Structural equation modeling approach to evaluate effects of socio-economic and built environment factors on trip generation rates of commercial land use

Abstract

Understanding trip generation at large-scale commercial land uses is essential for effective urban transport planning, particularly in rapidly growing cities in developing countries. This study investigates trip-generation patterns at extra-regional commercial centers in Mashhad, Iran, using an exploratory structural equation modeling (SEM) framework that jointly examines physical attributes, built-environment characteristics, and socio-economic factors. Using observed trip counts from 33 extra-regional commercial sites, two dependent measures were analyzed: the total number of trips and trip rates normalized by floor area (trips per 100 m²). The findings suggest that, within the analyzed sample and under the current exploratory model specification, physical scale and surrounding built-environment characteristics are associated with higher total trip volumes, whereas the examined socio-economic indicators did not show statistically significant associations. In contrast, no statistically significant relationships were identified between the examined explanatory factors and trip rates normalized by floor area.Given the limited sample size and model-fit constraints, the results should be interpreted as exploratory rather than confirmatory. This study provides preliminary empirical evidence on commercial trip generation in a data-constrained context and underscores the importance of distinguishing between total trip volumes and density-based trip measures. Future research using larger samples and alternative modeling approaches is needed to validate and extend these findings.

Introduction

Changes in land use exert substantial influence on urban traffic conditions, making their analysis essential for effective transportation management. Research has consistently shown that land-use transformation can intensify traffic congestion, alter travel demand, and reshape urban development patterns¹. In many developing countries, revisions to comprehensive and detailed urban plans often result in increased demand for commercial land uses. This rising demand is primarily driven by higher profitability compared to other land uses, economic incentives², urban sprawl³, gentrification processes⁴, and persistent planning challenges⁵.In traffic impact assessment, accurately estimating trip generation and trip attraction rates for commercial developments before construction is critical. Improved accuracy at this stage enhances modeling quality⁶, forecasting precision⁷, and overall traffic management outcomes⁸. Traditional evaluation approaches have relied heavily on the physical and structural characteristics of land uses—such as plot size, infrastructure capacity, and building density—to derive coefficients for estimating trip production and attraction. The Institute of Transportation Engineers (ITE) Trip Generation Handbook⁹ remains the foundational reference for these practices, providing extensive statistical categorizations. However, it is important to note that the handbook’s recommended rates are not suitable for mixed-use complexes, including large shopping malls¹⁰. Globally, significant efforts have been made to refine trip-generation coefficients to better match real-world conditions. One example is the 2013 study conducted in Mashhad¹¹], which applied detailed statistical techniques across various land-use groups to improve reliability. Findings across multiple studies indicate that, beyond physical factors, built-environment characteristics and socio-economic conditions also strongly influence the number of trips a land use generates. Yet, most previous research investigates these factors in isolation rather than through integrated frameworks.

Schneider, et al.¹² proposed methodologies to refine ITE estimates by incorporating factors such as residential and employment density, transit service quality, and building setbacks during peak hours, improving prediction accuracy in smart-growth environments. Currans, et al.¹³ criticized conventional forecasts for lacking sensitivity to socio-urban contexts, leading to inflated estimates that can distort impact-fee calculations and complicate development processes. Their research, based on affordable housing projects in Los Angeles and San Francisco, highlighted the importance of variables such as parking supply and job density.Other work further supports the role of local characteristics: De Gruyter, et al.¹⁴ examined multimodal trip generation in the UK and Ireland using regression models that included car ownership, housing type, and amenity proximity. Chen, et al.¹⁵ explored multimodal trip generation using smart-card data to integrate socio-economic attributes and built-environment factors. In the context of developing countriesAhmed, et al.¹⁶ identified significant deviations from ITE rates along Dhaka’s Mirpur Road and emphasized the influence of economic conditions and adjacent land uses. Mirzaei and Mignot¹⁷ used logistic modeling to examine the influence of attitudes, built-environment factors, and travel purposes on mode choice. Mohammadpour and Mehrjou¹⁸ applied the WASPAS decision-making model to analyze socio-economic variables affecting intra-city trips in Rasht, Iran. Machine learning approaches have also been introduced in recent years. Xiao, et al.¹⁹ developed a multidimensional CNN–LSTM framework to analyze spatial interactions between land-use features and trip generation, considering attributes such as housing type and proximity to central retail clusters. Doulabi, et al.²⁰ applied multiple linear regression to small and medium-sized cities in Louisiana, incorporating factors such as floor area and population densityAboelenen, et al.²¹ created regression models for residential trip generation based on household size, income, and vehicle ownership. Because trip-generation research in developing countries is still limited, several studies have focused on specific land uses. Altaher, et al.²² estimated trip-attraction rates for banks in Zagazig, Egypt using metrics such as gross area and employee count. Anand and George²³ proposed a robust modeling framework suitable for contexts with sparse or noisy data.Collectively, these studies confirm that built-environment factors (e.g., road-network density, transit availability, land-use mix) and socio-economic factors (e.g., population density, household income, car ownership) jointly shape urban travel behaviorRith, et al.²⁴ argue that overlooking these integrated factors leads to inaccurate trip-generation forecasts in developing countries, potentially misaligning predicted demand across travel modes and prompting unsuitable design solutions.

