Skip to main content

2023 | Buch

Recent Advances in Modeling and Forecasting Kaiyu

Tools for Predicting and Verifying the Effects of Urban Revitalization Policy

herausgegeben von: Saburo Saito, Kenichi Ishibashi, Kosuke Yamashiro

Verlag: Springer Nature Singapore

Buchreihe : New Frontiers in Regional Science: Asian Perspectives


Über dieses Buch

This book is the first comprehensive presentation of a Kaiyu Markov model with covariates and a multivariate Poisson model with competitive destinations. These two models are core techniques when the authors and colleagues conduct their Kaiyu studies. The two models are usually used to forecast the effects of specific urban redevelopment on both the number of visitors and consumer shop-around or Kaiyu movements. Their Kaiyu studies originated from the constructions of a Kaiyu Markov model and the disaggregated hierarchical decision Huff model almost simultaneously around the early 1980s. This book retrospectively reviews how these models have evolved from the start to the present state, and previews the ongoing efforts to make further extensions of these models. The extension of the Huff model started from the disaggregated hierarchical decision Huff model with shop-arounds. In retrospect, the model formulated the consumer’s simultaneous choice of destinations as a joint probability. The mechanism to determine this joint probability was a recursive conditional probability system. Now the Huff model has shifted from joint probability to multivariate frequency Poisson with competitive destinations. On the other hand, the Kaiyu Markov model started from a descriptive model. Because it cannot forecast changes in shop-arounds or consumer Kaiyu behaviors, the Kaiyu Markov model with covariates was developed in which entrance and shop-around choice probabilities are explained by the respective two logit models with covariates such as distances and shop-floor areas. The noticeable point is that it can explain consumers’ probability of quitting their shop-arounds. Thus, the model enables one to evaluate the effects of urban revitalization policy that promotes consumers’ shop-arounds or Kaiyu behaviors. Furthermore, if the Kaiyu Markov model can estimate the actual numbers of flows of consumers’ shop-arounds among shopping sites, the corresponding money flows also can be estimated as economic effects. This book discusses from scratch the evolution of all these topics. Thus this book provides the basics of the Kaiyu Markov model, a tutorial for the theory and estimation of the conditional logit model, and a chapter serving as a practical research manual for forecasting changes caused by urban development based on consumers’ Kaiyu behaviors.



Retail Models

Disaggregate Hierarchical Decision Huff Model Incorporating Consumer Kaiyu Choices Among Shopping Sites
A classical Huff model is extended to a disaggregated hierarchical Huff model, which incorporates the consumers’ multi-stage decisions concerning shopping destinations and shop-arounds. We represent the consumers’ multi-stage decisions as four stages; the first stage is a choice among destination cities, the second is between two alternatives of the set of large stores and small shops, the third is a choice of an individual destination shop, and the last is a choice to shop-around from those large stores chosen to the next destination shop. We formulated the extended model as a fully recursive joint probability model expressed as a multivariate log-linear model and estimated as disaggregate conditional logit models.
Saburo Saito
A Dynamical Huff Model: Computing the Competitive Equilibrium Distribution of Shop Floor Areas over a City Center Commercial District Using a Fixed-Point Algorithm
The equilibrium point of the dynamical retail model for Saga City is computed by Merrill’s method. The dynamical retail model endogenizes the shop floor areas of shopping centers so that its equilibrium point is considered a competitive spatial equilibrium of shop floor areas. The recent revision of the Large-Scale Retail Store Law in Japan, which follows the Japan-US Structural Impediments Initiative, suggests that a new urban commercial policy is required to establish some guidelines to allocate shop floor areas among city retailers from the viewpoint of city formation. The competitive spatial equilibrium of shop floor areas might serve as a guideline. This study compares the actual and the computed equilibrium of shop floor area distribution and suggests further research.
Saburo Saito, Hiroyuki Motomura

