Does Accessibility Affect Retail Prices and Competition? An Empirical Application

Does Accessibility Affect Retail Prices and Competition?An Empirical Application

Juan Luis Jiménez & Jordi Perdiguero

Published online: 1 September 2010# Springer Science+Business Media, LLC 2010

Abstract This paper attempts to link the concepts of accessibility and the firm’sconduct in the regional retail market in Spain. We use a database that includessale price, service station location, level of traffic and type of road. We show thataccessibility has two main effects on final prices. The accessibility of own-brandgas stations increased their prices while the accessibility of rival gas stationscauses price reductions. If we include the value of time, then no rationalconsumer should travel further than his nearest petrol station in search of lowerprices. Finally, our paper shows that service stations can establish a dominantposition if consumers do not have access to other retailers within a 17-minradius.

Keywords Accessibility . Location . Petrol stations . Oligopoly

JEL Classification R40 . L13 . L81

Netw Spat Econ (2011) 11:677–699DOI 10.1007/s11067-010-9144-5

J. L. Jiménez (*)Grupo de Economía de las Infraestructuras y el Transporte (EIT), Facultad de Economía,Empresa y Turismo, Universidad de Las Palmas de Gran Canaria,Despacho D. 2-12 Campus de Tafira, 35017 Las Palmas, Spaine-mail: [email protected]

J. PerdigueroDepartamento de Política Económica, Grup de Governs i Mercats (GIM), Institut de Recerca en EconomíaAplicada (IREA), Universitat de Barcelona, Av. Diagonal 690,08034 Barcelona, Spaine-mail: [email protected]

1 Introduction

Accessibility is a widely used concept in many spheres of transport economy.During the last 40 years it has been increasingly used as a reference term in thefields of transport economics and land planning, and increasing the accessibilitylevels has become nowadays a common objective for these plans (van Wee etal. 2001).1

In broad terms, accessibility measures—through different methods—the easeof reaching valued destinations; more consumers will reach those firms locatedin areas with higher accessibility in the same amount of time, thus increasingtheir business opportunities.2 For this reason, transport planning has traditionallyfocused on the provision of infrastructures, since a direct relationship was assumedto exist between the quality of infrastructure, accessibility and welfare. However,this premise could bring about a vicious circle with considerable long-termenvironmental repercussions.

A parallel branch of the literature has instead analysed the relationshipbetween retailing prices and accessibility. Kaufman et al. (1997), for example,study for the US the relationship between food prices in different types ofestablishments and their accessibility in terms of family incomes. They show thatlow-income families with a low probability of living in the vicinity of supermarketsmay have to buy the same highly priced products as families with greaterpurchasing power.

According to this idea, MacDonald and Nelson (1991) demonstrate that the pricesof a selected basket of goods are 4% lower in suburbs than in city centres. By usinga database that includes the locations of fast-food restaurants, Stewart and Davis(2005) determine how differences in location affect their final prices. Theseexamples show that accessibility can be analysed from a different perspective ofdirect transport economics.3

This paper attempts to demonstrate the direct relationship between accessibilityand the level of competition in the retail petrol market. In fact, a possible motivationis that petrol consumption is unlikely to be a programmed decision and, for thisreason, they can exercise market power by applying high prices, and they may evencreate “social exclusion” (if we consider petrol to be a “basic commodity”). This ismore serious for car users in remote areas where public transport is not normally asfrequent as in city centres.4

This paper combines the classic analysis of accessibility with a closerapproximation of the industrial organization methodologies. This second

1 Handy (2002) explicitly states that in most U.S. transport plans There are constant references to theaccessibility issue and its improvement defined as a major objective.2 See Handy and Niemeier (1997); Baradaran and Ramjerdi (2001); El-Geneidy and Levinson (2006).3 For example, how the access facility generates opportunities for individuals in terms of improving theiremployment prospects (see van Wee et al. 2001; Srour et al. 2002; Franklin and Waddell 2003). For pricevariation of goods such as housing, see El-Geneidy and Levinson (2006); and for the possibilities whenlocating new sales outlets, see Ritsema van Eck and de Jong (1999).4 Nevertheless, we do not estimate the number of consumers who will be excluded if monopolizationoccurs, and similar. Due that, it is only a motivation, not an issue of the paper.

678 J. L. Jiménez, J. Perdiguero

approach analyses how accessibility to the different stations changes the marketstructure and influences the prices set by firms. There is an extensive amount ofliterature analyzing the influence of market structure on the prices set by firms.However, not as many articles consider the role that the distance to the othercompetitors plays in business strategies.

One of the first articles to take into account the effect that the distance betweenthe competitors has on their pricing strategies is that of Spiller and Huang (1986).They show how wholesale gasoline markets in the northeast U.S. that are closer aremore integrated because proximity lowers the cost of arbitration. Those wholesalersthat are further apart and therefore not connected to any relevant market have agreater ability to exercise market power and would be candidates for analysis by thecompetition authorities.

The wholesale market in the U.S. has also been analysed by Pinske et al.(2002). The authors propose a semiparametric estimator that allows them to discernwhether the behaviour of the various operators is close to global competitionmodels, where all the operators are competing with everyone, or models of localcompetition, where the players compete only with their neighbours. Theeconometric results show how competition in this industry is basically local.Therefore, agents only compete with closer rival operators, and not with those whoare geographically distant.

Contrary to previous studies show empirical evidence that wholesale gasolinemarkets are local, being the geographically closest competitors that significantlyaffect the pricing strategies of firms, Bromiley et al. (2002) note several pointsat which the gasoline retail market would also have a strong local character.This fact can explain why there are service stations offering different priceseven within the same city. One of the main reasons the author put forward forthe relevant market in which competition is reduced is that petrol stations havehigh search costs to be borne by consumers in finding a cheaper station. Ifthese high search costs are added to the individual savings that are normallyexpected, gas stations only have incentives to compete strongly against theirclosest rivals.

The proposals outlined by Bromiley et al. (2002) seem to find empirical evidencein the study by Barron et al. (2004), where they show that a greater number ofcompetitors within a distance of a mile and a half leads to a lower level of prices inthe market and a lower price dispersion. Therefore, the more competitors there arewithin a close distance, the lower the price fixed by the operators. When taking intoaccount the distance to the nearest competitor for the number of competitors within amile and a half, the results did not vary significantly. The closer the competitors, thelower the price and the lower its dispersion. The econometric results show that thepetrol retail market in the United States is affected mainly by the local competition.However, the authors do not analyse whether this effect of local competition mayhave some kind of domino effect, spreading the competitive effect to a widergeographical area.