Despite this evidence, only a limited number of studies have examined these factors simultaneously within comprehensive modeling frameworks. Moreover, the majority of available research stems from developed countries, leaving cities in developing nations—where socio-economic conditions, infrastructure, and travel patterns differ significantly—understudied.

The present study aims to address this gap by analyzing the determinants of trip-generation rates for commercial land uses in Mashhad, Iran. Unlike earlier work relying primarily on linear regression, this research employs Structural Equation Modeling (SEM) to analyze complex relationships between observed and latent variables. SEM enables the simultaneous assessment of socio-economic, environmental, and physical variables and their direct and indirect interdependencies. The core research question investigates which of these factors exert the strongest influence on commercial trip generation, and whether SEM provides better explanatory power compared to traditional regression approaches. The findings aim to support planners and policymakers in formulating effective traffic-management strategies and refining traditional trip-generation estimation methods.

Literature review

Land-use patterns exert profound effects on urban systems, especially transportation networks, and are fundamental to sustainable metropolitan planning. Urban mobility is strongly shaped by variables such as land-use density, activity distribution, and the spatial arrangement of services²⁵. Integrating transportation and land-use planning yields major benefits, including reduced environmental impacts, enhanced economic efficiency, improved public spaces, lower congestion, and more livable urban environments. This integration is particularly important in rapidly motorizing nations like China and India²⁶. Trip generation is widely regarded as a key parameter linking land use and transportation development, forming the foundation for transportation planning and traffic-impact analysis²⁷. Trip production refers to the generation of trips from a specific location, while trip attraction refers to trips destined toward that location. Estimating both relies on analyzing regional, demographic, and network-related variables²⁸. Several established methods exist for computing trip production and attraction. Among them, the ITE Trip Generation Manual is the most widely used, offering rates for over 100 land-use categories and multiple transport modes²⁹. Other American models include URBEMIS for air-quality regulation³⁰, the EPA MXD tool for mixed-use trip estimation³¹, and California’s Smart Growth Trip Generation Model, which incorporates land-use mix, density, and street connectivity³². Additionally, Clifton and Currans have developed context-specific models for residential and retail developments³³. Factors influencing travel behavior and mode choice fall into several categories: private-vehicle use³⁴, active transportation³⁵, and public transport³⁶. Key determinants include built-environment variables (distance, land-use mix, network connectivity, safety, pedestrian and cycling infrastructure)³⁷, travel time and cost³⁸, psychological factors (attitudes and habits)³⁹, socio-economic characteristics (age, income, employment) and service indicators such as parking supply and transit frequency⁴⁰. Car ownership and usage are influenced by vehicle cost⁴¹, fuel cost⁴², household income⁴³, age⁴⁴, presence of children⁴⁵, household size and number of employed members⁴⁶, and student or worker status⁴⁷. Built-environment factors, including density, proximity to public transport, mixed-use development, walkability, bicycle infrastructure, and commercial-center accessibility, can reduce dependence on private cars^48–51. Shopping-trip generation is shaped by demographic and socio-economic characteristics such as age, gender, income, and household structure^52–55. For example, being female increases shopping trips in some U.S. cities but reduces them in others⁵³. Physical attributes such as gross floor area, number of employees, parking supply, and number of commercial units significantly affect trip attraction in retail centers^{6,7,10,56–59}.

Recent regional studies in Iran and the Middle East confirm that localizing trip-generation relationships is essential. In Mashhad, MGWR models reveal spatial heterogeneity in built-environment effects⁶⁰. In Tehran, multilayer perceptron networks have shown strong predictive capability for active travel using built-environment and accessibility variables⁶¹. Comparative work in Cairo, Istanbul, and Tehran highlights major differences in work and non-work travel patterns⁶². Research in Rasht confirms the critical role of urban-form characteristics in shaping pedestrian-trip shares⁶³. Parallel studies in Palestinian cities demonstrate that household-based trip-generation rates differ substantially from those in developed countries⁶⁴. Globally, recent research emphasizes enhanced models for shopping-center trip generation⁶⁵, the use of mobile-phone location data for improved accuracy⁶⁶, and strengthened evidence linking transit accessibility to reduced car dependence, especially in TOD contexts⁶⁷.

Despite these advancements, previous studies rarely integrate socio-economic, physical, and built-environment factors simultaneously when modeling commercial trip generation. Addressing this limitation, the current study develops a comprehensive model that incorporates all these factor groups using Structural Equation Modeling.

