Consumer Search and Automobile DealerCo-Location

Charles Murry and Yiyi Zhou∗

June, 2016

Abstract

Retailers co-locate with rivals in order to take advantage of economiesof agglomeration when consumers have limited information and engagein costly search, even though co-location implies fiercer price competi-tion. We estimate a structural model of consumer search for spatiallydifferentiated products using rich data on new car transactions. We usethe model to separately disentangle the competition and agglomerationeffects of retail co-location by simulating retail closures. A full infor-mation model that ignores the agglomeration effect would overstate thegains to incumbent rivals and the welfare loss to consumers due to cardealer closures.

Keywords: retail agglomeration, spatial competition, car dealers, retail exitJEL Classification: D83, L13, L62

∗Charles Murry: Department of Economics, Pennsylvania State University, Uni-versity Park, PA, 16802, cmurry@psu.edu. Yiyi Zhou: Department of Economicsand College of Business, Stony Brook University, Stony Brook, NY 11794-4384, USA,yiyi.zhou@stonybrook.edu. We thank Simon Anderson, Paul Grieco, Peter Newberry, RegisRenault, Henry Schneider, Steven Stern, Matthijs Wildenbeest, Mo Xiao, and participantsat IIOC 2014 in Chicago for useful discussions and comments. The authors are solely re-sponsible for all errors.

1

1 Introduction

Economists have long sought to understand the location decisions of firmsand these decisions’ effect on industry profits and consumer welfare. The co-location of firms is especially ubiquitous in many retail industries, and therehas been special attention paid as to why some firms tend to locate near eachother even though this would typically imply greater price competition.1 Aclassic explanation for the co-location of retail stores has to do with limitedconsumer information, see Stahl (1982) and Wolinsky (1983). The basic ideais that if consumers must engage in costly search in order to resolve infor-mational problems before purchase, then consumers are more likely to searchareas where there is a concentration of stores in order to limit search costs.This agglomeration effect encourages co-location of stores. However, if storesare close to each other then price competition may be fierce, potentially out-weighing benefits to stores from co-location.2

Understanding the agglomeration and competition effects of co-location hasimportant implications for evaluating the consequences of retail closures. Onthe one hand, the agglomeration effect implies that a nearby rival’s exit wouldreduce the total attraction of that geographic area and force the incumbentfirms to lower their prices in order to continue attracting searching consumers.This would decrease the surplus of incumbent firms and potentially increaseconsumer welfare. On the other hand, the competition effect implies that anearby rival’s exit would increase the market power of incumbent firms andlead to higher prices. This would increase the surplus of incumbent firms

1For example, Hotelling (1929) studied the location decisions of firms selling to ge-ographically disperse consumers and how these decisions influence consumer substitutionacross products and geography in an attempt to explain the co-location of firms. However,d’Aspremont et al. (1979) showed that Hotelling’s principle of minimum differentiation wasinvalid and suggested that firms would want to locate far from each other using a variationof the same model. However, the optimal location decisions of firms are very sensitive tochanges in the set up of Hotelling’s model.

2There is a large literature on agglomeration economies that focuses on productiondriven reasons for co-location of manufacturing, see Rosenthal and Strange (2004) for anoverview of empirical evidence from the urban economics literature. We focus on demanddriven reasons for co-location because of our focus on retailing, not manufacturing.

2

and reduce consumer welfare. The agglomerative effects of retail closures areparticularly salient given the recent U.S. financial crisis of 2007-2009 that sawmany retail firm exits due to bankruptcy and other reasons.

In this paper we have two main goals. The first goal is to quantify theagglomeration and competition effects of retail co-location and to evaluatehow much of these effects are related to limited consumer information. Thesecond goal is to evaluate the welfare effects of retail closures in light of retailagglomeration economies. To accomplish these two goals, we estimate a modelof consumer search for spatially differentiated products in the new car retailindustry and price competition among new car dealers. This is an ideal settingto examine issues of retail co-location for two reasons. First, retail co-locationis ubiquitous in this industry. For example, in Virginia (the geographic regionof the data we use in our formal analysis) about 90% of car dealers are locatedwithin one-half mile of a competitor (see Figure 1). Second, this industry hasbeen the setting of massive retail closures over the past half century, whichwas exacerbated by the recent financial crisis. We estimate the model usingdetailed transactions level data on new car purchases, which includes the pop-ulation of new car transactions, including the transaction price and distancefrom the consumers home to the retail transaction location, a key determinantof search costs.

The model we present is a parametric version of the optimal portfolio choiceproblem described in Chade and Smith (2006), very similar to the specificationdeveloped in Anderson et al. (1992) (Chapter 7) and recently extended to em-pirical applications by De los Santos et al. (2012) and Moraga-González et al.(2015).3 In the model, we split the overall market into separate geographicareas, with each area representing a cluster of multiple car dealers. We assumeconsumers pay a search cost to visit a dealer cluster, and this cost is a functionof the distance between the consumers’ home and the cluster. After they paythe cost, consumers are able to inspect all products within a dealer cluster at

3Moraga-González et al. (2015) also study the new car market. However, they do notstudy the effects of dealer co-location. Instead, their focus is on the effects of consumersearch on prices and market power, and how full information models might lead to biasedpredictions.

3

no additional cost. Consumers simultaneously decide the set of areas they willsearch, and conditional on that set, they choose to purchase the best option.As in Stahl (1982) and Wolinsky (1983), the model implies that co-locationhas two effects, a price competition effect and an agglomeration effect of co-location. To validate the use of the search model, we first present empiricalevidence that consumer demand is influenced by clusters of co-located dealersby capturing the effects of co-location in a simpler demand framework.

Figure 1: Distance to the Nearest New Car Dealer

Distance to the nearest Dealer (in miles)0 2 4 6 8 10

Num

ber

of

De

ale

rs

0

50

100

150

200

250

300

350

400

450

500

Note: This figure plots the distribution of the distance of new car dealers’ closest rival. Datafrom Virginia Department of Motor Vehicles and includes all new car dealers in the state ofVirginia in 2008.

To estimate the model, we use detailed car transactions data that includeall new car transactions from all dealers in a single large market, the priceof each transaction, and the distance between the dealer and the consumer’shome. Unlike other studies of retail agglomeration, the detailed spatial natureof our data allow us to accurately capture spatial demand substitution patternswhich underline the effects of co-location. We estimate a substantially smallerdollar per mile of disutility from traveling to purchase a car than previous

4

studies for this industry, $56 per mile. However, we still find that consumerinformation, and therefore consumer search, is limited. For example, the modelpredicts that nearly all consumers search less than four geographic areas whenpurchasing a new car, and the median consumer searches just one geographicarea. These results are in line with survey data from industry reports ofnew car buying habits. We quantify the importance of search frictions bysimulating what equilibrium prices would be if consumer’s search cost werezero. Our simulation results predict that the average retail price would be $422lower, which in turn suggests that dealers use consumers’ limited informationto exercise market power.

We conduct counterfactual exercises that simulate the closings of incum-bent car dealers. Specifically, we close a single dealer and then re-calculateequilibrium prices and consumer demand. We then do this for every dealer,one at a time, for both our search model and a standard model that assumesfull information. We find that for both our search model, and a full informationmodel, dealer closure results in a decrease in consumer surplus because pricesrise and consumers have fewer options. We also find that the total surplusof unclosed dealers increases after dealer closure for both models. However,consumer surplus falls by less in the search model compared to the full infor-mation model and the total surplus of unclosed dealers increases by less for thesearch model compared to the full information model. The main reason is thatthe search model implies a smaller price increase because incumbent dealersin the same geographic area have an extra incentive to keep prices low, thatis to continue to attract consumers to their geographic area. Additionally, ifthe closed dealer was not in a particular consumer’s search/choice set in thefirst place, closure will have no direct effect on that consumer’s surplus.

Our analysis of dealer closures is related to recent literature. Furthermore,our dealer closure counterfactuals are particularly relevant to understandingthe effects of massive dealer closures sparked by the financial instability of UScar manufacturers over the past decade. Both Benmelech et al. (2014) and Oz-turk et al. (forthcoming) study how retail agglomeration effects retail closures,and both papers find evidence of positive agglomeration effects. Benmelech

5

et al. (2014) use data across retail industries to estimate the effect of closuresdue to chain level financial problems on the closure decisions of close-by in-cumbent retail outlets. They find that nearby retail outlets are more likelyto close after rival’s closure. Ozturk et al. (forthcoming) look at the effectof Chrysler dealer closings on the prices of nearby dealers using a nationalsample of new car transactions in a differences-in-differences framework. Theyfind that, although prices go up after a closure, the effect of closures on pricesmoderates with distance. This implies that a co-location agglomeration effectexists, but that it is dominated by a competition effect, a similar result toours.

Most of prior work on retail co-location has focused on inferring the effectof agglomeration economies through firm entry and location decisions. Someof these studies have found closing rivals would have a net negative effect, forexample Seim (2006), Jia (2008), and Zhu and Singh (2009). On the otherhand, Vitorino (2012) finds evidence of an agglomeration effect of co-locationdominates in a shopping mall setting, and Ellickson et al. (2013) find thatagglomeration effect is a function of local market size in the big-box retailindustry. We distinctly depart from this literature by estimating a structuralmodel of consumer search for spatially differentiated products. By modelingthe explicit mechanism of the agglomeration benefit (i.e. how co-location af-fects consumer demand), we can separately quantify the effects of competitionversus agglomeration on firm and consumer behavior. Furthermore, we use theestimated model to evaluate the welfare effects of retail closures, somethingthat is not possible with the optimal entry models in the papers mentioned.In contrast, we must assume that locations are exogenous to demand shocksin the consumer utility function. Our justification, which we explain in moredetail later, is that there are many regulations governing the entry and exit ofdealers.

