Accessibility Barriers in Urban Public Transport for Disabled Users: An AHP-Based Severity Index and Behavioral Regression Analysis

Abstract

This study examines accessibility barriers experienced by individuals with disabilities in urban public transportation and analyzes how these barriers influence their travel behavior. Survey data were collected from 450 participants with different disability types in Alanya, Turkey, a tourism-oriented city characterized by pronounced seasonal mobility fluctuations. To ensure internal consistency and analytical robustness, the Analytic Hierarchy Process (AHP) was applied to prioritize seven accessibility criteria, and the consistency of pairwise comparisons was verified prior to analysis. Based on the AHP-derived weights, a composite accessibility-based Problem Severity Index (PSI) was constructed and integrated into regression models to quantify behavioral effects. The results show that the Problem Severity Index (PSI) is strongly associated with satisfaction (R² = 0.895), frequency of public transport use (R² = 0.924), and perceived travel difficulty (R² = 0.924), reflecting constrained mobility conditions and limited modal alternatives rather than improved service quality. Deficiencies in bus stop design and vehicle accessibility equipment were identified as the most influential barriers affecting public transport experience. Beyond the case study context, the proposed AHP–regression framework provides a structured analytical approach for evaluating accessibility performance and generating empirical evidence to inform inclusive and sustainable urban mobility planning. The findings offer empirical evidence on the relative importance of accessibility barriers and highlight critical infrastructure and service deficiencies. Rather than constituting a decision-support tool themselves, these results provide structured information that, when appropriately contextualized, can inform and guide transport authorities and urban planners in prioritizing accessibility improvements and enhancing inclusive public transport performance over time.

1. Introduction

Ensuring equitable access to urban public transportation is a central objective of inclusive mobility and sustainable development agendas and is recognized as a fundamental right of persons with disabilities [1]. Despite the expansion of transport services and the adoption of accessibility frameworks in many cities, individuals with disabilities continue to encounter barriers that restrict independent mobility, social participation, and travel decision-making [2]. Accessibility extends beyond physical infrastructure to encompass operational conditions and behavioral determinants that influence the actual usability of public transport systems for individuals with disabilities [3,4]. Recent research further emphasizes that accessibility is shaped not only by physical design but also by social and behavioral dynamics affecting travel satisfaction and participation in urban life among disadvantaged groups [5].

According to official statistics from the Turkish Ministry of Family and Social Services, approximately 6.9% of the population in Türkiye—corresponding to nearly 4.9 million people—live with at least one disability that limits daily activities and mobility. Among them, 3.3% report difficulty in walking or climbing stairs, and 4.1% report difficulty carrying objects, indicating significant physical mobility constraints that directly affect independent travel and public transport use [6]. At the national level, a considerable proportion of persons with disabilities report reducing or canceling trips due to accessibility barriers in public transport systems. These findings illustrate the magnitude of mobility-related exclusion and reinforce the policy relevance of accessibility-oriented transport planning. Urban-level studies similarly indicate lower rates of regular public transport usage among disabled users, particularly in cities with limited barrier-free infrastructure. In addition to mobility limitations, inadequate accessibility is associated with increased social exclusion and indirect economic costs, whereas accessibility improvements have been shown to enhance social participation and reduce dependency on private transport.

More broadly, the decision to use public transport for daily trips is influenced not only by accessibility conditions but also by sociodemographic and behavioral factors such as age, income, household characteristics, and perceived service quality. Discrete choice and logit-based modeling studies demonstrate that these determinants significantly shape public transport usage patterns in urban contexts [7,8]. Within this broader behavioral framework, accessibility barriers experienced by persons with disabilities—including physical, informational, and systemic obstacles throughout the travel chain—represent an additional layer of constraints that negatively affect travel behavior and participation in urban life [9].

2. Literature Review

Existing literature indicates that accessibility barriers are shaped by multiple elements, including infrastructure limitations at bus stops, lack of accessibility equipment on vehicles, inadequate pedestrian environments, and insufficient signalization [10,11]. Although research increasingly highlights the importance of linking accessibility to user satisfaction and public transport engagement, relatively few studies quantitatively integrate accessibility prioritization with behavioral analysis, particularly in developing and tourism-based urban contexts [12,13]. Furthermore, most studies evaluate accessibility descriptively rather than using decision-support frameworks capable of identifying which barriers have the greatest behavioral implications [14,15]. Recent sustainability research further highlights the need for decision-support approaches that move beyond descriptive assessments by linking accessibility conditions to observable behavioral outcomes among people with disabilities [16].

Empirical research over the last five years has increasingly examined how accessibility barriers influence travel behavior, participation, and perceived safety among people with disabilities in urban transport systems. Studies emphasize that accessibility is shaped not only by infrastructure-related factors, such as stop design and vehicle boarding conditions, but also by service attributes and behavioral responses, including perceived service quality, reduced trip frequency, and public transport satisfaction [17,18]. This emerging evidence supports the need for integrated frameworks that link measured accessibility conditions to behavioral outcomes among disabled users.

Recent research has also investigated barriers and facilitators encountered by people with disabilities in public transport use [19], explored transport and mobility decision-making patterns among this group [20], and examined route preferences and accessibility tradeoffs in urban contexts [21].