Research method and variables

This research employs a Structural Equation Modeling (SEM) approach to investigate the intricate relationships between the trip generation rates of commercial uses with extra-regional performance and a set of explanatory variables, encompassing physical characteristics of the land uses, built environment factors, and socio-economic characteristics of their locations in Mashhad, Iran. SEM is a highly flexible linear multivariate statistical modeling technique used to test hypotheses framed for interactions between observed and unobserved variables^68,69. This method has been used in many studies in the field of travel behavior since 1980 and has been effective in this field^52,69–74. The ability to evaluate indirect effects on the latent variables is the advantage of the SEM method over some other modeling methods, such as regression analysis⁶⁸. The availability of this method and its features in AMOS and LISREL software packages have also led to its widespread use in various scientific fields^75,76. In this study, in addition to evaluating the effect of various factors on the trip generation rate of commercial land uses, we are also looking to examine the correlation of the variables of each category of socio-economic, physical, and built environment factors. Therefore, SEM is particularly appropriate for this analysis due to its robustness in handling complex variable relationships and its ability to model latent constructs, providing a detailed understanding of both direct and indirect effects within the urban transport system.

Study area and data collection

Mashhad, the second-largest metropolis in Iran, presents a complex urban structure with a population of about 3.4 million people residing in 353 urban neighborhoods as of 2021⁷⁷. The Transportation Network Management and Engineering Organization’s extensive database underpins this study, offering rich, granular data on urban transportation and travel behaviors, crucial for robust SEM analysis. This database indicates a total of 6.8 million weekday trips, with shopping trips accounting for approximately 14% of the daily total, highlighting their significance in urban travel patterns as depicted in Fig. 1. As observed, the share of commercial trips from the total daily trips is 14%. Therefore, paying attention to commercial trips and understanding their characteristics is of particular importance. As discussed in previous sections, this issue is the reason for defining the subject of the present study.

Fig. 1.

Share of daily trips by trip purpose in Mashhad city.

Data collection in a large and complex city such as Mashhad is constrained by access, cost, and administrative limitations. The selection of commercial sites followed a multi-stage screening process. First, an initial list of commercial land uses was extracted from municipal land-use maps and official commercial registry records. This list included all sites classified as retail, service-oriented, or mixed commercial activities within the urban area.

In the second stage, sites were screened based on functional scale and service area. Commercial uses primarily serving neighborhood-level demand, such as small local shops or convenience retail embedded within residential blocks, were excluded. Sites exhibiting dominant mixed residential–commercial functions were also excluded to avoid confounding trip-generation patterns. Finally, candidate sites were evaluated based on data availability. Only those commercial land uses for which reliable trip-count data and complete spatial information for built-environment analysis were available during the study period were retained. This screening process resulted in a final sample of 33 extra-regional commercial land uses. Nonetheless, the sample reflects a diverse distribution of commercial uses across the city. The 33 sites were selected based on three main criteria: (i) commercial user type (e.g., retail, wholesale, services), (ii) geographic coverage to represent different neighborhoods, and (iii) variations in traffic volume. This strategy ensured that the dataset captured both spatial and functional diversity of commercial land uses in Mashhad.

Trip generation was measured using observed trip counts conducted at each selected commercial site. A trip was defined as a person-based inbound or outbound movement directly associated with the commercial activity, regardless of travel mode (e.g., private car, public transport, walking). Trip counts were conducted manually by trained observers following standardized traffic-count procedures used by the municipal transportation authority. Observation periods focused on peak activity hours, which were identified through preliminary field observations and consultation with local transport experts. For each site, counts were conducted during one weekday and covered at least two consecutive peak-hour periods. When multiple entrances were present, observers were stationed at each access point and counts were aggregated. A subset of sites was double-counted by independent observers to check consistency, and discrepancies were resolved through reconciliation. Hourly trip generation for each site was calculated as the average number of observed trips during the identified peak hour(s).

Although the dataset comprises only 33 commercial land uses, this study does not aim to conduct a fully confirmatory structural equation modeling (SEM) analysis. Instead, a simplified and exploratory SEM framework is adopted to investigate preliminary patterns of association among physical, built-environment, and socio-economic factors in an under-studied developing-country context. Given the limited availability of detailed trip-generation data for extra-regional commercial uses in Mashhad, this sample size reflects practical data constraints rather than an idealized statistical design.In this study, extra-regional commercial land uses are defined as large-scale retail or service-oriented establishments whose catchment area extends beyond their immediate surrounding neighborhood and attracts trips from multiple urban districts. Operationally, candidate sites were identified based on three criteria: (i) floor area exceeding the threshold defined by the local urban development plan, (ii) functional classification as city-level or supra-district commercial use, and (iii) evidence of trip attraction beyond the adjacent traffic analysis zone, based on observed access patterns and expert consultation with municipal transport planners.

These observed trip-generation measures served as the dependent variables in the subsequent modeling framework.