Kaiyu Markov Models

Kaiyu Markov Model and Evaluation of Retail Spatial Structures
Using an absorbing Markov chain model, we formulate consumers’ shop-around behaviors as the Kaiyu Markov model. Based on the Kaiyu Markov model, we conceptualize the shop-around effect to enable us to evaluate retail spatial structures, which we define as an increase in the frequency of visits to each commercial district due to consumers’ shop-around behaviors. We improved the traditional survey of consumers’ purchasing behaviors to collect the actual consumers’ shop-around behaviors and applied the Kaiyu Markov model to the data obtained from the improved survey conducted for the metropolitan area of Nobeoka City, Japan. We estimated the actual shop-around effects for the 17 commercial districts in the city and showed that the concept of the shop-around effects serves as a helpful tool for evaluating and planning the retail spatial structure of the city.
Saburo Saito, Toru Sakamoto
Basics of Kaiyu Markov Models: Reproducibility Theorems—A Validation of the Infinite Kaiyu Representation
The Kaiyu Markov model is formulated as an absorbing stationary Markov chain model in which consumers’ Kaiyu (shop-around) behaviors are expressed as an infinite process of state transitions in a stationary Markov chain. This chapter gives the basics of mathematical formulations of the Kaiyu Markov model. It proves reproducibility theorems that provide validity for representing consumers’ Kaiyu behaviors as infinite stationary state transitions.
Saburo Saito, Kenichi Ishibashi, Kosuke Yamashiro, Masakuni Iwami

Estimation of Huff Model and Kaiyu Markov Models with Covariates

Kaiyu Markov Model with Covariates to Forecast the Change of Consumer Kaiyu Behaviors Caused by a Large-Scale City Center Retail Redevelopment
It is common for a person with several purposes to start a trip from home and return home after visiting several places. This phenomenon is called a trip chain, which is likely to occur, for instance, in leisure, sightseeing travel trips, sales, or commodity transport trips. Among others, a shopping trip chaining behavior of a consumer occurs ubiquitously in an agglomerated commercial district. We call it consumer’s shop-around or Kaiyu behavior. The apparent cause is in the district’s accumulated and proximate locations of retail facilities. Thus, the consumer’s shop-around behavior can be considered the agglomeration effect of the locational configuration of retail facilities. Hence, their actual locational arrangement can be evaluated by such a criterion as what amount of the agglomeration effect, equivalently, the consumer’s shop-around or Kaiyu behavior the arrangement induces. Based on this standpoint, this study proposes an evaluative framework for assessing retail redevelopment programs in the city center retail environment. This study develops a stationary Markov chain model with covariates to forecast consumers’ shop-around or Kaiyu behaviors. The model was applied to the city center of Fukuoka City, Japan, and used to evaluate redevelopment programs there from its forecasts. Meanwhile, this study proves the observed aggregate stationarity theorem or reproducibility theorem to show that the aggregate stationary Markov chain modeling has a rigorous validity even if any arbitrary non-stationary process rules each disaggregate process.
Saburo Saito, Kenichi Ishibashi
Estimation of Disaggregate Huff and Kaiyu Markov Model: A Lecture Note on Conditional Logit Model
This chapter provides the fundamentals of theory and estimation of conditional logit models. Previous chapters have referred to the conditional logit model when estimating the hierarchical choice Huff model and the Kaiyu Markov model with covariates. However, those chapters focus on the new aspects of modeling and estimating the conditional logit models, such as a multivariate extension of logits, the equal treatment of destination choice, and quitting Kaiyu. Thus, understanding this uniqueness requires some background in the theory of conditional logit models. This chapter intends to fill this gap for those new to the theory and estimation of the conditional logit models. The method we employ to present the conditional logit model follows the original idea deeply rooted in random utility models. Some pedagogical devices for presentation are also included. Another characteristic of our presentation is to prove that the mean and variance of the double exponential distribution become Euler’s constant and one-sixth of π squared, which is rarely provided in standard texts.
Saburo Saito, Kosuke Yamashiro, Masakuni Iwami

