The Costs of Zoning Regulations in Retail Chains: The Case of the City Planning Act of 1968 in Japan Mitsukuni Nishiday The Johns Hopkins Carey Business School

January 30, 2014

Abstract The deregulation of zoning restrictions on retailers has been at the forefront of urban policy debates in recent years. Despite the increasing attention, however, we know surprisingly little about how the zoning regulations, which impose a constraint on the supply of land for a particular use, a¤ect retailers in urban areas. This paper examines the e¤ect of the zoning regulations introduced in 1968 in Japan on entry of convenience-store outlets. The act speci…es two zoning restrictions for retail outlets. First, the law prohibits building commercial outlets in speci…ed residential and industrial areas (“type 1 zoning”). Second, the law requires a developer to obtain permission from the local government to develop an outlet in speci…ed areas (“type 2 zoning”). To account for the spatial dependence in demand, costs, and zoning regulation status across markets, this paper employs an equilibrium model of store-network choice by multistore …rms. Using the cross-sectional data of entry and zoning in Okinawa in 2002, the paper …nds hypothetically eliminating the type 1 and type 2 regulations would increase the total number of convenience-store outlets by 10% 14% and 2% 3%, respectively. Although the magnitude of the increase in store counts and sales are similar across two national chains in Okinawa, the geographical markets in which each …rm increases its outlets after the deregulation di¤er across these two chains. Keywords: zoning; entry; regulation; chain; retailing JEL codes: L1; L5; R3

An earlier version of the paper was circulated under the title “The E¤ect of Zoning Regulations on Entry in the Retail Industry.” I would like to thank Jeremy Fox, Ali Hortaçsu, Chad Syverson, and Jean-Pierre Dubé for their detailed comments. I have also bene…ted from discussions with Panle Jia, Steve Levitt, Ryo Nakajima, Ye¸sim Orhun, Zhu Wang, Ting Zhu, and seminar participants at various places. I would also like to thank Mr. Tomita from the development division at Prefecture O¢ ce of Okinawa for sharing the details on how the prefecture government implements the regulation. I am grateful for the editor and an anonymous referee for providing valuable suggestions. I gratefully acknowledge …nancial support from the NET Institute (www.netinst.org), the Kau¤man Foundation, and the Center of East Asian Studies. All remaining errors are my own. y 100 International Drive Baltimore, MD 21202 USA. E-mail: [email protected]

1

1

Introduction

Since Hotelling’s (1929) seminal article, economic models typically posit …rms can freely locate to maximize their pro…ts. In reality, however, various government regulations restrict where …rms can enter and a¤ect the costliness of developing a retailing outlet or plant. For instance, Boylaud and Nicoletti (2001) report that in many economies, “The regulations concerning commercial real estate and zoning are among the greatest barriers to the development of retail services.” Zoning regulations generally impose a constraint on the supply of land for a particular use. Although zoning regulations aim to achieve some desirable goals, such as protecting residential and agricultural spaces by isolating them from commercial spaces, they also introduce an e¢ ciency loss, such as a forgone gain from a trade due to the non-existence of outlets or increased opportunity costs associated with complying with the regulation when developing an outlet. Indeed, the deregulation of zoning restrictions on retailers in urban areas has been at the forefront of urban policy debates in recent years in various economies. For example, in Japan, although the zoning regulations enacted in 1968 have promoted city planning and provided neighborhood amenities, such as open space, less disturbances, and less tra¢ c, mounting public opinion has been calling for deregulating the land-use laws on the grounds that the regulations are overly stringent for retail outlets in zoned areas (Keidanren 2002; Council for Regulatory Reform 2004). The goal of this paper is to examine whether and to what extent the existing zoning regulations a¤ect market outcome, such as sales, pro…ts, and location choice, by using zoning and outlet location data from the convenience-store industry in 2002 in Okinawa, Japan. Although the industrial organization literature recognizes the importance of zoning regulations, relatively little is known about how zoning regulations a¤ect retailing market outcome for two reasons. The …rst reason is data limitations. The past literature did not incorporate the zoning information in the empirical model until recently due to lack of direct zoning data. For instance, Seim (2006) drops some of the detailed demographic variables from the estimation because the zoning ordinances, which are unobserved, may confound them (see also Netz and Taylor 2002). Such analysis will miss the contribution of the e¤ect of zoning or land-use regulation on entry, and may lead to an omitted variable bias. Only recently have economists started to investigate the role of zoning on …rm entry, with the direct data on zoning and an empirical framework based on …rm entry games. This paper aims to …ll that gap in the literature by incorporating the zoning information into the model of …rm entry as in Gri¢ th and Harmgart (2008), Ridley, Sloan, and

2

Song (2011), Datta and Sudhir (2013), and Suzuki (2013). Second, and more fundamentally, the scarcity in research arises from the lack of appropriate methodology. In urban areas, geographically adjacent markets are certainly related through demand, costs, or zoning regulation status. Measuring the e¤ectiveness of zoning regulation under interdependent markets would be a challenging task for …rm entry models, however, because the dimension of the choice set for a …rm grows exponentially in the number of markets. Empirical entry models, such as Bresnahan and Reiss (1990, 1991a,b), typically focus on geographically isolated markets, which are often rural markets, so we can conveniently treat the outcome from each market as a separate observation generated from the model when estimating the model. The literature has neglected the role of interdependence, largely because of the computational burden of solving multi-store …rms’store location choice problem. Accordingly, most of what we know is the aforementioned recent work on the e¤ect of zoning regulations on …rm entry that use data from isolated and local markets. This neglect is unfortunate because zoning regulations, by their nature, often cover urban markets, and unlike rural markets, urban markets are rarely isolated either in demand or costs. This paper is the …rst to study how regulations a¤ect entry in the context of multi-store …rms’entry into geographical markets, allowing for the decision in a market regarding the number of outlets to depend on decisions in the adjacent markets. Although some of the previous work captures the industry dynamics this paper abstracts, such as Suzuki (2013), this paper incorporates interdependent choices across markets that previous work abstracts away. This paper tackles this issue by formulating the game as supermodular, so we can reduce the computational burden of computing a Nash equilibrium. This paper focuses on the e¤ect of entry restriction, one of the key instruments that regulation authorities exercise, along with price and quantity, on market outcome (Viscusi, Vernon, and Harrington 2000). In particular, this work investigates how the City Planning Act in Japan enacted in 1968 a¤ects the entry of retail chains’outlets. The unique feature of this act in Japan is that the law imposes two distinct types of zoning restriction on the development of small-size retail outlets. Previous work has focused on either the …rst or the second type. The …rst zoning restriction of the City Planning Act is active in areas under “Category 1: Exclusive Low Rise Residential Zone”and “Category 12: Restricted Industrial Zone.” Category 1 applies to a part of urban areas, and is for residential spaces. Accordingly, developing a retail outlet, such as a convenience store, is prohibited to maintain the quietness in these residential areas. Category 12 applies to another part of urban areas, which are solely for industrial activities. Hereafter this paper terms the regulation under 3