Precisely one of the objectives of the article by Atkinson et al. (2009) is toanalyse the existence of domino effects in price movements of the stations inGuelph (Ontario). These effects seem to dominate when there are price reductions,

Does Accessibility Affect Retail Prices and Competition?... 679

but are not seen when the movements are on the rise. The results also show that,although the station only seems to react to the movements of a small group ofcompetitors, it is not always the closest geographically.5

The goal of this paper is to analyse how the difference in accessibility, measuredthrough the distance to other petrol stations affects the prices charged by retailers inthe market. That is, to what extent do service stations whose consumers have easyaccess to stations of rival brands suffer increased competitive pressure and are forcedto set a lower price? The paper also examines whether service stations whoseconsumers have better access to stations of the same brand have greater marketpower and thus set a higher price.

Although there is evidence in the literature on the relationship between the marketstructure and the prices set by companies, and more specifically the geographicalproximity of rivals, this article presents significant contributions. Firstly, it analysesthe competitive effect exerted by rivals in terms of geographical distance. Inprevious articles the authors assume a certain distance to define the geographicalmarket in which companies compete, to analyse subsequently the effect of thenumber of competitors in that market. Our approach nevertheless allows us toanalyse how the effect varies in the number of opponents depending on location andto verify the extent to which rivals located further away exert less competitivepressure. This evidence may be useful for defining the relevant geographic marketsin which operators compete.

Secondly, the article also analyses how the presence of establishments of the samebrand at different geographical distances allows operators to increase their marketpower and thus impose higher prices. To our knowledge, there is no empiricalevidence in the gasoline market that shows the effect of own brand petrol stations inthe market to the prices fixed by the retailer.

Thirdly, and taking into account the effects generated by rivals and the stations ofthe same brand in different geographical markets, we can analyse the effect of amonopoly by a single company in a particular market, that is, how prices would varyif a company acquired all the service stations within a certain radius. This result maybe of broad utility to anticipate the possible effects of mergers in the market.

The main results of the econometric analysis show us how rivals that aregeographically distant exert lower competitive pressure. Similarly, stations of thesame brand make it possible to increase the market power and thus the price,although it again decreases as it becomes further away geographically. The servicestations will increase their margin by more than 5% if they monopolize the servicestations within 17 or so minutes of the posting.

The paper is structured as follows. In Section 2 we present the data used for theempirical application and the main descriptive statistics, which we develop and solveeconometrically in Sections 3 and 4. Finally, in Section 5 we show the mainconclusions.

5 The negative relationship between the proximity of competitors and the price level is not unique to thegasoline market. A good example is provided by the article by Mazzeo (2002) on the market for roadmotels in the United States.


2 Data

The Spanish petrol market has undergone various gradual changes during the last20 years, at both wholesale and retail levels. Basically, it passed from being a statemonopoly run by CAMPSA, now Repsol, to liberalization of prices and generalsupply, although it is currently characterized at all levels by notable degrees ofconcentration.6

The empirical application has been made in an autonomous region of Spain,Galicia. The Spanish petrol market has undergone an intense process of restructuringin recent decades. This liberalization process contains two main packages: one ofliberalization and privatization measures, aiming to increase competition and henceefficiency in the sector; and, secondly, measures to facilitate vertical and horizontalintegration with the intention of creating a “national champion”.

The result of this restructuring process is to obtain a fully liberalized market inwhich the state does not have any involvement, but which is in turn extremelyconcentrated, both horizontally and vertically. This means that in almost all cases thefinal price set by the service station is determined by the large oil company. This isbecause many stations are vertically integrated companies with refining capacity inSpain (Repsol, Cepsa and BP) and the rest have exclusive contracts that cause long-term indirect fixation of the final price as these contracts provide pricerecommendations. This has already been pointed out by the competition authorityin Spain, Comisión Nacional de Competencia (CNC), where a recent report, CNC(2009) indicates that the existing high vertical integration in the Spanish gas marketis causing the price decision to be taken by a limited number of agents, the major oilcompanies, which can reduce the level of competition in the market and increaseprices as well as their uniformity.

The improvements in the global supply of land transport infrastructure, which arelinked to this liberalization process, have not enabled Spain to achieve the sameindicators as the rest of the European Union. This is despite the transport networkbeing dominated by road usage, both for people and for merchandise; see Camposand de Rus (2002). Between the 1970s and the 1990s Spain reduced its average percapita funding, in spite of the aforementioned global increase in the level ofinfrastructure.

To summarize, in absolute terms the development of infrastructure through moreand better roads and the possibility of reaching a more favourable situation regardingpetrol, the sector’s most important input market, mean that it is possible to influencepositively the levels of accessibility in advance. These should be understood aspotential opportunities to achieve a determined objective, i.e. lower prices and ashorter time period for the acquisition of goods.

In respect of the database, Galicia is an area of Spain with a high degree of bothrural and urban areas. These data enable us to compare the results between theinfrastructure and demand typologies. In this region, in 2007, there were 604 retail

6 See Perdiguero and Borrell (2007) for a more detailed view of the sector’s liberalization process and itscurrent situation in Spain. Alternatively, see Perdiguero and Jiménez (2009) for a differentiated regionalanalysis.


service stations representing different companies and locations. Both these factorsdetermine market behaviour.7

The database contains the sale price for unleaded petrol 95, set by thecompanies in the week 5 to 11 November 2007 and obtained from the SpanishGovernment’s Ministry of Industry, Trade and Tourism website. As not all theservice stations provided information, the database we have used contains 511out of 604 stations. The 93 service stations with missing values do not show anyspecific pattern, either in terms of brand or location, so we do not believe thatour results will be affected. In fact, petrol shows the same percentages for thetotal population (604) and those 511 petrol stations analysed, using data onbrand, traffic intensity and location. For this reason, we can consider our sampleas representative of the total population. In sum, no sample bias has beenincluded. As we have mentioned, the brand and road location of each servicestation that provided data are also available.

Using the location of each service station, we were able to allow the followingdemand and access facility indicators: the category of the road the station is sited onand the traffic intensity; these variables show a low correlation of 0.17. For the roadsthe classifications are: rural, urban transient, industry/office and, with a lesserpresence, residential and motorway. For access, our classification uses four levelsranging from lower to higher volumes of traffic, and we have assigned themascending numerical values: poor (1), medium (2), good (3) and very good (4).Finally, using Spain’s annual abstract of statistics (Anuario Estadístico de España,Caixa L, 2007) for 2007 prepared by La Caixa’s Research Department, we haveestimated the local population; we have also calculated the international wholesaleprice rates for unleaded petrol 95 in the Rotterdam spot market, by using OPEC’swebsite to calculate the gross margin.

Table 1 shows some descriptive indicators, grouped into petrol brands. From thiswe can draw attention to the market structure. Three companies, the Repsol Group,Cepsa and Galp, account for over 67% of all the region’s stations.