Frequency-Based Retail Models for Forecasting the Number of Visitors and Sales of Commercial Districts

A Disaggregate Kaiyu Markov Model to Forecast the Sales of Retail Establishments Based on the Consumers’ Frequency of Visits
We have already proposed the frequency-based consumer Kaiyu Markov model to evaluate redevelopment programs in a city center retail environment from the viewpoint of consumers’ Kaiyu behaviors (The model referred to is the one proposed by Saburo Saito, Yoshinobu Kumata, and Kenichi Ishibashi, “A choice-based Poisson regression model to forecast the number of shoppers: Its application to evaluating changes of the number and shop-around pattern of shoppers after city center redevelopment at Kitakyushu City,” Papers on City Planning 30, 1995). This frequency-based model was an excellent one in the sense that the model enables one to account for the combined effects of the increase of attracted visitors and the change of consumers’ Kaiyu movements caused by redevelopment in terms of how many consumers visit and move among retail facilities. However, redevelopment programs require a significant investment, so the frequency-based Kaiyu Markov model needs to incorporate a money-based retail sales forecast. This study aims to construct such a model by disaggregating and introducing a consumer’s purchasing model in the previous model and show it with its application to the case of Kitakyushu City.
Kenichi Ishibashi, Saburo Saito

Multivariate Poisson Models with Hub Functions and Intervening Opportunities

How Would the Kyushu Super-Express Railway Opening Change the Flow of Tourists from the Kansai Region within the Kyushu Wide Area, Japan?: A Micro Behavior Analysis of the Destination’s Hub Function
This study predicts whether residents in the Kansai region, the area around Osaka, Kyoto, and Kobe, will increase their sightseeing trips to Fukuoka, Kumamoto, and Kagoshima in the Kyushu region, the southernmost island among four major islands of Japan due to the Opening of the whole Kyushu Shinkansen line, super-express railway line, in March 2011, in terms of the number of people. We also predict how people’s flow among three regions, Fukuoka, Kumamoto, and Kagoshima, will change before and after the Opening of the Kyushu Shinkansen.
The Kyusyu Shinkansen Line partially opened in 2004 between Shin-Yatsushiro and Kagoshima, the south terminal, which composes almost half the southern end of the entire route. This time, the Kyushu Shinkansen line will open in its entire path from Fukuoka through Kumamoto down south to Kagoshima. We note that the location of the Kansai region is 600 km west of Kyushu island. Thus, if the Kansai region residents take the railway to Kyushu, they must take the existing super-express railway up to Fukuoka and then go through Kumamoto down south to Kagoshima using the new Kyushu Shinkansen line. Therefore, for the Kansai residents, since the change will take place beyond Fukuoka, nothing will change to Fukuoka by Opening the Kyushu Shinkansen so that their travel time and cost are the same as before. In this case, the prediction of the increase in the number of visitors to Fukuoka by the traditional models should be zero since Fukuoka’s travel time and cost are the same before and after the Opening of the Kyushu Shinkansen.
In contrast to the traditional models that assume Fukuoka’s attractiveness or value would not change before and after the Opening of Kyushu Shinkansen, this study thinks Fukuoka’s attractiveness or value would increase after the Opening of Kyushu Shinkansen. The reason why we think this is so is the following. Suppose that Fukuoka, Kumamoto, and Kagoshima constitute the composite sightseeing services and that the sightseeing resources are the same before and after the Opening of the Kyushu Shinkansen. Also, assume that visitors must pay the generalized travel cost or the price to consume the composite sightseeing travel services. The Opening of the Kyushu Shinkansen drastically reduces the generalized travel cost in terms of travel time. Thus the visitors can enjoy the same composite sightseeing services at lower prices. Hence, this study considers the Opening of the Kyushu Shinkansen to increase sightseeing visitors’ visit value to Kyushu.
