Spatial Emerging Patterns of Vacant Land in a Japanese City Experiencing Urban Shrinkage a Case Study of Tottori City

Urban and Regional Planning Review Vol. 4, 2017 | 111

Spatial Emerging Patterns of Vacant Land in a Japanese City Experiencing Urban Shrinkage A Case Study of Tottori City

Keisuke SAKAMOTO, Akiko IIDA, Makoto YOKOHARI

Abstract In response to the phenomenon of urban shrinkage that affects many Japanese cities, and its various associated social issues, the Japanese Government has implemented compact city policies as a part of urban planning. A compact city is characterized in Japan by dense and proximate urban areas that are linked by public transformation systems, but the feasibility of applying a single compact city model to different regional or local circumstances is susceptible. To understand the actual state of urban shrinkage, this study was aimed at examining spatial emerging patterns of vacant land in residential areas of mid-sized cities called “regional urban centers” in Japan where such patterns are still unknown. The study was conducted using a three-step procedure: a) identifying the overall trends of urban shrinkage in all Japanese regional urban centers and selecting a case study site b) constructing a detailed database of vacant land and five characteristics of the residential areas: age, orderliness, distance from the central area, convenience of public buses, and road width, and c) identifying the relations between the number of vacant lands and the characteristics of the residential areas by using a Poisson regression model. Tottori City, a regional urban center of Chugoku Region, was selected as the case study site. The results show that the convenience of public buses, one of the emphasized factors when the government made a compact city plan, had no relationship with the distribution of vacant land. It might be because Tottori residents highly depend on private cars and are little concerned with public transportations. Instead, a narrow road width appears to be one of the crucial hindrances of residential replacement. A high rate of vacant land then should be regarded not as problematic space but as a resource for reconstruction of residential areas, such as the widening of narrow roads.

Keywords: urban shrinkage, vacant land, regional urban center, residential area, compact city

Graduate School of Engineering, The University of Tokyo E-Mail: [email protected]

1. Introduction

Population decline in urban areas has been said to be a major factor associated with the process of urban shrinkage (Rybczynski and Linneman, 1999). A typical symptom of urban shrinkage is an increase in the number of vacant houses and the plots of vacant land. For example, in Philadelphia, U.S.A the population declined by approximately 200,000 residents while there was a parallel gain of more than 20,000 plots of vacant land between 1992 - 2010 (Wiechmann and Pallagst, 2012). Increases in vacant houses and vacant land can lead to social issues such as deterioration of the residential environment and inefficiency of public services (Wiechmann and Pallagst, 2012). In industrialized countries that have been facing population decline since the 1950’s many cities have suffered from urban shrinkage (Oswalt and Rieniets, 2007). In Japanese cities other than the metropolises: Tokyo, Osaka, and Nagoya, population loss has been a major concern too. Because these large cities offer better opportunities for education and employment, large numbers of people tend to migrate from smaller regional cities to the metropolises. In addition, population decline in the cities has been affected by a steady decrease in birthrates. The total fertility rate (1) in Japan was 1.41 in 2012, which is relatively low compared with other industrialized countries such as the U.S., U.K., and France (MHLW, 2014). Unless these two causes of depopulation are addressed, it is inevitable that regional cities in Japan will continue to suffer from urban shrinkage. This study is focused on urban shrinkage in mid-sized cities called “regional urban centers (2).” Of 578 cities in Japan (this total excludes the three metropolises), there are 42 regional-urban-center cities. These cities are a political and economic hub for their particular region and they have developed and expanded due to the inflow of population from neighboring villages, towns, and smaller cities. As neighboring municipalities have suffered depopulation and urban decay, the role of the regional urban centers have become increasingly important. However, these cities have also begun to experience population decline. In order to sustain the vitality of the regions, urban shrinkage in regional urban centers needs to be addressed. Responding to the challenge of urban shrinkage in the regions, the Japanese Government revised the Act on Special Measures concerning Urban Reconstruction in 2014 in order to promote a compact city policy. A compact city has three key characteristics in general: dense and proximate development, urban areas linked by public transport, and accessibility to local services and jobs (OECD, 2012), which are followed in the Japanese compact city policy (MLIT, 2014). Several Japanese researchers have suggested that a compact city should also provide environmental and financial benefits to its residents. Some cities in Japan have already incorporated compact city policies in their urban planning. However, there has still been doubt in the U.S., an advanced country suffering urban shrinkage, as to whether a single compact city model is applicable and feasible for all cities, because conditions vary with city and locality (Neuman, 2005). Moreover, the use of a single, comprehensive compact city model lacks empirical support (Gordon and Richardson, 2007). For example, Hollander (2010) suggests that rather than apply a general model, policies should be tailored to the specific environment or neighborhood. Considering this and recent discussion about the compact city policy in Japan (Nozawa, 2014), the approach adopted by the Japanese Government seems somewhat problematic in that it is insensitive to the specific context and conditions of all cities. To address the urban form of Japanese shrinking cities in the future, it is necessary to understand the actual state of urban shrinkage in neighborhoods of such cities.

