Abstract
Correct understanding of the residential parking characteristics and parking behavioral regularities is important to the planning of parking facilities, the selection of parking equipment, and the formulation of parking management measures in residential areas. There are few in-depth reports on the time domain characteristics of the residential parking and the parking choice behavior of developers and drivers. Besides, the influence of developers’ potential demands and drivers’ mental latent variables on the results of parking choice has been ignored. To address the gap, in this paper, we propose the concept of latent variables for parking choice based on the analysis of the time domain characteristics of residential parking in different cities and areas. We construct the SEM-BL integrated model for developers and drivers that considers the combined effect of both manifest and latent variables for the actual residential situation and analyze the factors influencing the choice of stereoscopic parking facilities from the perspective of developers and drivers in residential areas. The results show that the type of residential area, floor space, and plot ratio have a critical impact on parking choice. Residential drivers are most concerned about factors such as walking distance after parking, access convenience, pick-up time, pedestrian access, and environment, while developers choose a stereoscopic parking garage with more concern for economy, ignoring the consideration of drivers for walking distance after parking and convenience of accessing the car. Therefore, the mismatch between the needs of both sides is one of the important factors that make it difficult to solve the problem of parking in residential areas. Based on these results, we put forward a proposal for residential parking planning and equipment selection that takes into account the common needs of both developers and drivers.
1. Introduction
With the rapid development of the urban economy and the continuing increase in household car consumption, the number of private cars in China has increased dramatically [1]. Take, for example, the latest data for Chongqing Municipality, a mega-city in China with a total population of 20.329 million. In 2020, 1.488 million out of the total 2.078 million urban car ownerships are private cars, with an annual growth rate of 8.5% and average daily growth of 446. In Chongqing, every 1,000 people have 74 private cars, which means every 13.7 people have a private car. Consequently, apart from the “inherent deficiencies” in old residential areas such as low parking facility standards and obstructed parking access, there is a widening mismatch between the drastically increasing demands for private cars and the shortage of available parking space in new districts. It is difficult to obtain a parking space in many residential areas, even though its price is rising faster than the house price, which has made residential parking the most difficult problem in urban parking [2, 3]. For one thing, land resources for residential use are already limited. For another, in addition to the necessary land use for greening, roads, and public constructions, developers always take the most advantage of government planning for more commercial housing areas, resulting in a very limited supply of parking space.
Compared with other types of parking, residential parking has some special characteristics. Parking methods are usually determined by developers according to the project conditions (manifest variables), and owners have only a few choices of renting or purchasing parking spaces after moving in. Rowe identified five categories of factors associated with these characteristics, including parking supply and price, property and development characteristics, neighborhood household characteristics, accessibility, and building form and development patterns [4]. Zhang’s study results demonstrate that parkers’ sensitivity to parking searching time, walking distances, the economy, and their environmental awareness, along with their commuter parking payment method, are relatively more significant and influential factors [5]. In this case, the traditional disaggregate theory, if used, can only introduce manifest variables concerned by developers while ignoring drivers’ psychological variables of parking choice behaviour and other latent variables that cannot be observed directly. This makes it impossible to obtain the real parking preferences that take into account the common needs of both drivers and developers, which reduces the explanatory power of the model.
To address the aforementioned issues, this study introduces the concept of drivers’ psychological latent variables, which creates conditions to study the behavior of parking choices from the perspective of drivers’ subjective psychology. However, the biggest difference between latent variables and manifest variables is that latent variables cannot be directly measured statistically, and other measurable variables (observed variables) need to be sought in the research process. For example, it is impossible to measure the latent variable of parking convenience in residential areas directly, but it can be measured by its observed variables such as walking distance after parking, satisfaction with parking guidance facilities, suitability of parking space size, and access convenience of parking facilities. These observed variables can be measured by objective indicators such as the Likert scale.
Residential parking choice behavior involves many variables, thus constituting a complex hierarchical structure. Traditional methods such as regression analysis can no longer explain the relationship between latent variables and the relationships between measured and latent variables. These problems are well handled by multivariate statistical models (SEM), which are widely used in management science and psychology, but less applied in transportation and never applied to the study of parking behavior in residential areas. The contributions of this study are summarized as follows: (1) By integrating the SEM model with the BL model in the disaggregate model and drawing on the data obtained from a residential parking survey, this paper establishes the parking choice SEM-BL integration models for developers and drivers, respectively. (2) The models are solved according to the solution steps provided in the paper [6], and the parking preference behavior is analyzed from the obtained model results. (3) Finally, the survey data and the mathematical model are combined to derive the most concerned indicators of residential drivers about the convenience, safety, and parking service environment of stereo parking facilities, which provides the basis for residential parking planning and the selection of stereo parking equipment.