This study aims to formulate this idea as the hub functions of three cities and construct a Poisson model to forecast the frequency of sightseeing trips incorporating the hub functions. Further, we show that the number of sightseeing visitors from the Kansai region to Fukuoka will also change due to the Opening of the whole Kyushu Shinkansen line and to predict how much the number of tourists for Kumamoto and Kagoshima will increase.
This study also applies the same framework to predict the change in people’s sightseeing flows among three regions, Fukuoka, Kumamoto, and Kagoshima, within the wide Kyushu area before and after the Opening of the Kyushu Shinkansen.
Saburo Saito, Kosuke Yamashiro, Masakuni Iwami
A Micro Behavior Approach to Estimating and Forecasting the Intervening Opportunity Effects with a Multivariate Poisson Model: A Case for the New Terminal Complex of Kyushu Super-Express Railway, JR Hakata City
Originally, Stouffer (Am Sociol Rev 5:845–867, 1940) introduced the notion of intervening opportunities to explain why the volume of people’s movement between regions decreases as the distance of the movement increases. Thus, almost all previous studies on modeling and verifying the intervening opportunities have dealt with aggregate data such as interregional migration flows. However, we can define the intervening opportunity at the consumers’ micro behavior level and estimate its effect from their micro behavior data by specifying a particular intervening opportunity. We have done this by using a special occasion of the development in Fukuoka City, Japan, a twin city composed of two core CBDs (Central Business Districts) with transport terminals: the JR (Japan Railway) Hakata Station district and the Tenjin district with NNR (Nishi-Nippon Railroad) Tenjin Station. The JR Hakata Station and NNR Tenjin Station are about 2 km apart and connected by subway and bus. Tenjin consists of many shops and three department stores. Its shop-floor area amounts to 260,000 m2. In 2011, to coincide with the entire operation of the Kyushu Shinkansen bullet train connecting JR Hakata Station and Kagoshima, the Hakata Station was wholly renovated, and the new JR Hakata terminal station building named “JR Hakata City” opened. It is a shopping complex with a total floor space of 200,000 m2, which increased by 100,000 m2 from the previous building.
Since many visitors seem to travel through Hakata to Tenjin without stopping at Hakata due to its less retail agglomeration than Tenjin, quite an interesting problem is how this situation would change if JR Hakata City opens. As suggested, if some shopping site is located midway from a visitor’s home to his/her destination, the shopping site is said to be an intervening opportunity to the destination. Therefore, we can define the shopping site’s intervening opportunity effect on the destination as the difference of the number of visitors to the destination between with and without this intervening opportunity.
This study aims to estimate and forecast this intervening opportunity effect by predicting the changes in the numbers of visitors to the Hakata and Tenjin districts caused by the opening of JR Hakata City before it opens. To address the purpose, we constructed a multivariate Poisson model with intervening opportunity effects based on the consumer’s micro behavior data obtained from visitors at the city center of Fukuoka City by the on-site survey conducted before its opening. We estimated how many visitors who have been visiting Tenjin through Hakata without stopping at Hakata would be intercepted by the new redevelopment of JR Hakata City. Conversely, we also estimated how many visitors who used to have been trapped by Tenjin on the way to Hakata would pass through Tenjin not intercepted after the opening of JR Hakata City.
Saburo Saito, Kosuke Yamashiro, Masakuni Iwami, Mamoru Imanishi, Masakuni Kakoi, Yasufumi Igarashi

Integrated Modeling of Weighted Multivariate Poisson Models with Competitive Destinations and Kaiyu Markov Models