Urban and Regional Planning Review Vol. 4, 2017 | 113

An increase in vacant land is a typical indicator of a neighborhood undergoing urban shrinkage. Previous research in Japan shows that the emerging patterns of vacant land are related to three particular residential area qualities: age, orderliness, and convenience of public train. Ohsawa, et al. (2009) examined the patterns of emergence and persistence of vacant land in residential areas of Kashiwa City, a suburban city of Tokyo, and found that both are related to the age of a residential area. Ujihara, et al. (2006) focused on Okayama City, a central city in the Chugoku Region, and found a relationship between the emergence of vacant land and disordered residential development. Sakai (2014) examined the causes of the increase/decrease in vacant land in suburban areas of Tokyo and Osaka, and suggested that good accessibility to urban centers by public train tend to decrease the plots of vacant land. However, previous research has been limited to metropolitan suburban cities and regional cities that still have population growth. Thus, it is unclear how the factors identified above impact on the increase in vacant lands in regional urban centers that continue to experience population decline. In particular, whereas public trains are a common means of transportation for intra-city traffic in metropolises or some large cities, they are mostly used for inter-city or urban–rural traffic in most regional urban centers, and public buses are instead the typical means of intra-city transportation there. On the contrary, unlike metropolises or other large cities where there is an adequate public transportation system and several central urban areas, daily transportation in most regional urban centers strongly depends on private cars and the centers have developed around a single central urban area. Thus, accessibility to the center is likely to be measured simply by the distance from the central area to neighborhoods. In relationships between the emergence of vacant land and neighborhood characteristics, road width has a possible influence, due to the difficulty in rebuilding. This is because the Japanese Building Standard Law stipulates that new housing developments must have a frontal road wider than 4.0 m, or there is difficulty in entrance and exit by cars. Hence, the purpose of this research is to identify the distribution of vacant land and the relationship between emerging patterns of vacant land and the five characteristics of residential areas: age, orderliness, distance from the central area, convenience of public buses, and road width of residential areas in a regional urban center that is undergoing urban shrinkage.

2. Method

First, the case study site was selected by using official statistical data. Urban shrinkage is identified by three key indicators: population growth/decline rate, the increase/decrease rate of the number of vacant houses, and the increase/decrease rate of the area of vacant land. These indicators were used for assessing which regional urban centers have been experiencing urban shrinkage as compared with the metropolises and other large cities (2). The cities were classified into eight categories according to these indicators. Second, a database of vacant land was constructed by using ArcGIS 10.2. For the purpose of this research vacant land was defined as: (1) unused land that attracts rubbish and debris or grows weeds, or (2) unused land with a ‘For sale’ sign or roped off. The location of all vacant land in residential use districts in the urbanization promotion area (UPA) was recorded by a portable GPS in September 2014. Every plot of vacant land was then matched with satellite images from Google Earth, and their locations were illustrated by ArcGIS 10.2. Vacant land was then classified into two types: abandoned and non-built. The

Urban and Regional Planning Review Vol. 4, 2017 | 114 abandoned-type vacant land was once occupied with housing, whereas the non-built-type vacant land has never been occupied after land development such as land readjustment project. The type of vacant land was determined by referring to time-series maps published by Zenrin Co. Ltd. and aerial photographs published by the Geographical Survey Institute in 1970, 1981, 1990, and 2004. The map published in 1970 were the oldest of all published by Zenrin. Other years were selected on the basis of the period in which a residential plot was not assumed to experience both building and demolition, and in which aerial photographs were taken by the government throughout the research area. The distribution of the two types of vacant land was illustrated by using ArcGIS 10.2. Third, a database of five characteristics of residential area, i.e., age, orderliness, convenience of public buses, distance from the central area, and road width was constructed using ArcGIS 10.2. The age of the residential areas being investigated in this study was identified by referring to all topographic maps in the research area published by the Geospatial Information Authority of Japan. Development in the residential areas was traced by using ArcGIS10.2 in order to specify when this took place. The orderliness of the residential areas was assessed based on the type of residential development project. These consist of two types: individual projects that are undertaken by the private sector (IP), and land readjustment project (LRP), which can be initiated by the public sector (LRPp) or by landloads (LRPl). Ujihara, et al. (2006) contends that IPs are disordered projects while LRPs are ordered. All the residential areas were classified according to the project types above by referring to the latest official data of the basic survey of city planning. Distance from the central area may be indicated by the distance from the main train station to the neighborhoods, because that station is the hub of both intra-city and intercity traffic and is usually near the main commercial and tertiary industrial area. The distance was then represented by concentric ring buffers in 100-meter intervals from the train station. The buffers were illustrated by using ArcGIS10.2. Considering accessibility to the central area, the convenience of public buses was assumed to be determined by one proportional and two inverse elements, by referring to the methods of a study applying the Huff model to the convenience of public trains (Muto & Okuda 2013). These elements are how many buses run to the main train station, how much the fare is, and how long it takes to travel from the bus stop to the station. If the attractiveness of bus stops is represented by multiplying these three elements, the convenience of bus transport might be formulated by summing the attractiveness of all bus stops within the neighborhoods:

n A = i (1) i f t i i

C A (2) j = ∑ i i

In Equation (1), Ai is the attractiveness of bus stop i in the study area, ni is the number of buses to the main train station per day from i, fi is the fare from i to the station, ti is the travel time from i to the station. In Equation (2), Cj is the convenience of public buses in neighborhood j. The location of bus stops was identified using Fundamental Geospatial Data provided by the Geospatial Information Authority of Japan. The numbers of buses, fares, and travel times were derived from information of the bus companies in the area.

