The Spatial Distribution of Taxi Stations in Bangkok

Abstract

Taxis play a crucial role as an on-demand transportation mode in developing countries due to perceived inefficiencies of cities’ public transportation systems. However, studies on the locational distribution of taxis in urban areas are limited, despite the need to improve passenger service quality by balancing the demand and supply of taxi services. Notably, taxi stations possess distinct characteristics compared with other public transport stations that serve passengers directly; in contrast, taxi stations solely support taxi drivers in locations where they begin and conclude their work. This study aims to determine the spatial distribution pattern and assess the agglomeration economies of taxi stations, using Bangkok as a case study, a city with a significant number of registered taxis and dispersed taxi stations. This research takes into account various spatial variables, including land price, land use mix index, population density, and gas station locations. The primary tool for analyzing the spatial distribution pattern was the spatial statistics model, employing ArcGIS 10.8 software. The analysis consisted of three steps: testing for clustered or dispersed patterns using Moran’s I, applying Anselin’s local Moran’s I (LISA) to examine the relationship between taxi station coordinates and spatial variables such as land price, land use mix index, and population density, and evaluating the relationship between taxi stations and energy service stations using the network analysis tool. The results revealed that taxi stations exhibited a spatially clustered pattern and were closely correlated with the land use mix index, land price, and population density, as indicated by Moran’s index values of 0.425, 0.328, and 0.373, respectively, especially those located within a 3000 m radius of gas stations. These findings elucidate the location selection of taxi stations, which tend to prioritize areas that can generate maximum profits for the taxi business rather than those with minimal location costs. They also tend to be situated in proximity to market areas. Additionally, the study recommends that the government consider promoting electric taxis as a sustainable mode of urban transport in the future to reduce the usage of natural gas (NGV) and liquefied petroleum gas (LPG).

1. Introduction

Bangkok serves as the central service area for taxis in Thailand, boasting 83,188 registered taxis within the city, which accounts for 96 percent of all taxis in the country [1]. In stark contrast, there are only 3264 public buses operating in the same region [2]. These figures underscore the vital role of taxis in Bangkok’s transportation landscape, with the number of taxis outnumbering public buses by a factor of 25.49. Taxis in Bangkok are typically organized into various taxi stations scattered throughout the city. Unlike other public transportation stations and hubs, taxi stations exclusively cater to taxi drivers, while clients can conveniently hail taxis directly from the roadside. Additionally, taxi stations can be situated anywhere within the city, as they do not need to factor in passenger accessibility. In contrast, passenger terminals for other public transportation modes must prioritize passengers’ convenience, particularly concerning connections with other modes of transportation [3].

Taxi operations in Bangkok are managed by various taxi companies, each of which operates its own taxi stations. These stations serve as multifunctional facilities, including offices, parking areas, and storage facilities for both the taxi companies and drivers. Furthermore, taxi stations affiliated with larger taxi companies often take on additional roles, such as handling taxi sales, repairs, and vehicle inspections. Consequently, these taxi stations in Bangkok serve as the central hubs for all taxis operated by these companies. Each taxi company maintains stations strategically positioned throughout the urban area, covering locations from the city center to suburban areas.

The location of taxi stations in Bangkok is intricately tied to the issue of passenger rejection, a significant challenge in the city’s taxi service. Late returns of taxis to the stations have been identified as a root cause of this problem [4]. Specifically, during the shift change period, taxi drivers aim to optimize their earnings by picking up passengers heading in the same direction as the taxi stations to avoid returning empty. Given that taxi drivers shoulder the majority of taxi operation costs and pay rent to the taxi station during this period, they are highly motivated to maximize profits within the constraints of limited time. This decision making process by drivers profoundly influences both the placement of taxi stations and the selection of passenger destinations.

Previous research has predominantly shown that the placement of public transport stations or passenger terminals is influenced by key physical factors. Zhengdong et al. (2009) [5] suggested that in Wuhan, public transport stations tend to be situated on major roads within rapidly developing areas of the city. Their research also revealed a positive and significant correlation between station locations and districts with high customer demand. Xuemei et al. (2010) [6] identified three pivotal factors influencing the placement of urban passenger transportation terminals: they should be positioned in areas with high travel demand, located at a distance from and connected to major roads, and integrated with public bus lines. Qu et al. (2019) [7] discovered that potential areas for establishing urban taxi stands are those with a high volume of pick-up and drop-off activity, particularly in public transportation hub areas. Additionally, Zhu et al. (2019) [8] proposed that the central location of public transport transfers plays a significant role in determining the placement of passenger transport hubs.

However, it is worth noting that taxi stations in Thailand exhibit distinct characteristics compared with other public transport stations that serve passengers directly. Taxi stations primarily support taxi drivers at locations where they commence and conclude their work. While the positioning of taxi stations in Bangkok is governed by the city’s planning laws and regulations, they are categorized as “office” types of land use since they serve as taxi offices rather than traditional transit stations. This classification grants taxi station owners the flexibility to choose locations for their taxi stations across various land use types in Bangkok, resulting in a widespread distribution of taxi stations throughout the city. This aspect sets Bangkok’s taxi service apart from those in other developed countries, where stringent zoning laws, building safety regulations, environmental regulations, air and noise pollution regulations, and parking quantity regulations are enforced to maintain urban orderliness.

While prior studies have focused on identifying the land use and transport factors that influence the spatial distribution of public transport stations or passenger terminals, the specific nuances of the relationship between taxi stations and land use characteristics, as well as the spatial distribution of taxis, have not been clearly elucidated.