these two categories as type 1 zoning. The second zoning restriction is imposed on an area called “Urbanization Control Area”(hereafter type 2 zoning), of which markets are mutually exclusive to those of type 1 zoning. In this area, each outlet needs to submit an application to its governor to obtain permission to show building an outlet is in accordance with the government’s urban planning, such as preservation of farm land, scenery, or natural environment. Given these two zoning restrictions, this paper evaluates the costs of zoning regulations by two measures. The …rst measure is to examine hypothetically how the revenue and pro…ts would change if the regulation authorities were to abolish these two zoning restrictions. The second measure is to identify from the data the additional compliance cost for developing an outlet in type 2 zoned markets. This paper uses the empirical model and the data from Nishida (2013), which extends Jia (2008) and estimates an empirical model of strategic network choice with post-entry outcome in the convenience-store industry in Okinawa. This paper re-estimates the model by incorporating the data on type 1 zoning restriction, which have recently become publicly available at the government’s Geographic Information System (GIS) website. Using the parameter estimates, this paper simulates the e¤ect that hypothetically eliminating these two zoning regulations would have on market outcome. The convenience-store industry is suitable for analyzing how zoning a¤ects market outcome of retail chains for two reasons. The …rst reason is policy oriented: eliminating or relaxing the existing zoning regulations introduced in 1968 has been an urban policy issue for commercial activities, including convenience stores. For instance, the Japan Business Federation (“Keidanren”) and the Council for Regulatory Reform have requested deregulation several times during the 2000s, emphasizing the bene…ts of having more convenience stores in type 1 and type 2 zoned areas are economically signi…cant (Keidanren 2002; Council for Regulatory Reform 2004).1 Second, both retail trade areas and zoned areas are usually geographically contiguous rather than discrete or isolated. This feature makes our focus on the convenience stores relevant because conveniencestore outlets are typically located densely and contiguously in urban (and some rural) markets. These outlet location decisions within a chain are dependent across markets both in costs and demand. This complexity of location choices and geographically contiguous zoning regulations call for an equilibrium analysis of multi-store …rms deciding entry into multiple markets, in contrast to single-store …rms. 1

For Council for Regulatory Reform kaikaku/old/minutes/wg/2004/1118/item041118_02.pdf

4

(2004),

see

http://www8.cao.go.jp/kisei-

The estimated model …nds that hypothetically removing the existing type 1 and type 2 regulations would increase the total number of outlets in Okinawa for each convenience-store chain by approximately 10%

14% and 2%

3%, respectively. The costs of type 2 zoning regulation for

an outlet entering a geographical market amount to approximately US$46; 000 per year and outlet. Although the increase in the total number of outlets due to deregulation is of similar magnitude across chains, the markets in which the number of outlets would increase due to removing type 2 zoning regulation would di¤er across chains. These …ndings are robust to plausible alternative speci…cations. Broadly speaking, this paper builds on the literature of the role of regulation, which has generated much attention in economics (see, for instance, Stigler 1971; Peltzman 1989). This paper contributes to the literature that examines the relation between regulation and competition (Joskow 1973; Samprone 1979). In particular, this paper is related to a strand of the literature that examines the e¤ect of regulation on the margin of …rm entry (Schaumans and Verboven 2008; Cohen and Mazzeo 2010; Ryan 2012; Nishida and Gil forthcoming).2 This paper is also linked to the literature on examining how zoning regulations a¤ect various economic outcomes, such as entry of retailers (Lewis 1945; Basker 2007), productivity of retailers (Schivardi and Viviano 2011; Haskel and Sadun 2012), employment in retailing (Bertrand and Kramarz 2002), and land and housing prices (Pogodzinski and Sass 1990; Glaeser and Gyourko 2002; Glaeser, Gyourko, and Saks 2005; Quigley and Raphael 2005; Quigley and Rosenthal 2005). The rest of the paper is organized as follows. Section 2 provides the details about the zoning regulations in Japan. Section 3 describes the data and the convenience-store industry. Section 4 outlines the equilibrium network-choice model and estimation method. Section 5 reports the parameter estimates and two counterfactuals: eliminating type 1 and 2 zoning regulations. Section 6 concludes.

2

The City Planning Act of 1968

This section provides the institutional details regarding the zoning regulations this paper exploits. In 1968, Japan’s government introduced the City Planning Act, comprehensive city planning regulations at the national level, including zoning regulations. The purpose of this act is to ensure cities’sound and orderly developments. Table 1 summarizes how this City Planning Act a¤ects the 2

See Ferrari and Verboben (2010) for the reference therein.

5

development of small-size retail outlets with ‡oor space of less than 150m2 , such as a conveniencestore outlet.34 The Urbanization Promotion Area is an area in which the government prioritizes urbanization by constructing public facilities, such as water, gas, and electricity. This Urbanization Promotion Area is broken into 12 mutually-exclusive geographical areas, and developing a small-size outlet with ‡oor space of 150m2 is per se illegal in geographical areas under Category 1 (Exclusive Low Rise Residential Zone) and Category 12 (Restricted Industrial Zone). By prohibiting the development of an outlet, the government aims to maintain the amenity of these areas, such as quietness for residential areas.5 This prohibition is the …rst type of zoning regulation this paper studies (type 1 zoning). By contrast, column 1 in Table 1 shows that in other areas in Okinawa, including the remaining 10 categories in the Urbanization Promotion Area, developing a small-size retail outlet is not per se illegal. The second zoning restriction imposes a development permission system. Column 2 in Table 1 shows that in the Urbanization Control Area, where the government aims to prevent urban sprawl and disorganized urbanization in accordance with the government’s urban planning, such as preservation of farm land, scenery, or natural environment, developing a retail outlet requires permission from the local government (type 2 zoning). To prevent undesirable development actions, the law requires a developer to apply for permission from the governor of the prefecture or the city to build a new residential home or a commercial facility, such as a convenience store, demanding the applicant prove the establishment will not violate the urban planning in that area. Note that this regulation, unlike type 1 zoning regulation, does not prohibit the development of convenience-store outlets; rather, the act permits the development of outlets in zoned areas, provided the submitted outlet development plan complies with strict construction requirements. To develop a small-size retail outlet in the Urbanization Control Area, such as a convenience store, the developer of the outlet needs to document three things: (1) the outlet serves local people, (2) the outlet provides daily necessities for the people living in that Urbanization Control Area, and (3) the outlet maintains minimum proximity to residential areas or maximum ‡oor space.6 Note type 1 and type 2 zoning regulations are geographically mutually exclusive as shown in Figure 1. 3

A typical convenience-store outlet has a ‡oor size of 110m2 on average. One might be concerned zoning regulations are endogenous: the appendix provides discussions on the assumption of exogeneity of zoning regulations. 5 Building Standard Law Article 48-1. 6 An exception available for small-size retail outlets in some prefectures in Japan is that under Article 34-8 of the City Planning Act of 1968, the development of the retail outlet is permitted if the store serves tra¢ c drivers on major roads at roadside rest facilities under some conditions. However, in Okinawa, the prefecture government does not allow this exception; therefore, I am not going to attend to this exception. 4