Urban and Regional Planning Review Vol. 4, 2017 | 115

The width of all the roads in the research area was calculated based on the methods of Okuaki (2012) by using Fundamental Geospatial Data provided by the Geospatial Information Authority of Japan. The roads were then classified according to five widths, less than 3.5 m, 4.0−5.5 m, 6.0−8.5 m, 9.0−14.5 m, and > 15.0 m by referring to the standards of road types constructed by development activity in the City Planning Act. Generally, road width is fixed based on 0.5-m or 1.0-m intervals. Therefore, if a road width is 3.9 m, it is classified as a 4.0−5.5-m type using 3.75 m as the standard of whether the road is < 3.5 m or 4.0–5.5 m. Other types of road width were also established in this way. Subsequently, to construct neighborhood units for calculation, polygons of three characteristics, age, orderliness, and distance from the main train station, were intersected using ArcGIS 10.2. The convenience of public buses and five road widths were standardized by dividing by the area of each neighborhood unit and assigning to each as an attribute value. The number of two types of vacant land was also summed for each unit by regarding the centroid of each plot of vacant land as the point representative of its location. Fourth, relationships between the distributions of two types of vacant land and the five aspects of residential neighborhood (characteristics of age, orderliness, access to central area, convenience of public buses, and road width) were identified using a Poisson regression model referring the method of Sakamoto and Yokohari (2016). This model is generally formulated as

ln( ) x (3) λi = ∑βk k,i k

y λ i exp(−λ ) p( y | λ ) = i i (4) i i y ! i th In Equation (3), xk,i is the k explanatory variable in unit i, βk is the parameter of xk,i, and λi is the expectation value of the response variable in unit i. In Equation (4), p(yi|λi) is the probability that the measured value of the response variable will be yi in the case where λi is already known through Equation (3). In this study, however, the response variable, the number of each of the two types of vacant land, depends on the size of a residential neighborhood. Equation (3) was therefore modified as follows using the area of residential neighborhood as an offset:

λ ln( i ) = β x a ∑ k k,i i k

ln( ) ln(a ) x (5) λi = i +∑βk k,i k where ai is the area of the residential neighborhood in unit i. Estimation of the parameter is based on the maximum likelihood estimation method, as follows.

L({ }|{y },{x }) ( y ln ) (6) βk i k,i = ∑ i λi − λi i where L is the likelihood function, and the estimated value vector of {βk} is acquired when L reaches a maximum. Furthermore, the set of explanatory variables in each model was optimized by referring to Akaike’s Information Criterion to start with the full set of explanatory variables, i.e., all the characteristics of the residential neighborhoods above.

Urban and Regional Planning Review Vol. 4, 2017 | 116

3. Results

3.1. Selection of the research site Table 1 shows the categories of the metropolises, other large cities, and regional urban centers classified according to the three indicators of urban shrinkage. Category Ⅰ includes the special wards of the metropolises, Tokyo, Yokohama, Nagoya, and Osaka, which are characterized by growth in population ((i) +) and decreases in the number of vacant houses and vacant land ((ii) − and (iii) −). In contrast to the growing cities in Category Ⅰ, Category Ⅷ includes cities with decreasing population ((i) −) and increased numbers of vacant houses and vacant land ((ii) + and (iii) +). This implies that the cities in Category Ⅷ are most likely experiencing urban shrinkage. The specified cities in Category Ⅷ are Yamagata, Niigata, Nagano, Gifu, Tottori, Yamaguchi, and Kochi. From these cities, this study selected Tottori as the case study site. Because Tottori faces urban shrinkage in its residential areas, the Government revised its master plan in 2014 in order to implement compact city polices. Tottori is a regional urban center in the Chugoku region, located in Western Japan and it is the capital city of Tottori prefecture in the northeastern part of the region (Figure 1). The total population of the city was 197,391 in 2010 including a population of 137,300 in the urbanization promotion area (UPA), which is 31.3 km2 in area. Tottori is known for its historical artefacts – in particular its 17th century castle, the south of which today’s central business district has been developed. The main train station of Tottori, JR Tottori station, opened in 1908 and is located in the central business district. The UPA has expanded around JR Tottori station. One third of the UPA (10.1km2) is residential: this is our research area.