In parallel, decision-support and behavioral-economics-informed approaches have gained prominence in accessibility research. Multi-criteria decision-making (MCDM) methods are widely used to prioritize accessibility investments under budget constraints, while Data Envelopment Analysis (DEA)-based models are increasingly applied to evaluate public transport performance, including transfer efficiency across network nodes. Recent empirical studies have also examined the equity of vertical transport system provision (e.g., elevators and escalators) in metro stations for mobility-impaired users, highlighting the importance of fair spatial distribution of accessibility infrastructure. These developments provide a strong methodological foundation for the present AHP-based severity index [22,23].

Türkiye presents a particularly relevant case for such an investigation. Despite national accessibility standards, implementation and performance vary across cities [24]. Alanya represents a unique urban environment due to its seasonal population fluctuations and multicultural resident profile, including long-term international residents and individuals with diverse accessibility needs. These conditions make Alanya a particularly relevant setting to examine how accessibility challenges interact with user mobility behavior. In tourism-oriented cities, accessibility challenges tend to intensify due to pronounced seasonal population fluctuations and the presence of international visitors with diverse mobility expectations, underlining the importance of accessibility for sustainable mobility in tourist destinations [25].

To address this research gap, this study investigates the accessibility barriers experienced by disabled public transport users and examines how these barriers influence travel behavior. The Analytic Hierarchy Process (AHP) is used to prioritize accessibility criteria, and the resulting accessibility index is integrated into regression models to analyze relationships with satisfaction, frequency of use, and perceived difficulty. The use of AHP is particularly valuable in this context, as it ensures internal consistency in subjective survey judgments before further behavioral modeling, a practice widely adopted in sustainable transport planning studies [26].

This study provides three main contributions:

• It operationalizes accessibility barriers using a structured multi-criteria decision-making approach rather than descriptive reporting.
• It empirically links accessibility levels to behavioral outcomes using survey data collected from 450 passengers with disabilities.
• It offers a transferable framework by combining AHP-derived criterion weights with a composite Problem Severity Index (PSI) and regression-based behavioral modeling, which can be replicated in other cities by re-estimating the weights and PSI using locally collected survey data.

The remainder of this article is structured as follows: Section 2 explains the study area, sampling, and methodological framework; Section 3 presents AHP results and regression findings; Section 4 discusses implications for policy and planning; and Section 5 provides conclusions and recommendations for future research.

Recent methodological advances have increasingly applied multi-criteria and efficiency-based approaches to evaluate accessibility and equity in public transport systems [22,23]. Iterative Data Envelopment Analysis (DEA) models have been employed to assess transfer efficiency across multimodal public transport networks, while recent equity-focused studies have examined the installation and spatial distribution of vertical transport systems in metro stations to improve accessibility for mobility-impaired users. These approaches demonstrate the growing integration of performance evaluation and equity considerations in contemporary accessibility research. Despite the growing body of research on transport accessibility for people with disabilities, most existing studies either provide descriptive assessments of accessibility barriers or examine behavioral outcomes in isolation. While recent empirical studies have highlighted the importance of accessibility for satisfaction, participation, and travel behavior [3,5,8,27], relatively few studies have integrated consistency-controlled multi-criteria prioritization of accessibility barriers with behavioral modeling within a unified analytical framework [12,13]. In particular, the linkage between prioritized accessibility deficiencies and multiple behavioral responses remains underexplored, especially in tourism-oriented urban contexts characterized by seasonal population fluctuations. In this regard, the present study contributes to the literature by combining an AHP-based accessibility severity assessment with regression-based behavioral analysis, thereby systematically linking perceived accessibility barriers to satisfaction, public transport use frequency, and perceived difficulty.

3. Methods

The methodological procedure of this study consisted of five sequential stages in order to ensure analytical consistency: defining the study context, collecting participant data, applying the Analytic Hierarchy Process (AHP) to determine the relative importance of accessibility criteria, constructing an accessibility index (PSI), and applying regression modeling to examine the relationship between accessibility and mobility-related behavioral outcomes. Figure 1 illustrates the overall methodological framework, integrating the AHP-based accessibility assessment with the regression analysis.

3.1. Study Area

Alanya is a tourism-driven coastal district located in southern Türkiye, characterized by significant seasonal fluctuations in population and transport demand. While the district has a permanent population of approximately 359,891 residents, this number increases substantially during peak tourism periods. Tourism-oriented cities are widely recognized as experiencing pronounced temporal variations in travel demand, which can place additional pressure on public transport systems and urban accessibility conditions [28]. In addition to short-term visitors, Alanya hosts an international community of long-term foreign residents, offering a unique socio-demographic context in which diverse mobility expectations and accessibility needs coexist. Previous studies have shown that urban transport systems serving heterogeneous and diverse user populations, particularly in tourism-oriented cities, often face greater challenges in ensuring inclusive and universally accessible transport services [27,28,29].

This dynamic and multicultural environment makes Alanya an appropriate case study for examining accessibility challenges, as varying user profiles and seasonal ridership pressures can exacerbate mobility barriers experienced by disabled public transport users. Such conditions are particularly relevant for accessibility-focused research, as seasonal demand surges and diverse user expectations may intensify existing mobility barriers if accessibility considerations are not systematically integrated into transport planning [30].

3.2. Participants and Data Collection

A total of 450 individuals with disabilities residing in Alanya participated in the study, with participants recruited from the central districts where public transport use is concentrated. The sample comprised 336 male and 114 female respondents. A detailed distribution of disability types by gender is presented in

Table 1. Distribution of study participants by gender and disability type (n = 450).

Disability Type	Male (n = 336)	Female (n = 114)	Total
Visual impairment	84	49	133
Hearing impairment	90	17	107
Physical disability	148	37	185
Speech and language impairment	14	11	25
Total	336	114	450

.