The location of taxi stations plays a pivotal role in the urban transportation network, especially in a city like Bangkok, where there is a disconnect between taxi service infrastructure and city transportation planning. This study has the potential to enhance urban mobility and alleviate urban traffic congestion. Notably, in Bangkok, taxi passengers predominantly exhibit behaviors characterized by using taxis for long-distance journeys, traveling directly from their point of origin to their destination. This behavior differs from that observed in developed countries, where taxis are often employed for first-mile and last-mile trips [9]. Furthermore, understanding the distribution pattern of taxi stations can provide valuable insights for urban planners and policymakers, aiding in the formulation of land use plans and policies that strike an appropriate balance between taxi supply and demand within the city.

The primary objective of this study is to ascertain the spatial distribution pattern of taxi stations in Bangkok and investigate the presence of agglomeration economies within this context. The research findings have the potential to shed light on the decision making processes of taxi station owners when selecting their locations and can serve as a valuable analytical tool for exploring agglomeration economies. By delving into the spatial distribution of taxi stations, this study aims to uncover valuable insights into its implications for urban mobility, areas requiring improvement, and its contributions to the development of more comprehensive urban and public transportation planning strategies.

2. Literature Review

2.1. Location Theory

Location theory primarily pertains to the selection of industrial locations, a domain within land economics that aims to identify potential lands with the lowest production costs and the highest profitability. Several factors influence the choice of industrial location, including labor, time, materials, distance from production sources to markets, and transportation costs. These factors can be categorized into spatial and non-spatial factors that influence the spatial clustering or dispersal of industries.

Alfred Weber’s Theory of the Location of Industries [10] recognizes the economic geography of locations, which is closely tied to transportation costs. Factors affecting transportation costs encompass the proximity to consuming places, access to material resources, the presence of power places, and the existence of central locations between material resources and power places. These factors enable theoretical analyses applicable to any situation influencing industrial location decisions, which can be categorized into three approaches: (1) analyzing cost savings in transportation by correlating transportation costs with the distance from the production site to resources and marketplaces, (2) investigating the accounting costs of production, and (3) considering the interaction between labor and space to minimize labor costs and ensure access to skilled labor.

2.2. The Theory of Industrial Location

2.2.1. The Theory of Geographical Industrial Location

According to Smith (1966) [11], this theory represents an extension of Alfred Weber’s Theory of the Location of Industries, adapted to accommodate evolving economic complexities and changing times. Smith’s analysis delves into the selection of industrial locations with a focus on achieving the lowest cost or highest profit, taking into consideration three fundamental components: economic activity size, factors influencing production, and marketing characteristics. These components encompass a total of nine economic factors, both direct and indirect, that play pivotal roles in the location selection process.

The six direct economic factors include production cost, which represents the monetary expenses associated with production, and the annual profit potential of a specific location. Transportation cost is another economic factor, which analyzes the distance and quantity of materials to be transported. Market accessibility plays a crucial role in market sharing and transportation costs. Labor cost varies across different industrial locations due to differences in labor skills, labor laws, and welfare provisions in each area. Land price, reflecting the availability of resources and infrastructure readiness, is a significant factor in location selection. Typically, urban areas have higher land prices due to their dense infrastructure and good accessibility. Finally, economies of scale can reduce transportation, labor, and communication costs for businesses of the same type.

Among the three indirect economic factors are government policies offering tax incentives and infrastructure support within designated special economic zones. Entrepreneurial skills also wield significant influence on the success and profitability of businesses, with high-profit enterprises often gravitating toward areas of high potential. Lastly, one must take into account the budgetary constraints faced by business owners.

2.2.2. Industrial Location Theory

Webber (1985) [12] developed this theory as an extension of Alfred Weber’s location theory. According to Webber, two primary factors influence location selection: scientific factors, such as economic status and labor costs, and the owner’s satisfaction with labor policies and welfare in a specific area. This theory aligns with classical location theory, which underscores the importance of selecting low-cost areas for industrial development. Decision making regarding the appropriate location for business establishment hinges on three key factors: comprehending production demand, considering production reduction or closure, and identifying the most cost-effective land areas.

Regarding the selection of the lowest-cost area, Webber identifies two influential factors: transportation cost and production cost. Transportation cost encompasses the expenses associated with the movement of resources, production materials, and finished goods between production areas and market areas. The proximity to infrastructure, resource-rich zones, or market areas plays a pivotal role in minimizing transportation costs. Additionally, the size of the business has a significant impact on location selection. Larger factories, in particular, stand to gain from situating themselves in close proximity to resource-rich or market areas, thereby reducing both time and transportation costs.

Production cost, encompassing both financial and non-financial components, significantly influences location selection. Non-labor costs, which encompass expenses related to establishing and equipping the business, such as machinery costs, contribute to variations in product costs, transportation costs, land costs, and taxes associated with each location. Labor costs exhibit variability across different regions, while infrastructure costs play a crucial role in determining transportation efficiency, particularly in urban areas characterized by high-density infrastructure and elevated land prices. Government policies pertaining to property tax, labor costs, environmental regulations, and welfare provisions also exert a significant influence on production costs. Agglomeration economies further contribute to the realization of economies of scale. Lastly, the personal preferences and needs of entrepreneurs, such as proximity to their residences or cost-saving considerations, exert notable influence on location choices.

2.3. The Spatial Distribution of Taxis

In recent years, there has been a burgeoning body of research spanning diverse disciplines that delves into the spatial distribution within urban areas. However, there has been a notable absence of studies specifically addressing the spatial distribution of taxi stations. While there is existing research examining the relationship between taxis and urban areas, these studies predominantly employ location theory to elucidate the spatial distribution of urban taxis and can be categorized into three distinct groups.