Table 1. Classification of Japanese cities according to three indicators of urban shrinkage

Categories Indicators* Metropolises and other large cities Regional urban centers

Special wards of Tokyo, Yokohama, Ⅰ (i)+ (ii)− (iii)− Nagoya, Osaka

Ⅱ (i)+ (ii)− (iii)+ Saitama, Kobe Otsu, Okayama

Mito, Utsunomiya, Toyama, Sapporo, Sendai, Fukuoka Ⅲ (i)+ (ii)+ (iii)− Kanazawa, Kumamoto, Miyazaki, Naha

Ⅳ (i)+ (ii)+ (iii)+ Chiba, Hiroshima Takamatsu, Matsuyama, Oita, Kagoshima

Ⅴ (i)− (ii)− (iii)− Nagasaki

Ⅵ (i)− (ii)− (iii)+ Fukui, Matsue

Aomori, Morioka, Akita, Maebashi, Nara, Kitakyushu Ⅶ (i)− (ii)+ (iii)− Kofu, Tsu, Wakayama, Tokushima, Saga

Yamagata, Niigata, Nagano, Gifu, Kyoto Ⅷ (i)− (ii)+ (iii)+ Shizuoka, Tottori, Yamaguchi, Kochi

* (i) Population growth/decline rates between 2005 and 2010 (Census in 2005 and 2010) (ii) Increase/decrease rates of vacant houses between 2008 and 2013 (Housing and Land Survey in 2008 and 2013) (iii) Increase/decrease rates of vacant lands between 2003 and 2008 (Basic survey on land in 2003 and 2008)

Urban and Regional Planning Review Vol. 4, 2017 | 117

Figure 1. Location of Tottori City

Composed of residential lots and vacant land, the area was categorized by land use and residential zones by referring to official land-use data. The data was verified by the research team by surveying the area to identify any gaps between the official data and the actual conditions (Policy Research Institute for Land, Infrastructure 2012).

3.2. Construction of a database An exhaustive survey revealed that the research area had 1,113 plots of vacant land. The average size of these plots is 3.4a and the total area of vacant land is approximately 37.6ha, which accounts for 4% of the total research area. Figure 2 shows images of plots of vacant land.

Figure 2. Plots of vacant land Figure 3. Distribution of vacant land

Urban and Regional Planning Review Vol. 4, 2017 | 118

The vacant land was classified as abandoned (once occupied), or non-built (land that has never been occupied). The total number of the abandoned-type vacant land is 412, that is, 37% of all vacant land, while there are 701 plots of non-built-type vacant land. This constitutes 63% of all plots of vacant land. The distribution of these two types of vacant land is illustrated in Figure 3. In the research area, the topographic maps were published by the Geospatial Information Authority of Japan in 1909, 1947, 1973, 1988 (partly 1985), and 2005 (partly 2004), so the UPA is classified into six areas according to the six periods of development: pre 1909, 1910−1947, 1948−1973, 1974−1988, 1989−2005, and 2006−2014 (see Figure 4). Figure 5 shows the area in hectares that was developed during each of these time frames. Prior to 1909, residential development centered around the castle built in the Edo era (1603− 1868) ― this now accounts for 22% of the total residential area in Tottori. From the end of the Meiji era until two years after the end of World War II, (1940−1945), there was minimal residential expansion with only a 0.37ha/year increase on average during this period. During 1948−1973, a period, which includes a time of rapid economic growth in Japan (1954−1973), the amount of residential land increased more than 30 times (11.3ha/year). The expansion of residential land intensified between 1974 and 1988, with the average increase rate of residential area per year measuring 15.9ha/year. The period from 1986−1991 was one of unprecedented economic growth in Japan although after the financial bubble burst in 1991 and was followed by a ten-year period of economic decline. This seems to be reflected in a decline in residential growth between 1989 and 2005, (13.4ha/year), with a further 50 percent reduction between 2006 and 2014 (7.1ha/year). Figure 6 shows that the research area was divided according to two types of development: IP and LRP. Residential areas developed by IPs account for 69% of the total research area while

Figure 4. Development project type in the research area Figure 5. Developed residential area of each period

Urban and Regional Planning Review Vol. 4, 2017 | 119

Figure 6. Development project type in the research area Figure 7. Developed residential area of each project

Figure 8. Distance from Tottori station Figure 9. Proportion of area according to distance from Tottori station

Urban and Regional Planning Review Vol. 4, 2017 | 120

Figure 10. Bus stops and routes in the residential district

Figure 11. Road width in the UPA Figure 12. Total road length and proportion of each road width

Urban and Regional Planning Review Vol. 4, 2017 | 121

31% have been developed by LRPs (Figure 7). Additionally, of the area developed by LRPs,