1.1. Literature Review
The breakthrough point to solve the difficulty of parking in residential areas is how to satisfy people’s demands regarding living comfort and living environment while maximizing land savings and improving space utilization. Therefore, the main measures to ease the conflicts between the existing parking supply and demand are to improve the parking allocation index and develop stereo parking equipment. In terms of research on parking in residential areas, the hot issues studied include the following:(1)The application of multilevel stereo parking garages in residential areas(2)The present situation of parking problems in residential areas(3)Parking environment in residential areas(4)Parking demands and parking layout planning in residential areas(5)Parking behavior and parking policy development.
Multilevel stereo parking garages are increasingly used in residential areas. Larionov and Larionova introduced the advantages and disadvantages of the arrangement of underground parking lots in residential construction and their research results proved the prospect of arranging underground parking lots in residential areas [7]. According to Duvanova et al., to relieve the difficult problem of parking in residential areas, attention should be paid to optimizing the use of parking space in residential areas on the criterion of maximizing the use of intrayard space [8]. They held that an underground garage is an effective measure to alleviate the lack of parking spaces, and a comprehensive assessment of the completed effect is necessary [9].
In order to alleviate the problem of difficult parking in residential areas, Wang et al. probed the countermeasures to the parking problems in residential areas in terms of parking allocation index and parking facility supply [10]. Molenda and Sieg established a model to investigate the trade-off between privileged parking for residents and economic vitality in terms of the product variety available in a vibrant city district and identified situations in which assigning on-street parking spaces to residential parking constitutes an optimal policy, both from a welfare and a resident perspective. They found, however, that privileged parking for residents is unlikely to result in a first-best allocation of on-street parking space if an efficient level of economic vitality is to be ensured at the same time [11]. Seo and Kim investigated the actual parking conditions in a low-rise residential area in Nangok-dong, Seoul, and proposed some ideas for solving the complex parking problem in low-rise residential neighborhoods [12]. Scheiner et al.’s study indicated that high-density downtown residential areas are often characterized by severe parking pressure and illegal parking, which can be significantly reduced by adopting measures like introducing parking fees for residents of private parking lots, issuing paid parking permits for residents who provide parking spaces, defining short-stay parking zones, providing parking spaces of different sizes, and increasing the level of enforcement [13].
Studies related to parking environment are particularly pertinent as this paper analyses the factors that are of most concern to drivers in residential areas from a quantitative perspective. Park et al. suggested that the green parking project has the potential to improve the thermal environment of outdoor spaces for pedestrians [14]. According to Ma et al., parking affects the living environment in residential areas. Therefore, it is a priority to address the parking problems of private automobiles and create a harmonious living environment between humans and automobiles in residential areas [15]. Won et al. proposed a residential parking environment evaluation model based on hierarchical analysis, and the results of the study show that the model can quantitatively determine the changes in the residential parking environment [16].
Li et al. gave some suggestions regarding parking methods and the layout of parking spaces in old residential quarters based on their empirical study [17]. Zhang et al. introduced a double-objective model that considers both utilizing rate and walking distance, and the results of their study demonstrate that the model increases the occupying rates of parking lots in residential zones while decreasing the walking distance [18]. The study by Chullabodhi et al. shows that the parking decision of developers is not determined by parking demand but by market demand. And the residential parking capacity depends on the total floor area and the number of residential units in the project, the distance to the nearest transit station, and the average unit price [19].
Since the end of 2019, COVID-19 has become a public health emergency of international concern, and the perceived severity of the outbreak and the rise of lifestyle patterns such as telecommuting have caused a shift in the established travel mode selection habits of the population. Through a survey, Jiang et al. found that 41% of the people in China choose to travel by private car, 35% up from the nonepidemic period [20]. The use of private cars increased significantly during the epidemic, so parking fees had a more significant negative effect than before the epidemic.