How Would the Opening of JR Hakata City, a New Terminal Complex of the Kyushu Super-Express Railway, Change the Number of Visitors, Retail Sales, and Consumers’ Kaiyu Flows in the City Center Commercial District of Fukuoka City?
In this chapter, we demonstrate the detailed record of our efforts to forecast changes in the number of visitors and sales of the city center at Fukuoka City, Japan, caused by a large-scale commercial redevelopment, JR Hakata City. The purpose of doing this is twofold; The first is to describe and record the present state of the art to forecast the effects of urban development policies based coherently on consumer behavior changes, particularly consumer Kaiyu behavior changes. The second is to provide some challenging problems we faced in the present methodology employed and suggest a direction to improve the methodology further.
The uniqueness of our forecasting efforts lies in the approach based on the consumers’ behavior models explaining their choices about the frequency of visits to destinations and, in particular, their choices about how they undertake Kaiyu behaviors or shop-around behaviors among shopping sites in the city center. Consequently, our models become not probability-based ones but frequency-based ones. We also deal with the Kaiyu flows among various districts within a city center in terms of the number of people, revealing the accompanied money flows within the city center.
Furthermore, Fukuoka City is a twin city with two core CBDs, Tenjin, the area with the largest retail agglomeration, and Hakata, having the railroad station terminal redeveloped as JR Hakata City this time. Thus the focus of our forecasting efforts is how Tenjin’s supremacy will change through the large-scale development carried out on one side of the twin cores. The novel feature of our efforts to explore this is that we dig deep into how the supremacy of Tenjin as a destination will change by exploring a causal path, the Hakata’s intervening opportunity effects on the destination Tenjin, from predicting the number of visitors to Tenjin intercepted by Hakata, the midway to the destination Tenjin, after the large-scale retail development at Hakata.
When carrying out our forecastings, we utilize several methods we developed ourselves, such as the weighted Poisson models for the on-site samples, the Kaiyu Markov model with covariates, and the consistent estimation method for the Kaiyu path density from the on-site samples. These methods correspond to respective aspects of a unique individual’s entire behavior. Thus the results obtained from these methods should have coherency. However, the most challenging problem we faced in our forecasting task was how to keep consistency between the results from different models and data. Therefore, while describing the detailed records of our forecasting efforts, we also indicate and discuss how to address the problem and improve the present methodology.
Saburo Saito, Kosuke Yamashiro, Masakuni Iwami
How Many Customers Would Be Brought Back from Suburban Shopping Malls to the City Center by Redeveloping the City Center Station Building, JR Oita City, Japan? A Multivariate Poisson Model with Competitive Destinations
The city center commercial district of Oita City, Japan, competes with two large-scale suburban shopping malls and is exhibiting a declining trend, a common phenomenon for almost central shopping streets of local cities. However, what is distinctive about the central shopping street in the city center of Oita City is that it neighbors JR (Japan Railways) Oita Station, just 200 m apart. The Oita station plans to redevelop its building at the same time as the elevation of railroad tracks and turn it into a large-scale commercial complex, JR Oita City. The JR Oita City will open in March 2015. Facing the development, people involved with the central shopping street seem to have given up on it as devastating impacts on their business.
On the contrary, we regard this development as an opportunity that the central shopping street can prevent customers from moving out to the suburban malls, bring them back to the city center and revitalize the city center of Oita City by enhancing visitors’ Kaiyu flows within the city center of Oita City. This study aims to verify and validate our claim in advance by forecasting.
More concretely, we first formulated a multivariate Poisson model to explain consumers’ choices about the frequency of visits to the competing destinations that can forecast the visit frequency to each destination in a way that can decompose it into how much the competitiveness of one destination affects to increase or decrease the visit frequency to the other destinations. Then, we estimated this model based on the data obtained from the on-site interview survey of consumer Kaiyu behaviors. Further, by setting up the Oita Metropolitan Area, we forecasted the increase in incoming visitors to the city center before and after JR Oita City’s opening while clarifying how many customers would return from the other two shopping malls. Moreover, by constructing the Kaiyu Markov model, we forecasted the changes in the actual number of visitors’ Kaiyu OD (Origin-Destination) flows between JR Oita City, the central shopping street, the local department store, and several other districts in the city center before and after the opening and also the changes in the retail sales of these commercial establishments and districts.
Saburo Saito, Masakuni Iwami, Kosuke Yamashiro