The first group encompasses studies that investigate the mobility patterns of urban taxis, analyzing the spatial distribution of these vehicles [13,14,15,16,17,18]. Taxis are often considered representative of travel demand within a city, effectively serving as substitutes for private vehicles. By examining the pick-up and drop-off locations of taxis through GPS data, researchers gain valuable insights into travel behavior and trip distribution. These studies collectively suggest that taxis predominantly operate within high-demand areas, resulting in their concentration within densely populated and diverse land use areas within urban regions.

The second group of studies focuses on the disparity between taxi demand and supply [19,20,21,22,23,24,25,26,27]. These investigations delve into various factors, including land use, and their impact on the travel demand experienced by taxis. They analyze pick-up and drop-off data, route assignments, travel time, and spatial data obtained from GPS sources. The findings collectively highlight a notable incongruity between the supply of taxis and the demand for their services within urban areas. Taxis tend to congregate in high-density land use areas and economic districts, revealing a lack of alignment between service provision and demand.

The third group of studies is characterized by its focus on taxi variables and the development of models to comprehend the spatial relationships of taxis [28,29,30,31,32,33]. These investigations make use of real-time GPS data, encompassing details regarding taxis’ pick-up and drop-off locations, travel routes, and land use activity variables. Researchers employ statistical models to meticulously analyze hourly taxi data, frequently categorizing extensive study areas into points of interest (POIs) for in-depth examination. These studies significantly contribute to an enhanced understanding of urban taxi behavior by pinpointing specific demand points within broader regions. Notably, the complexity of these studies lies in the intricacies of model development, encompassing the input and processing of variables, thus extending beyond the scope of the first two groups.

Based on prior research, it becomes evident that taxis predominantly display travel patterns that are centered around the central business district (CBD), primarily due to the significant travel demand generated within this area, effectively establishing it as a primary market for taxis. The concentration of taxi services within the CBD is advantageous, as it leads to time- and cost savings in travel and presents taxi drivers with the prospect of earning a higher income when compared with other areas.

Based on the reviewed location theory and related research, it becomes evident that crucial factors influencing location selection revolve around cost considerations and the potential for profit associated with specific locations. These factors are often reflected in variables such as proximity to market areas, land use types, and population density. Additionally, the selection of locations close to energy resources, represented by variables like gas station locations, also plays a significant role in the context of this research. Furthermore, variables related to land prices are among the key factors under investigation in this study.

3. Methodology

3.1. Study Area

The study area, the Bangkok Metropolitan region, was chosen due to its substantial taxi population, rendering it the largest taxi market in Thailand. To facilitate effective analysis of this area, the researchers divided it into 50 administrative districts. These districts were subsequently categorized into three distinct classes based on their respective characteristics.

The first class comprises 18 administrative districts, known as CBD areas, which function as pivotal financial and commercial centers within Bangkok. These districts are characterized by dense infrastructure and are Bang Kho Laem, Bang Rak, Bang Sue, Chatuchak, Din Daeng, Dusit, Huai Khwang, Khlong Toei, Klong San, Pathum Wan, Phaya Thai, Phra Nakhon, Pom Prap Sattru Phai, Ratchathewi, Samphanthawong, Sathon, Watthana, and Yan Nawa.

The second class encompasses 18 administrative districts, recognized as inner city areas, which fulfill dual roles as both residential and commercial zones. These districts are Bang Kapi, Bang Khen, Bang Na, Bang Phlat, Bangkok Noi, Bangkok Yai, Bueng Kum, Don Mueang, Khan Na Yao, Lak Si, Lat Phrao, Phra Khanong, Prawet, Sai Mai, Saphan Sung, Suan Luang, Thon Buri, and Wang Thonglang.

The third class comprises 14 administrative districts, categorized as suburban areas, predominantly characterized by their residential and agricultural zones. These districts are Bang Bon, Bang Khae, Bang Khun Thian, Chom Thong, Khlong Sam Wa, Lat Krabang, Min Buri, Nong Chok, Nong Khaem, Phasi Charoen, Rat Burana, Taling Chan, Thawi Watthana, and Thung Khru. The locations and boundaries of all 50 administrative districts are visually represented in Figure 1.

3.2. Data

The secondary data employed for the analysis of the spatial distribution of taxi stations in Bangkok encompassed various datasets, including information on taxi stations, land use, land prices, population, and gas stations dedicated to taxis. These datasets were procured from public organizations. The verification and determination of the coordinates of taxi stations and gas stations were accomplished using Google Earth Pro.

Subsequently, the data underwent transformation into geographic information system (GIS) data, which encompassed both attribute information and coordinate data. Following this transformation, the data were imported into the ArcGIS Pro 10.8 software for the purpose of conducting spatial statistical analysis.

3.2.1. Taxi Stations Data

The data employed in this study were sourced from the Land Transport Department of Bangkok and pertained to the year 2020. It comprised 118 addresses representing taxi stations, which were affiliated with three distinct categories of taxi companies: cooperative taxis, taxi companies, and taxi limited partnerships. Notably, private taxis were excluded from this dataset. It is essential to highlight that all three categories of taxi stations possess the autonomy to select their respective locations, provided they are adhering to the city planning laws and regulations of Bangkok.