6

In Okinawa, 7% of the total population lives in the Urbanization Control Areas, 85% lives in other City Planning Area, and the remaining 8% lives in the Non-City Planning Area. The extent of coverage of the population by the City Planning Area is substantial: this area accounts for roughly 90% of the population in Japan.7

3

Data and Descriptive Analysis on E¤ect of Zoning Market De…nition. Unlike the traditional market de…nition in entry models based on iso-

lated markets, this paper adopts a framework in which adjacent markets a¤ect the pro…ts of a market through the demand and cost spillovers and the business-stealing e¤ect from outlets in adjacent markets. To deal with contiguous markets often observed in urban areas, this paper de…nes a market as a uniform grid and allows eight adjacent grids to a¤ect a grid’s pro…t. This paper chooses the 1kilometer (= 0:6mile) square grid that the Census Bureau sets for two reasons. First, the data variables from the government that are relevant for this study, such as demographics and outlet sales information, use this 1km grid and border as their unit of the data. Second, various surveys suggest a consumer’s average travel time to a convenience store is around 10 to 20 minutes by walking, and a trade area for a typical convenience store has a radius of about 500 meters (=546 yards) to 700m. The next section discusses how the model deals with the demand spillovers, namely, how the population in adjacent markets a¤ects the revenue in a given market through people traveling across grid borders. Applying this grid de…nition, Okinawa island has 1; 201 mutually exclusive grids. I exclude 367 markets (i.e., 367 grids) with zero population both in the number of residents and workers, leaving us 834 markets. Of 834 markets, the number of type 1 zoning grids (markets) and type 2 zoning grids (markets) is 18 and 140, respectively, and the remaining 676 (=834

18

140) grids are non-zoned

markets. These type 1 zoning markets and type 2 zoning markets are mutually exclusive; there is no market in which both type 1 and type 2 zoning regulations apply. When estimating parameters, I exclude these 18 markets under type 1 zoning regulation, because developing a convenience-store outlet in those markets is prohibited. This exclusion leaves us 816 (= 834

18) markets for

estimation. If a grid contains both non-zoned and zoned areas (either type 1 or type 2), the market is non-zoned if the size of the zoned area is less than 50% that of the 1km square grid. As Figures 7

The Urbanization Promotion Area, Urbanization Control Area, and Undelineated Area account for 15%, 37%, and 48% of the City Planning Area in Japan, respectively.

7

1 and 2 show, whereas type 1 zoned markets are near the city center where population density is high, type 2 zoned markets are more likely to be suburban areas surrounding the city center. The Convenience-Store Industry in Okinawa and Data Sources. The conveniencestore industry in Okinawa has two players, Family Mart and LAWSON, that design optimal store networks taking into account their competitor’s store-network con…gurations. These two chains are nationwide convenience-store chains, and employ uniform pricing across outlets in Japan. Within the 816 markets used for estimation, Family Mart and LAWSON had 127 and 95 outlets in 2002 in Okinawa. The number of local stores is 125.8 In this industry, increasing the density of outlets in a given market reduces costs, because supply of items to each outlet is heavily dependent on delivery trucks that deliver perishables two to three times a day to maintain the food as fresh as possible. Accordingly, the industry invests heavily in sophisticated distribution networks, and each chain has its own distribution center in Okinawa, from which a chain supplies the items for an outlet. The average sales per store, computed from each company’s …nancial statements in 2002, are US$1:43 million for Family Mart and US$1:45 million for LAWSON in Okinawa.9 I collect the 1km grid-level cross-sectional data from a variety of sources. The geographical map concerning type 1 and type 2 zoning regulations is available from the Ministry of Land, Infrastructure, Transport and Tourism.10 This paper obtains the number of outlets in a grid from the 2002 Convenience Store Almanac (TBC 2002), which provides the outlet-level street address information for every convenience store, which I converted into longitude and latitude coordinates. The aggregate revenue data at the 1km grid level are available from the 2002 Census of Commerce from the Ministry of Economy, Trade and Industry. The number of residents and workers at the 1km grid level come from the 2001 Census of Population and Establishment and Enterprise Census, respectively, both of which are provided by the Census Bureau. The data on land prices come from the Ministry of Land, Infrastructure, Transport and Tourism. The data contain location and land price of 215 geographically distinct points, and I treat the closest point from the centroid of a square grid (=market) as the market’s land price.11 8 The local stores are outlets from another chain, Hot Spar. In this study, I treat the Hot Spar stores as non-chain local stores for which locations store owners choose. I do so because this Hot Spar company originally started as a voluntary chain in Okinawa, and the company did not make coordinated store-location decisions. 9 I calculate these numbers by dividing the aggregate sales for each chain in Okinawa by the number of outlets for each chain in Okinawa. 10 http://nlftp.mlit.go.jp/ksj/jpgis/datalist/KsjTmplt-A09.html and http://nlftp.mlit.go.jp/ksj/gml/datalist/KsjTmpltA29.html, respectively. 11 Please refer to Nishida (2013) for a more detailed description of the variables used in this paper.

8

Figure 3 shows the location choices of both chains’ outlets and non-zoned areas and type 1 zoning areas at Central Naha, the capital city of Okinawa Prefecture, and the region surrounding Naha.12 This …gure suggests that outlets tend to locate randomly within a 1km square grids (i.e., markets), regardless of whether a grid is close to type 1 zoning areas.13 Descriptive Analysis on Type 2 Zoning Regulation. I now present a reduced-form analysis, which examines how demographics a¤ect outlet-opening decisions and measures whether the zoning regulation has a large in‡uence on market structure in the retail industry. Table 2 provides the results from the ordinary least-square regressions of the total number of multi-store …rm outlets in a market, both Family Mart and LAWSON brands, on the log of the market’s number of residents, and a zoning indicator that is one if the market is type 2 zoned and zero otherwise. Column 2 adds the log of the number of workers as a control. In both speci…cations, as expected, population either during the day (workers) or night (residents) in a market is positively associated with the number of outlets in the market. The results show the number of convenience-store outlets is negatively associated with type 2 zoning indicator variable. For example, being located in a type 2 zoned market is associated with the reduction in the number of outlets by around 0:3. Although reduced-form regressions are informative about the likely direction and strength of the e¤ect of type 2 zoning on entry, this paper employs an equilibrium model of entry for three reasons. First, a structural model allows us to recover the primitives of the cost structure, such as the cost of type 2 zoning regulation, which is the crucial variable for policy analysis. A regulator would like to quantify the costs due to type 2 zoning when the government wants to evaluate whether the costs are of an economically meaningful magnitude to o¤set the bene…ts from the regulation. Second, reduced-form regressions can be inadequate approaches for modeling …rm entry in concentrated industries, such as retailing. Examples of the features of the industry that the reduced-form analysis misses include the strategic interactions between Family Mart and LAWSON, headquarters’decision on the store networks, and so on. The number of outlets is an endogenous variable determined in the game because “Nature”does not randomly give the number of outlets in the right-hand side of the regressions; rather, …rms maximizing their pro…ts determine the number of outlets. A failure to account for all of these key features that characterize the …rms and the industry can lead to biased estimates. Finally, an advantage of structural modeling is that it allows 12

I am indebted to an unknown referee for suggesting to map outlet con…gurations and type 1 zoning areas. Given that the grid borders are exogenously imposed by the Bureau of Census, this seemingly random location choice within a grid seems natural. 13

9

us to conduct realistic out-of-sample predictions about changes in market structure due to zoning policy changes once we uncover basic model parameters. By predicting the change in geographic patterns of market con…gurations by two chains due to a change in zoning regulations, we can look at distribution consequences of policy interventions, such as who does and does not bene…t from a change in zoning. We can also ask practical questions a regulator may …nd relevant, such as in which markets we would …nd the impact of deregulation to be most e¤ective on entry behavior.