Structural equation modeling (SEM) methodology

Structural equation modeling (SEM) was employed to examine the interrelationships among physical attributes, built-environment characteristics, socio-economic factors, and trip-generation measures within a unified analytical framework. SEM was selected due to its ability to simultaneously model multiple relationships and to account for interdependencies among explanatory variables commonly observed in urban transport systems.

Given the limited sample size and the constraints associated with detailed trip-generation data for extra-regional commercial land uses, the SEM application in this study is explicitly exploratory rather than confirmatory. The objective is not to test a fully specified causal theory or to achieve model generalization, but to identify preliminary patterns of association that may inform hypothesis development and future large-sample studies. The constructs representing Physical factors, Built-environment factors, and Socio-economic factors are specified as formative composite constructs rather than reflective latent variables. Each indicator captures a distinct dimension of the underlying concept, and changes in the indicators are assumed to contribute to—rather than result from—the construct. Accordingly, the measurement specification emphasizes theoretical coverage and parsimony rather than internal consistency among indicators, and traditional reliability and convergent validity metrics such as Cronbach’s alpha, composite reliability, and average variance extracted (AVE) are not directly applicable. In AMOS, formative constructs were implemented by specifying directional paths from observed indicators to their corresponding composite constructs. Model specification and refinement were guided by both theoretical considerations and estimation stability under small-sample conditions.

Indicators exhibiting weak contributions, high multicollinearity, or unstable parameter estimates were removed to reduce model complexity and improve numerical stability. This refinement process does not imply the theoretical irrelevance of excluded indicators (e.g., parking supply or public transport accessibility), but reflects empirical limitations associated with the available data and the exploratory objective of the analysis.Model estimation was conducted with careful attention to uncertainty and robustness. Standard errors and statistical significance levels are reported for all estimated structural paths, and results are interpreted cautiously. Sensitivity checks were performed to ensure that key directional relationships were not driven by individual observations. In addition, Bootstrap resampling with 2000 samples and bias-corrected 95% confidence intervals was used to quantify uncertainty and examine the stability of key structural paths under small-sample conditions. Bootstrap-based confidence intervals were inspected to verify the consistency of coefficient signs and significance patterns. Given the exploratory focus of the study, detailed bootstrap standard errors and confidence intervals are not reported in tabular form, and results are summarized using standardized coefficients and p-values.Variable selection and data analysis.

The study focuses on 33 commercial uses that have significant extra-regional impacts, influencing travel behavior beyond their immediate geographic locales. The selection of these areas was based on criteria such as regional performance level, the number of generated trips, and geographic location within Mashhad.It is important to note that, in urban planning studies, the term ‘region’ refers to a broad geographical subarea within the city’s study area that shares common characteristics. These shared characteristics can encompass various factors such as topography, climate, population density, settlement patterns, dominant economic activities, or cultural identity. In urban transportation studies, each region is further subdivided into smaller zones, which serve as trip production and attraction centers. Ideally, each zone is designated for a specific land use, such as residential, commercial, industrial, or recreational. However, in practice, the distribution of land uses throughout the city often results in a mixture within each zone due to the absence of regulated zoning policies. For instance, in Mashhad’s traffic area, specifically, there are 12 regions and 353 traffic zones, as depicted in Fig. 2. The figure also shows the locations of the selected 33 selected commercial uses within the city.

Fig. 2.

Mashhad’s zoning map and locations of selected commercial uses.

Structural equation modeling approach

In the present study, the constructs labeled as Physical factors, Built-environment factors, and Socio-economic factors are treated as conceptually defined composite groupings rather than fully validated latent variables in a confirmatory measurement sense. The included indicators represent formative contributions to each construct, reflecting distinct dimensions of land-use characteristics, urban infrastructure, and contextual socio-economic conditions. As such, the measurement specification emphasizes theoretical coverage and parsimony rather than internal consistency among indicators.Structural Equation Modeling (SEM) facilitates the exploration of hypothesized relationships between observed and latent variables through a series of structural (regression) equations. It is acknowledged that covariance-based SEM is typically recommended for large samples, particularly when models involve multiple latent constructs and indicators. However, recent methodological literature recognizes the applicability of simplified and exploratory SEM models in small-sample urban and transportation studies when the primary objective is pattern exploration rather than confirmatory inference or theory testing^78,79. Accordingly, the SEM framework in this study is intentionally simplified by limiting the number of latent constructs, reducing indicator redundancy, and focusing on the estimation of robust directional relationships rather than model generalization. SEM has several advantages over linear regression modeling, These methodological characteristics motivate the use of SEM in the present exploratory analysis but do not imply superiority over alternative modeling approaches. some of which are mentioned below:

• Examining the Complex Relationships Between Variables: SEM allows for the simultaneous examination of complex relationships between multiple independent and dependent variables, whereas linear regression models primarily focus on bivariate relationships⁷⁸.