This distribution demonstrates that the sample captures a diverse range of accessibility needs across multiple disability categories, even though the subsequent analysis focuses on aggregate behavioral associations rather than subgroup-specific effects. The final sample size exceeds the statistically required minimum sample size determined based on a 95% confidence level and a 5% margin of error, relative to the estimated disabled population of approximately 3000 residents in Alanya. This population estimate is derived from the officially registered 2542 individuals reported in the Alanya Disability Database Project and accounts for potential underrepresentation of individuals not included in administrative records [31]. A purposive sampling strategy was adopted to ensure that respondents had direct and recent experience with public transport use, which is essential for reliably assessing perceived accessibility barriers and behavioral responses. Participation in the survey was entirely voluntary, and all respondents were informed about the purpose of the study prior to completing the questionnaire. Although the selected accessibility criteria are not exclusive to tourism-oriented cities, their perceived severity and relative importance are shaped by seasonal population fluctuations, increased pedestrian density, and the interaction between residents and visitors that characterize tourism destinations such as Alanya.

Data collection was conducted through face-to-face surveys at bus stops, disability coordination centers, and public spaces frequently used by disabled individuals. The questionnaire contained two sections: (i) demographic information and (ii) an evaluation of the perceived severity of accessibility-related problems based on a five-point Likert scale, ranging from 1 (“very low severity”) to 5 (“very high severity”). These severity ratings provide the input for deriving the Problem Severity Index (PSI), as described in Section 3.4.

Participants evaluated seven accessibility criteria, which were later used in the modeling framework:

• bus stop conditions (Criterion A),
• accessibility equipment in vehicles (Criterion B),
• sidewalk slope (Criterion C),
• sidewalk surface quality (Criterion D),
• sidewalk width and curb height (i.e., the vertical level difference between the sidewalk and the carriageway, which directly affects step-free accessibility) (Criterion E),
• illegal roadside parking (Criterion F),
• intersection signalization (Criterion G).

These accessibility scores formed the empirical basis for the Analytic Hierarchy Process (AHP) weighting and were subsequently used as inputs in the regression analysis.

3.3. Analytic Hierarchy Process (AHP)

The Analytic Hierarchy Process (AHP) was employed to determine the relative importance of seven accessibility criteria representing the main barriers faced by disabled public transport users. AHP is widely applied in transport and accessibility research, as it enables qualitative judgments to be systematically structured and converted into quantitative weights [32,33,34].

The evaluated criteria included bus stop condition, accessibility equipment in vehicles, sidewalk slope, sidewalk surface quality, sidewalk width and curb height, illegal roadside parking, and intersection signalization. Pairwise comparisons were conducted using Saaty’s 1–9 preference scale to capture the relative influence of each criterion on perceived accessibility. Individual respondent judgments were aggregated into a single group comparison matrix, from which the final criterion weights were derived using the eigenvector method.

The internal consistency of the pairwise comparisons was assessed using the Consistency Ratio (CR). The obtained value (CR = 0.0926) was below the commonly accepted threshold of 0.10, indicating satisfactory consistency of the aggregated judgments. The resulting AHP weights are reported in

Table 2. Pairwise comparison matrix for accessibility criteria.

	A	B	C	D	E	F	G
A	1.00	1.00	3.00	9.00	7.00	7.00	1.00
B	1.00	1.00	1.00	5.00	9.00	9.00	1.00
C	0.33	1.00	1.00	1.00	9.00	7.00	1.00
D	0.11	0.20	1.00	1.00	3.00	1.00	1.00
E	0.14	0.11	0.11	0.33	1.00	0.33	0.33
F	0.14	0.11	0.14	1.00	3.00	1.00	0.33
G	1.00	1.00	1.00	1.00	3.00	3.00	1.00
TOTAL	3.73	4.42	7.25	18.33	35.00	28.33	5.67

and

Table 3. Normalized Comparison Matrix and Priority Vector.

	A	B	C	D	E	F	G	Total	Total/7
A	0.26	0.22	0.41	0.49	0.20	0.25	0.18	2.02	0.29
B	0.27	0.22	0.13	0.27	0.26	0.32	0.18	1.66	0.24
C	0.09	0.22	0.13	0.05	0.26	0.25	0.18	1.19	0.17
D	0.02	0.04	0.13	0.05	0.09	0.04	0.18	0.56	0.08
E	0.03	0.02	0.01	0.01	0.02	0.01	0.06	0.20	0.03
F	0.03	0.02	0.01	0.05	0.08	0.03	0.05	0.32	0.05
G	0.26	0.22	0.13	0.05	0.08	0.10	0.17	1.05	0.15
									1.00

. Bus stop inadequacy received the highest weight (0.29), followed by insufficient vehicle accessibility equipment (0.24), whereas sidewalk slope (0.03) and sidewalk surface quality (0.05) were assigned comparatively lower importance. These AHP-derived weights were subsequently used to construct the Problem Severity Index (PSI), which served as the main explanatory variable in the regression analysis.

3.4. Accessibility Index Development

The AHP-derived weights were operationalized into a composite accessibility index to quantify the severity of mobility barriers experienced by each participant. For each criterion, individual Likert-scale ratings (1–5) were combined with the corresponding AHP weights and aggregated to obtain a single Problem Severity Index (PSI) score. Since the Likert scale was defined such that higher values represent greater problem severity (1 = very low severity, 5 = very high severity), the original severity ratings were directly used in the index construction, and no reverse coding was required. The PSI for participant i was calculated as

$$\mathrm{P}\mathrm{S}{I}_i= {\sum }_{k=1}^7\left(w_k\cdot s_{ik}\right)$$

(1)

where wk denotes the AHP-derived weight of criterion k and $s_{ik}$ represents the severity rating assigned by participant i.