To ensure the precision of the provided addresses, Google Earth Pro was utilized to cross-verify them, given that certain registered addresses did not align with the actual operational locations of the taxi stations. Google Earth Pro is renowned for delivering coordinates of acceptable accuracy for research purposes [34,35]. Furthermore, it utilizes the World Geodetic System 1984 (WGS84) datum to generate latitude and longitude data [36], which seamlessly integrates with the ArcGIS Pro 10.8 software. Consequently, based on the results obtained from Google Earth Pro, it was determined that only 80 taxi stations were eligible for inclusion in the subsequent analysis, as depicted in Figure 2.

3.2.2. Land Use Data

The land use data for Bangkok, dating back to 2013, were sourced from the Department of Public Works and Town & Country Planning. This dataset encompassed nine distinct land use categories, each of which was attributed to the 50 administrative districts. These land use categories were as follows: residential zones, commercial zones, industrial zones, rural and agricultural zones, government institutes, public utilities and amenities zones, educational institutes zones, religious institutes zones, and recreational zones. To analyze the land use data effectively, the Entropy model, as introduced by Zhang et al. (2012) [16], was employed. This model facilitated the calculation of the land use mix index for each of the 50 administrative districts.

$$\mathrm{Entropy}=-{\Sigma }_j \frac{P_j\times \mathit{ln}\left(P_j\right)}{ln\left(J\right)}$$

(1)

The land use mix index (P_j) was computed individually for each land use type, with ‘J’ representing the total count of land use types present within the area. The resulting entropy values fall within a range of 0 to 1. In this context, a value of 0 signifies a scenario characterized by homogeneous land use with no mix, whereas a value of 1 denotes the highest level of diversity of land use.

The mixed land use degrees observed across the 50 administrative districts in Bangkok exhibited a spectrum, ranging from 0.23 to 0.70. These values signify varying degrees of diversity in land use types, encompassing both low and high diversity scenarios. Calculations yielded an average mixed land use degree for Bangkok, which was determined to be 0.46. Among these districts, Nong Chok, situated in the suburban periphery of Bangkok, displayed the lowest mixed land use degree, while Phra Nakhon, located within the central core of Bangkok, exhibited the highest mixed land use degree.

Notably, the districts characterized by the highest degrees of land use diversity were predominantly clustered within the Bangkok Central Business District (CBD). These areas function as primary centers for commerce and employment, experiencing substantial traffic and attraction levels. In contrast, land use diversity degrees gradually diminished as one moved away from the CBD. Figure 3 visually portrays the distribution of taxi stations based on the mixed land use degree.

3.2.3. Land Price Data

The data concerning land prices at the 80 taxi station locations in Bangkok were sourced from the National Land Appraisal Price Report, which covered the years 2016 to 2019. This report was generously made available by the Thai Treasury Department. It is noteworthy that land prices in Thailand are conventionally quantified in “Baht per Square Wah,” with one square wah being equivalent to four square meters.

As per the analysis conducted, the average land appraisal price associated with taxi stations in Bangkok exhibited a range spanning from THB 10,000 to 400,000 per square wah. In approximate terms, this translates to a range of USD 7.25 to 2900 per square meter. The comprehensive average land appraisal price for taxi stations within Bangkok was ascertained to be THB 74,168.75 per square wah, which is roughly equivalent to USD 537.72 per square meter.

The lowest recorded land appraisal price for a taxi station in Bangkok stood at THB 10,000 per square wah, which equates to approximately USD 7.25 per square meter. This particular price was associated with two taxi stations situated within the Bang Bon and Sai Mai districts. In contrast, the Ratchathewi district boasted the highest land appraisal price for a taxi station, reaching THB 400,000 per square wah, or approximately USD 2900 per square meter.

Figure 4 provides a visual representation of the distribution of taxi stations throughout Bangkok, spanning from the central business district (CBD) to suburban areas. The figure prominently showcases a concentration of high land prices in the heart of Bangkok’s central core, while lower land prices are notably dispersed throughout the suburban regions.

3.2.4. Population Density

The population density data for Bangkok, acquired from the Ministry of Interior in Thailand for the year 2021, were computed by dividing the population statistics by the respective sizes of the 50 administrative districts within Bangkok. This analysis unveiled a diverse spectrum of population densities, spanning from 757 to 20,931 persons per square kilometer. On average, the population density for Bangkok was calculated to be 6189 persons per square kilometer.

The Nong Chok district showcased the lowest population density, with a mere 757 persons per square kilometer. Situated in the suburban periphery of Bangkok, this district is characterized predominantly by agricultural zones and vacant lands. In stark contrast, the Pom Prap Sattru Phai district exhibited the highest population density, soaring to 20,931 persons per square kilometer. Remarkably, this district ranks as the second smallest in terms of size within Bangkok, covering a mere 1.93 square kilometers.

Despite the fluctuations in population density, taxi stations were identified to be distributed throughout areas characterized by both high and low population densities, as illustrated in Figure 5. This observation underscores the fact that taxi stations are not exclusively concentrated within densely populated regions. For additional insights, descriptive statistics pertaining to the land use mix index, land appraisal price, and population density are provided in

Table 1. The descriptive statistics of taxi station distribution data.

Variables	Mean	Median	Std. Dev.	Min	Max
Land use mix index	0.462	0.450	0.097	0.230	0.700
Land appraisal price	74,168.750	46,000.000	74,903.649	10,000.000	400,000.000
Population density	6188.680	5202.000	3709.308	757.000	20,931.000

.