One item of data obtained from the service station location is the distance tothe other stations, in minutes, which depends on the road speed, and metres.This allows us to demarcate, for each point of sale, the possibilities forcompeting with the others. The data demonstrate that, apart from having agreater number of petrol stations, these three companies have an averageminimum distance from other service stations, regardless of brand, whichexceeds the 12.6 min mean. As we will see in Section 5, this is a significantadvantage with respect to avoiding competition. Finally, the distribution of salespoints in Galicia, in both rural and urban areas, is very similar to the average forthe sector; Agip and the Repsol Group focus more on built-up than on countryareas.

7 All the following data excluding prices were supplied by Catalist (www.catalist.com), which is acompany dedicated to the sale of information on the petrol sector. Annex I gives a more in-depthexplanation of some of the variables. The subsequent analysis was carried out by the programming ofcomputer codes, using the Matlab statistical program and the SAS package; these permitted us to managea broad database.


http://www.catalist.com

The data bring up one result that we will see in later sections. Almost 50%of the points of sale are located on roads of average intensity, and only 4% arelocated on low-intensity roads. The greater the intensity of traffic on the road,then the higher the price of petrol 95 is. This is a clear effect of demand ineach local market. For the same aforementioned motive, service stations arefurther apart from one another when the intensity is lower. These results areshown in Table 2.

As for the relationship according to the type of road, we initially commentedthat the economic, social and geographical characteristics of Galicia determinethe type of retailer. A total of 63% of stations are situated on rural roads,

Table 1 Some descriptive statistics (by brand)

Brand Number ofretailers

Averageprice

Average trafficintensity

Average minimumtime (mins) to otherpetrol stations

% petrol stations in …

Rural Urban Industry/office

Motorway

AGIP 18 1.0957 (0.01) 2.5 9.0 (5.7) 11.1 77.7 5.6 5.6

Alcampo 2 1.0965 (0.004) 2.5 1.0 (0.4) 0 50 50 0

Avanti 2 1.0935 (0.006) 2 6.3 (5.9) 100 0 0 0

Avia 1 1.105 3 9.5 100 0 0 0

Campaño 1 1.121 3 23.7 100 0 0 0

Campsa 144 1.098 (0.004) 2.5 16.6 (15.8) 55.6 42.4 2 0

Carrefour 1 1.097 3 5.8 0 100 0 0

Ceao 2 1.085 (0.01) 3 2.3 (0.7) 50 0 50 0

Cepsa 110 1.102 (0.006) 2.6 12.0 (12.1) 67.3 27.3 4.5 0.9

Erg 12 1.094 (0.008) 2.7 8.4 (8.3) 66.6 16.7 16.7 0

Eroski 3 1.055 (0.005) 2.3 3.6 (3.2) 66.7 0 33.3 0

Ertoil 1 1.097 2 5.3 100 0 0 0

Fegoblan 1 1.101 (0) 4 6.1 100 0 0 0

Fina 3 1.098 (0.002) 2 19.5 (15.0) 66.7 33.3 0 0

Galpa 41 1.103 (0.019) 2.7 9.4 (7.8) 73.2 12.2 2.4 7.3

Meroil 15 1.099 (0.007) 2.2 8.3 (8.7) 73.3 20 6.7 0

Ortegal oil 3 1.098 (0) 2 16.4 (8.7) 33.3 66.7 0 0

Petronora 43 1.100 (0.006) 2.5 15.1 (10.4) 67.4 27.9 2.3 0

Repsolb 113 1.099 (0.009) 2.8 13.0 (11.9) 59.3 28.3 8.0 4.4

Sayar 1 1.09 2 13.0 100 0 0 0

Sertuy 1 1.101 4 6.3 100 0 0 0

Shell 14 1.100 (0.009) 2.8 6.7 (4.5) 50 42.9 7.1 0

Top oil 1 1.109 3 2.7 0 100 0 0

Unbrandeda 66 1.096 (0.008) 2.3 9.4 (8.8) 65.2 21.2 10.6 0

Valcarce 2 1.101 (0) 2.5 20.5 (11.0) 100 0 0 0

Average (ortotal data)a

604 1.099 (0.00009) 2.6 12.6 (17.5) 63.2 28.6 5.6 1.7

Own elaboration from Catalist database information. Standard deviation for average price in bracketsa Some missing values are includedb The Repsol Group consists of Campsa, Repsol and Petronor


although the highest average price for these products is in the city where 28%of all petrol stations are located; motorways make up 1.6% of the total. Asmight be expected, petrol stations are further apart on rural roads andmotorways than in urban centres, business and work centres and even industrialareas (Table 3).

The data seem to show us that prices are affected both by accessibility anddemand. On the one hand, consumers have less accessibility, as do competitors,and this means that retailers may increase their prices; on the other hand, theprice equilibrium may be lower if retailers show lower levels of demand.

3 Empirical strategy

With these market characteristics, the issue is whether some types of determiningfactor derived from the accessibility affect the prices of the product analysed. Doesimproving accessibility increase consumers’ chances of finding better prices? Tounderstand this we have focused on isochronic or cumulative opportunity

Table 3 Descriptive statistics (by type of road)

Type of road Retailers Price Average minimum time (mins)to other petrol stations

Rural 382 1.099 (0.009) 13.8 (11.5)

Industry/office 34 1.099 (0.019) 4.3 (4.0)

Urban transient 173 1.100 (0.009) 11.3 (13.7)

Motorway 10 1.100 (0.001) 21.0 (15.1)

Others 5 1.090 (0.004) 4.6 (2.5)

Own elaboration

Standard deviation in brackets

Table 2 Descriptive statistics (by average traffic intensity)

Average traffic intensity Retailers Price Average minimum time (mins) toother petrol stations

Poor 25 1.096 (0.009) 18.7 (14.8)

Medium 302 1.098 (0.007) 14.8 (12.6)

Good 157 1.100 (0.012) 8.7 (10.7)

Very good 118 1.100 (0.009) 10.8 (10.5)

Unsurveyed 2 1.089 (0.04) 4.8 (1.0)

Own elaboration



measurement, which is a traditional approach developed by Wachs and Kumagai(1973) and Vickerman (1974). This indicator counts the number of potentialopportunities that may be reached within a predetermined travel time period ordistance. It is expressed as:

Ai ¼XJ

j¼1

Bjaj ð1Þ

where Ai is accessibility measured from point i to potential activity in zone j, Bj is abinary value equal to 1 if zone j is within a predetermined threshold and aj are theopportunities in zone j.