For comparability and interpretability, the resulting PSI values were linearly rescaled to a 0–100 range as follows:

$$\mathrm{P}\mathrm{S}{I}_{i\left(0–100\right)}={\displaystyle \frac{\left(PS{I}_i-1\right)}{4}}\times 100$$

(2)

Higher PSI scores indicate greater accessibility barriers.

3.5. Population and Sample Size Determination

The target population consisted of disabled public transport users residing in the central districts of Alanya. According to official records from the Alanya Disability Database Project maintained by the Alanya Municipality, the estimated population size (N) was approximately 3000 individuals [31]. To determine the minimum required sample size for statistical reliability, the standard formula for finite population sampling was applied:

$$n:{\displaystyle \frac{N\times Z^2\times p\times \left(1-p\right)}{E^2\times \left(N-1\right)+Z^2\times p\times \left(1-p\right)}}$$

(3)

where

• N = 3000 (population size)
• z = 1.96 (Z-value for 95% confidence level)
• p = 0.5 (maximum variability assumption)
• E = 0.05 (margin of error)

The minimum required sample size was therefore n = 341 participants at a 95% confidence level with a ±5% margin of error. However, to increase statistical robustness, minimize sampling bias, and enhance the reliability of the regression models, data was collected from 450 respondents, which exceeds the minimum threshold.

The enlarged sample improves the precision of the AHP-derived weights, reduces variance in PSI estimates, and increases the explanatory power of behavioral regression models. As a result, the analytical findings are more representative of disabled mobility patterns in Alanya and carry stronger inferential validity.

4. Results

This section presents the results of the Analytic Hierarchy Process (AHP), the distribution of the Problem Severity Index (PSI), and the regression models used to examine how accessibility barriers influence the mobility behavior of disabled public transport users in Alanya.

AHP Results

The AHP results are presented in three steps: (i) construction of the pairwise comparison matrix, (ii) derivation of normalized weights, and (iii) consistency verification. To identify the relative importance of accessibility barriers, a two-stage evaluation process was used. In the first stage, participants rated each of the seven accessibility criteria using a five-point Likert scale, where

• 1 = Very low severity
• 2 = Low severity
• 3 = Moderate severity
• 4 = High severity
• 5 = Very high severity

This initial scoring captured the absolute severity of each barrier based on participants’ personal mobility experiences, providing a perceptual foundation for the second stage of the analysis.

In the second stage, participants compared the criteria pairwise using Saaty’s 1–9 fundamental scale [32] to express relative importance.

For example:

• 2 reflects importance between “equal” and “moderate,”
• 4 between “moderate” and “strong,”
• 6 between “strong” and “very strong,”
• 8 between “very strong” and “extreme.”

Thus, Likert scores captured experienced severity, whereas the Saaty scale (with intermediate values) captured fine-grained comparative judgments. Together, they ensured that the AHP weighting process was both perceptually grounded and analytically robust. The aggregated 7 × 7 pairwise comparison matrix derived from participants’ comparative judgments is presented in Table 2. Since the pairwise comparisons were obtained from multiple respondents, individual judgment matrices were aggregated to derive a single group comparison matrix. In accordance with standard AHP practice, the geometric mean was used to combine individual pairwise judgments for each element of the matrix. This aggregation method preserves the reciprocal properties of the pairwise comparison matrix and is widely recommended for group decision-making applications in AHP. The resulting aggregated 7 × 7 comparison matrix was subsequently used to compute the priority weights and consistency indicators.

The pairwise comparison matrix assigns numerical values to the criteria and enables the calculation of their relative importance levels. Once the matrix is constructed, it becomes possible to quantify the priority relationships among the criteria and obtain a structured representation of their comparative weights.

Following the development of the comparison matrix, the next step involves generating the normalized matrix. Normalization is applied to standardize the values and eliminate the potential computational issues caused by extremely large or small numbers. This transformation brings all entries into a comparable scale, thereby simplifying subsequent calculations and rendering the relationships between criteria more interpretable and consistent. Matrix normalization is a widely adopted procedure in AHP-based analyses, as it ensures numerical stability and facilitates the derivation of reliable priority weights from pairwise comparisons [35].

The normalized matrix obtained from the pairwise comparisons is presented in Table 3. To determine the relative priority levels of the criteria, Saaty’s eigenvector method [32] was employed. In practice, the priority vector was derived by averaging the values across each row of the normalized matrix, which represents a widely accepted approximation of the principal eigenvector in AHP computations.

This approximation method has been extensively used in applied AHP studies due to its computational simplicity and its ability to produce consistent priority estimates that closely match the principal eigenvector solution [33]. Table 3 presents the normalized pairwise comparison matrix and the corresponding priority vector used to derive the relative weights of the accessibility criteria within the AHP framework.

The internal consistency of the pairwise comparison matrix was evaluated using the Consistency Ratio (CR), which assesses whether the comparative judgments provided by respondents are logically coherent. A CR value below the commonly accepted threshold of 0.10 indicates an acceptable level of consistency in the pairwise comparisons.