• Latent Variables: SEM can model variables that are not directly observable but are measured through observed indicators. Linear regression typically works with observable variable⁸⁰.

• Handling Measurement Errors: SEM explicitly accounts for measurement errors in its models, which helps to accurately analyze and reduce the impact of measurement errors. This issue is not given as much attention in linear regression as it is in SEM⁸¹.

• Theoretical Models Test: With SEM, researchers can test complex theoretical models and evaluate whether the data collected fits the theoretical model, which is possible in a more limited form of linear regression⁸².

• Analyzing Direct and Indirect Relationships: SEM allows for the examination of direct, indirect, and overall effects, whereas linear regression is typically limited to direct effects⁸³.

The SEM model in this study was structured to examine paths from physical characteristics, built environment attributes, and socio-economic factors to the dependent variable—hourly trip rates to commercial uses. Trip-rate normalization is defined as the average hourly number of trips divided by gross floor area (GFA) and expressed as trips per 100 m² of GFA. Table 1 presents the detailed variables included in the SEM, categorized into groups such as physical, built environment, and socio-economic. While SEM provides a framework for accounting for measurement error, the present application focuses primarily on structural relationships among composite indicators. Measurement error is therefore acknowledged implicitly through cautious interpretation and uncertainty reporting rather than through a fully specified confirmatory measurement model.

Table 1.

Descriptive variables for trip generation rate of each commercial use.

Variable group	Variable name	Description			Unit	Symbol
Physical	Land area	Total land area of the commercial use			m2	Land_Ar
	Gross floor area	Total building area of the commercial use, including external walls			m2	GFA
Hourly trip	Number of trips	Average hourly trips produced or attracted by the commercial use			Trips	Trip
	Number of trips per 100 m2	Average hourly trips produced or attracted per 100 square meters of gross floor area (GFA)			trips/100 m2 GFA	Trip_pr_100sqm
Built environment	Road length density	Average total length of roads per one hectare of the zone of the commercial use			m/ha	Rd_Ln_Densi
	Road surface ratio	Ratio of total road surface in the zone of the commercial use to the zone area			m2/m2	Rd_Srf_rati
	Asphalt surface ratio	Ratio of total asphalt road surface in the zone of the commercial use to the zone area			m2/m2	Asph_Serf_rati
	Number of parking lots	Number of on-street parking lots in the zone of the commercial use			lots	Park_Lot
	Parking area ratio	Ratio of total on-street parking area in the zone of the commercial use to the zone area			m2/ hectares	Park_Ar_rati
	Parking density	Average number of on-street parking lots per one hectare of the zone of the commercial use			lots/ha	Park_Lot_Densi
	Number of bus stops	Number of bus stops in the zone of the commercial use			Stops	Bus_Stop
	Bus stop density	Average number of bus stops per one kilometer of line length in the zone of the commercial use			Buses	Bus_Stop_pr_km
	Number of shopping units	Average number of shopping units per one hectare of the zone of the commercial use			Units	Shop_Unit_pr_ha
Socio-economic	Population	Total population of the zone of the commercial use			People	Zone _Pop
	Female population	Total Female population of the zone of the commercial use			People	Female_Pop
	Male population	Total Male population of the zone of the commercial use			People	Male _Pop
	Population density	population per one hectare of the zone of the commercial use			People/ha	Pop_Densi
	Female to male ratio	Ratio of female to male population in the zone of the commercial use			–	Feml_Male_rati
	Male population ratio	Ratio of male population to total population of the zone of the commercial use			–	Male_Pop_rati
	Female population ratio	Ratio of female population to total population of the zone of the commercial use			–	Feml_Pop_rati
	Ratio of employees	Ratio of employees to population of the zone of the commercial use			–	Empl_Pop_rati
	Car ownership	Car ownership per capita in the zone of the commercial use			Cars/person	Car_Own

Geographical and statistical considerations

Defining the appropriate spatial scale for SEM analysis is crucial for capturing the true variability and interdependencies among urban factors. Gehrke and Wang⁸⁴ highlight the importance of selecting an independent and separate geographical area that accurately represents the characteristics of the built environment in studies related to transportation and land use. This approach ensures that spatial heterogeneities in the built environment, which can significantly influence travel behavior, are appropriately modeled.

Following these guidelines, this study utilizes statistical areas similar to censuses, aligning with practices recommended by Mitchell Hess, et al.⁸⁵ for defining environmental conditions. This methodological choice is supported by the use of neighborhood units, which are often used as approximations for environmental conditions in travel behavior research^86,87. By employing these defined areas, the SEM analysis can effectively reflect realistic urban configurations and interaction patterns, essential for accurately modeling the impact of land use and socio-economic factors on trip generation.