We also contribute to the growing literature on consumer information, suchas Sovinsky Goeree (2008), Hortaçsu and Syverson (2004), and Hong andShum (2006) among others. Like in those studies, we find evidence that lim-ited consumer information can bias demand results and counterfactuals in full

6

information models. There are many theoretical studies that recognize thatlimited consumer information and search leads to agglomeration benefits ofco-location, such as Stahl (1982), Wolinsky (1983), Wolinsky (1986), Dudey(1990), Fischer and Harrington Jr (1996), among others. However, this ideahas not been explored empirically using a consumer search model that cap-tures the demand mechanisms from the theory literature. As such, our papercontributes to more recent literature on the structural estimation of consumersearch by explicitly studying the agglomerative benefit of search for firms. Inparticular, there are numerous recent studies that also nest a simultaneoussearch framework in a differentiated products demand framework, for exampleWildenbeest (2011), De los Santos et al. (2012), Seiler (2013), Honka (2014),and Moraga-González et al. (2015).

2 Data and Overview of the Market

In this section, we first present a detailed description on the data used in theempirical analysis. Second, we present a set of descriptive statistics document-ing consumer travel distance to purchase new cars and the spatial distributionof new car dealers.

2.1 Data

We combine several data sets for our analysis. The first data source providesdetailed records of all new vehicle transactions in Richmond, Virginia for fouryears. The second data source provides general information on characteristicsand prices of all vehicles sold during this period, and the third data sourceprovides information on all dealerships. We also use data from the Census forconsumer location and demographic characteristics.

The primary data are obtained from the Virginia Department of MotorVehicles, henceforth DMV, and consist of all new vehicle transactions initiallyregistered in Virginia from 2007 through 2010. For each transaction we knowthe make, model, and transaction price of the car. We also know the iden-

7

tification number assigned to the dealer by the DMV and the name of thedealer. Finally, the data include either the nine-digit or 5-digit zip code of thepurchaser for each transaction.

Table 1: Descriptive Statistics, Car-Dealer Levelvariable Mean SD Min Median Max

Avg. Price 24,987.25 8,490.18 11,102.84 22,674.95 76,866.16HP/Weight*1000 0.55 0.09 0.30 0.53 1.36MPG Highway 2.82 0.46 1.50 2.80 3.70Passengers 5.20 0.84 2.00 5.00 10.00US Brand 0.26 0.44 0.00 0.00 1.00Luxury 0.28 0.45 0.00 0.00 1.00

Note: New car transactions in Richmond, Virginia from our selected sample from the VADMV described in the text.

We make a number of sample selection decisions for the raw data in orderto focus on the market for new retail cars. We remove all commercial vehicles,motorcycles, trailers, and consumer pickup trucks.4 We also dismiss observa-tions with prices near or at zero, which likely represent something else besidesa typical consumer transaction, for example fleet sales, or some error in thedata recording process.

We also have general information on car characteristics and pricing fromintellichoice.com. This includes characteristics of each trim level of each modelof car, invoice prices, manufacturer suggested retail price, and other fees as-sessed at the time of sale. Intellichoice.com also furnished us with a list of allcustomer incentives provided by manufacturers during the time period, whichis crucial to constructing a correct transaction price for a car, as dealers inVirginia report the transaction price less manufacturer rebates to the DMVfor tax purposes.

In order to focus on our research questions, we limit our study to trans-actions with buyers and sellers both located in the Richmond metropolitan

4It is common in the literature to consider pickup trucks a different market. Additionally,some models of pickup trucks have dozens of trim levels that vary widely in price andcharacteristics, making it problematic to aggregate to the model level.

8

area.5 In total, we are left with 79,097 transactions. Richmond is a relativelyisolated market. For example, among all transactions with buyer’s locationin Richmond, 93.3% of the associated dealers are located in Richmond area.In other words, Richmond habitants rarely purchase cars from outside thearea. Among all transactions with dealer’s location in Richmond, 82.5% ofthe buyers live in Richmond.

We merge the transaction data with a list of all active and recently closedautomobile dealers provided to us by the Virginia DMV. To merge the data, weaggregate transactions to the year-model level.6 Our data used in estimationcontain 56 dealerships selling 32 different brands and a total of 2,280 uniquedealer-model-year combinations. Among them, 22 dealers sell multiple brands,and 6 dealers sell vehicles produced by multiple manufacturers.

We complement the data with U.S. Census demographic data on the incomeand population at the level of “block groups”, which, on average, contain about1100 people. In total, the Richmond area contains 575 block groups.

We present descriptive statistics of the automobile data in Table 1. Theaverage (sales weighted) price of a car in our sample is about $25,000, 26% ofthe sales are US brand cars, and 28% are cars that are classified as luxury byintellichoice.com.

2.2 Consumer Travel Distance to Purchase

To calculate the distance of each transaction, we geo-code the consumer loca-tions and the dealer addresses into decimal longitude and latitude coordinates.We observe the zip code of each consumer, and using the zip code centroid pro-vided by the US Census, we assign zip codes to the nearest block group usingthe Census centroid for each block group. Figure 2 illustrates how far con-sumers travel to purchase a car. The mean travel distance is about 11.75 miles,the median is 9.74 miles, and the standard deviation is 8.21 miles. However, a

5We will refer to the greater Richmond metropolitan area as Richmond.6It would be ideal to use trim information, but unfortunately the transaction is not

always recorded in such detail. We assign the base 4-door trim characteristics to eachmodel.

9

more telling statistic that shows how important distance is in the consumer’schoice problem is the distance traveled past the nearest dealer. Figure 2 alsoindicates the distribution of the distances traveled past the nearest dealer.The mean of the extra distance traveled is 6.12 miles, and the median is 5.05miles. 27% of the buyers bought from the nearest dealer, 52% traveled lessthan 10 miles, 18% traveled less than 20 miles and only 3% traveled more than20 miles.

Figure 2: Distribution of Buyers’ Travel Distance

Distance0 5 10 15 20 25 30 35 40

Fre

qu

en

cy

0

0.1

0.2

0.3

0.4

0.5Purchase DistancePurchase Distance Past Closest Dealer

Note: Distribution of consumer purchase travel distance. Data from Virginia DMV trans-actions data. Sample selection described in the text.

2.3 Spatial Distribution of Car Dealers

Dealers in Richmond locate in clusters in a small number of geographic lo-cations, mainly along primary roads and commercial areas. We group the56 dealers into 9 geographic areas. Figure A.1 in the Appendix A shows thedistribution of car dealers in Richmond numerically coded by the geographicareas we assigned each dealer. The areas range in size from a few dealers toover a dozen. The most common type of area is a suburban commercial centercommon on the outskirts of most U.S. cities. Richmond has many relativelylargely populated suburban areas, whose growth is in part driven by the fact

10

that the city and surrounding area is served by three major interstates, I95,I64, and I295. The city of Richmond is about 62 square miles and has a pop-ulation of a little over 200,000 residents, but the Richmond metropolitan areais home to more than one million residents, and is the 43rd most populousmetropolitan area in the United States. We present a very brief description ofeach dealer area in Table 2. The two largest areas are two suburban centers,Short Pump (a suburb 15 miles west of downtown) and Midlothian (a suburb15 miles southwest of downtown). Each of these areas has a large mall andother commercial activity along a major thoroughfare surrounded by suburbanand exurban residential development.

Table 2: Description of Dealer Selling Areas

Area Dealers Description

1 6 Ashland: Suburb 20 miles north of downtown Richmond on I952 24 Short Pump: Suburb 15 miles west of downtown3 5 Mechanicsville : Historic town 10 miles northwest of downtown4 22 Midlothian: Large suburb, 15 miles west/southwest of downtown5 2 Downtown Richmond6 8 Airport: limited development; 10 miles east of downtown Richmond7 3 Woodlake: Suburb 20 miles SW of downtown8 6 Chester: Suburb 15 miles south of downtown9 5 Colonial Heights: Suburb and military base 25 miles south of downtown

Note: A dealer refers to a single franchise location of a brand. For instance, a singlelocation that sells Dodge and Jeep counts as two dealers. Milages are approximate milesusing Google Maps driving directions using the geographical centers of each area.

To further illustrate that co-location is a dominant feature of this industry,we display the total sales by dealer US Census Tract in Figure A.2 in AppendixA. In this figure, we shade each Census Tract in the Richmond metro area witha shade of red corresponding to how many car sales originate from dealers inthat Census Tract in 2008. Most Census Tracts are not shaded, and thereforenot visible on the map. However, the Tracts that are shaded clearly illustratethe geographic clusters of dealers that are present in Richmond. There areclear “pockets” of sales with no dealers in between each shaded pocket. This

11

corresponds to the geographically separated dealer clusters in the Richmondarea.

Table 3: Sales and Prices by Dealer Area, 2008Dealer US Brands Non-US Brands

Area Dealers Sales/Dealer Avg. Price Dealers Sales/Dealer Avg. Price

1 6 150 26,426 0 – –2 7 168 29,903 17 469 33,1343 1 103 24,576 4 526 24,0774 9 278 26,437 13 334 31,4235 2 344 25,037 0 – –6 7 161 26,160 1 374 20,1597 0 – – 3 956 22,0788 2 320 26,463 4 743 23,0159 5 130 25,820 0 – –

Note: New car transactions from Richmond, Virginia new car transactions from our selected sampledescribed in the text.

3 Demand Model

We consider a market where differentiated cars sold by many different geo-graphically dispersed dealers are sold to geographically dispersed consumers.We use subscript i to denote consumer, subscript j to denote car model (forexample Ford Fusion), subscript f to denote dealer (ie “Bob’s Honda Sales”),and subscript t to denote year. Consumer i makes a discrete choice, either topurchase a new car model from a dealer or to consume an outside option. Theindirect utility that consumer i derives from purchasing the car model j fromdealer f in year t is

uijft = xjftβ − αipijft + ξjft + εijft, (1)

where xjft is a vector of observed product-dealer attributes, pijft is the pricecharged by dealer f , ξjft captures the unobserved product-dealer-year at-tribute, εijft captures an idiosyncratic match value that can be ascertained

12

only upon visiting the dealer, such as the fit and personal comfortability ofthe car, or a personal image in the car, or the specific way dealer salespeoplesell particular cars. αi captures consumer heterogeneity in tastes for price. Weassume

αi = exp(α0 + α1hi + α2ςi), (2)

where hi is the log of household i’s yearly income and ςi follow a standardnormal distribution.