Consistency was controlled at the response-set level during data collection. After each completed questionnaire, the consistency ratio (CR) of the pairwise comparison matrix was calculated. Response sets yielding CR values greater than 0.10 were excluded, and the questionnaire was re-administered at approximately the same locations on different days until a consistency-acceptable response set (CR ≤ 0.10) was obtained. Only response sets satisfying the consistency criterion were included in the aggregation and subsequent analysis.

In this study, the calculated CR value was 0.0926, confirming that the aggregated judgments satisfied the consistency requirement and that the derived AHP weights are reliable for subsequent analysis.

Based on the final AHP weights (Table 3), bus stop condition (Criterion A) received the highest relative importance (0.29), followed by accessibility equipment in vehicles (Criterion B, 0.24) and intersection signalization (Criterion G, 0.17). Sidewalk width and curb height (Criterion E, 0.15) showed moderate importance, whereas illegal roadside parking (Criterion F, 0.08), sidewalk surface quality (Criterion D, 0.05), and sidewalk slope (Criterion C, 0.03) had comparatively lower weights.

In the study area, where the urban topography is predominantly flat, sidewalk slope was assigned a relatively low priority weight (approximately 3%) in the AHP results, indicating that it represents a less critical accessibility barrier compared to other evaluated factors. This outcome reflects local geographical conditions and may differ in cities with more complex or uneven terrain.

In the AHP analysis, inappropriate on-street parking received a relatively high priority weight, indicating its importance as an accessibility barrier, particularly during the summer tourism season when increased vehicle volumes may intensify accessibility constraints. Similarly, sidewalk width and height were assigned notable priority weights, reflecting their relative importance under conditions of seasonal population growth that increases both vehicular and pedestrian traffic and may reduce the functional adequacy of existing sidewalk infrastructure.

Unsignalized intersections were also assigned notable priority weights, highlighting their relevance as accessibility and safety concerns, particularly for individuals with visual or auditory impairments. These findings provide empirical evidence on infrastructure elements that may require consideration in accessibility-oriented transport planning and improvement efforts.

In addition to infrastructure-related factors, vehicle-related accessibility barriers emerged as a highly influential dimension in the AHP results. Accessibility equipment and vehicle design received a substantial relative importance weight (0.24), indicating that deficiencies associated with door width, entrance height, boarding assistance, and operational conditions constitute a critical component of perceived accessibility problems in public transport. These findings suggest that accessibility challenges are not limited to the physical characteristics of stops and sidewalks but also extend to the design and operational features of public transport vehicles and staff-related support mechanisms.

Finally, survey responses frequently emphasized deficiencies in bus stop design, particularly the lack of enclosed, shaded, and climate-responsive waiting areas. These concerns were strongly associated with Alanya’s climatic conditions and underline the importance of improving bus stop infrastructure to reduce climate-induced accessibility barriers.

5. Integrating AHP Results with Regression Analysis

Through the AHP analysis, the relative importance levels of accessibility problems encountered by disabled individuals in urban public transport were identified, and for each participant, a composite accessibility score called the Problem Severity Index (PSI) was computed. However, AHP alone does not reveal the quantitative behavioral impact of these problems. Therefore, a second analytical stage was conducted using regression models, where the PSI variable derived from AHP was statistically tested against behavioral outcomes such as satisfaction level, frequency of public transport use, and perceived difficulty.

Recent methodological studies have emphasized that coupling multi-criteria decision-making techniques with regression-based statistical analysis enables a more comprehensive evaluation of how prioritized service deficiencies influence observed user behavior [34,36].

This integrated approach not only determines which criteria are more significant but also demonstrates the extent to which these criteria, when combined into a total severity score, translate into actual user behavior. By linking multi-criteria decision-making outputs with statistical behavioral modeling, the study quantifies how accessibility constraints shape transportation experience and mobility decisions among disabled users. Such hybrid analytical frameworks have been increasingly adopted in transport and accessibility research to move beyond descriptive rankings and establish empirical relationships between accessibility conditions and behavioral responses [37].

5.1. Independent Variable: Problem Severity Index (PSI)

In this study, the weights of the seven accessibility criteria were derived from an aggregated (group-level) Analytic Hierarchy Process (AHP) comparison matrix, representing the collective relative importance of the barriers faced by disabled individuals in public transportation. To construct an individual-level severity measure, these group-level criterion weights (w_k) were multiplied by each participant’s corresponding severity ratings (s_ik), and the resulting products were summed to obtain the Problem Severity Index (PSI). This approach ensures that PSI values are comparable across participants, as differences in PSI reflect variation in individual severity perceptions rather than differences in weighting structures.

PSI was calculated as defined in Equation (1).

PSI represents a composite severity score that quantitatively reflects the extent of accessibility barriers experienced by each individual.

5.2. Dependent Variables

Three dependent variables were defined for the behavioral regression models:

• Y₁ (Satisfaction): “How satisfied are you with public transport services?” (measured on a 1–5 Likert scale and converted to a 0–100 interval)
• Y₂ (Usage Frequency): “How many times did you use public transport in the last month?” (measured on a 1–5 Likert scale and converted to a 0–100 interval)
• Y₃ (Difficulty Perception): “How frequently do you experience difficulties while using public transport?” (measured on a 1–5 Likert scale, converted to a 0–100 interval). The original Likert-scale responses reflect the perceived severity of accessibility-related problems, where higher values correspond to greater perceived severity. Since the Likert scale was defined such that higher values represent greater problem severity, no reverse coding was applied. The original severity ratings were directly multiplied by the AHP-derived weights and aggregated to compute the Problem Severity Index (PSI), ensuring a consistent and interpretable severity measure. Higher PSI values therefore unambiguously indicate greater accessibility-related problem severity.

$$Y_{\left(0–100\right)}=\left({\displaystyle \frac{Y-1}{\max \left(Y\right)-1}}\right)\times 100$$

(4)

5.3. Regression Model

For each dependent variable, the following linear regression model was constructed:

$$Y_i=a+\beta \times PS{I}_i+{\epsilon }_i$$

(5)

-

Y_i: Dependent variable (satisfaction, usage frequency, or perceived difficulty)

-

PSI_i: Independent variable

-

α: Intercept term

-

β: Slope coefficient representing the effect of the Problem Severity Index on the dependent variable

-

ε_i: Random error term

This functional form allows the statistical estimation of how accessibility-related problem severity influences user satisfaction, mobility frequency, and perceived difficulty.