In terms of parking policy development research, Stewart et al. investigated behaviors during a common task of residential parking for drivers, and the results reveal that individuals show relatively consistent parking behaviors under the same geospatial context, and the standard deviation of the deceleration threshold has a larger discrepancy between couples parking at different residences than within couples when parking occurs at the same place [21]. Chen et al. compared commercial vehicle parking regulations and violations in commercial, mixed-use, and residential land use areas in New York City [22]. Christiansen et al. found that reducing access to free workplace parking stands out as one of the most effective ways of reducing car use on work trips, and parking restrictions will have the greatest effect in compact cities [23]. Ommeren et al.’s study shows that in the city center, the waiting time for a permit is up to four years. The results indicate that one additional year of waiting for a parking permit reduces car ownership by 2 percentage points, corresponding to price elasticity of car demand of −0.8 [24]. Zhan’s study shows that providing free on-street parking around residential areas induces an increase in private car ownership of nearly 9 percent [25]. Seya et al. analyzed the residential parking rent price elasticity of car ownership in Japan and the result indicates the net effect of a price increase may be that noncar ownership increases in mega cities and one-car ownership increases in other cities [26]. Ommeren et al. proved that the provision of residential parking permits in shopping districts induces a yearly deadweight loss of at least euro 500 per permit, which is about 30% of the supply cost of a parking place in shopping districts [27]. The study of driver parking in Doha, Qatar, shows that “Intelligent Parking Space Detection” was chosen as one of the factors affecting people’s choice of parking significantly more often than amenities such as “Wider parking spot.” The findings indicate that future parking investments may be better directed towards smart parking solutions [28]. Antonson analyzed changes in car use, car ownership, spatial parking patterns, and the consequences for the everyday life of residents in a housing area with a relatively restrictive parking requirement in Gothenburg, the second-largest city in Sweden. The findings suggest the need for a comprehensive approach to parking policy by introducing more restrictive parking requirements in parallel with other measures, such as raising parking charges and decreasing the number of public parking spaces [29].
The above research achievements have played a positive role in relieving urban residential parking problems and formulating residential parking development plans. However, as different countries have different situations, it is inadequate for the existing research to focus mainly on residential on-street and internal parking. Besides, this body of research ignores the latent variables such as parking feelings and parking services that cannot be directly measured or accurately quantified, resulting in mathematical models that fail to truly reflect the parking choice behavior and process of drivers, which in the end, weakens the explanatory power of these models.
In view of the above problems, this study selects different types of residential areas from representative cities in China, such as Chongqing, Guangzhou, and Xi’an, to carry out parking data collection and parking behavior survey. It then establishes the SEM-BL integrated model of residential parking behavior by applying SEM-logit theory and explores the preference behavior of developers and drivers regarding stereo parking facilities through the structural model and measurement model of the established integrated model so as to provide theoretical support for the planning of stereo parking facilities and the selection of parking equipment in residential areas.
2. Methodology
This section explains the integration process of the SEM model and the BL model in the nonset counting model, followed by a detailed explanation of the SEM-BL mathematical model.
Generally, the structural equation model (SEM) is a comprehensive statistical method to describe the relationship between variables based on the analysis of the covariance matrix of variables, which is essentially an extension of the general linear model. It has been widely used in traffic research since 1980 [30]. This paper applies the SEM model to the study of residential parking behavior, using SEM to analyze the factors influencing residential parking patterns. SEM is formed with measurement models and a structural model. Measurement models depict the relationship between observed and latent variables, while the structural model depicts the relationship among latent variables [31].
The Binary logit model (BL) model refers to a set of only two options where the developer and the driver choose a certain parking method among the options. In this paper, the BL model is used to explain the intention and the probability of occurrence of parking options for the developer and the driver.
To illustrate the behavior of residential parking method choice well, SEM and BL models are combined to construct an integrated SEM-BL model. Simićević developed a BL model to identify and quantify the characteristics of users and trips, on which the acceptance of parking price is dependent [32]. Peng quantitatively analyzed the influence of latent variables, such as parking convenience, safety, and parking environment, on the parking choice behavior in old residential areas and developed an SEM-logit integration model with latent variables and explicit variables [33]. Chen constructed an SEM-BL model of elderly people’s public transportation travel decision-making behavior considering four aspects: personal attributes, travel attributes, environmental attributes, and psychological latent variables [34]. Si introduced latent variables into the logit model, setting up an SEM-Logit model to explore the mode choice patterns between taxis and online car services. The results showed that the SEM-Logit model with the latent variables is better than a general Logit model in terms of model precision and hit ratio [35]. Lin considered the traveler’s psychological factors and constructed an SEM-Logit integration model of travel choice behaviors. The results indicated that the SEM-Logit integrated model has higher accuracy and explanatory ability, and latent variables have a significant impact on travel choice behavior [36].