Note that in the auto industry, prices depend on consumer characteristicsand dealer characteristics. As common in the literature on search, we assumethat, even though consumers do not know the prices each specific dealer wouldcharge them, they know the average price charged by a specific dealer. Thistype of information is available on a plethora of car buying websites. Also,advertisements may communicate this information, along with informationabout dealer specific prices, for example the willingness of each dealer to giveprice discounts. We assume

pijft = pjft + ϑijft,

where pjft is the average price charged by dealer f and ϑijft is consumer i’sprice deviated from the average. Here, ϑijft captures consumer i’s and dealerf ’s bargaining power that can be ascertained upon visiting the dealer.

We assume that consumers must search to find out the exact utility theyderive from each car sold by each dealer. To be more specific, we assume thatbefore searching consumers know the product attributes xjft and ξjft, and theaverage price charged by each dealer pjft. However, consumers do not knowthe exact values of their own price deviation from the average ϑijft and theirown match value εijft. As common in the literature, we assume that thatconsumers know their distributions before search and costly search reveals theexact values to consumers. Let

εijft = −αiϑijft + εijft

13

to capture the information that consumer i does not know before search. Weassume εijft follow an EV Type I distribution with location parameter 0 andscale parameter 1.7

Let δjft denote the mean utility that is common to all consumers

δjft = xjftβ + ξjft, (3)

and µijft denote consumer heterogeneous utility from average price

µijft = −exp(α0 + α1hi + α2ςi)pjft. (4)

Then, we can write the utility specified in equation (1) as

uijft = δjft + µijft + εijft. (5)

Consumers have an outside option, including purchasing from a dealer outsideRichmond, or non-purchase, or purchase of a used car. We model the utilityfrom the outside choice as ui0t = εi0t.

3.1 Search Mechanism

Existing theoretical literature typically model consumer search strategies intwo ways. One strand of the literature assumes non-sequential, or simul-taneous, search strategy, where consumers sample a fixed number of sellersand choose to purchase from the most preferred seller among those they havesearched.8 The other strand of the literature assumes sequential search strat-egy, where after each search consumers choose to purchase from the lowestprice observed so far or to make an additional search. Both search strategieshave been adopted by empirical researchers. There are two studies we areaware of that compare both strategies in a retail goods setting. De los Santos

7Although we are specific about how individual prices are related to average pricesand uncertainty in the model, we do not explicitly model a bargaining protocol betweenconsumers and dealers. This would severely complicate the model, and it is likely orthogonalto the mechanisms of dealer co-location that is our primary focus in this paper.

8See Stigler (1961), Burdett and Judd (1983), and Janssen and Moraga-González (2004).

14

et al. (2012) who use detailed data on the browsing and purchasing behaviorof a large panel of consumers to empirically test to what extent consumers areindeed using sequential and non-sequential search strategies. They found thatin their setting, the non-sequential search strategy outperforms the sequentialsearch model. Honka and Chintagunta (2014) use variation in actual prices inconsumers’ observed considerations sets to conclude that simultaneous searchbetter matches their data on the demand for auto insurance. Because we donot observe consumers’ consideration sets in our data, we are unable to let thedata tell us which search strategy better represents our empirical setting. Inthis paper, therefore, we assume that consumers engage in costly simultaneoussearch in order to learn the exact match values. This would be consistent witha consumer learning about the locations and general offerings of all dealers inthe market in some pre-search stage, and then planning a shopping trip thatincludes all of the geographic areas that make it into the consumers search set.

In the model, consumers have the choice to search cars at one or more ofnine possible dealer areas. Each dealer area represents a geographic area wheredealers are clustered. Consumers pay a fixed cost to search each area, and oncethe cost is incurred they learn εijft for every car in the area. The choice setof a particular consumer is made up of only those cars from those areas thatshe has searched. We normalize the search cost of the outside good to zero,so that every consumer choice set includes the outside option. Consumerssimultaneously decide the set of areas they will search, and conditional onthat set, they choose the best option. The model is a parametric version ofthe optimal portfolio choice problem discussed in Chade and Smith (2006),very similar to the specifications of De los Santos et al. (2012) and Moraga-González et al. (2015).9

Let Fmt be the set of dealers located in area m, and Jft be the set ofproducts sold by dealer f . To obtain a closed-form expression for choice prob-

9However, we do not adopt the Marginal Improvement Algorithm (MIA) proposed byChade and Smith (2006) to find the optimal choice set of each consumer, because theconditions of MIA are not satisfied in our context. Instead, we follow De los Santos et al.(2012) and Moraga-González et al. (2015) by analytically computing optimal search setprobabilities.

15

abilities from the search model, we follow the literature in assuming that εijftand εi0t follow a standard type-I extreme value distribution i.i.d. across con-sumers, car models, dealers and time. Hence, consumer i’s expected gain fromvisiting a subset of areas S is

Uit(S) = Eε

[maxui0t, max

∀j∈Jft,f∈Fmt,m∈Suijft

]

= ln

1 +∑

j∈Jft,f∈Fmt,m∈S

exp(δjft + µijft)

.Let cim denote the search cost of visiting area m. We assume that the searchcost is linear in the distance from her home address to the dealer location.That is, cim = γdim, where dim is the distance from consumer i’s location to aparticular area m. We define the value of visiting a subset S as

Vit(S) = Uit(S)−∑m∈S

cim + ωiSt,

where ωiSt is an individual and choice set specific term that captures theunobserved search cost shocks, such as the traffic patterns from visiting aparticular set of areas. If consumer i does not search any area, we normalizeUit(Ø) = 0 and

∑m∈Øcim = 0.

To solve for the optimal search set of the consumer we obtain a closed-form

expression for E[maxS∈S

Vit(S)

], by assuming that ωiSt follows a standard type

I extreme value distribution. Then, we can get the probability that consumeri visits a subset S∗:

PiS∗t = Pr(Vit(S∗) ≥ Vit(S) for ∀S)

=exp[Uit(S

*)−∑

m∈S* cim]∑S∈S exp[Uit(S)−

∑m∈S cim]

,

where S is the set of all possible search sets.

16

3.2 Purchase Decision

The probability that consumer i purchases product j from dealer f conditionalon a search set S follows the familiar analytical expression:

Pijft|S =exp(δjft + µijft)

1 +∑

j′∈Jft,f ′ ∈Fmt,m∈S exp(δj′f ′t + µij′f ′t), ∀mf ∈ S

= 0, ∀mf /∈ S,

where mf is the area where dealer f is located. Then, the unconditionalprobability that consumer i purchases product j from dealer f is

Pijft =∑S∈S

Pijft|SPiSt. (6)

Let consumer locations (in our application US Census block-groups) be in-dexed l = 1..L, and let Fl(·) be the area-specific distribution of consumerincome hi. Additionally let nlt be the number of potential consumers in areal. The predicted demand for each vehicle model from each dealer is obtainedby aggregating individual choice probabilities over all areas:

qjft(δt, pt; θ) =∑l

nlt

ˆPijft(δt, pt, hi; θ)dFl(hi), (7)

where θ = (α0, α1, α2, γ) represents all “non-linear” parameters of the model.

3.3 Dealer Agglomeration and Consumer Demand

The effects of dealer co-location are straightforward in the model. Dealer areaswith more dealers will, all else equal, have a greater value of Uit(S). The moreproducts in a search set, the greater the chance that a high utility product isfound. This comes through increased variation in the observed characteristicsof cars as choice sets increase, along with the fact that more draws of idiosyn-cratic shock, εijft, increase the maximum order statistic.10 In turn, the higher

10This is a well known property of variants of the logit discrete choice model and has theflavor of “love of variety” in representative consumer models. See Anderson et al. (1992) for

17

the value of Uit(S) the more likely the consumer will choose to search thatdealer area. Variation in chosen search sets across consumers is generated bydifferent draws of search set specific idiosyncratic cost shocks, ωiSt, and differ-ences in distances to different dealer areas. Holding travel distances constant,dealer areas with more products offered will be searched with greater prob-ability. The size of search costs will ultimately determine how many dealerareas are searched by each consumer. However, as is pointed out by Chadeand Smith (2006), the optimal set of dealer areas for each consumer will notnecessarily follow a cut off rule of an ordering of Uit(S)’s from highest to lowest.

4 Estimation Methodology

Our estimation follows recent literature that has combined the methods ofBLP with micro level data, for example Berry et al. (2004), Petrin (2002),and Sovinsky Goeree (2008), in a General Method of Simulated Momentsframework. The method employs the nested fixed point structure of BLP thathas an outer and inner loop, but uses both the aggregate moments suggestedby BLP as well as moments derived from the individual nature of the data.In the outer loop, we search for all non-linear parameters, θ = (α0, α1, α2, γ),that minimize a GMM objective function. For any guess of θ we solve for δt(θ)from the contraction mapping suggested in BLP in the inner loop. During theinner loop we recover those fixed effects in the mean utility equation (3) usinglinear regression.

For any candidate value of θ, we calculate the aggregate market share ofeach product from each dealer:

sjf(δt, pt; θ) = qjf (δt, pt)/Mt, (8)

where Mt is the total number of potential consumers Mt =∑

l nlt.Calculating sjf involves computing the choice probabilities which involves

a multi-dimensional integral which cannot be computed analytically. We use

details of welfare in discrete choice demand models.

18

simulation methods to approximate the integral, using the empirical distribu-tion of individual characteristics by Census block-group for the case of con-sumer demographics. We simulate ten consumers from each of the 574 Censusblock-groups. We then weight each individual based on their block-group pop-ulation. Let sjft be the observed market share of product j sold by dealer fin year t. The solution to sjf (δt, pt; θ) = sjft exists and is unique (see Moraga-González et al. (2015)). The solution is denoted by δt(θ). We are able to usethe contraction mapping suggested by BLP to solve for δt(θ).

After inverting demand using the contraction mapping, we solve for themodel-dealer-year specific demand unobservables as

ξjft(θ) = δt(θ)− xjftβ. (9)

To control for the correlation of price pjft with the unobserved product at-tribute ξjft, we use the competing product characteristics as the instrumentalvariables. Hence, the first set of empirical moment conditions are given by

G(1)N (θ) =

1

N1

∑j,f,t

ξjft(θ)Z(1)jft, (10)

where N1 is the number of model-dealer-year-level observations and Z(1)jft is a

set of instrument variables. Notice that because of the car style, location, andyear fixed effects, the moment conditions are expressed over the transitorycomponent of the unobserved quality.