4

Empirical Approach

This section describes the empirical model and methodology. The main objective in estimating the …rm entry model is to understand what drives store-network choice by two multi-store retailers. Empirical Model. The market structure is determined by the strategic actions of two players choosing a player’s store network in equilibrium. The empirical model is a static simultaneousmove game of complete information by two players.

Player i and player j, i; j 2 fF amily

M art; LAW SON g, choose their number of outlets in every market, Ni and Nj , respectively. Namely, a store network for a chain is a vector of the number of markets where the dimension M is the total number of markets. Player i maximizes the total pro…t function,

i (Ni ; Nj ),

by

choosing its store network, Ni = (Ni;1 ; :::; Ni;M ). The total pro…t function is the sum of the market-level pro…ts,

i;m (Ni ; Nj ),

per-store pro…ts in market m;

i;m ,

across markets. The market-level pro…ts in market m are the multiplied by the number of stores in market m, Ni;m . To sum

up, i (Ni ; Nj )

=

M m=1 [Ni;m

i;m (Ni ; Nj )]:

We decompose …rm i’s per-outlet pro…t function into per-outlet revenue and costs as ri;m (Ni ; Nj )

i;m (Ni ; Nj )

=

ci;m (Ni ), where ri;m (Ni ; Nj ) is the per-outlet revenues and ci;m (Ni ) is the per-outlet

costs for …rm i stores in market m. Firm i’s per-outlet costs in market m are modeled as a function of the type 2 zoning status in the market, the number of …rm i’s outlets, distance to each …rm’s

10

distribution center, market characteristics, and unobserved cost shocks: ci;m (Ni ) =

|

1(market m is type 2 zoned) {z }

(1)

compliance costs due to type 2 zoning regulation

+

within log(max(Ni;m ; 1))

|

+

across

X

{z

+

Di;l

l6=m

}

| distance {z

costs due to distance to distribution center

cost savings from stores within a market and adjacent markets

+

| land

pland;m + {z }

cost of rent

cost |{z}

common …xed costs

+

|

q 1

2(

di;m }

2 c 2 m

{z

+

c 2 i;m ) :

}

cost shocks

The parameter of interest is the …xed costs due to type 2 zoning regulation, parameterized by . This parameter captures the increase in the compliance costs the outlet may incur in obtaining permission to develop an outlet in a type 2 zoned market. In contrast to Schaumans and Verboven (2008), who analyze the restricted entry in pharmacies and physicians by modeling the entry restrictions as strictly binding constraints to outcome, this paper builds on a free-entry model and incorporates the e¤ect of type 2 zoning regulation by an increase in the …xed costs. I do so because the law does not impose a stringent upper bound on how many outlets can exist in a market. and

across

within

capture how the presence of outlets in the same market and eight adjacent markets

reduces the outlet-level costs in market m through economies of density, respectively. Note that the speci…cation re‡ects two cost spillovers that make a chain’s decision regarding the number of outlets in a market dependent on the presence of outlets in adjacent markets: First, having more outlets in adjacent markets reduces a …rm’s costs in a market because the …rm can save on costs in several dimensions, including gas for delivery trucks or the costs of advertising in newspapers (“economies of density” parameter

within ).

Second, the as the log distance to the distribution

center increases, the outlet-level costs also increase (

distance ).

Di;l is a dummy variable that equals

one if at least one …rm i’s store is in market l and 0 otherwise.

land

captures how rent in market

m, proxied by the land price in market m, a¤ects the market-level costs of an outlet in market m.

c m

and

c i;m

are the market-level and chain-market-level cost shocks, respectively, which this

work assumes i.i.d. across markets and markets and chains, respectively.14

2

and

2

measure the

correlation of combined unobservables across …rms in a given market and the magnitude of the sum of the cost shocks. 14

I maintain this restrictive distributional assumption to identify the model parameters regarding demand and cost spillovers.

11

Firm i’s per-outlet revenues in market m are modeled as a function of demographics, the number of outlets of own and rival chain, and unobservable revenue shocks: ri;m (Ni ; Nj ) =

Xm | {z }

+

demographics

+

+

+

own;within log(max(Ni;m ; 1))

|

+ 1) +

{z

rival;across

local;within log(Nlocal;m

+ 1) +

local;across

;within

and

;across

Dj;l

Di;l }

{z

X

}

Dlocal;l

l6=m

business-stealing e¤ect from local stores

q 1(i is LAWSON) + 1 ( 1 {z } |

brand …xed e¤ect, LAWSON

In this speci…cation,

X

l6=m

{z

| LAW SON

X

l6=m

business-stealing e¤ect from rival chain stores

|

own;across

cannibalization, or business-stealing e¤ect from own chain stores

rival;within log(Nj;m

|

+

2 r 1 m

{z

+

}

r 1 i;m ) :

revenue shocks

}

captures how demographics a¤ect the outlet-level revenues in market m. capture how the presence of outlets in the same market and adjacent markets

a¤ects the outlet-level revenues in market m, respectively. I use max(Ni;m ; 1) for the cannibalization e¤ect within a market for own chain because an outlet would have no own business-stealing e¤ect within a market when there is no other outlet of the same chain brand in the market. Dj;l and Dlocal;l are de…ned similarly to Di;l .15

LAW SON

is the brand …xed e¤ect for LAWSON. I apply

the same assumptions for the market-level and chain-market-level revenue shocks did for cost equation.

1

and

1

r m

and

r i;m

as I

measure the correlation of combined unobservables across …rms in

a given market and the magnitude of the sum of the revenue shocks, respectively. Computation of Nash Equilibrium and Estimation.

The solution concept is pure-

strategy Nash equilibrium, which is a pair of store networks that are best responses. However, searching for a Nash equilibrium is often computationally intractable because the total number of possible outcomes of the game is at the order of K 2M , where K is the number of choices (i.e., how many outlets a …rm can develop in a market), and M is the total number of markets. In this empirical application, K = 5 and M = 816, so the number becomes 51632 . To deal with this 15

I use this non-log speci…cation for competition e¤ects from adjacent markets because one of the necessary conditions for supermodularity of the game is not met in the speci…cation in which the number of stores in adjacent markets a¤ect the outlet-level revenue in logs. I tested an alternative speci…cation in which the per-store sales decline in the total number of stores in adjacent markets. Under this speci…cation, the game is supermodular and the parameter estimates are similar to the baseline speci…cation of this paper.