Using this measure as a starting point, and adapting the consumer’s decision toacquire petrol in one service station or another, always assuming that the purchasingdecision always occurs in the same location (in a 20 min radius), we propose thefollowing analysis. We have stipulated beforehand, for the group of petrol stationsstudied, how many of them have a competitor, regardless of the brand, within20 min’ travelling distance. We have obtained the probability of a consumer atstation i (Pri) obtaining a lower price in petrol stations located within this radius(20 min) than the station of origin. It is as follows:

Pri ¼ niN

ð2Þ

where N is the total number of petrol stations situated within 20 min of the consumerlocated at i and ni is the number of petrol stations in that zone whose prices are lowerthan those of station i.

We present, in the following Tables 4 and 5, the results by type of location androad traffic intensity (and in Map I, Annex II). We have taken three variables intoaccount: the average number of petrol stations that make up this exogenous“radius” of attraction, the average discount that would be obtained from thesecompetitors in euros per litre and, in the last column, contains the probabilityaccording to Eq. (2).

A total of 91% of petrol stations are located either in rural or in urban areas. Thefirst classification shows that the retailers located on rural roads have a lowernumber of competitors in a given radius than the urban ones do—7.8 as comparedwith 10.5—despite the average discount being greater in the rural environment.

Table 4 Probability of obtaining better prices (by type of road)

Type of road Average number of petrol stationswithin 20 minutes, with better prices

Average discount in petrol stationswithin 20 minutes (€ per litre)

Pri

Rural 7.82 0.0030 0.204

Industry/office 10.32 0.0013 0.176

Urban transient 10.49 0.0021 0.242

Motorway 12.2 0.0004 0.109

Others 3.66 0.0483 0.507

Own elaboration


Nevertheless, the probability that the individual must consider is greater in the citythan in the country, 0.24 versus 0.20. This indicates that, in a Galician city,consumers of petrol 95 have a higher probability of obtaining a lower price inanother petrol station if they continue their search; this is because there is a greaterconcentration of population and activity.

Table 5 shows a similar result but uses road traffic intensity instead. Here theprobability of finding a petrol station with a lower price compared with theprevious grouping is 12% for those with average intensity, which is 50% of allthe petrol stations; it is 7% for those with good intensity, and for those withvery good intensity it is 6%. Consumers who use a poor intensity road onlyhave a 4% possibility of finding a lower price at other petrol stations locatedwithin 20 min.

Despite the aforementioned, a consumer who knows the petrol station prices inthis radius8 does not have to travel to that service station. In order to define moreclearly the decision to change, we use the traditional sequential search modelsdeveloped by Stigler (1961). This model is based on an individual “cost–benefitanalysis” at the consumer level: why do consumers have to move from one petrolstation to other if there is a low probability of obtaining a discount and they have tospend money to travel there?9

In our case any consumer near or at a station would have to choose betweenfilling up there or evaluating the “benefit” of moving to another petrol station to seeka lower price, while taking the distance and time involved into account. For this lastpoint, we have updated and used time value data from the European HEATCO(2006) project, for business passengers using cars by the hour. In Spain in 2002, itwas equal to €25.95, which would be €30.05 in 2007 if we update the figure usingthe Retail Price Index.

Table 5 Probability of obtaining better prices (by average traffic intensity)

Type of road Average number of petrol stations within20 minutes, with better prices

Average discount in petrolstations within 20 minutes

Pri

Poor 11.64 0.0026 (0.005) 0.042 (0.084)

Medium 8.68 0.0024 (0.007) 0.122 (0.206)

Good 8.41 0.0035 (0.012) 0.072 (0.131)

Very good 9.09 0.0020 (0.005) 0.060 (0.109)

Unsurveyed 2 0.0022 (0.003) 0.166 (0.192)

Own elaboration


8 It should be borne in mind that service stations in Spain are obliged to display their prices on panels attheir entrances; they must also make their prices available to the Ministry of Industry, Trade and Tourism,which publish them daily on its webpage.9 As we indicate in the introductory paragraph of this Section 3, we are assuming that most people livenear their jobs. This assumption induces us not to take into account commuters, i.e., people who live faraway from their jobs. In this case, the transport cost would be reduced and the analysis in Eq. (3) wouldnot be effective.


So, a consumer would seek a product until the expected discount was lowerthan or equal to the search costs. In this case the proposed equilibrium equationwould be:

Di»Pri ¼ VT»Distance ð3ÞThe left-hand part of the equation is the average discount received from buying

from a cheaper station (Di), multiplied by the probability of obtaining a discount ifwe move to the next petrol station. This gives us the marginal revenue. VT is thevalue of time and Distance is the number of minutes between petrol station i and thelower priced j, the lower price being a necessary condition. The latter yields themarginal cost. The difference between them is the marginal profit.

From (3), assuming forty litres for each fill-up and using the Galician database,the results indicate that no rational consumer using this type of analysis would travelfrom the nearest petrol station to another in search of this discount, motivated by avalue of time that is greater than the potential benefits. If we consider the value oftime for leisure trips, which is €9.60 when updated to 2007, in only three out of 511petrol stations would a consumer be interested in travelling to another.

To sum up, some price differences exist, and the probability of obtaining cheaperprices is greater when the ease of access to the points of sale is better. Accessimproves in line with a greater number of retailers, higher urban concentration andmore intense traffic. Despite this, the time costs would mean that no rationalconsumer would travel to a different petrol station to find an expected discount thatdoes not compensate for the costs incurred. Authors such as Borenstein (1991) havealready indicated that these scarce benefits in searching for a cheaper petrol stationare one possible reason for fixing prices above the perfect competition level. Wenow pass on to Section 4, with the objective of econometrically establishing therelationships among prices, type of road, population, brands and the accessibility ingeneral.

4 Econometrical analysis

In this section the alternative methodology we use follows on from the paperby Borrell and Perdiguero (2007), and we have borne in mind the information itprovides on service station location. This methodology deals with the priceapplied by stations, in the context of the distance to the other surroundingpetrol stations.

There is a long history of competition analysis in industrial economics, especiallywhen companies are located in different points of the territory. The first theoreticalreferences were by Hotelling (1929) and Salop (1979), who respectively developedthe linear city model and the unitary circular spatial market theories. These modelshave already demonstrated the importance that the proximity of the competitors canhave on the prices applied by the different points of sale. Competition in terms ofspace is by its very nature a paradigm of product differentiation. Each operator firstchooses the physical location and given the competitors’ location it then sets theprice at which to offer its products.


In recent years a whole set of methodologies has been developed to deal withproduct differentiation model data. The foundation for the empirical development ofthese models was laid by Pakes et al. (1993), Berry (1994) and Berry et al. (1995);however, more recently Ivaldi and Verboven (2005) have shown its growingusefulness in analysing competition policy cases.

The equilibrium equations for the aforementioned models show how the prices atdifferent points of sale depend on the elasticity and cross elasticity of other points ofsale. One direct way of estimating these elasticities is to calculate the demandequations, while bearing the product differentiation in mind. However, such a directestimate would require detailed information, which is seldom available, on aspectslike the characteristics that differentiate the products and quantities sold by eachpetrol station.