Structural equation modeling and correlation analysis

Structural equation modeling (SEM)

This study employs Structural Equation Modeling (SEM) to assess the complex interrelations between the trip generation rates of commercial uses and the influencing factors. Three categories of latent variables have been identified based on the theoretical framework:

1. Physical factors: These variables represent the physical characteristics of the commercial uses.

2. Socio-economic factors: These variables capture the socio-economic conditions of the areas surrounding the commercial uses.

3. Built environment factors: These encompass elements of the urban design and infrastructure that impact usage patterns.

Model 1, as depicted in Fig. 3, initially tests the relationships hypothesized in the theoretical framework. This model integrates the observed variables into the corresponding latent categories, providing a structured approach to analyze how each group of factors influences the dependent variables of trip generation rates.

Fig. 3.

Initial structural model (model 1).

Investigating variable correlations

Due to the extensive number of variables involved in this research, not all correlations are reported individually. Instead, the analysis focuses on the correlation within the subgroups of physical, socio-economic, and built environment factors, each examined separately to determine the impact on the dependent variables. Histograms along with density curves illustrate the distribution and correlation of these variables, with Figs. 4, 5, and 6 showcasing these relationships.

Fig. 4.

Correlation analysis for physical variables.

Fig. 5.

Correlation analysis for socio-economic variables.

Fig. 6.

Correlation analysis for built environment variables.

Figure 4 presents a detailed visual analysis of the correlations between physical factors, such as land area and gross floor area. The histogram and density curve provide insights into the distribution patterns of these variables, while the Pearson’s pairwise correlation values above the diagrams offer quantitative evidence of the relationships among these factors Fig. 5 reveals significant correlations between socio-economic factors. Specifically, Fig. 5 shows the full correlation of three variables “Male population”, “Female population”, and “Population”, as well as between three variables “Female to male ratio”, “Female population ratio”, and “Male population ratio” with each other. This indicates strong collinearity, suggesting that only one variable from each correlated group should be included in the final model to avoid redundancy. Consequently, “Population” and “Female to male ratio” were chosen for their greater availability and comprehensive representation of their respective categories. Figure 6 examines the relationships among built environment factors like road length density, parking area ratio, and asphalt surface ratio. Figure 6 illustrates the existence of a full correlation between “Parking area ratio” and “Parking density,” necessitating the selection of one over the other in the final model due to their collinearity. “Parking area ratio” was retained for its broader applicability and data availability.

This integrated approach to SEM and correlation analysis not only validates the model structure but also refines the variable selection, ensuring that the final SEM analysis is robust and reflective of the underlying theoretical constructs. This rigorous methodological framework facilitates a deep understanding of the multifaceted interactions between urban form, socio-economic context, and trip generation, providing insights critical for urban planning and policy-making.

Calibration and refinement of the SEM model structure

Because the indicators associated with each construct are specified as formative rather than reflective, traditional reliability and convergent validity metrics such as Cronbach’s alpha, composite reliability (CR), and average variance extracted (AVE) are not directly applicable. The relevance of indicators was instead assessed through theoretical justification, multicollinearity diagnostics, and the stability of estimated path coefficients under model refinement.Model calibration in this study was guided not only by statistical criteria but also by the need to maintain parsimony under small-sample conditions. variables exhibiting weak empirical contributions, high multicollinearity, or unstable parameter estimates were removed to reduce model complexity and improve estimation stability. This refinement process resulted in a simplified SEM structure suitable for exploratory analysis, while preserving the core conceptual distinctions among physical, built-environment, and socio-economic factors. Before finalizing the structural model, it is crucial to evaluate and refine the model based on the estimated indicator weights and their stability within the formative measurement specification. These indicator weights represent the relative contribution of each observed indicator to the formation of its corresponding composite construct. A high indicator weight indicates a stronger contribution of the observed variable to the formation of the corresponding composite construct. Statistical significance of indicator weights was examined to identify indicators with weak or unstable contributions under the formative specification.Evaluation all of models was conducted using IBM SPSS 26 and Amos 24 software and involved assessing various fit indices that measure how well the model conforms to the empirical data. The exclusion of certain indicators (e.g., number of parking lots and bus stop density) does not imply their theoretical irrelevance to trip generation. Rather, their removal reflects empirical instability under the present data constraints, including weak factor loadings, high collinearity with retained indicators, and limited contribution to model identification. Given the exploratory objective of the study, priority was given to indicators that demonstrated both theoretical relevance and estimation stability within the simplified model structure.

Table 2 reports the initial indicator weights for each construct along with their corresponding P-values. Several indicators exhibited low absolute weights and/or statistically insignificant contributions, indicating weak roles in forming their respective constructs. Additionally, variables such as the ratio of employees, the number of parking lots, and bus stop density showed P-values exceeding 0.05, marking them as statistically insignificant. These variables were removed to enhance the model’s structural integrity.

Table 2.

Indicator weights and statistical significance for the initial formative measurement specification (Model 1).