5.4. Example Calculation

For illustration purposes, PSI values were computed by multiplying the severity ratings by the corresponding AHP weights and summing the results, following Equations (1) and (2).

For illustration purposes, Participant K001 is presented as an example. The value 2.63 represents the participant’s raw PSI score obtained by summing the products of the group-level AHP weights (w_k) and the participant’s corresponding severity ratings (s_ik) across all criteria.

Transformation to the 0–100 interval.

$$PS{I}_{\left(0–100\right)}=\left({\displaystyle \frac{2.63-1}{4}}\right)\times 100=40.8$$

Thus, Participant K001 has a PSI score of 40.8, indicating a moderate level of perceived accessibility problems.

5.5. Detailed Presentation of R² Calculations for Regression Models

In this section, the linear regression relationships between the independent variable, the Problem Severity Index (PSI, standardized to a 0–100 scale), and three behavioral outcome variables—(i) satisfaction (Y₁), (ii) public transport usage frequency (Y₂), and (iii) perceived difficulty (Y₃)—are examined. All regression estimations were conducted using the Ordinary Least Squares (OLS) method.

The regression analysis follows a simple linear specification in which each behavioral outcome is modeled as a function of PSI and an error term. The dependent variables represent perceived behavioral responses, while PSI serves as a composite indicator of accessibility-related problem severity. This reduced-form specification is employed to assess the direction and strength of the associations between accessibility barriers and behavioral outcomes, rather than to establish causal relationships.

Model parameters, including the intercept and slope coefficient, were estimated using standard OLS procedures. Model performance was evaluated using the coefficient of determination (R²), which indicates the proportion of variation in each behavioral outcome explained by PSI. Higher R² values therefore reflect a stronger linear association between problem severity and the corresponding behavioral response.

5.6. Path to R2: Computational Steps for Model 1 (Satisfaction)

Model 1 is presented in full computational detail for transparency and replicability, while Models 2 and 3 report only estimation results to avoid redundancy. This section examines the linear relationship between the Problem Severity Index (PSI, X) and satisfaction level (Y₁) using data collected from 450 respondents.

The distribution of the satisfaction variable indicates that 62 respondents reported a satisfaction level of 25, 162 respondents reported a level of 50, 147 respondents reported a level of 75, and 79 respondents reported the highest satisfaction level of 100. This distribution provides sufficient variability for regression analysis.

For each participant, two main variables were defined: the Problem Severity Index (PSI), rescaled to a 0–100 interval, and the satisfaction score (Y₁), also expressed on a 0–100 scale. Although satisfaction was observed at four discrete levels, the rescaled values allow the assessment of aggregate behavioral associations within a linear regression framework.

The regression coefficients were estimated using standard Ordinary Least Squares (OLS) procedures. The estimated slope coefficient (β = 1.44) indicates a positive association between accessibility-related problem severity and satisfaction levels. Specifically, a one-point increase in PSI is associated with an average increase of 1.44 points in satisfaction. The intercept of the model was estimated as −7.67; while this value does not carry a direct behavioral interpretation, it determines the vertical positioning of the regression line within the observed data range.

Model performance was evaluated using the coefficient of determination. For Model 1, the resulting value (R² = 0.895) indicates that approximately 89.5% of the variation in satisfaction is explained by the Problem Severity Index, demonstrating a strong linear association between perceived accessibility problems and satisfaction outcomes. The relationship between PSI and satisfaction is illustrated in Figure 2.

Overall, these results indicate that the linear regression model provides a robust reduced-form representation of the association between accessibility-related problem severity and satisfaction levels among disabled public transport users.

Figure 2 shows the relationship between the Problem Severity Index (PSI) and satisfaction level (Model 1). Colored markers represent discrete satisfaction levels (25, 50, 75, and 100), while the solid line shows the fitted linear regression model (Ŷ = −7.67 + 1.44·PSI). The coefficient of determination (R² = 0.895) indicates a strong explanatory power of the model.

As illustrated in Figure 2, there is a strong linear relationship between the Problem Severity Index (PSI) and user satisfaction. The high coefficient of determination (R² = 0.895) indicates that PSI explains a substantial proportion of the variability in satisfaction levels among disabled public transport users.

5.7. Model 2: Effect of Problem Severity Index on Public Transport Use Frequency (Y₂)

Model 2 examines the relationship between the Problem Severity Index (PSI) and the frequency of public transport use (Y₂). The dependent variable represents public transport use frequency measured on a Likert-type scale and subsequently transformed to a 0–100 scale. The distribution of responses indicates sufficient variability for regression analysis.