The SEM-BL integrated model for residential parking mode selection is an integrated model with SEM as the superstructure and BL as the substructure, and its solution process is completed by the following two-stage construction method:(1)SEM stage. Based on the valid data, the loading coefficients between the latent variables and the selection intention and the loading coefficients between the latent variables and their observed variables are solved.(2)BL stage. Based on the structural metric equations of the latent variables obtained in stage 1, the standard values are obtained through the functional processing of the structural metric equation results, and then the characteristic variables of the same driver are uniformly assigned to the logit function and solved to obtain the parameter estimation results of the characteristic variables of the model.
3. Prerequisite Hypothesis
The framework of the SEM-BL integration model is shown in Figure 1. The assumed condition of the model is given as follows:(1)The residential area developers rationally choose what they think is the optimum parking method according to the location conditions, residential area type, relevant planning requirements, economic and technical indicators, etc.(2)According to the actual development of modern residential areas, the parking method can be divided into self-propelled stereo parking garages and mechanical stereo parking garages.(3)According to random utility theory, drivers’ preferences for parking methods depend on a utility function U containing manifest and latent variables.(4)The result of the stochastic error term of the function divided by the Utility Function is submissive to the Gumbel distribution (independent of the distribution mean value of 0), and the rest functions are submissive to the normal distribution.
4. Mathematical Formulation
Based on the random utility theory, residential developers and drivers (hereinafter generally referred to as samples) pursue the maximum “utility” when choosing parking methods. The parking method collection preferred by sample n is Ai (i = 1, i = 2). Let us take Uin in equation (1) as a utility of the nth parking method chosen by the ith driver.
In equation (1), Vin is the fixed term in the utility function of the ith scheme chosen by the nth driver, εin is the probability item, and assume that variables contained by Vin are linear relation.
In equation (2), is the vector of unknown parameters, is the vector of factors influencing sample n’s preference for parking method i.
In equation (3), m is the number of manifest variables in the preference scheme, q is the number of driver’s manifest variables, l is the number of latent variables, pimn is the manifest variables in an alternative plan, diqn is car drivers’ manifest variables, ηiln is latent variables, , and are unknown parameters.
The fitted relation of latent variables ηiln is determined by SEM, ηiln can be expressed by related manifest variable yirn in equation (4) or its corresponding measurement index xitn in equation (5).where yitn is the related manifest variable associated with a latent variable, αiln is the parameter to be estimated, xi,r,n are a series of measured variables corresponding to latent variables, δiln, ζirn are errors-in-variable, t is the number of manifest variables related with latent variables. r is the number of the measured variables corresponding to latent variables.
U1n, U2n are the utility of the preference of plan 1 and plan 2, respectively. The Binary Variable is introduced to explain the parking method preference behavior:
According to random utility theory, this paper derives the probability of the nth driver choosing the ith parking method in the SEM-ML model in the case of considering latent variables.
According to utility maximization theory, the selection of alternative scheme 1 means that U1n exceeds the utility U2n:
Based on probability theory, selective probability Pin of alternative scheme i is deduced further when the random error term εin is Gumbel distribution.where Pin is further deduced and simplified into the following:
4.1. Residential Parking Survey
The paper selected typical large cities in China, such as Chongqing, Guangzhou, and Xi’an, to collect parking data in ten residential areas in April 2017 and investigated parking behavior by inquiring residents one-to-one and face-to-face in June 2019 (Table 1, Figure 2). 150 questionnaires (150 valid questionnaires, 100% effective rate) were distributed to the developers, the property management companies, and parking equipment suppliers. 320 questionnaires were sent to residential drivers and 300 valid questionnaires were collected (93.75% effective rate). The ratio of male to female in the driver behavior survey was about 3 : 1, with 36% aged less than 30, 58% aged between 30 and 50, and 6% aged over 50.