Variable group	Variable name	Indicator weight	Significance value (P-value)
Physical	Land area	1.027	–
	Gross floor area	0.878	< 0.001
Built environment	Road length density	0.843	< 0.001
	Road surface ratio	0.922	–
	Asphalt surface ratio	0.939	< 0.001
	Number of parking lots	0.270	0.126
	Parking area ratio	0.714	< 0.001
	Number of bus stops	− 0.155	0.390
	Number of shopping units	0.478	0.004
Socio-economic	Population	0.679	0.120
	Population density	0.564	0.024
	Female to male ratio	0.651	0.014
	Ratio of employees	− 0.027	0.894
	Car ownership	0.575	0.004

Following the adjustments, the model was recalibrated., Table 3 shows the recalibrated indicator weights of the retained indicatorsensuring that all retained indicators exhibit statistically significant and stable contributions In AMOS, formative constructs were implemented by specifying directional paths from observed indicators to their corresponding composite constructs.and exhibit P-values below 0.05, confirming their significant contributions to the latent constructs. Figure 7 visually depicts the adjusted structural model (model 2) after the removal of statistically insignificant variables. It highlights a streamlined and focused set of variables that adequately represent the latent constructs within the study.

Table 3.

Indicator weights and statistical significance for the refined formative measurement specification (Model 2).

Variable group	Variable name	Indicator weight	Significance value (P-value)
Physical	Land area	1.026	–
	Gross floor area	0.878	< 0.001
Built environment	Road length density	0.843	< 0.001
	Road surface ratio	0.914	–
	Asphalt surface ratio	0.949	< 0.001
	Parking area ratio	0.701	< 0.001
	Number of shopping units	0.486	0.003
Socio-economic	Population	0.690	0.022
	Population density	0.622	0.027
	Female to male ratio	0.606	0.029
	Car ownership	0.514	–

Fig. 7.

Refined structural model (model 2).

Evaluation of the fitted structural model

Following model refinement, overall model fit was assessed using a standard set of global fit indices, including χ², degrees of freedom (df), Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). Given the exploratory purpose of the analysis and the limited sample size, these indices are interpreted as descriptive indicators of approximate model adequacy rather than strict accept/reject criteria (Table 4).

Table 4.

Global fit indices for the initial SEM model.

Test name		Description	values
		Chi-square	135.97
df			58
TLI		Tucker-Lewis index	0.635
CFI		Comparative fit index	0.729
RMSEA		Root mean square error approximation	0.205

ommonly cited cut-off values for fit indices (e.g., CFI/TLI ≥ 0.90, RMSEA ≤ 0.08) are derived mainly from confirmatory SEM literature and are provided only as general guidance. In the present study, which adopts an exploratory SEM approach with a small sample size, fit indices are interpreted cautiously.

The initial model exhibited limited global fit (CFI = 0.729, TLI = 0.635, RMSEA = 0.205), indicating substantial scope for model improvement. Consequently, a refinement process was undertaken by removing non-significant socio-economic paths and simplifying the structural specification.

The refined model demonstrated improved descriptive fit, with CFI = 0.894, TLI = 0.840, χ²/df = 1.935, and a substantial reduction in AIC (from 201.97 to 88.43). Given the exploratory nature of the study and the small sample size, these fit indices are interpreted as descriptive indicators of approximate model adequacy rather than strict accept/reject criteria.

. Given the exploratory nature of the study and the small sample size, these fit indices are interpreted as descriptive indicators of approximate model adequacy rather than strict accept/reject criteria., along with a substantial reduction in AIC (from 201.97 to 88.43). These improvements suggest better descriptive adequacy of the refined model, while all results are interpreted as exploratory (Table 5).

Table 5.

Global fit indices for the final refined SEM model.

Test name		Description	values
		Relative chi-square	1.935
GFI		Goodness of fit index	0.777
NFI		Norm fitness index	0.812
RFI		Relative fit index	0.718
TLI		Tucker-Lewis index	0.840
CFI		Comparative fit index	0.894
RMSEA		Root mean square error approximation	0.171
RMSEA 90% CI			0.095–0.244

Discussion

Further analysis examined the regression coefficients to explore the associations between predictor constructs and trip-generation measures. Table 6 reports standardized coefficients and p-values for each tested path. Within the analyzed sample, physical and built-environment factors show statistically detectable associations with total trip counts, whereas socio-economic factors do not exhibit statistically significant relationships. These results should be interpreted as exploratory and descriptive rather than confirmatory. Bootstrap inspection indicated that only the path from Physical factors to total number of trips showed a stable and statistically detectable association, while confidence intervals for paths involving built-environment factors and trip rates per 100 m² included zero.

Table 6.

Standardized structural path estimates with bootstrap 95% CI (final model).