77 % was initiated by the public sector (LRPp), and 23% was initiated by landlords (LRPl). The distances from JR Tottori train station to various points in the research area are illustrated in Figure 8. These vary from 500m at the nearest point to the farthest point at 9,200m from the station. Figure 9 shows the proportion of the research area according to distance from the station. Approximately 90% of the research area is situated within 6,000m of JR Tottori station, with the exception of two isolated residential areas that lie between 6,000m and 9,000m. Figure 10 shows all the bus stops in the residential district and all direct bus routes to JR Tottori station. The total number of bus stops is 142, 119 of which pertain to at least one bus route to the main station. The travel time from a bus stop to JR Tottori station is 4–46 minutes, and the fare is 160–550 yen. The most frequent bus service at a stop is 88 buses per day, and the least frequent is two buses per day. All roads in the UPA were classified according to the five types of width (Figure 11 shows a graphic representation of this). Figure 12 shows the total kilometers for each road width as a proportion of the total number of roads. Roads narrower than 3.5 m, in front of which rebuilding is unauthorized based on the Japanese Building Standard Law, account for 23% of total road length, and approximately half of the roads in the research area are less than 5.5m wide (46%). The most common road width is between 6.0 and 8.5 m, which accounts for 36% of the total road length. The three layers of residential characteristics, age, orderliness, and distance from JR Tottori station, were intersected. Then, the research area was divided into 595 neighborhood units. The convenience of public buses and five road widths were standardized by dividing by the area of each neighborhood unit and assigned to each unit as an attribute value.

3.3. Parameter estimation by Poisson regression model Table 2 shows descriptive statistics of the variables based on all 595 neighborhood units. The following results were obtained by parameter estimation of the Poisson regression model on the basis of the dataset in Table 2. Table 3 shows results of parameter estimation for the abandoned-type vacant land as the response variable. The explanatory variables were the age of the area, distance from JR Tottori station, and density of road widths. According to the Wald 99% confidence interval, the area developed before 1909 and between 1948 and 1973, and road widths < 3.5 m and 4.0−5.5 m had relatively strong positive relationships with the number of the abandoned-type vacant lands, and relationships with distance from JR Tottori station and the road widths 6.0−8.5 m were negative. Table 4 shows the results of parameter estimation for the non-built-type vacant land as the response variable. The explanatory variables were age of the area, project type, distance from JR Tottori station, and density of road widths. According to the Wald 99% confidence interval, land readjustment projects by landlords and the distance from JR Tottori station had relatively strong positive relationships with the number of the non-built-type vacant lands, and relationships with all periods of development, land readjustment projects by the public sector, and road widths 4.0−5.5 m were negative.

Urban and Regional Planning Review Vol. 4, 2017 | 122

Table 2. Descriptive statistics of variables Variable Mean Std. Dev. Min Max Response Variable number of abandoned-type vacant land [place] 0.7 1.7 0.0 14.0 number of non-built-type vacant land [place] 1.2 2.6 0.0 28.0 Offset residential area [ha] 1.7 2.1 0.0 11.4 Explanatory Variable area developed before 1909 (dummy) - - - - area developed between 1910 and 1947 (dummy) - - - - area developed between 1948 and 1973 (dummy) - - - - area developed between 1974 and 1988 (dummy) - - - - area developed between 1989 and 2005 (dummy) - - - - land readjustment project by private sector (dummy) - - - - land readjustment project by landlords (dummy) - - - - distance from main train station [km] 3.6 2.1 0.5 9.2 convenience of public bus [service*10-4/yen*minute*ha] 0.4 1.5 0.0 15.4 density of road narrower than 3.5m [m/ha] 0.7 1.0 0.0 9.0 density of road between 4.0m and 5.5m width [m/ha] 0.7 0.9 0.0 8.6 density of road between 6.0m and 8.5m width [m/ha] 1.2 1.3 0.0 10.4 density of road between 9.0m and 14.5m width [m/ha] 0.3 0.6 0.0 8.7 density of road broader than 15.0m width road [m/ha] 0.1 0.3 0.0 2.0

4. Discussion

4.1. Relationship between rate of vacant land and neighborhood characteristics The results of Table 3 indicate that abandoned-type vacant land tends to be more predominate in residential areas which are older and which have a high proportion of narrow roads. With respect to residential area age, development before 1973 is probably related to a high rate of abandoned-type vacant land, and this appears consistent with a previous report that Japanese residential neighborhoods have a typical replacement date of about 50 years after development (Konagaya, 2004). This is plausible because in such old residential areas, landlords are likely to be older or have already died and house demolition is more likely. A possible explanation of the low estimated Z-Value of area developed between 1910 and 1947 is that the development area in this period was extremely small compared with other periods (Figure 5), which might be insufficient to accurately estimate the influence of age. Road width < 3.5 m, in front of which rebuilding is unauthorized based on the Japanese Building Standard Law, appeared remarkably related to a high rate of abandoned-type vacant land. However, widths of 4.0−5.5 m were likely also to have a positive relationship, which may indicate that residential areas with a high density of such narrow roads, where entry and exit by car is difficult, tends to have a low demand for residential replacement. In contrast, the distance from JR Tottori station and road widths 6.0−8.5 m had negative influences on the rate of abandoned-type vacant land. For widths 6.0−8.5 m, which permits resident entry and exit by car, this negative relationship possibly reflects its favorable impacts on the demand for residential replacement. The proximity to JR Tottori station may also have