We will calculate the effect of the demand and cross elasticities using theeffect of the analysed service station and of rival brands upon the current priceit fixes. Having relatively close rivals should lower the price, as the demandelasticity will be higher. In other words this is the quantity the seller wouldcease to sell if the price increase is greater than that of its nearby rivals,although this depends on whether those rivals sell the same brand. Ourempirical estimate is the following:

Pi ¼ a0 þ a1N rivals ðm minutesÞi þ a2N own ðm minutesÞi þ "i ð4Þ

i is each one of the 511 service stations and μ varies between 1 min and 30 min,which is a greater range than the previous 20 min analysis. We carry out 30econometric estimations beginning with 1 min of travel time, where we relate theprice of the service station i to the number of rivals and the number of same-brandpetrol stations within 1 min and so on.

We break down the data in order to analyse the more detailed market structure.Table 6 shows the number of same-brand and rival petrol stations within the 1–30 min time interval for each service station. The information in this table shouldaffect the econometric analysis, as we would expect the competitive effect of rivalpetrol stations to influence prices negatively; conversely, same-brand stations couldaffect them positively: this is the domino effect within the market. This latterrelationship should be greater whenever this station is less accessible or moreisolated from other stations.

The prices set by the stations may depend on the number of rivals and the numberof same-brand service stations, and these two variables may in turn depend on themarket price. A high price equilibrium is a sign for other firms to enter the market.The variables for the number of rival and same-brand petrol stations could createproblems of endogeneity.

The instruments used to solve the endogeneity problem of both variables (N rivalsand N own) are: the population of the municipality where the petrol station islocated; a set of dummies that take the value one if the nearest rival is of a particularbrand and zero otherwise; a set of dummies that determine the type of roadway onwhich the petrol station is located (rural, urban, industry/office and motorway), aswell as measuring the intensity of traffic on the road on which the service station islocated.


The estimation includes a geographical cluster that takes into account possibleheteroskedasticity among service stations on different roads. Including this clusterfor each road we take into account that all the service stations on a particular routecan have a different unobserved effect of the petrol stations on other avenues, and sothe estimate is not biased.

In the following Table 7 we can see the econometric results.As we can see in the previous table the signs for the variables are what we

expected, although on occasions there are no significant results in the first few

Table 6 Market Structure in influence zones by trip time (minutes)

Minutes Number of petrolstations competing

Number of petrolstations in monopoly

Average number ofcompetitors incompetition markets

Average number ofown petrol stations

1 15 589 1 0.14

2 54 550 1.26 0.17

3 111 493 1.54 0.25

4 141 463 1.88 0.32

5 186 418 2.17 0.41

6 215 389 2.51 0.50

7 240 364 2.76 0.59

8 274 330 3.01 0.73

9 306 298 3.21 0.84

10 335 269 3.45 1

11 351 253 3.76 1.14

12 370 234 4.07 1.30

13 386 218 4.43 1.40

14 406 198 4.78 1.54

15 427 177 5.04 1.65

16 442 162 5.40 1.79

17 455 149 5.66 1.92

18 470 134 5.97 2.06

19 477 127 6.36 2.23

20 487 117 6.67 2.38

21 496 108 6.99 2.48

22 508 96 7.23 2.65

23 514 90 7.61 2.79

24 520 84 7.97 2.98

25 524 80 8.28 3.17

26 530 74 8.58 3.33

27 533 71 9.00 3.50

28 536 68 9.41 3.63

29 549 55 9.69 3.79

30 551 53 10.17 3.95

Own elaboration


Tab

le7

Econometricalresults

Eq.

(4).Instrumentalvariablesanalysis

1minute

2minutes

3minutes

4minutes

5minutes

6minutes

7minutes

8minutes

Constant

1.09

90***

1.0990

***

1.0988

***

1.0989

***

1.0992

***

1.09

93***

1.09

94***

1.0995

***

(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)

Nrivals(.)

−0.0045

−0.002

3−0

.0012

−0.0014*

−0.0015*

*−0

.0012*

*−0

.000

9*−0

.000

7*

(0.495

)(0.376

)(0.263

)(0.080

)(0.044

)(0.049

)(0.073

)(0.079

)

Now

n(.)i

0.00

020.0035

0.0037

0.0032

0.0029

*0.00

24*

0.00

160.0011

(0.249

)(0.258

)(0.119

)(0.119

)(0.074

)(0.075

)(0.113

)(0.132

)

Sargantest

19.289

19.003

17.552

17.296

16.341

16.313

17.120

17.399

(0.201

)(0.214

)(0.287

)(0.301

)(0.360

)(0.362

)(0.312

)(0.296

)

9minutes

10minutes

11minutes

12minutes

13minutes

14minutes

15minutes

16minutes

Constant

1.09

95***

1.0996

***

1.0997

***

1.0997

***

1.0998

***

1.09

98***

1.09

99***

1.0999

***

(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)

Nrivals(.)

−0.0006*

−0.000

6*−0

.0005*

−0.0005*

−0.0004*

−0.0004

−0.000

4*−0

.000

4*

(0.067

)(0.077

)(0.096

)(0.086

)(0.069

)(0.102

)(0.081

)(0.060

)

Now

n(.)i

0.00

110.0009

0.0007

0.0007

0.0006

0.00

050.00

050.0005

(0.105

)(0.129

)(0.165

)(0.145

)(0.123

)(0.175

)(0.146

)(0.112

)

Sargantest

16.987

17.528

18.036

17.760

17.387

18.189

17.604

17.039

(0.320

)(0.288

)(0.261

)(0.276

)(0.296

)(0.253

)(0.284

)(0.317

)


17minutes

18minutes

19minutes

20minutes

21minutes

22minutes

23minutes

24minutes

Constant

1.10

00***

1.1001

***

1.1001

***

1.1001

***

1.1001

***

1.10

01***

1.10

01***

1.1001

***

(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)

Nrivals(.)

−0.0004

−0.000

4**

−0.0003*

*−0

.0003*

*−0

.0003*

*−0

.0003*

*−0

.000

3**

−0.000

3**

(0.055

)(0.039

)(0.037

)(0.039

)(0.034

)(0.032

)(0.028

)(0.029

)

Now

n(.)i

0.00

050.0005

*0.0004

*0.0004

*0.0004

*0.00

04*

0.00

04*

0.0004

*

(0.102

)(0.077

)(0.079

)(0.084

)(0.075

)(0.072

)(0.068

)(0.071

)

Sargantest

16.930

16.342

16.254

16.269

15.900

15.857

15.657

15.767

(0.323

)(0.360

)(0.365

)(0.364

)(0.389

)(0.392

)(0.405

)(0.398

)

25minutes

26minutes

27minutes

28minutes

29minutes

30minutes

Constant

1.10

01***

1.1001

***

1.1001

***

1.1001

***

1.1001

***

1.10

01***

(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)(0.000

)

Nrivals(.)