3.2.5. Gas Station Data

In Thailand, the use of natural gas for vehicles (NGV) and liquified petroleum gas (LPG) is widespread as alternative fuels in order to replace conventional fossil energy sources in taxis, primarily driven by the cost savings associated with these fuels. Bangkok, in particular, boasts an extensive network of NGV and LPG stations, effectively meeting the demand for these alternative fuels. The data pertaining to the addresses of NGV and LPG stations in Bangkok were procured from the Department of Energy Business for the year 2015 and subsequently verified utilizing Google Earth Pro. The analysis unveiled a total of 268 gas stations providing both NGV and LPG services, with numerous stations situated in close proximity to one another, as visually represented in Figure 6.

3.3. Spatial Autocorrelation

Spatial autocorrelation served as a valuable tool for investigating the spatial connections between the positions of taxi stations and three key spatial variables: the land use mix index, land appraisal price, and population density. To carry out this analysis, the global Moran’s I model, available within the ArcGIS Pro 10.8 software, was employed. This model yielded insights into the spatial distribution pattern of taxi stations, shedding light on whether they manifested a dispersed, random, or clustered arrangement. The depiction of the global Moran’s I model is presented in the second model.

$$I=\frac{n}{S_{\mathrm{o}}} \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{i,j }{z}_i{z}_j}{{\sum }_{i=1}^n{z}_i^2}$$

(2)

where z_i is the deviation of an attribute for feature i from its mean (x_i − $\overline{x}$), w_i,j is the spatial weight between feature i and j, n is equal to the total number of features, and So is the aggregate of all the spatial weights:

$$S_o={\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{i,j}$$

(3)

The z_I score for the statistic is computed as:

$$z_I=\frac{I-\mathrm{E}\left[I\right]}{\surd \mathrm{V}\left[I\right]}$$

(4)

$$\mathrm{E}\left[I\right]=-1/(n-1)$$

(5)

$$\mathrm{V}\left[I\right]=\mathrm{E}\left[I^2\right]-\mathrm{E}\left[I^2\right]$$

(6)

3.4. Anselin Local Moran’s I (LISA)

The Anselin local Moran’s I model was deployed to analyze the degree of intensity related to high values, low values, and outlier values in the distribution pattern of taxi stations throughout Bangkok. In contrast to the global Moran’s I model, which solely presents the spatial distribution pattern, this model enables the evaluation of the statistical significance of the relationship between taxi station locations and the three previously mentioned factors. The outcomes of the local indicators of spatial association (LISA) model are presented in the seventh model.

$$I_i=\frac{x_i-\overline{X}}{S_i^2} {\sum }_{j=1,j\ne i}^n{w}_{i,j}(x_j-\overline{X})$$

(7)

where x_i is an attribute for the feature I, $\overline{X}$ is the mean of the corresponding attribute, w_i,j is the spatial weight between feature i and j, and:

$$S_i^2=\frac{{\sum }_{j=1, i\ne j}^n{(x}_j-\overline{X}{)}^2 }{n-1}$$

(8)

with n equating to the total number of features.

The z_Ii score for the statistics is computed as:

$$z_{Ii}=\frac{I_{i }-\mathrm{E}\left[I_i\right]}{\surd \mathrm{V}\left[I_i\right]}$$

(9)

where:

$$\mathrm{E}\left[I_i\right]=-\frac{{\sum }_{j=1, i\ne j}^n{w}_{ij}}{n-1}$$

(10)

$$\mathrm{V}\left[I_i\right]=\mathrm{E}\left[I_i^2\right]-\mathrm{E}[I_{i }{]}^2$$

(11)

3.5. Network Analysis

The network analysis tool available in ArcGIS Pro was harnessed to evaluate the relationship between the locations of taxi stations and NGV and LPG stations, with the objective of examining the agglomeration economies associated with taxi stations in Bangkok. This method facilitates an assessment of road network accessibility between taxi stations and NGV/LPG stations, thereby optimizing routes and distances to enhance operational efficiency.

4. Research Findings

4.1. The Taxi Stations’ Distribution Pattern

The outcomes of the global Moran’s I analysis are displayed in

Table 2. The global Moran’s I result.

Variables	Moran’s Index	Expected Index	z-Score	p-Value
Land use mix index	0.425	−0.013	6.396	0.000
Land appraisal price	0.328	−0.013	5.113	0.000
Population density	0.373	−0.013	8.440	0.000

. The table reveals a noteworthy association between the placement of taxi stations and the land use mix index, land appraisal price, and population density. Furthermore, all these factors exhibit positive z-scores. These findings signify that the spatial distribution of taxi stations in Bangkok adheres to a clustered pattern. However, it is worth noting that the extent of clustering is not highly pronounced, as indicated by the Moran’s indexes, which fall below 0.500 out of a maximum score of 1.000.

4.2. The Spatial Distribution of Taxi Stations

The spatial distribution of taxi stations, as determined by the land use mix index degree, is depicted in Figure 7. Among the 80 taxi stations in total, 29 stations (equivalent to 36.25 percent) exhibited notable spatial clustering based on the land use mix index degree. These clusters can be categorized into three distinct patterns.

The first pattern is distinguished by a pronounced spatial clustering in regions marked by a high land use mix index degree, falling within the range of 0.48 to 0.70, thus exceeding the average land use mix index degree in Bangkok. In this pattern, 16 taxi stations (constituting 20.00 percent) were identified. Notably, these stations were predominantly clustered in the central business district (CBD) and inner-city areas.

The second pattern is characterized by a limited spatial clustering within regions exhibiting a low land use mix index degree, falling within the range of 0.36 to 0.42, which is below the average land use mix index degree observed in Bangkok. This pattern encompasses 12 taxi stations, accounting for 15.00 percent of the total. Notably, these stations were primarily clustered in suburban areas.