Urban and Regional Planning Review Vol. 4, 2017 | 123

Table 3. Parameter estimation of abandoned type vacant land

Variable Coefficient Std. Error Z-Value (intercept) -6.381 ** 0.368 -17.353 area developed before 1909 (dummy) 1.164 ** 0.335 3.472 area developed between 1910 and 1947 (dummy) 0.542 0.500 1.085 area developed between 1948 and 1973 (dummy) 0.969 ** 0.341 2.841 area developed between 1974 and 1988 (dummy) 0.367 0.373 0.984 area developed between 1989 and 2005 (dummy) -0.581 0.420 -1.384 distance from main train station [km] -0.102 ** 0.039 -2.602 density of road narrower than 3.5m [m/a] 0.351 ** 0.063 5.586 density of road between 4.0m and 5.5m width [m/a] 0.223 * 0.102 2.195 density of road between 6.0m and 8.5m width [m/a] -0.556 * 0.269 -2.070 density of road between 9.0m and 14.5m width [m/a] 0.682 0.354 1.924 degrees of freedom 584 residual deviance 438.66 ** 0 is not included in 99.9% Wald confidence interval * 0 is not included in 99% Wald confidence interval

Table 4. Parameter estimation of non-built type vacant land

Variable Coefficient Std. Error Z-Value (intercept) -3.628 ** 0.126 -28.893 area developed before 1909 (dummy) -1.990 ** 0.157 -12.681 area developed between 1910 and 1947 (dummy) -1.915 ** 0.456 -4.200 area developed between 1948 and 1973 (dummy) -1.489 ** 0.148 -10.061 area developed between 1974 and 1988 (dummy) -1.914 ** 0.149 -12.849 area developed between 1989 and 2005 (dummy) -1.026 ** 0.128 -8.033 land readjustment project by public sector (dummy) -0.287 * 0.117 -2.460 land readjustment project by private sector (dummy) 1.026 ** 0.114 9.035 distance from main train station [km] 0.069 ** 0.022 3.165 density of road between 4.0m and 5.5m width [m/a] -0.477 ** 0.108 -4.422 density of road between 9.0m and 14.5m width [m/a] -0.057 0.058 -0.993 degrees of freedom 584 residual deviance 941.64 ** 0 is not included in 99.9% Wald confidence interval * 0 is not included in 99% Wald confidence interval plus influences on residential demand and might cause the negative relationship, but another explanation is possible. That is, the price of land near the station is much higher than that of other areas, so vacant houses are rarely demolished. This interpretation is based on the Japanese property tax system, which reduces that tax on land to about one sixth (at most) if there is an existing house. It is then possible that the tax system promotes the persistence of vacant houses, especially in areas with a relatively high land price, and vacant land rarely emerges. In fact, referring to a supplement of the compact city plan in Tottori City (2016), vacant houses are densely located in the area near JR Tottori station. In order to verify the above assumption on the relationship between the emergence of vacant land and distance from the main station, it is necessary to analyze the distribution of vacant houses in the same way as the present research.

Urban and Regional Planning Review Vol. 4, 2017 | 124

The results of the parameter estimation imply that project types and the convenience of public buses have no relationship with the distribution of abandoned-type vacant land. Regarding project types, previous study has reported that the rate of this type tends to be low in residential areas as developed by LRPs, but such relationships were not observed in our study. The possible reason is that the earlier research took little consideration of the multicollinearity of the variables. It is evident that LRPs are related to the age of development or high density of road widths 6.0−8.5 m, because most LRPs were conducted after the Land Readjustment Act was established in 1954. This law requires that the roads developed by LRPs must be wider than 6.0 m in principle, so LRPs may have an indirect relationship with the low rate of vacant land because of the impacts of other related variables. Regarding the convenience of public buses, the results might reflect the difference of principal means of transportation. In metropolises like Tokyo, public trains are very important, and this explained the low rate of vacant land near train stations in some studies, but most mid-sized cities like Tottori are very dependent on private car transportation, so the convenience of public buses barely influenced the rate of abandoned-type vacant land. The results of Table 4 show that non-built-type vacant land was more common in new residential areas, because the coefficients tend to be larger in accord with the progress of the periods, and it seems natural considering new residential development is likely to be conducted after old residential areas were filled up. The distance from JR Tottori station also had a positive relationship with the rate of non-built-type vacant land, which may be explained as follows. The more distant an area is from the urban central area, the lower the land price is, so the scale of development tends to enlarge. It therefore takes a relatively long time for such distant areas to be filled in. The negative relationship with road widths 4.0−5.5 m may be connected with the low probability of building such relatively narrow roads in a developed area. This appears true of road widths < 3.5 m, but this variable is not shown in Table 4. This is likely because of a problem of the data used. That is, in some new developed areas, where there was a high rate of non-built-type vacant land, widening of roads has not been reflected yet by the latest official road data, so some widened roads were treated as original narrow roads of < 3.5 m in former agricultural areas. This might have affected the results of the analysis. Concerning the results about project types in Table 4, LRPs appear to have influenced the rate of non-built-type vacant land. However, the relationship with LRPp is negative, whereas the relationship with LRPl is positive. This result is probably explained as follows. The public sector has managed to sell out the developed area as much as possible to collect expenses for the project, so it is difficult for land to remain non-built, whereas some plots of land developed by landlords could be reserved for land uses other than housing in residential areas. This leads to the positive relationship with the rate of non-built-type vacant land. There were no relationships between the number of non-built-type vacant lands and the convenience of public buses, as with the abandoned type. This may be because developers consider bus transportation little in their development plans for some areas, because they probably know that Tottori residents strongly depend on private cars and are indifferent to public transportation.