Path name	Std. β	Lower 95% CI	Upper 95% CI	P
Physical → Number of trips per 100 m2	− 0.023	− 0.387	0.303	0.869
Physical → Number of trips	0.84	0.459	1.051	0.004
Built environment → Number of trips per 100 m2	0.206	− 0.294	0.592	0.365
Built environment → Number of trips	0.289	− 0.078	0.681	0.13

Bootstrap diagnostics indicate that the direction of paths from physical and built-environment constructs to total trips remains stable across resamples, whereas confidence intervals for paths to trips per 100 m² include zero.

Conclusion

This research developed a framework to examine the simultaneous associations between physical attributes, built-environment characteristics, and socio-economic factors and trip generation at extra-regional commercial centers in Mashhad, Iran. Based on the results of the refined structural model (Fig. 8) and the regression analysis (Table 6—Standardized structural path estimates with bootstrap 95% CI (final model).), several key conclusions can be drawn.The findings suggest that, within the analyzed sample and under the current exploratory model specification, trip attraction at large-scale extra-regional commercial centers is more strongly associated with physical scale and surrounding built-environment characteristics than with the socio-economic profile of the local catchment area. The exclusion of non-significant socio-economic paths reflects empirical limitations rather than conceptual irrelevance and improved model parsimony under small-sample conditions. A notable distinction emerges between total trip volumes and trip rates normalized by floor area. The results reported in Table 6 indicate statistically detectable associations between physical and built-environment factors and the total number of attracted trips within the analyzed sample, whereas no statistically significant associations were observed for trips per 100 m². This pattern should be interpreted cautiously, as normalization by floor area may reduce variability in trip-generation measures, and context-specific characteristics of large-scale commercial complexes may attenuate density-based effects.Despite the limited sample size (n = 33), the study provides an exploratory and parsimonious representation of trip-generation patterns for extra-regional commercial land uses. The results should be interpreted as indicative rather than confirmatory, and they highlight the need for future research using larger samples and alternative modeling approaches. For planning practice, the findings suggest that physical site attributes and infrastructural capacity may warrant particular attention in exploratory assessments of trip generation.

Fig. 8.

The final refined structural model for trip generation (after excluding non-significant socio-economic paths and model trimming.

Key findings and insights

• Physical and Land Use Factors: The study integrated the physical size of the land and the project’s foundation area, allowing for a nuanced understanding of how these physical attributes influence commercial uses’ travel behaviors and interactions within their environments.

• Socio-Economic Factors: Initially, nine socio-economic factors were considered. After extensive analysis, only four factors—population, female to male ratio, population density, and car ownership—remained in the model, indicating their relevance in influencing urban travel patterns.

• Built Environment Impact: Among the nine built environment parameters examined, five were identified as primary factors: road length density, road surface ratio, asphalt surface ratio, parking area ratio, and the number of shopping units. These parameters significantly shape the traffic dynamics associated with commercial uses.

• Model Fit and Implementation: The implementation of the final indicated that physical features of commercial uses and built-environment factors are associated with variations in the total number of trips, whereas socio-economic factors did not show statistically detectable associations in the present sample., socio-economic factors do not show a substantial impact on trip generation.

Implications for urban planning

The results underscore the importance of considering both physical and built environment factors in urban planning and traffic management. The findings suggest that, in similar extra-regional commercial contexts, physical and built-environment characteristics may warrant greater attention alongside socio-economic factors when evaluating trip-generation patterns.

Suggestions for future research

• Expanding Sample Size and Diversity: Future studies could explore a broader geographical scope and a more diverse set of commercial uses to enhance the generalizability of the findings.

• Longitudinal Analysis: Implementing longitudinal studies could help in understanding the temporal changes in how built environment and socio-economic factors affect trip generation over time.

• Incorporating Additional Socio-Economic Variables: While this study limited its focus to a few socio-economic factors, future research could include additional variables such as educational level, income variance, and employment types to deepen the socio-economic analysis.

• Advanced Traffic Simulation Models: Employing advanced simulation models could provide more detailed insights into the specific contributions of built environment configurations to traffic patterns.

• Impact of Public Policy: Examining the impact of specific urban planning policies on the effectiveness of commercial land use planning and its influence on traffic could provide actionable insights for policymakers.

These insights contribute to a more informed approach to urban development, helping policymakers design sustainable urban environments that effectively address the complexities of traffic congestion and land use planning. As cities continue to grow and evolve, such research is crucial for developing integrated strategies that consider the intricate relationships between land use characteristics and travel behavior.

Author contributions

A.B. (Abbas Babazadeh) supervised the project and provided guidance throughout the research.R.A. (Ramin Ahooee) conducted the research, collected and analyzed the data, and prepared the manuscript.A.N. (Ali Naderan) acted as a co-supervisor, advising on methodology and data interpretation.All authors reviewed and approved the final manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

Data will be made available on request.

Declarations

Conflict of interest

The authors declare no competing interests.

Generative AI and AI‑assisted technologies in the writing process

No AI or AI-assisted technologies were used in the writing of this manuscript.