(a)
(b)
(c)
(d)
(e)
(f)
The survey found that the parking resources in the residential areas that were built and occupied earlier were basically tense to varying degrees, which was due to the high occupancy rate of the residential areas built earlier and the small number of parking spaces allocated at that time. The parking resources in the residential areas occupied after 2011 can basically meet the demand, with some areas even having parking space utilization rates of less than 40% due to low occupancy rates. However, it was found through the property management staff and residents that despite the low utilization rate of some of the existing parking spaces, the planned parking spaces have been sold out. Therefore, as the occupancy rate continues to improve, the newly built areas will still face parking problems.
5. Factors Influencing Residential Parking Methods
Parking characteristics and behaviors in different regions and different residential areas are quite different and are influenced mainly by the following latent variables.
5.1. Location of Residential Area
In residential parking design, residential area location is a factor that should be considered according to the local conditions. The demand for private cars is more urgent in the suburbs and new urban areas because the surrounding living and service facilities are relatively imperfect, and the travel distance is longer. In contrast, the central areas of cities have perfect living and service facilities, and public transportation is more convenient, and the central areas usually adopt certain restrictions or price measures to regulate the number of private cars by increasing the cost of using cars. Therefore, the parking allocation standards in most cities take the location factor into consideration, such as the “Technical Regulations on Urban Planning Management in Chongqing” in China, where the parking allocation standards are divided into the central area and the area outside the central area.
5.2. Residential Plot Ratio and Population Density
The plot ratio is the main indicator of the development intensity of residential areas. China’s real estate development is gradually developing from the original bungalow to multistory, small high-rise and high-rise, and high-rise residential areas with high development intensity are the general trend of real estate development. There is a positive correlation between population density and plot ratio, and the increase in plot ratio means the number of residents per unit of land area increases (Table 2). Europe is more mature in terms of residential parking development methods. For instance, the relationship between population density, plot ratio, and parking options in residential areas is projected in the British Housing Design in Practice [37].
5.3. Type of Residential Area
Based on the service characteristics, the residential area can be divided into high-level residences (villas, garden houses), mid-level residences (high-quality residences), ordinary residences (small dwelling-size residences or apartments), public rental housing, temporary dwelling, low-rental house, and so on. The type of residence is related to the economic level of residents. The higher the economic level, the more people are willing to choose higher level residences, and the corresponding ability to purchase cars is stronger. In addition, due to the large differences in the architectural characteristics of the old and new residential areas and the living standards of the residents, they need to be studied separately in the selection of parking methods and layout planning. Therefore, different types of residential areas should choose the parking method according to the needs of the owners.
5.4. Household Type in Residential Area
Median and high-level residential areas are mainly made up of large units, while apartments, public rental houses, and low-rent houses are mainly made up of small and medium-size units. The higher the economic level, the more people are willing to choose larger residential units, and the stronger the corresponding ability to purchase cars. In Chongqing, different household types correspond to different parking space allocation indexes, and 100 m2 is taken as the basis for distinguishing middle and high-level residential buildings from ordinary residential buildings (Table 3). In Shanghai, different household types and different residential areas correspond to different parking space allocation indexes (Table 4).
5.5. Traffic Environment Surrounding Residential Area
The traffic environment surrounding residential areas mainly refers to the accessibility of urban roads and public transport environment. When the road network around the residential area is of strong accessibility, and compared with other transportation means, traveling by private cars is more economical and convenient, the residents’ desire for the car, the proportion of car -traveling, and parking space demand will increase, and the parking method changes accordingly.
The above analysis shows the main manifest variable that affects the choice of parking methods in residential areas are location condition, plot ratio, population density, population type, household type, and surrounding traffic conditions.
6. Space and Time Characteristics of Residential Parking
Considering the influence of the above manifest variables and drawing on the results of the questionnaires, the following laws of parking characteristics and parking behaviors are obtained.
6.1. Saturation
Residential parking saturation means the ratio of the actual number of parked vehicles to the number of parking spaces at a certain moment, and it is greatly influenced by the factors such as weekdays/holidays and the type of residential area. By comparing the data of Longhu Nanyuan (a high-level residence) and the Ao Yuan District (a median-level apartment), the Sweet Central Department (a small-size apartment) on April 11 (Wednesday) and April 14 (Saturday) 2018, the following analysis is obtained.
Figure 3 reveals the statistical pattern of allocated parking facilities in residential areas on working days. Regardless of the type of residential area and work, the residents’ parking behavior shows peak and trough effects due to similar commuting times. In terms of the saturation of parking facilities, it shows the characteristics of concentrating on driving out of the residential area in the morning peak and returning to the residential area in the evening peak, with relatively low daytime saturation, close to saturation or supersaturation in the night-time utilization, and small peak fluctuations at noon, showing the characteristics of strong, rigid demand.