−0.0003*

*−0

.000

3**

−0.0003*

*−0

.0003*

*−0

.0003*

*−0

.0003*

*

(0.029

)(0.030

)(0.024

)(0.024

)(0.026

)(0.025

)

Now

n(.)i

0.00

04*

0.0004

*0.0004

*0.0004

*0.0004

*0.00

04*

(0.070

)(0.072

)(0.060

)(0.062

)(0.066

)(0.062

)

Sargantest

15.840

15.868

15.386

15.388

15.513

15.400

(0.393

)(0.391

)(0.424

)(0.424

)(0.415

)(0.423

)

Stand

arderrorin

brackets(*

10%,**

5%,**

*1%

)


minutes. Also, we can see how the negative competitive effect of the rival decreases;i.e. the further away the rivals are then the less competition there is, and the effect onthe price the petrol stations set is lower. The variable for the effect caused by thesame-brand stations is equally reduced. This possible market domination effect isdiluted as we widen the market; the further away the same-brand petrol stations are,then the lower is the possibility of setting higher prices.

As shown in the table above, the Sargan test does not reject in any case thatthe first stage of the estimation is overidentifying. This fact indicates that theinstruments are valid and that therefore the coefficients obtained in the secondstage are not biased. Regarding the strength of the instruments, Staiger andStock (1997) recommend using the F test in the first stage to establish the extent towhich the instruments are correlated with endogenous variables. In our case, all theestimates in the F test are rejected at 1% in the first stage and all the coefficients are 0,obtaining a partial R2, one that only takes into account the instruments in all casessuperior to 0.20. Therefore, it seems that our instruments are valid and strong enoughnot to bias our estimation.

It is worth noting two aspects of the previous estimation: the possibility that thereis multicollinearity between the two regressors, and the adjustment of the linearmodel. With respect to multicollinearity, the correlation between the number of rivalsand the number of own petrol stations is not high, with a maximum of 30%. Thisfact indicates that the problem of multicollinearity, if any, is not important.Regarding the adjustment of the linear model, estimating a semilogarithmic modeldoes not alter the results significantly.

The above results indicate that competition is primarily local and is diluted, asthere is a greater distance between a petrol station and its rivals. This result fits withthe findings of Barron et al. (2004) in the U.S. gasoline retail market, where a greaternumber of competitors generated a smaller 1.5 mile equilibrium price in the market,or by Spiller and Huang (1986) and Pinske et al. (2002), where traders watched aswholesale gasoline in the United States competed with rivals mainly located closergeographically.

We must also note that the relationship is statistically significant in smallermarkets (4–6 min), and wider markets (18 min or more). One possible interpretationof these results is the existence of two different groups of consumers, who consumepetrol in the municipality where they reside and consumers who travel long distanceregularly.

After estimating the effect of the number of rivals and the same-brand petrolstations on the prices that service stations set, we continue by studying whatpercentage of the distribution margin these effects cause (Table 8).

In Table 8 we can see the effect that the monopolization of the differentmarkets has on the gross distribution margin. Monopolization of the market byunderstanding the effect would occur if a particular service station acquirescontrol of all the petrol stations that are competing in the market and movingthem to operate under its brand. For example, suppose a service station ofRepsol has four petrol stations of others brands less than 10 min away. Themonopolization of the market would lead to the acquisition of these fourstations by Repsol, so that they would operate under the same brand. Thedistribution margin has been calculated as the final price less taxes and the


Tab

le8

Price

increase

ofahy

pothetical

mon

opolistusingdaily

averageintensity

astheexplicativefactor

ofthenu

mberof

petrol

stations

Minute

(1)Rival’spriceeffect

onmarket(€)

(2)Ownpriceeffect

onmarket(€)

(1)+(2)=(3)Sum

ofeffects

(4)Average

number

ofrivals

(3)*(4)=(5)Average

effect

ofmarket

mon

opolization

(6)Average

who

lesale

marginof

distribu

tion

(com

petitivemarkets)

(5)/(6)=(7)Price

effect

ofshiftin

gfrom

competition

tomonopoly(com

petitive

markets,%)

1minute

0.0043

603

0.00

3895

50.00

8255

80.03

0.00

020.07

3139

30.32

2minutes

0.002242

50.00

3431

90.00

5674

40.12

0.00

070.07

1214

10.94

3minutes

0.001231

70.00

3648

80.00

4880

50.29

0.00

140.07

2025

61.99

4minutes

0.001341

70.00

3220

70.00

4562

40.45

0.00

210.07

2391

2.86

5minutes

0.001488

10.00

2856

30.00

4344

40.69

0.00

300.07

2497

94.12

6minutes

0.001248

60.00

2378

40.00

3627

0.91

0.00

330.07

2413

64.57

7minutes

0.000854

40.00

1599

50.00

2453

91.13

0.00

280.07

2302

43.83

8minutes

0.000683

70.00

1128

60.00

1812

31.40

0.00

250.07

2127

53.53

9minutes

0.000638

70.00

1061

90.00

1700

61.67

0.00

280.07

2450

83.91

10minutes

0.0005

822

0.00

0876

50.00

1458

71.97

0.00

290.07

2653

3.96

11minutes

0.0004

803

0.00

0680

10.00

1160

42.25

0.00

260.07

2688

93.59

12minutes

0.0004

692

0.00

0649

40.00

1118

62.57

0.00

290.07

2689

73.95

13minutes

0.0004

349

0.00

0612

90.00

1047

82.91

0.00

300.07

2633

74.20

14minutes

0.0003

587

0.00

0481

50.00

0840

23.29

0.00

280.07

2726

63.80

15minutes

0.0003

676

0.00

0495

50.00

0863

13.65

0.00

310.07

2772

34.33

16minutes

0.0003

712

0.00

0503

40.00

0874

64.04

0.00

350.07

2817

94.86

17minutes

0.0003

642

0.00

0499

80.00

0864

4.36

0.00

380.07

2786

95.18

18minutes

0.0003

666

0.00

0496

30.00

0862

94.75

0.00

410.07

2739

55.63

19minutes

0.0003

380.00

0446

60.00

0784

65.14

0.00

400.07

2740

35.54

20minutes

0.0003

161

0.00

0418

0.00

0734

15.50

0.00

400.07

2817

55.55

21minutes

0.0003

119

0.00

0420

30.00

0732

25.88

0.00

430.07

2851

95.91

22minutes

0.0002

907

0.00

0390

20.00

0680

96.22

0.00

420.07

2909

95.81


Tab

le8

(con

tinued)