The second set of moment conditions matches the predicted consumer pur-chase distance with the observed purchase distance from our individual levelchoice data. Because we include individual heterogeneity, we also use simula-tion at this step. We integrate out the random shock and income heterogeneityin αi by taking ten draws over the corresponding distributions for each indi-vidual in our sample. In particular, let yijft equal to one if consumer i boughtthe vehicle j from dealer f in year t and equal to zero otherwise. We compute

19

the following moment conditions:

G2N2

(θ) =1

N2

∑i,j,f,t

[Pijft(θ)

1−∑

j,f Pijft(θ)− yijft

]Z

(2)ijft, (11)

where Z(2)ijft includes the distance traveled by consumer i to buy the vehicle j

from dealer f in year t, and Pijft(θ) is the simulated choice probability for eachconsumer. The number N2 is the number of consumer-model-dealer-year-levelobservations in the sample.11

We stack the micro and macro moment conditions and then minimize theirweighted distance from zero by choosing θ. The estimates of θ are given by

θ = arg minθG(θ)′WG(θ), (12)

where G(θ) = [G(1)N (θ)G

(2)N2

(θ)]′ and W is a block diagonal weighting matrix.12

4.1 Identification

In this section we provide an informal discussion on the identification of the keyparameters of the model, the price coefficients α and the transportation costparameter γ. To identify the price coefficient, we need to find relevant and validinstruments to solve the simultaneity problem between price and unobservedcharacteristics, ξjft. Here, the simultaneity problem arises because dealers andconsumers observe the unobserved attribute ξjft when making their decisionsand so prices will adjust in the short-run to changes in ξjft. Following theliterature, we use the own product characteristics and the average exogenouscharacteristics of competing products as instruments for price. Here, we haverich variation in these instruments because we define competing firms as those

11In practice we construct the micro-moments using a randomly selected 20% sample ofthe individual transactions each time period.

12We use a two stage procedure. In the first stage we use the 2SLS weight matrix for thefirst set of moments, and the identity matrix for the second set of moments. Then we usethe first stage parameters to calculate the optimal weight matrix for each type of momentand re-estimate the model.

20

dealers within 5 miles. In addition, we use the number of car models with thesame body style and the the number of all car models within 5 miles.

The specific search mechanism that we model is not identified per se. Be-cause we do not observe choice sets or search behavior, we cannot reject an-other search model, or non-search model, in favor of our model. Our analysisin section 2.4 is designed to validate our choice of search model. In turn, thesearch cost parameters are identified conditional on our parametric assump-tions about search. The parameter on distance, γ, is identified from variationin distance and choice probabilities in the data. The identification of thesearch cost parameters relies on the assumption that dealer entry, exit andlocation are not correlated with the unobserved demand shocks, ξjft. Sincethe utility function includes dealers’ location fixed effect and time fixed effect,ξjft captures only transitory local demand shocks. Hence, this assumption isvalid if dealers’ entry, exit or location decisions are based on the long-run localconditions that are allowed to be contained in the location fixed effects andaggregate economic shocks that are captured by the time fixed effects, butnot on the realization of the transitory shock ξjft. This assumption is reason-able in our context for the following reasons. First, the sunk cost involvingthe entry, exit and location change of a dealer is large. This is partly dueto regulations that limit entry and exit. Hence, there is very little entry ofdealers in the industry, and when there is entry it is often a new brand enter-ing at an existing dealer location. Also, to the extent the local demographicsand population change over time, initial decisions about entry may not reflectcurrent demographics, population, or other transitory factors. Third, forcedexit of dealers by the manufacturer is very difficult in this industry becauseof state laws requiring payments to dealers for termination of franchise con-tracts. Lastly, there are other state laws that make it difficult for entry andexit, including mandated exclusive territories for brands. For a discussion ofthe regulatory environment see Lafontaine and Morton (2010) and Murry andSchnieder (2015). If our argument is invalid and dealers endogenously maketheir location choices based on the realization of the transitory shocks ξjft,our estimates of search cost parameter would be upward biased.

21

5 Discussion of Results

In this section we discuss the parameter estimates of the demand model. Firstwe present results from a simplified model, and then we present the resultsfrom the search model described in Section 4. We also simulate the estimatedmodel to learn about the extent of consumer information the model implies.Lastly, we present a model of pricing, and use the model to infer marginalcosts, which will be important for counterfactual simulations.

5.1 Evidence of the Effect of Dealer Co-location

To better understand the covariation in the data, we estimate a simple versionof our model where we make the following restrictions: (i) consumers havefull information and zero search costs and (ii) consumer heterogeneity comesthrough the idiosyncratic shock only, εifjt (see Berry, 1994). We present theresults in Table 4. Generally the results look sensible and are in line withother studies of demand for new cars. For example, the results imply an av-erage demand elasticity of either -3.27 or -4.43, consumers like acceleration(HP/Weight), more passenger seats, and dislike US brand cars. The coeffi-cients on miles per gallon (MPG) and luxury are not significant.13

We include an additional variable in consumer utility specifications in Table4 to understand how dealer co-location affects the consumers choice decision.Specifically, we add two product “characteristics,” separately in two specifi-cations, to the consumer’s indirect utility function: the number of availablecars within two miles of the product (Njft), and the number of cars avail-able within two miles of the same body style (N style

jft ). In a typical consumerchoice problem the existence of nearby competitors should not directly affectthe consumer’s utility. Nearby competitors would effect the consumer choiceprobability only though the denominator of the standard logit choice probabil-ity formula. However, we find these two variables have positive and significantcoefficients. Furthermore, the coefficient on N style

jft is larger than the coefficient

13The result on MPG is not an uncommon finding. For example see BLP and Petrin(2002).

22

on N , which makes sense because consumers likely care more about searchingfor similar cars. We take this as evidence that dealer co-location effects theconsumer choice problem in this market. In other words, there is a positiveeffect of nearby competitors in consumer utility.14

Table 4: Simple Demand Parameter Estimates(1) (2)

Variable Estimate s.e. Estimate s.e.

Price ($10,000) -1.160 (0.57) -1.571 (0.69)Constant -10.134 (1.68) -9.030 (2.00)HP/Weight*100 3.453 (2.16) 4.987 (2.63)MPG/10 (hwy) -0.185 (0.77) -0.707 (0.93)Passenger Seats 0.233 (0.09) 0.291 (0.11)U.S. Brand -0.503 (0.12) -0.590 (0.15)Luxury -0.359 (0.38) -0.391 (0.38)N 0.006 (0.003) – –N style – – 0.012 (0.005)

Mean Elasticity -3.27 -4.43Note: The sample includes 79,097 transactions and 2,280 model/dealer/year observations.We use the sum of attributes of all other products of the same types of cars within 5 miles asinstruments for the price. N is the number of products offered within 2 miles and N style isthe number within 2 miles of the same style. Both specificaitons include body style dummies(ie SUV, sedan etc.), dealer area dummies, and year dummies.

5.2 Search Model Estimates

The parameter estimates for the search model are presented in Table 5. Theestimates for horsepower/size, MPG, seats, and US brand are similar to the

14Endogeneity of these two variables might be a concern if the unobserved demand shock,ξjft, is correlated with firms location decisions. However, we include dealer area dummies(where the areas are those defined in Section 2.3) in the model to capture location specificunobserved heterogeneity. Identification of these parameters comes from changes in productsets across time within a 2 mile radius of each dealer. In general, this would be entry andexit of dealers that is not correlated with ξ, and changes to the products that manufacturersoffer, which is likely not a function of contemporaneous demand shocks in a single marketbecause it is a national decision by the manufacturer.

23

simpler specifications. Recall that price coefficients enter the consumers’ util-ity function in a non-linear way, see equation (2). Hence, a positive estimateof α1 implies that price has a negative impact on consumers’ mean utility,and a negative estimate of α2 implies that consumers with higher income areless price sensitive. The implied consumer-model-dealer-year level own-priceelasticities of demand are between -9.46 and -2.12, with a sales weighted av-erage of -4.10. This suggests that consumers are price sensitive on average,but there is substantial heterogeneity. Overall, our estimates of price elastici-ties are generally consistent with previous studies of automobile demand. Forexample, the average own-price elasticity is equal to -4.1 in Albuquerque andBrooenberg (2012), -5.3 in Murry (2015), and -3.14 in Nurski and Verboven(2015).

The distance search cost parameter is large in magnitude and preciselyestimated. It implies that consumers are very sensitive to distance. Next, weexplore the implications of the estimated search costs in detail.

Table 5: Demand Parameter Estimates from the Search Model

Variable Parameter Coefficient Std. ErrorDistance (100 miles) γ 16.949 (0.5689)Price, Mean ($10,000) α0 4.5657 (0.5867)Price*log(inc) α1 -0.3505 (0.0380)Price*random α2 0.1161 (0.0248)Horsepower/Size β1 4.8244 (0.2892)MPG/10 (hwy) β2 -0.7091 (0.1116)Passenger Seats β3 0.2594 (0.0348)U.S. Brand β4 -0.5783 (0.0778)Constant β0 -4.9825 (0.4912)

Note: The sample includes 79,097 transactions and 2,280 model/dealer/year observations.We use the mean of attributes of all other products of the same types of cars within 5 milesas instruments for the price. We include body style dummies (ie SUV, sedan etc.), dealerarea dummies, and year dummies. Standard errors are computed directly.