Figure 4 analyzes the statistical pattern of parking facilities in residential areas on holidays. It also shows the characteristic of greater night-time parking saturation. However, because of the sharp decline of the proportion of residents traveling for commuting purposes, residents’ purpose of travel is random and diversified, the peak and trough of parking saturation change more gently compared to weekdays, and some small apartments show a random distribution.
From the comparative analysis of the type of residential area and the time of occupancy (old and new district), it is found that the difficulty of parking and rigid demand characteristics are positively correlated with the level of the district. Longhu Nanyuan, a high-level district, has a total household of 1176, 156 of which are located in the villa area. According to a Nanyuan district staff, the proportion of households with two cars in the district is about 50%, and the proportion of households with three cars is about 30%. The saturation curve shows that the night peak saturation is >125% in Longhu Nanyuan (high-level district), >98% in Ao Yuan District (mid-level district), and >76% in Sweet Central Department (a small-size apartment).
The peak and valley effects of parking saturation in the high and mid-level residential areas have a strong regularity of change, with the peak coefficients being concentrated in the morning and evening peaks on weekdays, while the small apartments are mostly randomly distributed. This is because the small apartments are mostly investment properties, with a large proportion of houses rented out for office or commercial use, and the complex composition of travel and parking purposes causes strong randomness of parking behavior characteristics.
The low utilization rate of parking facilities in residential areas during the day results in a great waste of parking resources. As the peak saturation of residential parking and daytime commuter parking is complimentary, the variability of parking saturation in residential areas makes it possible to develop reasonable parking space sharing measures.
6.2. Concentration Factor in Morning-Evening Peak (PCF)
Peak traffic volume is an important indicator of the traffic environment in residential areas, reflecting the concentration of vehicle access in residential areas. The above analysis shows that the type of residential areas and time periods have a large impact on parking saturation. Therefore, the ratio that is used to define the traffic volume in early peak and in late peak is defined as the early peak factor (PCFa) and late peak factor (PCFp), respectively. The data of six residential areas, including Longhu Nanyuan, Ao Yuan District, on April 11 (Wednesday) and April 14 (Saturday) 2018, are used as an example to compare and analyze the early peak (7:00-9:00) and late peak (17:00-19:00) concentration factor of morning and evening peaks in residential areas.
Table 5 shows that regardless of weekdays and holidays, the morning and evening peak concentration coefficients show a pattern of PCF mid-level neighborhood > PCF high-level neighborhood > PCF ordinary apartment, which is consistent with the peak and trough effect change curves of different categories of neighborhoods in Figures 3 and 4. This is partly due to the large proportion of households with more than two cars in high-level neighborhoods and the existence of certain flexible working hours for high-income families. Besides, most residents who live in mid-level communities are working class, and their parking behaviors are influenced by their strict commuting time rules. In ordinary apartments, because of the diversity of its residents, the complexity of their parking purpose weakens the peak effect. Therefore, the residential parking pressure in morning-evening peak is the mid-level district > high-level district > ordinary apartment, which is the basis of the selection of parking facilities.
6.3. Parking Duration (H) and Parking Turnover Rate (V)
A set of relative parameter-parking time and parking turnover rate determines the efficiency of the parking lots. The data of Longhu Nanyuan (high-level residence) and Ao Yuan District (mid-level residence), Sweet Central Apartment (small apartment) on April 11(Wednesday) 2018 are used to analyze the parking duration in different types of residential areas (Figure 5).
The analysis of parking duration and turnover rate in residential areas reveals the existence of both long and short time parking, with long time parking dominating. In the case of Chongqing, the parking duration longer than 10 hours accounts for 31.8% ∼ 62.5%. The reasons for the variability of parking duration are that the households with more than two cars in high-level residential areas account for a larger proportion, there are vacant vehicles in households occupying parking spaces for a long time, the travel for commuting purposes in mid-level residential areas account for a larger proportion, the parking time of vehicles is more regular, and the parking length distribution of ordinary apartments is more balanced due to the diversity of travel purposes. Specifically, it shows the feature of H high-level district > Vmid-level district > V ordinary apartment, V old district > Vnew district.