Minute

(1)Rival’spriceeffect

onmarket(€)

(2)Ownpriceeffect

onmarket(€)

(1)+(2)=(3)Sum

ofeffects

(4)Average

number

ofrivals

(3)*(4)=(5)Average

effect

ofmarket

mon

opolization

(6)Average

who

lesale

marginof

distribu

tion

(com

petitivemarkets)

(5)/(6)=(7)Price

effect

ofshiftin

gfrom

competition

tomonopoly(com

petitive

markets,%)

23minutes

0.0002

839

0.00

0381

50.00

0665

46.63

0.00

440.07

2920

36.05

24minutes

0.0002

718

0.00

0355

60.00

0627

47.03

0.00

440.07

2998

46.04

25minutes

0.0002

633

0.00

0344

90.00

0608

27.03

0.00

430.07

2988

75.86

26minutes

0.0002

532

0.00

0326

0.00

0579

27.72

0.00

450.07

2999

56.13

27minutes

0.0002

503

0.00

0326

0.00

0576

38.13

0.00

470.07

3050

16.42

28minutes

0.0002

395

0.00

0314

40.00

0553

98.56

0.00

470.07

3051

96.49

29minutes

0.0002

310.00

0303

70.00

0534

79.04

0.00

480.07

3059

96.61

30minutes

0.0002

260.00

0303

0.00

0529

9.52

0.00

500.07

30511

6.90

Ownelaboration


international wholesale price.10 The taxes may include special taxes, retail taxes,state and regional taxes and VAT. The average retail margin for petrol stations withcompetitors in different markets is found in column 6. The average margin is calculatedas the average price before tax set by the gas stations that have a rival in each of themarkets, less the spot wholesale price set in the Rotterdam market. Because the numberof petrol stations with rivals in the different markets varies the average retail margin alsovaries.

The effect of monopolization is the sum of the effects of the rivals plus the effectof the own-brand petrol stations, multiplied by the average number of rival petrolstations in the market. Column 1 presents the effect generated by rivals, whilecolumn 2 shows the positive effect on prices generated by the coordination of theown-brand petrol stations in the market. The effect of monopolization is the sum ofboth the price increase that occurs when rivals are eliminated and the increase inprices that occurs when the coordination between stations of the same brandincreases. Once multiplied by the average number of stations competing in themarket (column 4), we obtain the total effect of monopolization of the market(column 5). The effect on the gross margin percentage (column 6) is presented incolumn 7.

The results of Table 8 show how in the monopoly as a broader market morecompanies can increase their prices. A 5 min monopolistic market leads to an increaseof a 4.12% margin; nevertheless, if it extends its dominance to monopolize the marketof 20 min, it would increase by a 5.55% margin. This positive relationship can be seenin Fig. 1.

The previous figure shows that this total effect on the distribution margin isbetween 0.32% and 6.90%. Upon being monopolized, that market would expect anaverage increase of at least 5%, in the margin when there is 17 min travel time. Thismargin is defined in the SSNIP test (Small but Significant Non-transitory Increase inPrices test). It has been used to define where, how and why petrol stations compete,and assumes a standard threshold used by competition authorities.

If we define the market as 17 min travelling distance from each retailer, then thereare 455 petrol stations with rivals in the market and 149 petrol stations that have norivals. This result may also be important for mergers occurring in the market. Thosemerger processes increase the number of service stations without rivals in 17 min ofthe posting, causing increases in margins of more than 5% in these affected stations.The competition authorities should take this into account.

Of these 149 service stations that have no rivals in the market, 99 of them have nosame-brand petrol stations nearby, and this could indicate that those markets may benatural monopolies. The demand in the existing market may not allow for thepresence of more than one operator, and thus there is no possibility of competition. Itshould also be stated that there might be entry barriers that prevent new operatorsfrom entering those markets, thus provoking a monopoly situation. Barriers that maylimit competition within those markets are the long time period needed for opening aservice station, the cost of land and municipal regulations that on occasion impedethe opening of such businesses. Eliminating these barriers could encourage newagents to enter and lower the market equilibrium price.

10 For a more detailed explanation of the components that make up the price of petrol, see Miras (2007).


The other 50 stations that do not have market competition still have one or moreservice stations in the same market. Market demand permits the existence of more thanone retailer, and if it is not part of the same company the price equilibrium is lower.Liberalizing and facilitating brand changes at the points of sale would improve thepresence of other companies within the markets. As in the previous case, the entrance ofnew operators would facilitate the existence of new brands in these monopolizedmarkets, and would lower the market equilibrium price. Measures for eliminating theentry barriers could play a role in this equally important case.

5 Conclusions

This paper has concentrated on demonstrating the direct relationship between accessibilityand the level of competition in a specific market. We have compiled a database for thepetrol retail sector in the Spanish autonomous region of Galicia, which possesses bothurban and rural areas. Among other factors this database contains price information,locations, brands, types of road and volume of traffic. Its fundamental objective is todetermine what affects the probability of finding a cheaper petrol station within a certaintime period, while controlling the characteristics of the road and the station’s location.

By using descriptive and econometric methodology, we can conclude thataccessibility has a main effect upon final prices: less accessibility fosters marketpower, because with less demand the price equilibrium may be less. But a demandeffect is that, consumers from rural areas, or areas with low traffic intensity, haveless chance of reaching lower-priced service stations within a 20-min drive. Furtherconclusions are that, third, if we include the value of time, no rational consumerwould deviate from the nearest service station on route to acquire petrol at anotherwith lower prices. Fourth, using the Agencies of Competition Policy’s methodology

Source: Own elaboration.Fig. 1 Monopolization effect as a percentage of the wholesale margin


(the SSNIP test), service stations can establish a dominant position, if there are nopetrol stations of other brands in a 17-min radius.

Contrary to the results by van Wee et al. (2001), city centres are not necessarilymore attractive to companies. The reduced accessibility increases the probability ofcompanies abusing their dominant position in the local geographical market. Thekey to pricing policy is not to be situated either in a periphery or in a city centre,unless access to the rival points of sale is low. The benefits will, nevertheless,depend on the level of demand at the point of sale and the price levels it can set and/or the accessibility of other rival petrol stations.

As a final recommendation for economic policy, we suggest that, when planningthe services that stations could offer, accessibility to establishments selling otherbrands should be borne in mind, as there is a significant effect on the priceequilibrium and consequently on social welfare.

Acknowledgements This paper has benefited from helpful comments and suggestions by Juan CarlosMartín, Javier Campos and two anonymous referees. We are also grateful for the database assistance by HéctorRodríguez and Adrià Botey. This research has received financial help from the SpanishMinistry of Science andTechnology (ECO 2009-06946/ECON). A previous version of this paper has been published asWorking Paperno. 456 in the Fundación de las Cajas de Ahorros (FUNCAS) collection. The usual disclaimer applies.