In the regression model, ${\mathrm{Y}}_{2\mathrm{i}}$ was calculated as defined in Equation (5), where $Y_{2i}$ denotes the public transport use frequency score, $PS{I}_i$ represents the Problem Severity Index, $\alpha$ is the intercept, β is the slope coefficient, and ${\epsilon }_i$ is the error term.

The estimated slope coefficient (β = 1.88) indicates a positive association between PSI and public transport use frequency, suggesting that individuals experiencing more severe accessibility barriers tend to rely more heavily on public transport. The intercept (α = −43.47) ensures appropriate positioning of the regression line within the observed PSI range. The coefficient of determination (R² = 0.924) indicates a strong linear association between accessibility-related problem severity and usage frequency. The relationship between the Problem Severity Index (PSI) and public transport use frequency is illustrated in Figure 3 (Model 2).

In the figure, the colored cross markers represent discrete usage frequency levels (0, 25, 50, 75, and 100), while the solid line indicates the fitted linear regression model. The estimated regression equation (Ŷ = −43.47 + 1.88·PSI) and the coefficient of determination (R² = 0.924) indicate a strong linear association between problem severity and public transport use frequency.

5.8. Model 3: Effect of Problem Severity Index on Perceived Difficulty (Y₃)

Model 3 examines the relationship between the Problem Severity Index (PSI) and perceived difficulty experienced during public transport use (Y₃). The dependent variable represents perceived difficulty measured on a Likert-type scale and subsequently transformed to a 0–100 scale. The distribution of responses indicates sufficient variability for regression analysis. In the regression model, $Y_{3i}$ was calculated as defined in Equation (5), where $Y_{3i}$ denotes the perceived difficulty score, $PS{I}_i$ represents the Problem Severity Index, $\alpha$ is the intercept, β is the slope coefficient, and εi is the error term.

The estimated slope coefficient (β = 1.88) indicates a strong linear association between accessibility-related problem severity and perceived difficulty. The intercept ($\alpha$ = −42.76) ensures appropriate positioning of the regression line within the observed PSI range. The coefficient of determination (R² = 0.924) suggests a strong linear association between problem severity and perceived difficulty.

The relationship between the Problem Severity Index (PSI) and perceived difficulty is illustrated in Figure 4 (Model 3).

In the figure, colored cross markers represent discrete difficulty levels (0, 25, 50, 75, and 100), while the solid line indicates the fitted linear regression model. The estimated regression equation and the coefficient of determination (R² = 0.924) demonstrate a strong positive association between problem severity and perceived difficulty.

The relatively high coefficients of determination observed in the regression models can be attributed to the use of a single composite explanatory variable (PSI) that aggregates multiple accessibility dimensions into a unified severity measure, as well as the transformation of behavioral responses into a standardized 0–100 scale. Given that accessibility barriers constitute a dominant determinant of satisfaction, usage frequency, and perceived difficulty among disabled public transport users, strong linear relationships are expected. Comparable levels of explanatory power have been reported in accessibility-focused behavioral studies that employ composite indices within unidimensional regression frameworks. To enable a direct comparison of model structure, coefficient estimates, and explanatory power across the three behavioral outcomes, the regression results are summarized in

Table 4. Comparison of 3 models.

Model	Dependent Variable	Intercept (α)	PSI Coefficient (β)	R2	Interpretation
Model 1	Satisfaction (Y1)	−7.67	1.44	0.895	Higher problem severity is associated with higher reported satisfaction scores, reflecting constrained choice and increased dependence on public transport despite accessibility barriers.
Model 2	Use Frequency (Y2)	−43.47	1.88	0.924	Problem severity strongly increases public transport usage frequency, indicating captive-user dependence and reliance on public transport under constrained mobility conditions.
Model 3	Perceived Difficulty (Y3)	−42.76	1.88	0.924	Increased problem severity significantly elevates perceived difficulty during public transport use.

.

The positive coefficients observed across the three regression models should not be interpreted as improved accessibility or service quality. Instead, these findings reflect a constrained choice and captive-user effect, whereby individuals experiencing more severe accessibility barriers continue to rely on public transport due to limited alternatives, affordability constraints, or the absence of accessible private mobility options. In tourism-oriented cities with pronounced seasonal fluctuations, dependence on public transport may remain high even under inadequate accessibility conditions, leading to seemingly paradoxical relationships between problem severity, usage frequency, and reported satisfaction measured on discrete scales. Therefore, the observed positive associations indicate behavioral dependence under constrained mobility rather than favorable accessibility performance. Although the behavioral outcome variables are measured on ordered categorical scales, linear regression models are employed to examine the association between accessibility-related problem severity and travel behavior. Ordered logit and ordered probit specifications were estimated as robustness checks; however, the exceptionally strong and monotonic relationship between the Problem Severity Index (PSI) and the ordered outcomes resulted in quasi-complete separation across threshold levels, leading to numerical instability and lack of convergence in maximum likelihood estimation. Such identification problems are well documented in the ordinal regression literature when explanatory variables exhibit very high discriminatory power. Consequently, linear regression is retained as the primary reduced-form specification, allowing for stable estimation and transparent interpretation of aggregate behavioral associations rather than structural or causal effects. It is important to note that the Problem Severity Index (PSI) and the behavioral outcome variables are derived from the same survey instrument and therefore reflect a shared perceptual data-generating process. As a result, the regression analysis is not designed to identify causal effects. Rather, the estimated coefficients should be interpreted as reduced-form behavioral associations that capture systematic co-variation between perceived accessibility-related problem severity and behavioral responses. Addressing potential endogeneity through instrumental variable strategies or structural equation modeling would require additional data, external instruments, or latent-variable specifications, which are beyond the scope of the present study. From a transportation planning perspective, the AHP–PSI framework provides a practical basis for prioritizing accessibility investments. The high relative importance of bus stop design and vehicle accessibility equipment suggests that improvements in these components are likely to generate the greatest behavioral and satisfaction-related benefits. Rather than simulating numerous hypothetical demand or satisfaction scenarios, the proposed framework enables planners to rank accessibility interventions based on empirically derived problem severity scores. This prioritization supports the targeted allocation of limited resources and facilitates evidence-based, cost-conscious design and retrofitting decisions in public transport systems. Future research may extend this approach by integrating detailed cost data and scenario-based simulations to further evaluate the efficiency and economic implications of alternative investment strategies. The empirical findings of this study demonstrate that accessibility-related barriers in public transport are not uniformly perceived but differ significantly in their relative severity and behavioral relevance. The AHP results indicate that bus stop conditions and vehicle-related accessibility features carry the highest relative importance, together accounting for more than half of the total problem severity weight. Furthermore, the regression analyses reveal strong linear associations between the Problem Severity Index (PSI) and key behavioral outcomes, including user satisfaction, usage frequency, and perceived travel difficulty, with high explanatory power across all models (R² > 0.85). These results confirm that perceived accessibility problem severity is a critical determinant of travel behavior among disabled public transport users.