Table 6 shows that the difference between weekday and holiday turnover rates in high and middle level residential areas is not significant, except for apartments. The weekday turnover rate is 0.96 ∼ 1.06 and the holiday turnover rate is 0.98 ∼ 1.12. It shows the feature of Vordinary apartment > Vmid-level district > Hhigh-level district, Vold district > Vnew district.
6.4. Walking Distance
Modern residential planning emphasizes the design idea of “separation of pedestrian and vehicular traffic, open surface space.” The researched residential parking facilities are divided into two types according to the construction form: attachments to the ground floor of the building and single-built centralized parking garages. For the attached parking facilities, the owner’s parking space is usually under the living unit, which can be easily reached by elevator, and there is a walking distance of 50–100 m inside the garage. For the single-built centralized garage, after parking vehicles, drivers need to walk home. Thus, there is a certain walking distance. Figure 6 shows the result of the survey on the walking distance acceptable to drivers.
The walking distance acceptable to drivers who park for residential purposes is greater than that of drivers for other parking purposes. A comparative analysis of drivers’ parking behavior from different neighborhood types shows that for 97-98% drivers, the acceptable walking distance is within 300 m, and drivers in high and middle level residential areas with better living environments require a shorter walking distance. The result shows that 41.68% of drivers in high-level residential areas have an acceptable walking distance of <50 m after parking, 32.61% of drivers in mid-level residential areas have an acceptable walking distance of 50∼100 m, and 28.64% of drivers in ordinary apartments have an acceptable walking distance of 100∼150 m.
6.5. Drivers’ Decision-making Factors
The survey of drivers’ decision factors in residential parking shows that although the residential area is constructed at different times and has different characteristics, drivers’ decisions on parking facility settings are basically the same: drivers are most concerned about parking safety, expecting pedestrian-vehicle separation, centralized parking, and short walking distance, and drivers are less concerned about parking fees due to the greater rigid demand for parking in residential areas.
7. Modelling and Analysis of Parking Preference Behaviour
7.1. Developers’ Parking Preference
Preference variables that affect residential parking methods should balance the common needs of drivers and developers. To this end, we designed an initial questionnaire before the survey, conducted a preliminary investigation and revised the questionnaire, tested its reliability, purification, and validity, and finally determined the model variables of developers and drivers.
The parking methods mainly include plane parking (centralized plane parking, decentralized on-street parking), and stereo parking garages (self-propelled and mechanical parking garages). Currently, reinforced concrete self-propelled parking is most popular, followed by mechanical parking, and plane parking is seldom used in newly built residential areas. In most cities in China, residential parking facility standards and specifications mandate that “residential construction should include indoor parking space, outdoor parking space is not included in the facility index of the residential building construction, but mechanical parking space is included in the index”.
Therefore, this study defines the choice of parking preference as stereo parking garage (n = 1) and plane parking garage (n = 2) and takes residential type, land area, and plot ratio as the manifest variables. Our choice of latent and observed variables takes into account the actual situation of the survey and the development trends of the times. As 82.95% of residential mechanical garage equipment is a lift-sliding garage by the end of 2020, self-propelled garage, lift-sliding garage, and storage garage (plane mobile, storage-retrieval) are taken as latent variables, and the applicable conditions for the construction of various parking garages are taken as observed variables, which are measured by Likert scale5 (Table 7).
As shown in Table 8, the results of the structural equation calculation consist of two parts: the causal relationship between manifest variables of residential areas and preference latent variables (structural modes) and the relationship between measured variables and preference latent variables (measurement modes). Residential type (TR) is positively related to developers’ preference of the self-propelled parking garage, with higher level residential areas preferring self-propelled garages.
Residential area (AC) is negatively correlated with developers’ preference for warehouse and vertical lift, indicating a preference for mechanical parking garages with small footprint and high automation for residential areas with tight land use. Plot ratio (PR) is negatively correlated with developers’ preference for self-propelled garages but positively correlated with developers’ preference for warehouse garages, indicating a preference for self-propelled garages in residential areas with small plot ratios and a preference for lift-sliding or warehouse mechanical stereo garages in residential areas with large volume ratios.
The calculation results of the measurement model show the most influential indicator for developers to choose self-propelled parking is land occupancy (PZZ1) and construction cost (PZZ2), and the least influential is walking distance after parking (PZZ3). The most influential indicators for developers to choose lift-sliding parking is meeting the planning index (PSH1), and the least influential is parking convenience (PZZ2). The most influential indicators for developers’ selection of warehouse parking are meeting the planning index requirements (PCC2) and equipment and maintenance costs (PCC3). The most influential indicators for developers to choose vertical-lift stereo garage indicators are the requirement to satisfy planning index (PCS2), design and maintenance cost (PCS4), preferential policy support (PCS5), and the least influential is parking convenience degree (PCS3).