The third pattern corresponds to outliers, particularly a significant spatial clustering within regions marked by a low land use mix index degree, falling below 0.43, which is beneath the average land use mix index degree in Bangkok. This pattern encompasses only one taxi station, representing a mere 1.25 percent, and it exhibited spatial clustering within the inner-city area.

In summary, land use, as measured by the land use mix index degree, appears to be a crucial determinant of taxi station locations in Bangkok. Areas with diverse land uses, especially in urban and commercial centers, tend to have more clustered taxi stations, while areas with less diverse land uses, particularly in suburban regions, have fewer taxi stations. This finding underscores the importance of considering land use patterns when planning and optimizing taxi services in urban environments.

The spatial distribution of taxi stations, as influenced by land appraisal price, is visualized in Figure 8. Among the 80 taxi stations in total, 29 stations (36.25 percent) displayed marked spatial clustering based on land appraisal price. These clusters can be categorized into three discernible patterns.

The first pattern denotes a conspicuous spatial clustering within regions characterized by a high land appraisal price, spanning from THB 150,000 to 400,000 per square wah (USD 1087.50 to 2900 per square meter). In this pattern, nine taxi stations were identified, comprising 11.25 percent of the total. Notably, this land appraisal price range exceeds the average for taxi stations in Bangkok. These stations were predominantly clustered in the central business district (CBD) and inner-city areas.

The second pattern corresponds to limited spatial clustering within regions characterized by a low land appraisal price, spanning from THB 10,000 to 60,000 per square wah (USD 7.25 to 435 per square meter). Within this pattern, there were 17 taxi stations, accounting for 21.25 percent of the total. Notably, this land appraisal price range falls below the average for taxi stations in Bangkok. These stations were primarily clustered in suburban areas.

The third pattern signifies an outlier, more precisely, a prominent spatial clustering within regions characterized by a low land appraisal price, ranging from THB 34,000 to 70,000 per square wah (USD 246.50 to 507.50 per square meter). Notably, this price range falls below the average land appraisal price for taxi stations in Bangkok. This pattern comprises only three taxi stations, accounting for a mere 3.75 percent. Spatially, these stations clustered in the middle city area and suburban areas.

In summary, land appraisal prices play a crucial role in determining the spatial distribution of taxi stations in Bangkok. Areas with higher land appraisal prices tend to have more taxi stations, particularly in urban and business-centric areas, while areas with lower land appraisal prices, especially in suburban regions, have fewer taxi stations. However, the presence of outliers in areas with low land prices suggests that other factors, such as transportation hubs or specific demand patterns, may also influence taxi station locations. Understanding the interplay between land prices and these additional factors is essential for urban planners and policymakers seeking to optimize taxi services and address the unique dynamics of transportation in different areas of the city.

The spatial distribution of taxi stations based on population density degree is depicted in Figure 9. Among the 80 taxi stations analyzed, 36 stations, constituting 45.00 percent, displayed a substantial spatial clustering contingent based on the population density degree. These clusters can be categorized into three distinct patterns.

The first pattern denotes a pronounced spatial clustering within regions characterized by a high population density degree, which ranges from 5540 to 13,504 persons per square kilometer. This pattern encompasses 20 taxi stations, comprising 25.00 percent of the total, and surpassing the average population density degree. These stations were spatially clustered in the central business district (CBD), inner-city area, and middle city area.

The second pattern is characterized by limited spatial clustering within regions characterized by a low population density degree, which ranges from 1441 to 4337 persons per square kilometer. This pattern encompasses 11 taxi stations, constituting 13.75 percent of the total, and falling below the average population density degree. These stations were exclusively clustered in suburban areas.

The third pattern can be characterized as an outlier, featuring notable spatial clustering within areas characterized by a relatively low population density degree, ranging from 3476 to 5187 persons per square kilometer. This range falls below the average population density degree. This pattern includes only five taxi stations, accounting for 6.25 percent of the total, and they were distinctly clustered within the middle city area.

In conclusion, the findings suggest that population density degrees play a significant role in determining the spatial distribution of taxi stations in Bangkok. Areas with higher population densities tend to have more clustered taxi stations, particularly in urban and central areas, while areas with lower population densities, primarily in suburban regions, have fewer taxi stations. The presence of an outlier pattern highlights the need for further investigation into the specific factors influencing taxi station locations in areas with relatively lower population densities. These findings have implications for urban transportation planning, service optimization, and addressing the transportation needs of residents in different population density contexts within the city.

4.3. The Agglomeration Economies of Taxi Stations in Bangkok

The results of the network analysis, as depicted in Figure 10 and

Table 3. The descriptive statistics of road accessibility of taxi stations to NGV and LPG stations.

Variables	Mean	Median	Std. Dev.	Min	Max
Road accessibility	1912.36	2052.02	748.34	23.10	2979.66

, demonstrate the agglomeration economies between taxi stations and NGV and LPG stations in Bangkok. The analysis revealed that all taxi stations in the study had road accessibility to NGV and LPG stations, with an average distance of 1912.36 m. The shortest distance between a taxi station and an NGV or LPG station was 23.10 m, indicating that some taxi stations were located adjacent to gas station areas. On the other hand, the longest distance recorded was 2979.66 m.

Table 4. The number of taxi stations that can access NGV and LPG stations classified by range of distance.