4.2. Implications for compact city policies Figure 13 shows a compact city plan under consideration in Tottori published by the government in 2016. This plan regards JR Tottori station as the most important urban core and

Urban and Regional Planning Review Vol. 4, 2017 | 125 removes some residential areas in UPA distant from the station from residential induction areas, which the basis law of the compact city plan (the Act on Special Measures Concerning Urban Reconstruction) obliges municipalities to establish and under which the government expects dense populations. When establishing the compact city plan, the government took much account of the public transportation network. According to the policies, some residential areas like A in Figure 13, where public transportation is inconvenient, were removed from the residential induction area, while areas near bus routes like B in that figure were included. However, the results of the above parameter estimation suggest that the convenience of public buses has little influence on the choice as to where people live. This suggests that the public transportation-oriented compact city plan is difficult to accomplish unless the convenience of public buses is so improved that bus transportation would affect preferences as to resident location. In fact, both areas A and B, which are old and with a high density of narrow roads, have similarly high rates of abandoned-type vacant land (Figure 14). If the government manages to persuade people to live in particular areas like B, it should first take the narrow roads into consideration rather than public transportation. In particular, referring to the result in Table 3, road widths < 3.5 m, in front of which rebuilding is unauthorized based on the Japanese Building Standard Law, appears to be one of the crucial hindrances to residential replacement. The high rate of vacant land should thus be regarded not as problematic space but as a resource for reconstruction of residential areas. For example, Tsuruoka City, which has a slightly smaller population than Tottori in the Tohoku region, established an organization called the Tsuruoka Land Bank in 2013, which has been gathering abandoned houses and land to use for widening of narrow roads and other purposes (Hayasaka, 2013).

Figure 13. Compact city plan in Tottori (Tottori, 2016; legends added by the authors)

Urban and Regional Planning Review Vol. 4, 2017 | 126

Figure 14. Old residential areas of high density of narrow roads

5. Conclusion This study investigates the emerging patterns of abandoned-type and non-built-type vacant land in the residential areas of a Japanese regional urban center, Tottori City, and identifies the characteristics of neighborhoods where high rate of vacant land tends to exist or not. The results show that the number of abandoned-type vacant lands has positive relationships with old development and a high density of roads narrower than 5.5m, and negative relationships with distance from JR Tottori station and road widths 6.0−8.5 m. In contrast, the number of non-built-type vacant lands is positively related to new development, LRPs initiated by the public sector, and distance from the main station; it has negative relationships with LRPs initiated by landlords and road widths 4.0−5.5 m. Regarding the convenience of public buses, which was one of the emphasized factors when the Tottori government made its compact city plan, there were no relationships with either abandoned and non-built-type vacant lands. This suggests that that plan is unable to function unless the convenience of public buses is so improved that bus transportation would affect the decision as to where to live. The analyses also indicate that narrow road widths (< 3.5 m), in front of which rebuilding is unauthorized based on the Japanese Building Standard Law, are one of the crucial hindrances to residential replacement. The high rate of vacant land may thus be regarded not as problematic space but as a resource for reconstruction of residential areas, such as the widening of narrow roads. This study focused only on the emerging patterns of vacant land, but urban shrinkage is also characterized by other issues including increases in the number of vacant houses, residents’ preferences for where to live, and an ultra-aging population, thus future research should examine these matters.

Urban and Regional Planning Review Vol. 4, 2017 | 127

Acknowledgements We would like to thank Yorikazu Okagaki, Ryo Motobe, and Masashi Kawahara, officials of Tottori City and Tottori Prefecture, for their cooperation in providing us essential materials and their knowledge of the areas.

Notes (1) “Total fertility rate” is a demographic indicator, which represents the average number of children a woman gives birth to. (2) The Fourth Comprehensive National Development Plan (MLIT, 1987) defined “regional urban center” as: the principal cities of the prefectural governments, and the cities with a population larger than 300,000. It excludes cities in the three metropolises and the large cities that have a population above 1 million, such as Sapporo, Sendai, Hiroshima, Fukuoka, and Kitakyusyu.