12

computational burden of computing an equilibrium, I formulate the game as a supermodular game, which has several convenient features. Two such features are the existence of pure-strategy Nash equilibria and a round-robin algorithm for computing a Nash equilibrium via iterations of myopic best responses. The algorithm can signi…cantly reduce the burden of the most computationally challenging step, which is to calculate the myopic best response given the competitor chain’s entry con…guration. The algorithm computes two equilibria, the most pro…table equilibrium for Family Mart and the most pro…table one for LAWSON, respectively. This paper adopts the equilibrium that maximizes the pro…ts for Family Mart because Family Mart is larger than LAWSON in sales and store counts. This paper estimates the model by choosing model parameters that minimize the di¤erence between the observed data and the outcomes the model predicts. Because the model does not yield a closed-form solution for the equilibrium number of outlets and revenues for a given parameter , I employ the method of simulated moments (MSM) to estimate the parameters of the model. Namely, for a given set of simulated draws regarding the pro…t shocks (

c , c , r , r ), m i;m m i;m

I solve the model

for the Nash equilibrium in the outlet network using the algorithms, and I repeat this computation process many times using di¤erent sets of the pro…ts shocks.16 I then treat the predictions of market outcome averaged across simulations as the corresponding moment conditions in the MSM.17

5

Results

This section discusses the parameter estimates and uses them to perform “what-if” experiments, namely, evaluating the impact of hypothetical changes in the zoning regulation status on the market structure. The model solves for each simulation using the same demographics and draw of the revenue and cost shocks that are used for estimating the model parameters. In each scenario, the model solves for the equilibrium that favors Family Mart. The prediction from the model is given by the equilibrium store network averaged across 200 simulations.

5.1

Parameter Estimates

Table 4 presents the parameter estimates from the two speci…cations of the model. Of interest is the coe¢ cient on the zoning status indicator in the …rst row in the upper panel, which is positive and 16 17

200 times in the empirical application. Nishida (2013) contains further details on the empirical model and estimation methodology.

13

precisely estimated. The sign implies being in type 2 zoned market increases the outlet’s …xed costs of operation, including the compliance costs and opportunity costs of going through the application and screenings. The monetary value of the compliance costs for an outlet translates to US$46; 000 per year, which is about 4% of the average annual sales of an outlet. The goodness of …t shown in Table 3 implies the model explains the data reasonably well.

5.2

Counterfactual: Eliminating Type 1 Zoning Regulation (Case 1)

To simulate the e¤ect of deregulating type 1 zoning regulation, I expand the markets available for chains to develop an outlet. Speci…cally, I let each chain choose the number of outlets not only in the 816 originally type 1 unzoned markets but also in the 18 markets that were originally type 1 zoned. I solve for the equilibrium number of outlets in 834(= 816 + 18) markets for these two chains using the baseline parameter estimates in Table 4. I then take the average of the outcomes from the model across simulations and treat that average as the prediction from the estimated model. The second column in Table 5 presents the results of this policy simulation (Case 1). Family Mart and LAWSON increase their total number of outlets by 13:5% and 10:0%, respectively.18 The sales increase due to eliminating type 1 zoning regulation is smaller than the increase in the store counts, namely, 14:6% and 12:3% for Family Mart and LAWSON, respectively, suggesting the average outlet-level sales in type 1 zoned markets (18 markets) are more than those in non-type 1 zoned markets (816 markets). The total pro…ts increase by 8:9% and 6:8% for Family Mart and LAWSON, respectively. The magnitude of the increase in the number of outlets in 18 type 1 zoned markets is signi…cant in that the resulting number of outlets exceeds the number of outlets in type 2 zoned markets, given that the 140 type 2 zoned markets are much larger in geographical space.19 Overall, the results con…rm the originally type 1 zoned markets have large sales and pro…ts potential when these markets are hypothetically deregulated.

5.3

Counterfactual: Eliminating Type 2 Zoning Regulation (Case 2 and 3)

To conduct the policy experiment of deregulating type 2 zoning regulation, I set the indicator function in Eq.(1), 1(market m is type 2 zoned), to zero in every market in the data. To conduct 18

The maximum number of outlets within the 18 originally type 1 zoned markets for Family Mart (LAWSON) is 1:70 (1:05), implying that the number of outlets for these 18 markets does not reach the upper bound for each chain, which is four outlets, after eliminating type 1 regulation. 19 Because a market is a 1km grid, the space of 140 type 2 zoned markets and 18 type 1 zoned markets translates into about 140km2 and 18km2 , respectively.

14

the opposite policy experiment in which a developer needs to apply for zoning approval in all 816 markets, I set this indicator function to one in every market. Columns 4 through 7 Table 5 summarize the key …ndings of these counterfactual experiments. Columns 4 and 5 present the results under the no-type-2-zoning-permission-system regime (Case 2). As one would expect from the positive sign and the magnitude of the zoning parameter

,

eliminating the current type 2 zoning regulation would moderately increase the number of outlets because opening an outlet would then be less costly: rows 1 and 4 of column 4 show that for Family Mart and LAWSON, we would expect about a 2:3%

2:6% increase in the total number

of outlets. Rows 3 and 6 focus on the change in the 140 originally zoned markets, and I …nd most of these increases in store counts are largely due to an increase in the number of outlets in the previously type 2 zoned markets. In fact, in those 140 type 2 zoned markets, the percentage increase in the total number of outlets is sizable: around 28:4% and 30:5% for Family Mart and LAWSON, respectively. The model also predicts the aggregate sales and pro…ts will increase by 1:1% to 1:6%, slightly smaller but overall proportional to the increase in the number of outlets. Regarding the total compliance costs due to the permission system, I de…ne the magnitude of the cost by multiplying type 2 zoning parameter

by the number of outlets in zoned markets. I …nd the

cost reduction associated with the deregulation for Family Mart and LAWSON is US$0:87 million, which is 1:3% of the aggregate pro…ts of Family Mart and LAWSON. The impact of type 2 zoning deregulation on costs is nontrivial, considering the number of outlets currently located in type 2 zoned markets is small: 11:5 outlets for Family Mart and 7:5 outlets for LAWSON. To illustrate this point, columns 6 and 7 in Table 5 show how much the opposite policy regime a¤ects the results (Case 3). Under the policy regime in which type 2 zoning regulation is in place in all 816 markets in Okinawa, I …nd the number of outlets, sales, and pro…ts would decrease substantially. For instance, the increase in the cost associated with regulatory compliance would be US$8:07 million, which constitutes 12:3% of the combined pro…ts of Family Mart and LAWSON. The number of outlets drop signi…cantly, too: around an 12

13% decrease for both …rms.