Distance from Taxi Station to Gas Station (Meter)	No. Taxi Station
0–1000	25
1001–2000	33
2001–3000	17
More than 3001	5
Total	80

further highlights that 58 taxi stations (72.50 percent) had road accessibility to NGV and LPG stations within a distance of less than 2000 m. When considering a distance range of 0–3000 m, a total of 75 taxi stations (93.75 percent) had access to NGV and LPG stations. However, only five taxi stations (6.25 percent) were situated at a distance exceeding 3001 m from NGV and LPG stations.

Furthermore,

Table 5. The number of routes available for taxi stations to access NGV and LPG stations classified by range of distance.

Distance from Taxi Station to Gas Station (Meter)	No. Routes
0–1000	45
1001–2000	106
2001–3000	159
Total	310

demonstrates that all 80 taxi stations had access to nearby gas stations within a distance of 0–1000 m, providing a total of 45 routes. As the access distance increased to 0–2000 m, the number of routes increased to 151. Moreover, by extending the access distance to 0–3000 m, the number of available routes expanded to 310.

These findings indicate that the majority of taxi stations in Bangkok have favorable road accessibility to NGV and LPG stations, with shorter distances yielding a higher percentage of accessibility.

In summary, these findings underscore that the majority of taxi stations in Bangkok enjoy favorable road accessibility to NGV and LPG stations, with shorter distances yielding a higher percentage of accessibility. This accessibility is crucial for the operational efficiency of taxi services and suggests that there are potential cost savings and environmental benefits associated with NGV and LPG adoption within the taxi fleet, given the convenient proximity to refueling infrastructure.

5. Discussion

In this research, location theory was applied to analyze the selection of taxi station locations in Bangkok, focusing on spatial factors. The findings indicated that taxi station locations were primarily chosen to maximize profits, prioritizing proximity to market areas and adjacency to energy sources. This observation aligns with the principles of location theory. However, there were specific outcomes that deviated from the theoretical expectations, as elaborated below.

Taxi stations situated in areas with the highest profit potential exhibited a pronounced degree of spatial clustering characterized by diverse land use, elevated land appraisal prices, and substantial population density. This behavior deviates from the conventional traits associated with taxi services, where taxi stations primarily serve the needs of taxi drivers rather than catering directly to passengers. Unlike public transit stations or bus stops that require placement in transport hubs within central business districts or inner-city areas to ensure passenger accessibility, taxi stations located in high-profit regions are strategically positioned within working zones, commercial districts, high-rise residential areas, and transportation hubs within the city. Consequently, taxis departing from these stations do not need to cover extensive distances to locate passengers, which contrasts with taxi stations situated in suburban areas. This spatial arrangement leads to fuel savings and mitigates the challenge of taxis having to return to the station promptly. These findings align with previous research on taxi distribution, as highlighted in the literature review section.

Table 6. The similarities and contradictions between the present study and past research.

Similarities	Contradictions
Taxi stations have the advantage of being situated in mixed land use areas.	Taxi stations primarily serve taxi drivers rather than directly catering to passengers, which is why they should not be situated in CBDs, high-value areas, or high-population areas.
Taxi stations are situated in areas where the cost of land is relatively high.
Taxi stations are positioned in regions where there is a concentration of people residing or frequenting the area.

illustrates the commonalities and disparities between the present study and previous research.

Additionally, all 80 taxi station locations analyzed in this study were found to have convenient access to fuel stations, specifically NGV and LPG stations, within distances of less than 3000 m. This observation indicates a strategic positioning of taxi stations to ensure efficient access to essential fuel sources.

In summary, this study’s findings shed light on the intricate decision making process behind the selection of taxi station locations in Bangkok. They underscore the intricate balance between profitability, accessibility, and the distinct characteristics inherent to taxi services.

The incongruent results identified in this study can be categorized into two distinct groups. Firstly, a limited number of taxi stations (only six in total) displayed a high level of spatial clustering in the lowest-cost areas. In practice, one would expect a more substantial concentration of taxi stations in such cost-effective locations. Secondly, 23 taxi stations, constituting approximately one-third of all stations, exhibited a low degree of clustering in the low-cost areas. Conversely, one would anticipate a higher degree of clustering in these areas. However, the results from the network analysis revealed that these taxi stations had convenient access to NGV and LPG stations within a 3000 m radius. This implies that these taxi stations prioritize proximity to energy sources, which aligns with the principles of location theory. This behavior could be primarily attributed to the rental conditions imposed on taxis, which mandate refueling before returning the vehicles to the taxi stations.

However, it is crucial to acknowledge that the factors taken into account in this study might be insufficient to comprehensively elucidate the outlier results. The location theory encompasses a multitude of variables that were not encompassed in this research but can exert significant influence on location selection. These variables encompass factors pertaining to the satisfaction of business owners and the presence of agglomeration economies. For example, the placement of maintenance facilities holds considerable importance for public transit stations, since buses and taxis often necessitate more maintenance in comparison to privately owned vehicles. Consequently, garages are typically situated in close proximity to taxi stations or public transit stations with the intention of economizing drivers’ time and expenses [37].

Moreover, the findings also unveiled that 28 taxi stations, constituting 35.00 percent of the total, did not manifest notable spatial clustering based on the three factors scrutinized in this investigation. Nonetheless, these stations were positioned within a 3000 m radius of NGV and LPG stations, with an average distance to fuel stations of merely 2000 m. This suggests that the positioning of NGV and LPG stations could exert a substantial impact on the determination of taxi station locations, irrespective of the existence of pronounced spatial clustering.

In summary, although certain findings deviated from the predictions of location theory, it is important to consider additional factors and influences, including business owners’ satisfaction and the proximity of related facilities such as garages, to gain a more comprehensive understanding of taxi station location selection. Furthermore, the proximity to NGV and LPG stations emerged as a critical determinant of taxi station locations, regardless of the presence or absence of significant spatial clustering related to other factors.