24

5.3 Search Cost and Search Frequency

To get a sense of the economic magnitude of the parameters, it is useful to con-sider how much a product’s price should be reduced to compensate consumersif they have to travel one more mile. Our results suggest that the averagevalue is $56 per mile. We also report the dollar-per-mile for each brand in thelast column of Table A.1 in Appendix A.15

The estimated travel cost is substantially lower than those reported in otherstudies that estimate consumer distance costs in the industry. For example,Moraga-González et al. (2015), the closest paper to ours, estimate a mediantravel cost of €107 per kilometer. This difference could come from differentreasons. First, Moraga-González et al. (2015) consider the market for cars inthe Netherlands. It is reasonable to think that search costs are much higher inEurope for multiple reasons, including congestion and the price of fuel. Second,the unit of observation in our data is different. We use individual purchase datawhereas use transaction data aggregated to a relatively fine geography. Ourparameter estimates could reflect important micro level information capturedin the co-variation of distance and purchase probabilities. Consider that aconsumer bought a Toyota Camry. Moraga-González et al. (2015) do notobserve which dealer she bought from, so they rely on the model to tell themwhat dealer she likely purchased from. However, we incorporate exact purchasedistance information, which may imply that consumers purchase from furtheraway dealer than their model predicts. Our search cost is also much lower thanstudies of consumer demand for cars that include distance in the indirect utilityfunction, like Nurski and Verboven (2015), Murry (2015), and Albuquerqueand Brooenberg (2012). For example Nurski and Verboven (2015) obtain anaverage travel cost of €112 per kilometer. It is possible search frictions mayhelp rationalize the purchase patterns with lower costs of distance.

Next, we consider what the estimates imply about how much consumers

15To calculate the cost per mile we first calculate the own distance elasticities, whichtells us how choice probabilities change if a dealer were to move one mile away from everyconsumer. The average distance elasticity is -0.92%. Next, we use the own price elasticityto calculate the price decrease that would increase choice probabilities by the same amount.We average this across all products to arrive at our average dollar per mile of $56.

25

search by using the results of the search model to simulate purchase behavior.In Figure 3, we plot the histogram of the number of searches per consumerfrom this simulation. In general, search is limited. We predict that conditionalon positive search, 49% of consumers search only one area and less than 1%search four areas. This results are generally consist with industry reports andprevious studies. For example, in a survey by DME Automotive, an industryconsulting group, 47% of all new car buyers report visiting a single dealerbefore purchasing a car. Moraga-González et al. (2015) report that 47% ofsurvey respondents in their data searched one dealer. Although our modeldoes not have empirical content regarding specifically how many dealershipsare searched because search happens at the area level, we can at least say thatsearching two or more areas implies searching at least two dealers, and so ourpredictions compare favorably to these other sources.16

Figure 3: Predicted Density of Search Intensity

Number of Searched Areas1 2 3 4 5 6 7 8 9

Density

0

0.1

0.2

0.3

0.4

0.5

0.6

Note: Model predicted distribution of the number of searches conditional on positive search.

16See http://www.dmeautomotive.com/announcements/1-in-6-car-buyers-skips-test-drive-nearly-half-visit-just-one-or-no-dealership-prior-to-purchase. In principle, we coulduse this information to add a moment inequality to our estimation objective function.However, we prefer to use survey data such as this to think about model validation.

26

6 Retail Markups

Although we do not use a model of car retail supply to estimate the demandparameters, we do use a supply model in counterfactuals, so it is importantto report the implications of the supply model given the estimated demandparameters. Below we describe our supply-side model.

6.1 A Supply Model

The profit of dealer f is defined as

πf (pt) =∑j∈Jft

(pjft −mcjft)qjf (pt)− FCft,

where mcjft is the constant marginal cost of product j sold by dealer f in yeart, and FCft represents fixed cost.

Dealers simultaneously maximize profit by setting price, taking into ac-count prices and attributes of competing dealers.17 The first order conditionfor a particular dealer that defines a Nash Equilibrium in prices is

qjf (pt) +∑j∈Jft

(pjft −mcjft)∂qjf (pt)

∂pjft= 0. (13)

Let 4 denote the price derivative matrix with the row j column k element

4jk =∂qj(pt)

∂pkt=

∑

l nlt´

(α0 + α1hi)Pijt(1− Pijt)dFl(hi), if j = k

−∑

l nlt´

(α0 + α1hi)PijtPiktdFl(hi), if j 6= k

We define an ownership matrix Ω∗, with Ω(j, k)∗ = 1 if product j and k are soldby the same dealer and zero otherwise. Let Ω = Ω∗ ×4(p) . Then, equation(13) can be written in matrix notation as the following markup equation

p−mc = Ω−1q(p). (14)

17Price here is the same concept as in the demand model: the average price for eachmodel at each dealer.

27

6.2 Recovering Price-Cost Markups

Given the estimated demand parameters, we compute the price cost markupsfrom equation (14). The weighted average price-cost markup, defined as thedifference between price and marginal cost, is $6,015 and the median is $6,051.We display the distribution of dealer margins, p−mc

p, in Figure 4. The weighted

average markup is 25.88% and the median is 23.75%. These are in line withother studies of the automobile industry, for example, 24% in Berry et al.(1995), 17% in Petrin (2002), $6,220 in Albuquerque and Brooenberg (2012),$5,238 in Murry (2015), and 43% in Nurski and Verboven (2015) and Moraga-González et al. (2015). The third column of Table A.1 in Appendix A shows theaverage markup in dollars for all brands. Among all brands, BMW’s markupis the highest and Suzuki’s markup is the lowest.

Figure 4: Dealer Margins (%)

Dealer Margin (%)0 10 20 30 40 50

Density

0

0.02

0.04

0.06

0.08

0.1

Note: Dealer margins predicted by the model. Dealer margins are defined as p−mcp .

6.3 Contribution of Search Frictions

The estimated model provides a way for us to understand the sources of dealermargins. The standard source of market power in discrete choice models of de-

28

mand is attributed to product differentiation. Additionally, our model impliesthat market power is also attributed to search frictions. Search frictions createincomplete choice sets for consumers as is seen in Figure 3, where the impliedsearch intensity of consumers is quite low – the median searching consumersearches only one geographic area. Search frictions can have two opposite ef-fects on price. On one hand, incomplete choice sets can lead to market powerfor firms because consumers are more captive to firms given they have fewerchoice options. Firms have the ability to raise prices because, conditional ona choice set, consumers do not have as many options available as they wouldif search costs were zero. This is the “competition effect.” Due to this effect,higher search cost would lead to higher price. On the other hand, becauseconsumers decide their search set of geographic areas by making a tradeoffbetween the expected utility of a search set and the search costs associatedwith that set, higher search cost would encourage dealers to lower prices toattract more consumers to visit their area. This is the “agglomeration effect.”Due to this effect, higher search cost would lead to lower price. Whether thecompetition effect dominates the agglomeration effect is an empirical question,depending on the model primitives.18

To quantify the impact of search cost on market power, we consider a sce-nario in which each consumer has full information and considers all products,the standard assumption in the literature. In Table 6 we present comparisonsof the price predicted by the full information model to the prices predicted bythe search model.19 The weighted average observed price (weighted by mar-ket share) is $24,940 and the weighted average simulated price from the fullinformation model (weighted by market share) is $24,518. Our result impliesthat search frictions contribute $422 to the price of a car on average.20

18Sovinsky Goeree (2008) examines a slightly different scenario where limited choice setsaffect the markup of firms. In her setting, firms advertise the existence of products toconsumers, which creates limited choice sets. She finds that her model implies markupsabout three times higher than a full information model. She does not have the second effectthat if prices rise too high then consumers will not include a product in their choice set.

19In estimation we match predicted shares to the data exactly, so prices in the data areidentical to model predictions for the search model.

20Moraga-González et al. (2015) perform a similar counterfactual. They find average

29

Table 6: Price Impacts of Search Frictions

Variables 2007 2008 2009 2010 all yearsaverage price

search $28,035 $27,897 $28,449 $28,521 $28,216full information $27,610 $27,462 $28,120 $28,066 $27,803difference $425 $434 $329 $455 $413

sales weighted average pricesearch $26,008 $23,966 $24,890 $24,798 $24,940full information $25,578 $23,538 $24,563 $24,353 $24,518difference $430 $428 $327 $445 $422

Note: price difference = price predicted by the search model minus price predicted by fullinformation model. Averages are weighted by market share.

Table A.1 in Appendix A compares the weighted average markups for eachbrand in the case of our estimated search cost to those in the case of zerosearch cost. This allows us to examine the additional markup firms earn asa result of search frictions. The weighted average markup predicted by thesearch model is $6,015 and the weighted average markup predicted by the fullinformation model is $5,593. Our results imply that about 8% of markupsare attributed to search frictions in this market. The results suggest that the“competition effect” dominates the “agglomeration effect” so that the pricesand hence markup predicted by the search model are in general higher thanthose predicted by the full information model. Currently the idiosyncraticshocks in our model generate product differentiation and thus market power.Although we are unable to estimate the variance of these shocks, if they hadless variance than what we assume, π2

6, then markups would be even more

attributable to search frictions, so we consider our finding conservative.

prices between their search model and a full information model very similar to ours. Thekey difference in the two models is that in their model, consumers search at the dealer levelas opposed to a geographic area. They also predict changes to manufacturer surplus afterdealer re-organization and find that manufacturers would prefer to consolidate brands underone dealership.

30

7 Retail Closures

Using our structural model we simulate dealer closures in order to tease out thecompetition and agglomeration effect of dealer co-location, and to understandwhat biases a full information model might have when considering such acounterfactual. A dealership closure generates two effects to the remainingdealers located in the same area. First, closing a dealer reduces the totalattraction of the whole area, and thus forces the dealers in this area to reducetheir prices. This is the “agglomeration effect”. Second, closing a dealer directlyreduces the price competition among dealers in the same area, and hencepushes the price higher. This is the “competition effect”.21

7.1 Background

The agglomerative effects of retail closures are particularly salient given therecent U.S. financial crisis of 2007-2009 that saw many retail firm exits dueto bankruptcy and other financial issues. Benmelech et al. (2014) documentmassive retail exits due to financial reasons, such as bankruptcy, during thefinancial crisis.22 They find evidence of agglomeration effects of closures ofbankrupt firms’ stores on non-bankrupt incumbent stores using data on thelocation and closures of multiple retail chains across the US. New car dealersalso saw a large swath of retail closures during and immediately following thefinancial crisis. For example, Chrysler, General Motors, and Ford all closeddealers for financial reasons in 2009 and 2010. However, as described in La-fontaine and Morton (2010), this industry has experienced large numbers of

21In complementary work, Ozturk et al. (forthcoming) examine the agglomeration versuscompetition effects of dealer closures. They use a national sample of car sales to infer atreatment effect of Chrysler dealership closures on rivals’ prices. They find that an average,rivals’ prices increase after a Chrysler closure, but prices of nearby dealers increase much lessthan distant dealer. They interpret their results as providing evidence that the competitioneffect dominates the agglomeration effect overall but the agglomeration effect is presentbecause nearby dealers experience a lower price increase than far away dealer after theclosures.