7.2. Drivers’ Parking Preference
The factors influencing drivers’ parking preferences in residential areas include not only manifest variables of driver characteristics (e.g., gender, education, vehicle price, and driving age) but also drivers’ perceptions and attitudes. The manifest, latent, and observed variables included are shown in Table 9. The latent variables are described by their observed variables, which are measured by the Likert scale 5.
The structural equation model calculated by AMOS and SPSS consists of two parts (Table 10): the causal relationship between drivers’ manifest variable and preference latent variable (structural modes) and the relationship between measured variables and preference latent variables (measurement modes).
The results of the structural model show that female drivers pay more attention to the convenience of stereo parking facilities than male drivers. The age of drivers is positively correlated with parking experience, which indicates that older drivers pay more attention to parking experience. Vehicle price is positively correlated with the safety of parking facilities, which indicates that high-income drivers are more serious about the safety of parking facilities. High-education drivers pay more attention to the service environment of parking facilities.
The results of the measurement model show that the most important indicator for the convenience of drivers in the residential area is walking distance after parking (PN1) and the access convince of the parking lot (PN4). The most important indicators that affect the feeling of drivers to access the car are the satisfaction with pick-up time (PF2) and the availability of multimedia information (PF3). The most important indexes for the safety of parking facilities are the supervision of the management (PS1) and pedestrian access settings (PN2). The most important indicators for the parking service environment are environment from the parking garage to residential area (PE3) and air condition (PN1) in the parking lot.
8. Conclusions
Based on the analysis of time-domain features of parking in urban residential areas, this paper constructs an SEM - BL integration model of developers and drivers under the coaction of latent variables and manifest variables. The model verifies that the type of residential area, land area, and plot ratio play a decisive role in developers’ parking preference. Developers’ choice of the stereo parking garage is to meet the planning requirements in the most economical way, but it ignores drivers’ consideration for walking distance and parking convenience. The main factors that affect the parking choice of drivers in the residential area are the walking distance after parking, the unimpeded passage of access, and the time to pick up the car. Therefore, the root cause of residential parking conflicts that makes stereo parking a “castle in the air” is the poor user experience caused by the mismatch of the needs of developers and drivers.
The results of our study can be replicated in any case. The quantitative analysis of the model brings the following insights to residential parking planning and equipment selection: the high-level district with a small plot ratio should adopt a self-propelled parking garage, while the apartment and ordinary residential area with a small floor area and large plot ratio can adopt the combination of self-propelled and mechanical or pure mechanical parking according to local conditions. It also shows that the preferential subsidy policy of encouraging residential areas to build mechanical stereo parking garages with high space utilization and a high degree of intelligence plays a positive role in the benign development of stereo parking facilities.
Our study also yields some other interesting results: the demands of female drivers for convenience and service environment are higher than that of male drivers; older drivers pay more attention to the parking feeling of stereo parking facilities; high-income drivers are more serious about the safety of stereo parking facilities; highly educated drivers pay more attention to the service environment of stereo parking facilities; walking distance after parking and the access convince of the parking facilities are the most important factors for the convenience of parking facilities in residential areas; management supervision and pedestrian access settings are the most concerning indicators for the safety of stereo parking facilities; environment from the parking garage to residential areas (PE3) and air condition (PN1) in the parking lot are the most important indicators for service environment of parking in residential areas. These results also provide the basis and theoretical support for residential parking planning and the selection of stereo parking equipment.
The SEM-BL integrated model of driver parking behavior developed in this paper has some portability, and its application and effectiveness in other areas of traffic behavior research, such as P&R interchange parking behavior, need to be further developed. It is also practical and reasonable to factor in such latent variable factors as price leverage and parking policy.
Data Availability
The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study.
Conflicts of Interest
The authors declare that they do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.
Acknowledgments
This work was supported by the Science and Technology Research Program of Chongqing Municipal Education Commission, “Research on Applicability Identification Mechanism and Site Selection Model of Public Stereo Parking Facilities for Multi-party Demand Balanced Optimization” (Grant no. KJQN201900706).