6. Conclusions

This study examined accessibility barriers in urban public transportation systems and demonstrated that accessibility is not merely a physical design issue but also a key behavioral determinant shaping satisfaction, usage frequency, and perceived difficulty among individuals with disabilities. By integrating the Analytic Hierarchy Process (AHP) with regression analysis, the study provided a consistency-controlled and empirically grounded framework to assess both the relative importance and the behavioral impact of accessibility criteria.

The use of the AHP ensured that subjective survey responses were internally consistent before further modeling, preventing unreliable prioritization of accessibility barriers. The subsequent regression analysis established a direct and robust link between accessibility conditions and user behavior, confirming that deficiencies in infrastructure and vehicle-related accessibility features significantly influence public transport experiences. This combined methodological approach moves beyond descriptive assessments and offers a quantitative basis for understanding how accessibility improvements translate into meaningful behavioral outcomes.

From a sustainability perspective, the findings highlight that improving accessibility contributes to long-term social inclusion, independent mobility, and equitable access to urban services. In cities facing seasonal population pressures, such as tourism-oriented urban areas, accessibility shortcomings can intensify transport exclusion if not addressed through evidence-based planning. Accordingly, prioritizing accessible bus stops and vehicle equipment emerges as a critical intervention for enhancing sustainable urban mobility.

Although the empirical analysis focused on Alanya, the proposed AHP–regression framework is transferable and can be adapted to different urban contexts, transport modes, and demographic structures. As a decision-support tool, the framework enables urban planners and policymakers to prioritize accessibility investments under budget constraints and varying mobility demands. Future research may extend this approach by incorporating longitudinal data or multimodal transport systems in order to further support inclusive and sustainable transport planning.

In the context of the United Nations Sustainable Development Goals, the findings of this study directly support SDG 10 (Reduced Inequalities) by highlighting the role of accessible public transport in promoting social inclusion and independent mobility for individuals with disabilities. At the same time, the results contribute to SDG 11 (Sustainable Cities and Communities) by demonstrating how targeted accessibility interventions in public transport systems can enhance equitable access, reduce transport-related exclusion, and improve the overall sustainability of urban mobility. By linking accessibility performance to behavioral outcomes, the proposed framework provides evidence-based guidance for inclusive and sustainable transport planning.

7. Policy Implications and Recommendations

Based on the empirical findings of this study, several policy-oriented and practice-driven recommendations can be proposed to enhance accessibility within sustainable urban public transport systems. First, priority should be given to improving bus stop infrastructure, including platform heights, ramps, tactile surfaces, and visual–auditory information systems, in line with universal design principles. The AHP results indicate that bus stop conditions carry the highest relative importance among all evaluated accessibility criteria, and the regression findings further demonstrate that deficiencies at the stop level significantly influence user satisfaction and perceived travel difficulty. These results suggest that relatively low-cost infrastructure interventions at stops can yield substantial behavioral benefits.

Second, vehicle-related accessibility features such as boarding ramps, interior circulation space, and secure wheelchair positioning areas should be systematically standardized and monitored. Given the high problem severity weight assigned to vehicle-related accessibility criteria, ensuring consistent accessibility performance across the vehicle fleet is essential for reducing uncertainty and promoting independent mobility among individuals with disabilities, particularly in cities experiencing fluctuating demand and seasonal mobility pressures.

Third, accessibility planning should be integrated into broader sustainable urban mobility strategies rather than treated as a standalone technical issue. The strong linear relationships identified between the Problem Severity Index (PSI) and behavioral outcomes in the regression models suggest that accessibility improvements can contribute to increased public transport use, reduced transport exclusion, and long-term social inclusion. Accordingly, municipalities and transport authorities are encouraged to incorporate accessibility-based indicators into performance evaluation and investment prioritization processes.

Finally, the decision-support framework proposed in this study can be adopted by urban planners and policymakers to guide evidence-based accessibility investments under budget constraints. By combining AHP-based severity prioritization with behavioral impact assessment, the framework enables the identification of interventions that maximize both social equity and sustainability outcomes. Future applications of this approach may extend to different transport modes and urban contexts, supporting inclusive and resilient mobility systems.