22For example, they document the complete liquidation of multiple large retailers, in-cluding Circuit City, Linens ’N Things, and The Sharper Image. Other large retail chainsthat experienced massive closings due to financial trouble include Kmart and Sears.

31

retail closures for many decades, from a high of around 60,000 outlets in theUS in the 1940’s, to less than 20,000 by 2010. Most recently a number of en-tire brands have dissolved, such as Oldsmobile, Pontiac, Hummer, and Saab,among others, making the issue of the agglomerative effects of retail closuresespecially important in new car retailing.

Recent dealer closures stemmed from two primary causes. First, Ameri-can manufacturers discontinued a number of brands in the mid to late 2000s,starting with Oldsmobile in 2004, and continuing with Saturn and Pontiac in2009, Mercury in 2010, and Saab in 2011. 23 These brands had seen steadydeclines in sales, and were reported as being unpopular and out of touch withconsumer needs in media and industry reports.24 For the case of GM ownedPontiac and Saab, there was also pressure from the U.S. government, whoprovided a large loan to GM in 2009 under the Troubled Asset Relief Program(TARP), to make the company “leaner” and more focused on core productsthat had a history of satisfactory sales and performance. The second cause ofthe dealer closures had to do with the financial crisis more directly. GM andChrysler received TARP U.S. government loans in 2009, and because of theirsubsequent reorganization were allowed to terminate dealers without answer-ing to state auto franchise laws. Both companies, along with Ford Motors,had a clear policy to create smaller dealer networks, but were generally unableto do so because state regulations prohibit dealer franchise contract termina-tion by manufacturers in the automobile industry. 25Dealers lobbied againstdealer closures, citing existing state regulations that prohibit closures. Manyof the proposed closures (from both of the reasons stated above) went intolegal arbitration. For example, when GM closed the Oldsmobile brand, theyreportedly paid over $1 billion to their dealers. In this section we examine theeffects of dealer closures, and offer an explanation of why even unclosed deal-ers might prefer other dealers not to close.26 Our results also suggest that the

23Saab, the major Swedish produced car brand, was owned by General Motors until 2011.After 2011, the company re-organized, and started producing cars again in 2014.

24http://tinyurl.com/cotzn925For example of popular press coverage of dealer closures, see

http://tinyurl.com/p7zvgys.26For an alternative explanation having to do with promotional activities by the manu-

32

loss to consumer welfare from a dealer closure, and the gain of rival dealers, isexaggerated by the standard full information model of demand.

7.2 Competition Effect and Agglomeration Effect

To separately examine the agglomeration effect and the competition effectwe calculate the equilibrium outcomes in the following two scenarios. In thefirst scenario, after the dealer closure we allow the consumers to adjust theirprobability of searching (PiS), but we hold constant the probability of purchaseconditional on each search set (Pijft|S). This quantifies the agglomerationeffect: firms will adjust their equilibrium prices only for the reason to attractnew searchers to their dealer area, not to price compete against a rival in thesame geographic area. In the second scenario, we hold constant the probabilityof searching (PiS) but adjust the probability of purchasing condition on eachsearch set (Pijft|S). This quantifies the competition effect: firms will adjusttheir price to compete with local rivals, not to attract more consumers to searchtheir geographic area. In both scenarios, we measure the separate effects on thedealers by comparing the difference in prices, sales and markups. Lastly, weallow the consumer to adjust both her search decisions and choice probabilitiesconditional on search to quantify the total effect.

To measure the welfare effects of dealer closure, we follow the literature(see Petrin (2002) and Fan (2013)) and define consumer welfare change as thecompensating variation. The compensating variation for consumer i is givenby

CVilt =V 1ilt − V 0

ilt

αi,

where αi < 0 is the negative of the consumer i’s marginal value of income, andV 0ilt−αi and V 1

ilt−αi are the expected maximum utility for consumer i beforeand after a dealer closure. Specifically,

V 0ilt = ln

[1 +

∑S

exp

(U0iSt −

∑m∈S

cim

)]

facturer, see Murry (2015).

33

where U0iSt = ln

[1 +

∑j∈Jft,f∈Fmt,m∈S exp(u

0ijlt − εijlt)

]. The post-closure util-

ity V 1ilt is analogously defined by replacing U0

iSt with U1iSt and u0ijlt with u1ijlt.

Given the compensating variation for a specific consumer, the change in theaverage consumer surplus in zip code l in year t is given by ∆CSlt = Eςi(CVilt).The total consumer welfare change is the sum of the welfare changes in all zipcodes, 4CS =

∑lt nlt∆CSlt, where nlt is the number of households in zip

code l in year t. The change in average per-consumer consumer surplus is4CS = 4CS∑

l nlt.

7.3 A Case Study

First, we simulate new equilibria as described above for the closing of a smalldealer, henceforth Small. Small terminated operation in 2009, and previousto closure was one of the worst performing dealers in terms of sales in oursample. Small sold 94 units of cars in 2007 and 69 units of cars in 2008. Weconsider a scenario in which this dealer was closed in 2007. Table 7 presentsthe agglomeration effect, competition effect and the total effect of closing thisparticular dealer.

In the first two rows, we present the results when we only allow adjustmentof the probability of searching (i.e., PiS ). This is the agglomeration effect ofclosing a dealer. We can see that other dealers in the same area would reducetheir price by $3 and their total sales in 2007 would suffer a loss of around 7units. Dealers located in other areas would increase their price by $1 and theirtotal sales would increase by 4 units.

In the third and fourth rows, we present results when we only allow adjust-ment of the probability of buying each product given a search set (i.e., Pijft|S). This is the competition effect of closing a dealer. We can see that otherdealers’ price would increase by $4 on average and total sales in 2007 wouldincrease by around 7 units. Dealers located in other areas would increase theirprice by $1 and total sales would increase by 10 units.

In the last two rows, we present results when we allow adjustment of bothPiS and Pijft|S. This is the total effect of closing a dealer. We can see that

34

other dealers’ average price in the same area stay almost the same and totalsales in 2007 increase by very slightly. This is because the agglomeration effectand competition effect offset with each other. Dealers located in other areaswould increase their price by $2 on average and total sales would increase byaround 13 units.

Table 7: Effects of Dealers Closing Dealer “Small”

Sales Average Pricebefore after change before after change

Agglomeration EffectSame area 1,831 1,824 -7 $26,153 $26,150 -$3Other areas 22,798 22,802 +4 $28,197 $28,198 +$1Competition EffectSame area 1,831 1,838 +7 $26,153 $26,157 +$4Other areas 22,798 22,808 +10 $28,197 $28,198 +$1Total EffectSame area 1,831 1,833 +2 $26,153 $26,153 $0Other areas 22,798 22,811 +13 $28,197 $28,198 +$1

Note: Dealers in the same area are referred to those dealers located in the same area asdealer "Small". Dealers in other areas are referred to those dealers located in different areaas dealer "Small". Sales refers to the total sales across all dealers in the relevant geographyfor 2007.

The welfare effects of closing Small is reported in Table 8. We compare theeffects implied by the search model to a model with full information. The firstrow shows that the welfare change for the search model. Overall, the consumersurplus declines by 0.027% of the pre-closure consumer surplus and the totalsurplus of all dealers decreases by 0.294% of the pre-closure dealer surplus.Among these dealer surplus change, the total surplus of unclosed dealers inthe same area increases by 0.104% whereas the total surplus of all uncloseddealers in other areas increases by 0.067%. The second row of Table 8 showsthat the consumer surplus change for full information model is -0.281%, andthe total surplus of unclosed dealers in the same area increases by 0.211%

35

whereas the total surplus of all unclosed dealers in other areas increases by0.213%.

The full information model over-predicts the consumer welfare loss. Themain reason is that in the full information model, a dealer closure lessensthe total choices available to all consumers. But in the search model, only afraction of consumers had the closed dealer in their choice set in the first place,so the closure does not impact the surplus of as many consumers. Additionally,the price changes are less severe in the search model because higher priceshave the added negative effect of decreasing search probabilities. The fullinformation model also over-predicts the surplus gain for unclosed dealers aftera dealer closure. The reason is that agglomeration effect a lower rise in pricesto offset the decreased attraction of the area due to the closure of a dealerin the area. Ignoring this effect would overestimate the consumer welfare lossand overestimate the benefits that other dealers can get from a closure of acompeting dealer.

Table 8: Welfare Effects of Closing Dealer “Small”

Consumer Dealer SurplusSurplus all dealers same area other areas

Search Model:change (m $) -0.578 -0.435 0.011 0.092percentage change (%) -0.027 -0.294 0.104 0.067

Full-Information Model:change (m $) -2.548 -1.088 0.095 1.120percentage change (%) -0.281 -0.190 0.211 0.213

Note: "all areas" refers to the total surplus change of all dealers in the sample including theclosed dealer, "same area" refers to the total surplus change of all other dealers located inthe same geographic area as the closed dealer, and "other areas" refers to the total surpluschange of all dealers located in all geographic areas other than the area of the closed dealer.

36

7.4 Average Effect of Dealer Closures

The previous section shows how the framework in this paper can be used tostudy the effects of a single dealer closure in one specific market. It givesinsight into the particular mechanisms at play in our model. However, it wasjust a case study, and this dealer may not be representative of all dealers inthe market we consider. Next, we investigate the general pattern of how thewelfare effect of dealer closure varies with dealer characteristics. To do this,we compute the welfare effects when we close each dealer, one at a time.