6. Conclusions

This research employed GIS tools to investigate the spatial distribution of taxi stations in Bangkok, with a primary emphasis on applying the location theory. The study effectively illustrated the spatial patterns of taxi stations and delved into the notion of agglomeration economies in the context of taxi stations and fuel resources. The network analysis technique proved to be a valuable tool for quantifying and examining this relationship, which is not explicitly addressed in the theory.

The study’s findings reveal that taxi stations follow a clustered spatial distribution pattern, with a significant concentration in Bangkok’s CBD and inner areas. Many of these stations are strategically positioned in close proximity to or adjoining fuel stations, a placement that offers advantages to both business proprietors and taxi drivers by reducing operational expenses. However, it is important to highlight that the conformity of taxi station locations with the land use regulations stipulated in Bangkok’s land use plan remains unclear. Taxi stations may be situated within diverse land use categories, encompassing residential and commercial zones, where their operations could potentially present challenges, including noise-related issues and other factors.

Moreover, the observation that taxi stations tend to spatially cluster in high-profit areas is noteworthy, underscoring the pivotal role of economic factors in shaping the decision making process regarding taxi station locations. This observation accentuates the entrepreneurial aspect of the taxi industry, wherein operators strategically position their stations to optimize their revenue potential.

A notable efficiency improvement is the reduction in taxis needing to return promptly to their base, thereby diminishing the practice of ‘deadheading’ where taxis return to their station without passengers. This practice is recognized for its inefficiency and negative environmental impact. The results suggest that taxi services in high-profit areas could potentially operate in a more environmentally sustainable manner.

These findings hold practical significance for urban planning, transportation policy, and the sustainable management of taxi services. They emphasize the economic motivations that lead taxi operators to strategically select their station locations, thereby enhancing efficiency for both drivers and passengers, while also aligning with urban mobility and sustainability goals.

The intricate relationship among profitability, logistical factors, and sustainability emerges as a central theme in this study, providing valuable insights into the multifaceted nature of urban transportation planning and the operational dynamics of taxi services.

Considering sustainability, it is evident that NGV and LPG are prominent fuel choices for taxis due to their cost-effectiveness compared with gasoline. However, it is essential to recognize that these fuels still contribute to carbon emissions and the release of fine particulate matter (PM 2.5), and their prices are steadily rising. Therefore, it is advisable for the government to contemplate the promotion of electric taxis as a prospective alternative to conventional taxis, taking into account environmental concerns and the long-term sustainability of the industry.

7. Recommendations

In light of the research findings, the significance of alternative fuel availability, particularly NGV and LPG stations, in influencing taxi station location decisions is apparent. It is imperative to advance sustainable transportation by fostering the expansion of infrastructure for cleaner and more environmentally friendly fuels. This support can contribute to the promotion of eco-conscious taxi operations and align with broader sustainability objectives in urban transport.

Moreover, when extrapolating these findings to developed nations, it is imperative to account for regulatory elements. Initiating this process involves an examination of the regulatory landscape governing the taxi sector in the developed nation. The goal is to ascertain how these regulatory factors might impact station location decisions and operational strategies. A comprehensive assessment should be conducted to determine whether any modifications to existing regulations are warranted in light of the research outcomes. This consideration is pivotal in ensuring the alignment of regulatory frameworks with the insights derived from this study.

Additionally, it is advisable to contemplate the integration of zoning regulations that facilitate adaptability in the placement of taxi stations. This approach can be customized to suit the distinct demands and operational dynamics of diverse regions within an urban area. By doing so, it becomes possible to accommodate discrepancies in demand patterns and logistical necessities effectively.

8. Research Limitations

This research primarily concentrated on examining the spatial arrangement of taxi stations, drawing from the foundational concepts of the location theory. The investigation relied solely on secondary data derived from public entities and Google Earth Pro, thus excluding any direct involvement of taxi station proprietors’ viewpoints or insights. Furthermore, it should be noted that the coordinates of taxi garages in Bangkok were not incorporated as variables when evaluating the agglomeration effects between taxi stations, fuel stations, and garages.

It is also essential to highlight that this research exclusively focused on exploring the connection between taxi stations and NGV/LPG stations, without delving into the dynamics of fuel station location selection. The study was primarily designed to analyze the spatial distribution patterns and agglomeration phenomena associated with taxi stations, rather than conducting an exhaustive examination of the factors that impact the decisions of fuel station placement.

Addressing these limitations could open up promising avenues for future research, allowing for a more in-depth understanding of the intricate relationships between taxi stations, fuel stations, and other relevant factors. Collecting primary data, including the viewpoints of taxi station owners, as well as incorporating garage coordinates, would enrich the analysis. Moreover, investigating the determinants behind fuel station location choices and their broader implications for the taxi industry could offer valuable insights and contribute to a more comprehensive exploration of this subject.

9. Recommendations for Future Research

Future research endeavors could consider the incorporation of a broader range of variables derived from location theory to gain a more comprehensive understanding of the factors influencing taxi station location choices. This might involve conducting interviews or surveys with taxi station owners to delve into their decision making processes, scrutinizing establishment policies to uncover specific criteria, leveraging garage coordinates data to explore the interplay between taxi stations and garages, and examining additional dimensions of taxi agglomeration economies. These avenues of investigation hold the potential to enrich our comprehension of taxi station location selection and shed light on the broader landscape of taxi agglomeration economies within Bangkok.