Annex I

Definition of variables (from Catalist)

Primary traffic

This is an estimate of the 24-h average two-way traffic flow on the primary street tothe nearest thousand. Guidelines for the various definitions used are:

Poor: Traffic levels are less than 5,000 vehicles per day.Medium: Traffic levels are between 5,000 and 15,000 vehicles per day.Good: Traffic levels are between 15,000 and 25,000 vehicles per day.Very good: Traffic levels are in excess of 25,000 vehicles per day.

Site location

Rural: Countryside background or low-density residential and industrial use, as wellas locations on long-distance commuter routes that experience consistent trafficflows, e.g. a site on a quiet stretch of trunk road.

Industry/office: Low residential back-up and much evidence of commercial unitse.g. industrial/office/retail. Applicable to business infrastructures such as port areas,manufacturing, distribution centres, shopping centres etc.

Residential: Tends to be located away from the commercial and industrial areasand surrounded by much private housing.

Urban transient: Characterized by high traffic volumes spread evenly throughoutthe day. Bypasses and ring roads are included in this class.

Motorway.


Annex II

Note: White points are petrol stations with high probability (more than 0.5). Lowgrey are those petrol stations with medium probability. Medium grey are those withlow probability. The black ones are petrol stations with a probability equal to 0.

References

Atkinson B, Eckert A, West DS (2009) Price matching and the domino effect in a retail gasoline market.Econ Inq 47(3):568–588

Baradaran S, Ramjerdi F (2001) Performance of accessibility measures in Europe. J Transp Stat 4(2/3):31–48Barron JM, Taylor BA, Umbeck JR (2004) Number of sellers, average prices, and price dispersion. Int J

Ind Organ 22:1041–1066Berry S (1994) Estimating discrete-choice models of product differentiation. Rand J Econ 25

(2):242–262Berry S, Levinshon J, Pakes A (1995) Automobile prices in market equilibrium. Econometrica 63(4):841–890

Map I Probabilities of obtaining petrol stations with lower prices


Borenstein S (1991) Selling costs and switching costs: explaining retail gasoline margins. Rand J Econ 22(3):354–369

Borrell JR, Perdiguero J (2007) La Competència en la distribució de gasolina a Catalunya. Tribunal Catalàde Defensa de la Competència (TCDC), Barcelona

Bromiley P, Papenhausen C, Borchert P (2002) Why do gas prices vary, or towards understanding themicro-structure of competition. Manage Decis Econ 23:171–186

Caixa L (2007) Anuario Económico de España 2007. Servicio de Estudios de La Caixa, BarcelonaCampos J, de Rus G (2002) Dotación de infraestructuras y política europea de transporte. Papeles de

Economía Española 91:169–181Comisión Nacional de Competencia (2009) Informe sobre la competencia en el sector de carburantes de

automoción. Comisión Nacional de Competencia, MadridEl-Geneidy AM, Levinson DM (2006) Access to destinations: development of accessibility measures,

Report 2006-16, Minnesota Department of TransportationFranklin J, Waddell P (2003) A hedonic regression of home prices in King County, Washington using

activity-specific accessibility measures. Paper presented at the Transportation Research Board 82ndAnnual Meeting, Washington DC

Handy SL (2002) Accessibility—vs mobility—enhancing strategies for addressing automobile dependence inthe US. Paper presented at the Transportation Research Board 82nd Annual Meeting, Washington DC

Handy SL, Niemeier DA (1997) Measuring accessibility: an exploration of issues and alternatives.Environ Plann A 29(7):1175–1194

HEATCO, Developing Harmonised European Approaches for Transport Costing and Project Assessment(2006): Deliverable 5: Proposal for Harmonised Guidelines. European Commission

Hotelling H (1929) Stability in competition. Econ J 39:41–57Ivaldi M, Verboven F (2005) Quantifying the effects from horizontal mergers in European competition

policy. Int J Ind Organ 23(9–10):669–691Kaufman PR, MacDonald JM, Lutz SM, Smallwood DM (1997) Do the poor pay more for food? Item

selection and price differences affect low-income household food costs. Agric Econ Rep Number 759.Economic Research Service, United States Department of Agriculture

MacDonald J, Nelson PE (1991) Do the poor still pay more? Food price variations in large metropolitanareas. J Urban Econ 30:344–359

Mazzeo MJ (2002) Competitive outcomes in product-differentiated oligopoly. Rev Econ Stat84:716–728

Miras P (2007) Los mercados de productos petrolíferos: Una panorámica. Econ Ind 365:69–78Pakes A, Berry S, Levinshon J (1993) Applications and limitations of some recent advances in empirical industrial

organization: price indexes and the analysis of environmental change. Am Econ Rev 83(2):241–246Perdiguero J, Borrell JR (2007) La difícil conducción de la competencia por el sector de las gasolinas en

España. Econ Ind 365:113–125Perdiguero J, Jiménez JL (2009) ¿Competencia o colusión en el mercado de gasolina?: una aproximación

a través del parámetro de conducta. Rev de Econ Apl XVII(50):27–45Pinske J, Slade ME, Brett C (2002) Spatial price competition: a semiparametric approach. Econometrica

70(3):1111–1153Ritsema van Eck JR, de Jong T (1999) Accessibility analysis and spatial competition effects in the context

of GIS-supported service location planning. Comput Environ Urban Syst 23:75–89Salop S (1979) Monopolistic competition with outside goods. Bell J Econ 10:141–156Spiller PT, Huang CF (1986) On the extent of the market: wholesale gasoline in the northeastern United

States. J Ind Econ XXXV:131–145Srour I, Kockelman K, Dunn T (2002) Accessibility indices: connection to residential land prices and

location choices. Transp Res Rec 1805:25–34Staiger D, Stock JH (1997) Instrumental variables regression with weak instruments. Econometrica 65

(3):557–586Stewart H, Davis D (2005) Price dispersion and accessibility: a case study of fast food. South Econ J 4

(71):784–799Stigler G (1961) The economics of information. J Polit Econ 69:213–225van Wee B, Hagoort M, Annema JA (2001) Accessibility measures with competition. J Transp Geogr

9:199–208Vickerman RW (1974) Accessibility, attraction and potential: a review of some concepts and their use in

determining mobility. Environ Plann A 6:675–691Wachs M, Kumagai T (1973) Physical accessibility as a social indicator. Socio-econ Plann Sci

7:327–456


Does Accessibility Affect Retail Prices and Competition? An Empirical Application

Documents

Transcript of Does Accessibility Affect Retail Prices and Competition? An Empirical Application