Figure 4 geographically illustrates the increase in the number of outlets before and after eliminating the current type 2 zoning policy. The right and left panels in this …gure con…rm that the increase in the number of outlets after eliminating the regulation occurs primarily at the border that divides type 2 zoned and unzoned markets. The markets predicted to have more outlets after the deregulation di¤er across Family Mart and LAWSON because their store networks before the deregulation and cost shocks di¤er. In particular, the …gure shows the previously type 2 zoned 15

markets in which the number of outlets increases due to removing type 2 zoning regulation tend to be adjacent to the markets in which each …rm has the existing outlets of its chain brand. Overall, the results con…rm that eliminating type 2 zoning regulation would provide a nonnegligible positive e¤ect on the number of outlets, sales, and pro…ts, and the estimated compliance costs are substantial compared to the total pro…ts.

5.4

Robustness Checks

This section explores two alternative speci…cations to examine the robustness of the results. Full Model with a Smaller Set of Variables.

One might be concerned some of the para-

meters from the baseline model in Table 4 that are not precisely estimated may drive the zoning policy exercise results in Table 5. To address this concern, I re-estimate the model by removing all variables from the baseline model that are not statistically signi…cant at the 5% con…dence level. The second column in Table 4 shows the estimated sign and the magnitude of type 2 zoning parameter

are reasonably similar to those of the baseline model in the …rst column. Probably

not too surprisingly given estimates are close across speci…cations, both speci…cations 1 and 2 yield results that are quantitatively and qualitatively similar to those in Table 5, indicating imprecisely estimated parameters do not drive the counterfactual results.20 I conduct the Wald test to check the signi…cance of the group of variables excluded in the robustness-check speci…cation. The test statistic is 44:66, which rejects the null hypothesis that the parameters of the excluded variables are jointly equal to zero at the 1% level, implying these excluded variables are jointly statistically signi…cant. E¤ects of Eliminating Type 2 Zoning Regulation: Non-Revenue Model. This robustness check evaluates how the conclusions from the baseline model with revenue regarding the e¤ects of type 2 zoning on entry hold up in a simpler model without revenue. Columns 1 and 2 in Table 6 provide the results of these counterfactual simulations. First, from rows 4 and 5 of column 1, in which I predict the …rst scenario of eliminating type 2 zoning regulation, we would expect roughly a 1:4% to 1:5% increase in the number of entering markets for both multi-store …rms. The direction of change in the number of outlets is reasonable if type 2 zoning is interpreted as an increase in …xed costs of entry. On the other hand, the last two rows in column 1 show the number of 20

Detailed results are available from the author upon request.

16

outlets in the second scenario, in which type 2 zoning restrictions are placed all over Okinawa. The model predicts the number of markets in which convenience stores are present decreases by about 7%. Although the non-revenue model predicts a smaller magnitude in the e¤ect of type 2 zoning policy on store con…gurations, the full- and non-revenue models are consistent in the directions of the predictions of the policy experiments.

6

Conclusions

This paper studies how zoning regulations a¤ect retail market outcome in the context of the convenience-store chains in Okinawa, Japan. Unlike previous work that uses independent and isolated markets to study the e¤ect of zoning regulation on market entry, this paper adopts gridtype market de…nition to allow for contiguous markets and employs an equilibrium model of storenetwork choice by multi-store retailers. The model’s feature of interdependency of decisions across markets allows researchers to estimate the model in urban markets where both demand and cost spillover exists across markets. The estimated model …nds eliminating type 1 zoning regulation signi…cantly increases the number of outlets (10%

14%), total sales (12%

15%), and pro…ts (7%

9%). This paper also …nds

eliminating type 2 zoning regulation would increase the total number of outlets for each chain by around 2:3%

2:6% in all markets, whereas the magnitude is somewhat larger for the total number

of outlets in originally type 2 zoned markets (28%

31%). The increase in the costs associated

with complying with type 2 zoning regulation translates to US$46; 000 per year and outlet. The results have several policy implications. First, the simulation analyses show eliminating the existing type 1 and 2 zoning regulations signi…cantly increases the sales and pro…ts through an increased number of outlets. Moreover, regulators can exploit the cost estimates as a lower bound on to which degree the existing regulations should provide bene…ts. Second, the markets in which we should expect the number of outlets to increase di¤er across chains, although the size of the increase will be similar. The government policies should recognize the di¤erence across chains in the change in the number of outlets has distributional policy implications, such as who does and does not bene…ts from a change in zoning regulations. This paper has two limitations. First, it focuses on the costs of the regulations and does not quantify the bene…ts in monetary units, such as the utility gained from quietness in residential areas (type 1 zoning) or the bene…ts from preserving farm land, scenery, or the natural environment (type

17

2 zoning). As Crew and Kleindorfer (2002) conclude, “In the eyes of economists, the objectives of deregulation are generally laudable if the idea is to obtain the bene…ts of competitive entry”and this paper serves as the …rst step toward obtaining overall welfare implications of zoning regulations. Second, this paper employs a static model of entry and lacks the dynamics of entry and exit. Although lack of the dynamics may not pose a serious issue for this particular application because in 2002, the industry was in a steady state in the number of outlets, the static model may miss some dynamic aspects of competition. Extending the store-network-choice model to accommodate dynamics is left for future research.

References Basker, E. (2007): “The Causes and Consequences of Wal-Mart’s Growth,” The Journal of Economic Perspectives, 21(3), 177–198. Bertrand, M., and F. Kramarz (2002): “Does Entry Regulation Hinder Job Creation? Evidence from the French Retail Industry,”The Quarterly Journal of Economics, 117(4), 1369–1413. Boylaud, O., and G. Nicoletti (2001): “Regulatory Reform in Retail Distribution,” OECD Economic studies, 32(1), 253–274. Bresnahan, T. F., and P. C. Reiss (1990): “Entry in Monopoly Markets,”Review of Economic Studies, 57(4), 531–553. (1991a): “Empirical Models of Discrete Games,”Journal of Econometrics, 48(1-2), 57–81. (1991b): “Entry and Competition in Concentrated Markets,” Journal of Political Economy, 99(5), 977–1009. Cohen, A., and M. J. Mazzeo (2010): “Investment Strategies and Market Structure: An Empirical Analysis of Bank Branching Decisions,” Journal of Financial Services Research, 38(1), 1–21. Council for Regulatory Reform (2004): Proposals for Deregulation of Zoning Regulation in Use District. Tokyo, Japan (in Japanese). Crew, M. A., and P. R. Kleindorfer (2002): “Regulatory Economics: Twenty Years of Progress?,” Journal of Regulatory Economics, 21(1), 5–22. Datta, S., and K. Sudhir (2013): “Does Reducing Spatial Di¤erentiation Increase Product Di¤erentiation? E¤ects of Zoning on Retail Entry and Format Variety,” Quantitative Marketing and Economics, 11(1), 83–116. Ferrari, S., and F. Verboven (2010): “Empirical Analysis of Markets with Free and Restricted Entry,” International Journal of Industrial Organization, 28(4), 403–406. Glaeser, E. L., and J. Gyourko (2002): “The Impact of Zoning on Housing A¤ordability,” Discussion paper, National Bureau of Economic Research. 18