Table 9: Welfare Effects of Dealer Closure

Mean SD Min Median MaxSearch Model:4CSft (m$) -2.430 2.300 -0.076 -1.683 -12.298

(%) -0.115 0.108 -0.004 -0.079 -0.5764DSft (m$) -1.832 1.727 -0.065 -1.247 -9.072

(%) -1.567 1.365 -0.094 -1.033 -6.2654DSsameft (m$) 0.108 0.144 2.43 ∗ 10−4 0.052 0.905

(%) 0.486 0.437 0.016 0.344 2.4514DSotherft (m$) 0.305 0.311 0.005 0.211 1.647

(%) 0.299 0.274 0.010 0.201 1.465

Full Information Model:4CSft (m$) -10.295 9.892 -0.364 -6.852 -46.553

(%) -1.393 1.277 -0.718 -0.984 -6.2964DSft (m$) -4.918 4.786 -0.232 -3.423 -22.235

(%) -1.033 1.014 -0.052 -0.691 -5.3764DSsameft (m$) 0.614 0.708 0.007 0.386 3.820

(%) 0.887 0.822 0.035 0.608 3.7864DSotherft (m$) 3.784 3.855 0.083 2.473 19.001

(%) 0.882 0.831 0.034 0.597 3.875

Characteristics of Closed Dealer:Qft 424.255 394.271 13 307 2131AreaShareft (%) 16.981 16.697 0.468 9.980 81.105

Note: Descriptive statistics from the results of counterfactual exercise of closing every dealer,one at a time. Variable definitions in the text.

37

Table 9 provides summary statistics on the distribution of welfare changesfor our search model and the full information model. Let 4CSft denote thepercentage change in average consumer surplus, 4DSsameft denote the percent-age change in total surplus of all unclosed dealers located in the same area asthe closed dealer, 4DSotherft denote the percentage change in total surplus ofall unclosed dealers located in the other areas, 4DSft denote the percentagechange in total surplus of all dealers, Qft denote the sales of the closed dealer,and AreaShareft denote the market share of the closed dealer in its geographicarea.

The search model predicts that consumer surplus declines by 0.115% on av-erage, across dealer closures, and the total surplus of all unclosed dealers in thesame area increases by around 0.486% on average. The full information modelpredicts that the mean consumer welfare loss is 1.393% and the total surplusgain of all unclosed dealers in the same area is 0.887% on average. Therefore,ignoring the search frictions and agglomeration effect would overestimate theconsumer welfare loss and overestimate the benefits that other dealers can getfrom a closure of a competing dealer. Looking at the percentiles of the wel-fare change distributions, the search model predicts very skewed distributions.This corresponds to the skewed distribution of the size of the closed dealer andthe skewed distribution of the market share of the closed dealer in its area.

As seen in Table 9, dealer surplus for incumbent dealers increases after asingle dealer closure (rows three and four in each panel). However, notableis that the full information model overstates the gain in dealer surplus aftera closure for those dealers in the same geographic area as the closed dealer.This is because incumbent dealers do not raise prices by too much after a rivalclosure in order to keep their area attractive to search, whereas in the no-searchmodel this mechanism does not exist. Change in total producer surplus, ∆DS,is strictly negative under the full information model, but this is not the casefor the search model. This is due to the different effects depending on the sizeof the closed dealer and the total sales and search intensity of the geographicarea of the closed dealer.

To understand the relationship between the welfare effects and the char-

38

acteristics of the closed dealer, we run regressions of welfare effects on thecharacteristics of the closed dealer, in particular, the size of the closed dealerand the market share of the closed dealer in its area. Notice that this regres-sion captures a correlation pattern rather than a causal effect. However, it isa useful way to summarize the results from the counterfactual simulations andto broadly confirm our intuition about the different mechanisms in the model.The results of the regressions are reported in Table 10.

The welfare effect of dealer closure depends on the size of the closed dealer.If a large dealer is closed, after controlling for its share in its area, this wouldincrease the market power of its neighboring competitors. As a result, otherunclosed dealers in the same area would increase their prices. Higher prices inthat area would also lead to higher prices of other areas due to the competitionamong areas. This explains why both the total surplus of unclosed dealersin the same area (4DSsameft ) and the total surplus of unclosed dealers inother areas (4DSotherft ) increase in the size of the closed dealer. Due to thehigher prices, consumers are worse off; in other words, consumer welfare change(4CSft) decreases in the size of the closed dealer.

Another factor is the market share of the closed dealer in its area. Thismeasures how important the dealer is to its area. If a dealer with a largeshare in its area is closed, after controlling for its size, this would significantlydecrease the attraction of that area. As a result, other unclosed dealers in thesame area would lose a substantial amount of consumer visits and thus sales.To offset the declined attraction of the whole area, they have to reduce theirprices. Lower price in that area would also lead to lower prices of other areasdue to the competition among areas. This explain why both the total surplusof unclosed dealers in the same area and the total surplus of unclosed dealersin other areas decrease in the area share of the closed dealer. Due to the lowerprices, consumers are better off ; in other words, 4CSft increases in the areashare of the closed dealer.

39

Table 10: Regression: Welfare Effects and Area Characteristics

4CSft 4DSsameft 4DSotherft

Qft -0.098 0.378 0.257(0.005) (0.021) (0.013)

AreaShareft 0.016 0.075 -0.106(0.026) (0.111) (0.067)

Constant 0.442 -1.665 -1.144(0.029) (0.122) (0.074)

Note: Estimates from regressions of the change in welfare from dealer closures versus areacharacteristics. Unit of observation is a single closure dealer closure in a given time period.∆CSft is the change in consumer surplus from the closure of dealer f in time t; ∆DSsame

ft

is the change in total surplus of the remaing dealers in the same area as the closed dealer;∆DSother

f t is the total change in surplus for dealers in other areas. Standard errors inparentheses.

Next, we examine the sources that drive the fact that the full informationmodels overestimates the consumer welfare loss and the surplus gain of un-closed dealers due to a dealer closing. We define the bias as the differencebetween the percentage change predicted by the full information model andthe percentage change predicted by the search model. We regress these biasterms on dealer and market characteristics. The two characteristics we use arethe total sales of the dealer, and the share of dealer sales in its area beforeclosure. We use both of these variables because we are interested in teasingout the contribution of dealer size conditional on the size of the dealer cluster,and also the relative importance in a cluster conditional on gross dealer size.The results of the regressions are reported in Table 11. Both factors are im-portant contributors to the bias between the full information model and thesearch model because these two factors are both positively related to how largethe agglomeration effect is implied by the model in a particular dealer cluster.The “larger” a dealer, both in raw terms and relative terms to nearby dealers,the more the full information model will overstate the effects of dealer closuresbecause it does not capture the countervailing agglomeration effect.

40

Table 11: Regression: Welfare Effects Bias and Area Characteristics

4CS Biasft 4DS Biassameft 4DS Biasotherft

Qft 0.921 0.266 0.378(0.058) (0.027) (0.035)

AreaShareft 1.848 0.940 1.216(0.306) (0.140) 0.184)

Constant -4.634 -1.341 -1.828(0.336) (0.154 (0.203)

Note: Estimates from regressions of the bias in the change in welfare from dealer closuresversus area characteristics. The bias is defined as the difference in the change between thesearch model and the full information model. The unit of observation is a single closuredealer closure in a given time period. ∆CSBiasft is the bias in the change in consumersurplus from the closure of dealer f in time t; ∆DSBiassame

ft is the bias in the change intotal surplus of the remaing dealers in the same area as the closed dealer; ∆DSBiasotherf tis the bias in the total change in surplus for dealers in other areas. Standard errors inparentheses.

8 Conclusion

In this paper, we estimate a structural model of consumer search for spatiallydifferentiated products in the new car market. Our approach contributes to theliterature on consumer demand with limited information and the literature onretail agglomeration by formally modeling consumer search for spatially differ-entiated products where the co-location of retail stores effects the consumers’search and purchase decisions.

We estimate a substantially lower dollar per mile of traveling to purchasethan other studies, both search and not search, for the automobile industry.Compared to the standard full-information model, our model implies greatermarket power for car dealers. After a dealer closure, our model predicts smallerconsumer welfare loss and smaller dealer surplus gain (for incumbent dealers)after a single dealer closure than the a full information model. Our resultssuggest that both the competition and agglomeration effects matter after adealer closure. Our results are consistent with Ozturk et al. (forthcoming),but are at odds with the findings of Benmelech et al. (2014), who find anegative overall effect on nearby rivals after closure. However, in the case of

41

Benmelech et al. (2014), the competition effect would be expected to be smallbecause they examine the effect across unrelated products.

We also validate our modeling assumptions by showing, in a simpler frame-work, that consumer purchasing decisions are partly a function of dealer ag-glomeration. We do this by estimating a simple model of demand where weinclude the number of nearby available products as a “characteristic” in con-sumer utility. Given the amount of co-location in different retail industries,this is an important finding on its own because it provides evidence of demandbased reasons for retail agglomeration, as opposed to cost side reasons.

To be sure, our analysis relies on particular assumptions, and althoughwe are confident that our model captures the major features of this industry,some caveats are worth mentioning. First, although we feel like the evidencewe present strongly suggests dealer agglomeration is a consumer considerationduring the car buying process, the search process may be more complicatedthan our model presents. In particular, the recent proliferation of car-buyingwebsites aimed at alleviating consumer information has likely started to changethe way consumers search for cars. However, cars are likely an experience good,so websites could never fully inform a consumer completely about the utilityfrom purchase like personal interaction can. Second, consumers may searchin a more complicated way, nesting geographical concerns with the searchfor a dealer (as in Moraga-González et al. (2015)) and the search for a cartype. Thirdly, although we present a demand driven reason for dealers to co-locate, there are likely cost driven reasons, for example land prices, zoning, andmanagement convenience for multi-dealership dealer conglomerates, amongothers. Our analysis can not be used to balance all the tradeoffs associatedwith the optimal location decision, only to quantify the demand side incentives.

ReferencesAlbuquerque, P. and B. Brooenberg, “Measuring the Impact of Nega-tive Demand Shocks on Car Dealer Networks,” Marketing Science 31 (2012),4–23.

