident s.
A s we will discuss furt her in det ail in Sect ion 5.3, Cot t erill (1986) argues t hat even when grocery ret ail st ores provide t he same good, each st ore’s real and perceived service levels vary, which is t ermed enterprise di erent iat ion by retailing economist s.15 Since t he het erogeneity occurs at t he store level, we can use t he aggregat e price level of a st ore for a market basket of goods rat her t han using t he individual prices of goods.
In order t o ident ify a comparable, homogeneous market basket of goods across all
15A grocery ret ail st ore is di erent iat ed by t he product -service-price mix it o ers (Cot t erill, 1986).