References 1) Gordon, P., and Richardson, H.W. (2007). Are compact cities a desirable planning goal? Journal of the American Planning Association, 63 (1): 95-106. 2) Hayasaka, S. (2013). Small-scale chain type division reorganization enterprise of a vacant room owner’s good will. City Planning Review, 62(3): 26-29. (in Japanese) 3) Hollander, J. B. (2010). Moving toward a shrinking cities metric: Analyzing land use changes associated with depopulation in Flint, Michigan. Cityscape, 12: 133-151. 4) Konagaya, K. (2004). Toshi Keizai Saisei no Machidukuri (Town development for revitalization of urban economy) : Kokon Shoin, Tokyo: 248pp. (in Japanese) 5) Ministry of Health, Labour and Welfare. (2014). Vital statistics in Japan: Trends up to 2012. Tokyo: Statistics and information department, minister’s secretariat, Ministry of health, Labour and Welfare. 6) Ministry of Land, Infrastructure and Transport. (1987). http://www.kokudokeikaku.go.jp/document_archives/ayumi/25.pdf (accessed January 2015). 7) Ministry of Land, Infrastructure, Transport and Tourism. (2014). http://www.mlit.go.jp/en/toshi/city_plan/compactcity_network.html (accessed January 2015). 8) Ministry of Land, Infrastructure, Transport, Land and Water Bureau. (2010) Household survey on land: Basic survey on land: Report: Ministry of Land, Infrastructure, Transport, Land and Water Bureau, Tokyo 9) Muto, M., and Okuda, D. (2013). A quantitative defining method of station’s territories in the competing area of multiple railway companies. RTRI report, 27(2). (in Japanese) 10) National Institute of Population and Social Security Research. (2013). “Population projection for Japan (March, 2013).” http://www.ipss.go.jp/pp-shicyoson/j/shicyoson13/t-page.asp (accessed January 2015). 11) Neuman, M. (2005). The compact city fallacy. Journal of Planning Education and Research, 25: 11-26.

Urban and Regional Planning Review Vol. 4, 2017 | 128

12) Nozawa, C. (2014). Compact city ha Kurashiyasui Machi ni narimasuka? (Is compact city livable?) : Minohara, K., Fujimura, R., Aiba, S., Ubaura, M., Nakajima, N., Nozawa, C., Hino, N., and Murakami, A. (Eds). Korekara no Nihon ni Toshikeikaku ha Hitsuyou desuka? (Is urban planning necessary for future Japan?), Gakugei Shuppannsha, Tokyo: 256pp. (in Japanese) 13) OECD. (2012). Compact city policies: A comparative assessment: OECD, Paris: 292pp. 14) Ohsawa, H., Yokohari, M., and Amemiya, M. (2009). Relationships between topography and emergence and persistence patterns of vacant lots in suburban residential zones. Journal of the Japanese Institute of Landscape Architecture, 72(5): 683-686. (in Japanese) 15) Okuaki, K. (2012). Method to generate the linedata of the road centerline automatically along with the attribute data of width of road. Transactions of GIS Association of Japan, 21. (in Japanese) 16) Oswalt, P., and Rieniets, T. (2007). “Global Context. Shrinking Cities.” http://www.shrinkingcities.com/globaler_kontext.0.html?&L=1 (accessed January 2015). 17) Policy Research Institute for Land, Infrastructure. (2012). A study on the amount, generation and extinction of vacant land and the future use of vacant land, PRILIT research reports, 106 18) Rybczynski, W., and Linneman, P. D. (1999). How to save our shrinking cities. Public Interest, 135 (spring): 30-44. 19) Sakai, A. (2014). A research of the reality of generation and diminishment of vacant parcels and the mechanism in suburban housing area: Case studies of 2 suburban housing areas in Capital area and Kinki area. Journal of the City Planning Institute of Japan, 49 (3): 1035-1040. (in Japanese) 20) Sakamoto, K., and Yokohari, M. (2016). Dynamic characteristics of emerging vacant houses and abandoned lands in residential neighborhoods: A case study in the residential neighborhoods of Utsunomiya, Tochigi. Journal of the City Planning Institute of Japan, 49 (3): 1035-1040. (in Japanese) 21) Statistics Bureau of Ministry of Internal Affairs and Communication. (2005). “Population Census. 2005 Population Census.” http://www.stat.go.jp/data/kokusei/2005/ (accessed January 2015). 22) Statistics Bureau of Ministry of Internal Affairs and Communication. (2008). “Housing and Land Survey. 2008 Survey.” http://www.stat.go.jp/english/data/jyutaku/index.htm (accessed January 2015). 23) Statistics Bureau of Ministry of Internal Affairs and Communication. (2010). “Population Census. 2010 Population Census.” http://www.stat.go.jp/english/data/kokusei/index.htm (accessed January 2015). 24) Tottori City (2016). “Compact city plan (a draft), summarized version.” https://www.city.tottori.lg.jp/www/contents/1438220986947/activesqr/common/other/570f 5908012.pdf 25) Ujihara, T., Taniguchi, M., and Matsunaka, R. (2006). The reality of urban retreat: Reverse sprawl and its relation with residential zone types. Journal of the City Planning Institute of Japan, 41 (3): 977-982. (in Japanese) 26) Wiechmann, T. and Pallagst, K. (2012). Urban shrinkage in Germany and the USA: A Comparison of Transformation Patterns and Local Strategies. International Journal of Urban and Regional Research, 36 (2): 261-280.