42

Anderson, S. P., A. D. Palma and J. F. Thisse, Discrete choice theoryof product differentiation (MIT press, 1992).

Benmelech, E., N. Bergman, A. Milanez and V. Mukharlyamov,“The Agglomeration of Bankruptcy,” Technical Report, National Bureau ofEconomic Research, 2014.

Berry, S., J. Levinsohn and A. Pakes, “Automobile Prices in MarketEquilibrium,” Econometrica 63 (July 1995), 841–890.

———, “Differentiated Products Demand Systems from a Combination of Mi-cro and Macro Data: The New Car Market,” Journal of Political Economy112 (February 2004), 68–105.

Berry, S. T., “Estimating discrete-choice models of product differentiation,”The RAND Journal of Economics (1994), 242–262.

Burdett, K. and K. L. Judd, “Equilibrium price dispersion,” Economet-rica: Journal of the Econometric Society (1983), 955–969.

Chade, H. and L. Smith, “Simultaneous search,” Econometrica 74 (2006),1293–1307.

d’Aspremont, C., J. J. Gabszewicz and J.-F. Thisse, “On Hotelling’s"Stability in competition",” Econometrica: Journal of the Econometric So-ciety (1979), 1145–1150.

De los Santos, B., A. Hortacsu and M. R. Wildenbeest, “Testingmodels of consumer search using data on web browsing and purchasing be-havior,” The American Economic Review 102 (2012), 2955–2980.

Dudey, M., “Competition by choice: the effect of consumer search on firmlocation decisions,” American Economic Review 80 (1990), 1092–1104.

Ellickson, P. B., S. Houghton and C. Timmins, “Estimating networkeconomies in retail chains: a revealed preference approach,” The RANDJournal of Economics 44 (2013), 169–193.

Fan, Y., “Ownership Consolidation and Product Characteristics: A Study ofthe US Daily Newspaper Market,” American Economic Review 103 (Auguest2013), 1598–1628.

Fischer, J. H. and J. E. Harrington Jr, “Product variety and firmagglomeration,” The RAND Journal of Economics (1996), 281–309.

43

Hong, H. and M. Shum, “Using price distributions to estimate search costs,”The RAND Journal of Economics 37 (2006), 257–275.

Honka, E., “Quantifying search and switching costs in the US auto insuranceindustry,” The Rand Journal of Economics 45 (2014), 847–884.

Honka, E. and P. K. Chintagunta, “Simultaneous or sequential? searchstrategies in the us auto insurance industry,” Search Strategies in the USAuto Insurance Industry (April 29, 2014) (2014).

Hortaçsu, A. and C. Syverson, “Product Differentiation, Search Costs,and Competition in the Mutual Fund Industry: A Case Study of S&P 500Index Funds,” The Quarterly Journal of Economics (2004), 403–456.

Hotelling, H., “Stability in Competition,” The Economic Journal 39 (1929),41–57.

Janssen, M. and J. L. Moraga-González, “Strategic Pricing, ConsumerSearch and the Number of Firms,” Review of Economic Studies 71 (2004),1089–1118.

Jia, P., “What Happens When Wal-Mart Comes to Town: An EmpiricalAnalysis of the Discount Retailing Industry,” Econometrica 76 (2008), 1263–1316.

Lafontaine, F. and F. S. Morton, “Markets: State Franchise Laws, DealerTerminations, and the Auto Crisis,” Journal of Economic Perspectives 24(Summer 2010), 233–50.

Moraga-González, J. L., Z. Sándor and M. R. Wildenbeest, “Con-sumer search and prices in the automobile market,” unpublished manuscript(2015).

Murry, C., “Advertising in Vertical Relationships: An Equilibrium Model ofthe Automobile Industry,” unpublished manuscript, Penn State University(2015).

Murry, C. and H. Schnieder, “The Economics of Retail Markets for Newand Used Cars,” unpublished manuscript, Cornell University (2015).

Nurski, L. and F. Verboven, “Exclusive Dealing as a Barrier to Entry? Ev-idence from Automobiles,” Review of Economic Studies forthcoming (2015).

44

Ozturk, O. C., S. Venkataraman and P. K. Chintagunta, “Price Re-actions to Rivals’ Local Channel Exits,” Marketing Science (forthcoming).

Petrin, A., “Quantifying the Benefits of New Products: The Case of theMinivan,” Journal of Political Economy 110 (2002), 705–729.

Rosenthal, S. S. and W. C. Strange, “Evidence on the nature andsources of agglomeration economies,” Handbook of regional and urban eco-nomics 4 (2004), 2119–2171.

Seiler, S., “The impact of search costs on consumer behavior: A dynamicapproach,” Quantitative Marketing and Economics 11 (2013), 155–203.

Seim, K., “An Empirical Model of Firm Entry with Endogenous Product-TypeChoices,” The RAND Journal of Economics 37 (Autumn 2006), 619–640.

Sovinsky Goeree, M., “Limited Information and Advertising in the USPersonal Computer Industry,” Econometrica 76 (September 2008), 1017–1074.

Stahl, K., “Location and spatial pricing theory with nonconvex transporta-tion cost schedules,” The Bell Journal of Economics (1982), 575–582.

Stigler, G. J., “The Economics of Information,” Journal of Political Econ-omy 69 (1961), 213–225.

Vitorino, M. A., “Empirical entry games with complementarities: An appli-cation to the shopping center industry,” Journal of Marketing Research 49(2012), 175–191.

Wildenbeest, M. R., “An Empirical Model of Search with Vertically Differ-entiated Products,” RAND Journal of Economics 42 (2011), 729–757.

Wolinsky, A., “Retail trade concentration due to consumers’ imperfect in-formation,” The Bell Journal of Economics (1983), 275–282.

———, “True Monopolistic Competition as a Result of Imperfect Competi-tion,” Quarterly Journal of Economics 101 (1986), 493–511.

Zhu, T. and V. Singh, “Spatial competition with endogenous locationchoices: An application to discount retailing,” Quantitative Marketing andEconomics 7 (2009), 1–35.

45

Appendix A: Additional Tables and Figures

Table A.1: Results by Brandaverage price ($) average markup ($) price elasticity dollar per

observed full info diff search full info diff (%) search full info mile ($)Acura 34,493 33,927 566 6,595 6,029 9.38 -5.23 -5.63 60.23Audi 39,851 39,280 571 6,879 6,309 9.04 -5.76 -6.19 62.78BMW 45,942 45,361 581 7,195 6,613 8.79 -6.40 -6.88 65.18Buick 30,125 29,637 488 6,301 5,813 8.39 -4.77 -5.10 57.94Cadillac 40,994 40,404 590 6,919 6,329 9.33 -5.93 -6.38 63.31Chevrolet 23,395 23,030 365 5,829 5,464 6.68 -3.93 -4.14 53.82Chrysler 27,298 26,865 433 6,129 5,696 7.60 -4.45 -4.73 56.03Dodge 23,420 23,013 407 5,872 5,465 7.46 -3.98 -4.21 54.35Ford 23,467 23,100 367 5,871 5,504 6.67 -3.98 -4.19 54.10GMC 38,190 37,708 482 5,582 6,201 7.76 -5.70 -6.07 60.93Honda 22,445 22,071 374 5,884 5,511 6.78 -3.82 -4.05 53.92Hyundai 19,699 19,259 440 5,739 5,399 8.30 -3.42 -3.63 52.17Infiniti 41,191 40,607 584 6,895 6,311 9.25 -5.96 -6.42 63.04Jeep 26,068 25,607 461 6,117 5,655 8.16 -4.26 -4.55 55.78Kia 18,674 18,301 373 5,604 5,231 7.14 -3.33 -3.51 50.89Lexus 42,024 41,460 564 6,994 6,430 8.77 -6.02 -6.46 62.74Lincoln 38,166 37,668 498 6,740 6,242 7.98 -5.67 -6.06 61.22Mazda 22,490 22,001 489 5,936 5,447 8.98 -3.78 -4.05 53.93Mercedes-Benz 45,190 44,612 578 7,108 6,530 8.84 -6.32 -6.78 64.51Mercury 24,355 23,869 486 6,029 5,544 8.77 -4.05 -4.33 54.64Mini 24,721 24,196 525 6,099 5,573 9.43 -4.07 -4.36 55.67Mitsubishi 22,177 21,646 531 5,896 5,364 9.90 -3.76 -4.03 53.27Nissan 22,736 22,291 445 5,910 5,466 8.13 -3.83 -4.07 53.71Pontiac 21,217 20,760 457 5,807 5,349 8.55 -3.65 -3.88 53.00Saab 32,625 32,077 548 6,393 5,846 9.36 -5.11 -5.50 58.43Saturn 24,063 23,534 529 5,994 5,465 9.68 -4.00 -4.29 54.13Scion 17,246 16,892 354 5,574 5,219 6.79 -3.13 -3.31 50.86Subaru 22,672 22,166 506 5,974 5,468 9.25 -3.81 -4.08 54.14Suzuki 18,253 17,520 733 5,633 5,100 10.44 -3.18 -3.41 50.89Toyota 23,341 22,973 368 5,941 5,573 6.60 -3.93 -4.16 54.28Volkswagen 22,136 21,621 515 5,929 5,415 9.50 -3.73 -3.99 53.87Volvo 34,019 33,456 563 6,561 5,998 9.38 -5.17 -5.57 59.78All 24,940 24,518 422 6,015 5,593 7.55 -4.10 -4.35 56.59

Note: All averages are weighted by the market shares. Difference = observed - predictedby full information model. The own price elasticities are measured in the absolute values.

46

Figure A.1: Dealers and Dealer Areas in Richmond, VA

Note: This maps displays the locaitons of dealers across the Richmond metropolitan area,color coded by assigned co-location area. Areas 1 and 2 are both shaded light purple andarea 9, not shown, is about 10 miles south of area 8 near Petersburg, VA.

47

Figure A.2: Sales by Dealer’s Census Tract, 2008

Note: This maps displays the sales of new car dealers in the Richmond metropolitan areafor the year 2008. Total sales are shaded by the US Census Tract of the dealer location.Darker red represents more total sales from that tract. As can be seen from Figure A.1,there are typically multiple dealers per tract.

48