Glaeser, E. L., J. Gyourko, and R. Saks (2005): “Why Is Manhattan So Expensive? Regulation and the Rise in Housing Prices,” Journal of Law and Economics, 48(2), 331–369. Griffith, R., and H. Harmgart (2008): “Supermarkets and Planning Regulation,” Discussion paper, CEPR Discussion Papers. Haskel, J., and R. Sadun (2012): “Regulation and UK Retailing Productivity: Evidence from Microdata,” Economica, 79(315), 425–448. Hotelling, H. (1929): “Stability in Competition,” The Economic Journal, 39(153), 41–57. Jia, P. (2008): “What Happens When Wal-Mart Comes to Town: An Empirical Analysis of the Discount Retailing Industry,” Econometrica, 76(6), 1263–1316. Joskow, P. L. (1973): “Cartels, Competition and Regulation in the Property-Liability Insurance Industry,” The Bell Journal of Economics and Management Science, 4(2), 375–427. Keidanren (Japan Business Federation) (2002): posal Reports for Economic Revitalization. Tokyo, http://www.keidanren.or.jp/japanese/policy/2002/026/part3.pdf.

Deregulation ProJapan (in Japanese),

Lewis, W. A. (1945): “Competition in Retail Trade,” Economica, 12(48), 202–234. Netz, J., and B. Taylor (2002): “Maximum or Minimum Di¤erentiation? Location Patterns of Retail Outlets,” Review of Economics and Statistics, 84(1), 162–175. Nishida, M. (2013): “Estimating a Model of Strategic Network Choice: The Convenience-Store Industry in Okinawa,” Discussion paper, Johns Hopkins University. Nishida, M., and R. Gil (forthcoming): “Regulation, Enforcement, and Competition: Evidence from the Spanish Local TV Industry,” International Journal of Industrial Organization. Peltzman, S. (1989): “The Economic Theory of Regulation after a Decade of Deregulation,” Brookings Papers on Economic Activity. Microeconomics, pp. 1–59. Pogodzinski, J. M., and T. R. Sass (1990): “The Economic Theory of Zoning: a Critical Review,” Land Economics, 66(3), 294–314. Quigley, J. M., and S. Raphael (2005): “Regulation and the High Cost of Housing in California,” American Economic Review, 95(2), 323–328. Quigley, J. M., and L. A. Rosenthal (2005): “The E¤ects of Land Use Regulation on the Price of Housing: What Do We Know? What Can We Learn?,” Cityscape, 8(1), 69–137. Ridley, D., F. Sloan, and Y. Song (2011): “Retail Zoning and Competition,”Discussion paper, Fuqua School of Business Working Paper. Ryan, S. P. (2012): “The Costs of Environmental Regulation in a Concentrated Industry,”Econometrica, 80(3), 1019–1061. Samprone Jr, J. C. (1979): “Rate Regulation and Nonprice Competition in the Property and Liability Insurance Industry,” Journal of Risk and Insurance, 46(4), 683–696. Schaumans, C., and F. Verboven (2008): “Entry and Regulation: Evidence from Health Care Professions,” The Rand Journal of Economics, 39(4), 949–972. 19

Schivardi, F., and E. Viviano (2011): “Entry Barriers in Retail Trade,”The Economic Journal, 121(551), 145–170. Seim, K. (2006): “An Empirical Model of Firm Entry with Endogenous Product-Type Choices,” RAND Journal of Economics, 37(3), 619–640. Stigler, G. J. (1971): “The Theory of Economic Regulation,” The Bell Journal of Economics and Management Science, 2(1), 3–21. Suzuki, J. (2013): “Land Use Regulation as a Barrier to Entry: Evidence from the Texas Lodging Industry,” International Economic Review, 54(2), 495–523. TBC (2002): Japan’s Convenience Store Almanac. Tokyo, Japan. Viscusi, W. K., J. M. Vernon, and J. E. Harrington (2000): Economics of Regulation and Antitrust. The MIT Press, Cambridge, MA.

Appendix: Potential Endogeneity of Zoning Regulations An ideal empirical model for measuring the impact of type 2 zoning regulation on entry would involve randomly assigning type 2 zoning restrictions to markets and comparing the outcomes across type 2 zoned and unzoned markets. In reality, however, conducting such social experiments is usually di¢ cult. This paper treats type 2 zoning regulation as exogenously given, which may be problematic if zoning decisions were made based on some unobserved (to the econometrician) market-speci…c factors, arising either from the demand or the cost sides, which also a¤ect the profitability of convenience-store outlets. Then one may mistakenly attribute observed outcomes, such as variations in the number of outlets across markets, to compliance costs of type 2 zoning regulation and not to systematic di¤erences in pro…tability across markets. As a result, the parameter estimates can su¤er from an omitted variable bias. A piece of anecdotal evidence mitigates the concern: conversations with a local regulator’s sta¤ has revealed that in practice, the decisions concerning where to assign type 2 zoned/unzoned area are based solely on conditions regarding population, and regulators do not consider the degree of commercial activity because such consideration would involve the di¢ cult task of predicting the amount of commercial sales in the near future. Including in the empirical model demographics at the market level, such as the number of residents and workers, will alleviate the concern regarding an omitted variable bias.

20

FIGURE 2 NUMBER OF RESIDENTS

FIGURE 1 TYPE 1 AND 2 ZONING AREAS IN OKINAWA

NOTE: Type 1 and 2 zoning areas are in black and

NOTE: Type 2 zoning area is marked with bold lines.

stripes, respectively.

FIGURE 3 OUTLET CONFIGURATIONS, 1KM SQUARE GRIDS, AND TYPE 1 ZONING AREAS

NOTE: The stars show Family Mart Stores and the circles show LAWSON stores. Type 1 zoning areas are in stripes. 21

FIGURE 4 INCREASE IN NUMBER OF OUTLETS AFTER TYPE 2 ZONING DEREGULATION FAMILY MART(LEFT) AND LAWSON (RIGHT)

NOTE: The counterfactual exercise employs the parameter estimates from the first column in Table 4.

22

TABLE 1 TWO ZONING REGULATIONS FOR DEVELOPING A RETAIL OUTLET WITH 150 SQUARE METERS Zoning Regulations

Area Category

Development of an outlet prohibited? ("Type 1 zoning'')

Development permission needed? ("Type 2 zoning")

Category 1: Exclusive Low-Rise Residential Zone

YES

NO

Category 12: Restricted Industrial Zone

YES

NO

(Other 10 categories in the Urbanization Promotion Area))

NO

NO

Urbanization Control Area

NO

YES

Undelineated Area

NO

NO

NO

NO

Urbanization Promotion Area

City Planning Area

Non-City Planning Area

NOTE: The City Planning Act of 1968. See Section 2 for the institutional details of the zoning regulations.

TABLE 2 DESCRIPTIVE REGRESSIONS

Variable

(1)

(2)

0.339

0.017

(0.018)***

(0.03)

Log number of residents (thousand people) Log number of workers (thousand people)

0.491 (0.038)***

Dummy for Urbanization Control Area ("Type 2 zoning")

-0.332

-0.264

(0 069)*** (0.069)***

(0.063)*** (0 063)***

0.300

0.420

R-squared

NOTE: * significant at the 10%; ** significant at the 5%; *** significant at the 1%. Standard errors in parentheses. Observations are 834 markets. The dependent variable is the aggregate of Family Mart and LAWSON outlets in a given market in 2002.