Other Applications of Kaiyu Markov and Related Models

An Opportunity Cost Approach to Valuation of the River in a City Center Retail Environment: Another Application of the Kaiyu Markov Model
Many cities in Japan face the decline of the city center commercial district due to motorization and locations of large shopping centers in suburban areas. This problem aroused the awareness that the loss of the city center commercial district means the loss of the prerequisite of subsistence of the city. Thus, many cities have launched redevelopment projects to revitalize their city center commercial districts. We have already proposed an evaluative framework for assessing the spatial structure of the city center commercial district based on consumers’ shop-around or Kaiyu behaviors. Furthermore, we constructed a frequency-based shop-around Markov model to forecast how redevelopment projects in a city center commercial district would change consumers’ shop-around or Kaiyu behaviors and consequently result in changes in retail sales of the city center commercial district.
On the other hand, evaluating public projects in terms of money has recently attracted severe concerns, those as harnessing exploding budgets for public investments and the consciousness of the environmental consequences of public projects. Thus, much discussion has been made on valuation methods of environmental resources such as CVM (contingent valuation method), travel cost method, user benefit method, and hedonic approach.
While the city center revitalization projects often accompany the improvement of natural environments such as parks and rivers, few studies have tried to evaluate them in terms of money. This study uses the above Kaiyu Markov model to evaluate the natural environment, such as the river in the city center district. More specifically, this study proposes an opportunity cost method for the valuation of natural resources in a city center district based on consumers’ shop-around or Kaiyu behaviors while applying to the Murasaki River in the city center commercial district of Kitakyushu City, Japan.
In the city center commercial district of Kitakyushu City, the Murasaki River flows from south to north, dividing the city center commercial district into east and west. The existence of the Murasaki River decreases consumers’ shop-arounds or Kaiyu movements between east and west so that retail turnover on each side decreases. Using the above model, we have estimated the total annual amount of the decrease in retail turnovers, which is the loss of retail sales that would have been obtained if it were not for the Murasaki River, which is the opportunity cost retailers are paying annually. To change the viewpoint, this can be regarded as willingness-to-pay that retailers are willing to spend every year for leaving the Murasaki River as it is. Hence if discounting, we can obtain the asset value of the Murasaki River. This study has done this.
Saburo Saito, Kenichi Ishibashi
Extraction of Long Sightseeing Kaiyu Routes in the Kyushu Wide Region, Japan
Traditional tourism studies have paid less attention to the problem of how tourism sites in a wide area should be linked to enhance the network effect of the tourism sites as a whole, effectively utilizing the respective sites’ inherent advantages. Recently in Kyushu, Japan, two big theme parks, Huis Ten Bosch and Space World, have opened, and the national park preserving Yoshinogari Ruins is now under construction. [The date of these statements are as of 1992.] Thus, we should strengthen some linkages among them. From these considerations, this study tries to extract long sightseeing tour routes in the Kyushu region and to estimate the network effect of the tourism sites as a whole from the viewpoint of tourists’ Kaiyu (travel-around) behaviors using data obtained from the Survey of Tourists who dropped at Fukuoka City.
Saburo Saito, Toru Sakamoto
A Bayesian Network Model of Consumers’ Kaiyu Behaviors
A consumer walking around a downtown shopping area usually visits several places and shops. From the point of view of city management, it is crucial to know and predict people’s movements in the area, for example, to plan to make better pathways to enhance mobility. From a shop manager’s point of view, it is also vital to grasp the tendency of people visiting other shops to identify whether those shops are competing or cooperating with his/her shop. In order to offer helpful information for these tasks, we used graphical models called Bayesian networks in this research. By applying the models to actual data obtained from questionnaire surveys of consumers who visited a downtown shopping area, we analyzed consumers’ shop-around or Kaiyu behaviors.
Rui Yamaguchi, Saburo Saito
Recent Advances in Modeling and Forecasting Kaiyu
herausgegeben von
Saburo Saito
Kenichi Ishibashi
Kosuke Yamashiro
Springer Nature Singapore
Electronic ISBN
Print ISBN