TABLE 3 GOODNESS OF THE FIT OF THE MODEL Variable

Data

Prediction

Std.dev

Number of Outlets Family Mart LAWSON

127

127.4

10.1

93

95 3 95.3

88 8.8

Number of Outlets in Adjacent Markets Family Mart LAWSON

867

834.6

72.3

620

629.5

57.0

1573863

1592421.8

112458.9

Annual Sales (thousand US dollars)

23

TABLE 4 PARAMETER ESTIMATES OF THE MODEL Specifications Variable in Cost Equation

Baseline

Sensitivity Check

Type 2 Zoned Market (γ )

45.82 (22.93)

46.98 (23.39)

Land Price (μ land )

1.03 ((1.49))

Constant in Cost Equation (μ cost )

828.44 (53.80)

816.70 (27.36)

Gross Cost Cost-Savings Savings Effect by Own Chain Store, Store within a Market (( α within )

126.61 126 61 (37.85)

132 98 132.98 (20.04)

Net N T Trade-off d ff ffrom Cl Clustering i Stores, S across Markets M k (κ across )

5 80 5.80 (1.60)

6.02 6 02 (1.78)

Distance from the Distribution Center (μ distance )

15.98 15 98 (4.15)

15 80 15.80 (1.63)

Correlation Parameter in Cost Shocks (ρ 2 )

0.02 (0.00)

0.01 (0.00)

Standard Deviation of the Unobserved Costs (λ2)

235.21 (18 36) (18.36)

216.06 (18.25) (18 25)

Baseline

Sensitivity Check

Nighttime Population

56.81 (3.16)

60.09 (3.24)

Daytime Population

69.36 (6 02) (6.02)

64.75 (8.28) (8 28)

Retail Sales

1.02 (0.42) (0 42)

1.00 (0.15) (0 15)

"Cannibalization" (Business-Stealing Effect by Own Chain Store), within a Market (δ own within )

-296.92 (35.14)

-311.41 (23.34)

"Cannibalization" (Business-Stealing Effect by Own Chain Store), across Markets (δ own across )

-27.13 (34.47) (34 47)

Business-Stealing Effect by Rival Chain Store, within a Market (δ rival within )

-260.06 (44 42) (44.42)

Business-Stealing Business Stealing Effect by Rival Chain Store, across Markets (δ rival across )

-1.22 1.22 (2.96)

Business-Stealing Effect by Local Store, Store within a Market (δ local within )

-26 23 -26.23 (31.52)

Business-Stealing B i St li Effect Eff t by b Local L l Store, St across Markets M k t (δ local l l across )

-3.18 3 18 (5.86)

Cost-Savings Effects

Other Parameteres in Cost Equation

Variable in Revenue Equation Demographics

Competition p Effects

-250.04 (23 46) (23.46)

Other Parameteres in Revenue Equation LAWSON Store Dummy (μ LAWSON )

4.51 (6.05)

Constant in Revenue Equation

293.95 293 95 (25.46)

297 11 297.11 (35.52)

Correlation C l i P Parameter iin R Revenue Sh Shocks k (ρ ( 1)

0 51 0.51 (0.01)

0.06 0 06 (0.01)

Standard Deviation of the Unobserved Revenues (λ 1 )

245.43 (27.92)

235.36 (12.92)

NOTE: Standard errors are in parentheses. Parameters are measured in thousand $US with the exception of ρ . Observations are 816 markets. The number of simulations used in the MSM estimation is 200. markets 200 24

TABLE 5 EFFECT OF TYPE 1 AND 2 ZONING REGULATIONS ON ENTRY, SALES, AND COSTS

Variable

Current: Out of 834 markets, 18 and 140 markets are type 1 and 2 zoned, respectively

Case 2: Eliminating type 2 zoning regulation in 140 markets

Prediction

%Δ

Prediction

%Δ

Prediction

%Δ

127.43 0 11.51

144.68 16.49

13.5%

130.76

2.6%

111.69

-12.4%

14.78

28.4%

11.50

-0.1%

95.29 0 7.47

104.78 10.07

97.52

2.3%

83.37

-12.5%

9.75

30.5%

7.43

-0.5%

$105.39 $0 $7.86

$120.83 $12.37

$107.10

1.6%

$97.63

-7.4%

$9.58

21.9%

$7.87

0.2%

$84.60 $0 $5.37

$95.02 $8.03

$85.83

1.4%

$78.24

-7.5%

$6.66

24.1%

$5.35

-0.4%

$34.96 $30.68

$38.08 $32.77

$35.44 $31.02

1.4% 1.1%

$31.42 $27.57

-10.1% -10.1%

$0.00

Δcosts -$0.87

$8.94

Δcosts $8.07

(in 18 originally type 1 zoned markets) (in 140 originally type 2 zoned markets)

Number of Outlets: LAWSON All 834 markets (in 18 originally type 1 zoned markets) (in 140 originally type 2 zoned markets)

(in 18 originally type 1 zoned markets) (in 140 originally type 2 zoned markets)

Type 2 Zoning Regulation Policy Simulation

Case 1: Eliminating type 1 zoning regulation in 18 markets

Number of Outlets: Family Mart All 834 markets

Annual Sales: Family Mart All 834 markets

Type 1 Zoning Regulation Policy Simulation

Annual Sales: LAWSON All 834 markets (in 18 originally type 1 zoned markets) (in 140 originally type 2 zoned markets)

Annual Profits Family Mart LAWSON

Costs of Type 2 Zoning Regulation (million US dollars) Family Mart and LAWSON $0.87

10.0%

14.6%

12.3%

8.9% 6.8%

Case 3: Imposing type 2 zoning regulation in all 816 markets

NOTE: Variables are aggregated to the level of Okinawa unless otherwise stated. For each simulation, I solve for an equilibrium number of outlets for each chain, using the parameters from the first column in Table 4. The number of local outlets and demographics for each market is held fixed throughout this counterfactual analysis. Annual sales and profits are in million US dollars.

TABLE 6 EFFECT OF CHANGING TYPE 2 ZONING REGULATION STATUS, PREDICTION FROM NON-REVENUE MODEL Model Prediction

Std.Dev

Number of Outlets, Family Mart (Data:127)

131.3

98.9

Number of Outlets, LAWSON (Data: 95)

96.2

138.7

133.2

%Δ 1.4%

97.6

1.5%

122.0

%Δ -7.1%

89.5

-7.0%

Policy 1: Eliminating type 2 zoning regulation in 140 markets Number of Outlets, Family Mart Number of Outlets, LAWSON Poilcy 2: Imposing type 2 zoning regulation in all 816 markets Number of Outlets, Family Mart Number of Outlets, LAWSON

NOTE: The number of simulations used in the MSM estimation is 200. 25