Association of Urban Form, Neighbourhood Characteristics, and Socioeconomic Factors with Travel Behaviour in Windhoek, Namibia

Abstract

This paper investigates the associations between urban form, neighbourhood characteristics, socioeconomic factors and commuting mode choice and neighbourhood-level active travel (walking and cycling) in Windhoek, Namibia. Despite growing interest in sustainable mobility, limited research has examined these relationships in medium-sized African cities, particularly in distinguishing between commuting and neighbourhood travel behaviour. To address this gap, the study explores three interrelated research questions: (1) In what ways are urban form, accessibility, and socioeconomic factors associated with residents’ choices between motorised and non-motorised commuting modes? (2) What factors determine the propensity of cycling within neighbourhoods? (3) How are similar factors associated with walking propensity at the neighbourhood level? Using survey data from 1000 residents across nine constituencies and spatial analysis through GIS, the study applies binary logistic and multiple linear regression models to analyse commuting and local travel patterns. The findings show that commuting mode choice is significantly associated with socioeconomic status, car ownership, commuting time, and urban sprawl around homes, all of which reduce the likelihood of walking or cycling. Neighbourhood walking, in contrast, is largely driven by necessity in underserved, high-density areas and is positively associated with population density, perceived safety, and community belonging but constrained by inadequate infrastructure and car access. Cycling, though less frequent, is associated with perceived security, access to local amenities, and cycling competence, while negatively constrained by inexperience and cultural norms. The study concludes that fragmented urban form and socioeconomic disparities reinforce mobility exclusion and calls for equity-oriented transport planning that integrates infrastructure and behavioural change.

1. Introduction

The selection of transport modes used by individuals to travel between home, work, or other daily destinations is a central concern in urban and transport planning. Globally, transport modes are categorised into motorised and non-motorised forms [1]. Motorised transport refers to vehicles powered by engines, such as private cars, buses, trains, and motorcycles, which provide speed, comfort, and convenience, particularly over long distances [2]. In contrast, non-motorised transport (NMT) includes human-powered forms of mobility, such as walking and cycling, which are not only cost-effective but also offer significant public health and environmental benefits [1]. These modes, often termed “active travel”, are gaining increasing policy attention for their potential to support sustainable urban development, mitigate climate impacts, and promote healthier lifestyles [3].

A growing body of international research has consistently demonstrated that commuting mode choice is shaped by variables such as income, age, education, car ownership, and gender [4,5]. High-income households tend to favour private motorised transport for speed and convenience, whereas lower-income populations often rely on walking, cycling, or public transport out of economic necessity [6]. However, personal preferences alone do not determine travel behaviour. The built environment, defined by features such as density, land-use mix, and street connectivity, plays a pivotal role in facilitating or constraining mobility [7,8]. Cities with compact, walkable neighbourhoods and accessible public services are more conducive to non-motorised travel, while low-density, car-centric urban forms tend to reinforce automobile dependency [9,10].

In addition to structural and socioeconomic factors, individual perceptions of safety, environmental consciousness, time efficiency, and social norms also shape travel choices [11]. For instance, perceived traffic risks or a lack of cycling infrastructure may deter active travel, particularly among women, children, and older adults [12]. As urban populations continue to grow, particularly in the Global South, understanding the dynamics of travel behaviour has become increasingly important for designing transport systems that are inclusive, equitable, and sustainable.

In Sub-Saharan Africa (SSA), the context is markedly different from high-income countries where much of the transport literature originates. Although a large proportion of trips in SSA cities are still undertaken by walking or informal transit, these modes are often not supported by safe infrastructure or coherent planning [13]. Car-oriented development patterns, weak investment in public transport, and spatial inequalities, particularly between formal and informal settlements, have exacerbated urban fragmentation and constrained mobility options for most residents [14]. Moreover, travel mode choices in African cities are shaped by local cultural perceptions and safety concerns [15]. In some contexts, active travel modes, such as cycling, may be stigmatised or viewed as indicators of poverty, further limiting their uptake despite their affordability and health benefits.

Despite these challenges, the empirical literature on commuting mode choice and active travel in African cities remains sparse, particularly when it comes to integrating the built environment, socioeconomic disparities, and behavioural factors into a unified analytical framework. Much of the existing research has focused on megacities, with limited attention given to smaller and medium-sized urban areas such as Windhoek, which experience similar but unique mobility challenges.

Windhoek, the capital city of Namibia, offers a compelling and under-researched case study. The population exceeds 483,926 and is growing rapidly [16]. Windhoek is characterised by sprawling, low-density development and stark spatial inequalities that reflect its colonial and apartheid-era planning legacies [17,18]. The city’s transportation system is dominated by private vehicles and informal taxis, whereas public transport remains limited, fragmented, and unreliable [17]. Non-motorised travel, although often the only option for low-income residents in peripheral settlements, is hindered by a lack of infrastructure, long distances between homes and services, and safety concerns for pedestrians and cyclists [19]. At the same time, there is little empirical evidence to inform urban mobility policy or explain how urban form, socioeconomic status, and individual perceptions are jointly associated with commuting decisions in this context.

Despite increased regional commitments to sustainable mobility, as articulated in frameworks such as the African Union’s Agenda 2063 [20] and the Southern African Development Community (SADC) Protocol on Transport [21], the practical integration of these policy objectives into urban transport systems across Sub-Saharan Africa remains limited. The case of Windhoek illustrates this disconnect. The city continues to face significant challenges associated with rapid urbanisation, entrenched socioeconomic inequality, and a high reliance on private motor vehicles. Although the Sustainable Urban Transport Master Plan (SUTMP) [22], introduced in 2013, provides a comprehensive vision for enhancing public transport and non-motorised mobility by 2032, its implementation has been constrained by institutional fragmentation, inadequate intergovernmental coordination, and insufficient financial resources. Furthermore, the plan’s emphasis on improving mobility, rather than accessibility, has limited its capacity to address the structural barriers to equitable transport provision. Windhoek’s low-density urban form and spatially marginalised informal settlements further exacerbate these challenges by restricting access to economic and social opportunities.

This study aims to fill a critical gap in urban mobility research by examining how socioeconomic factors, urban form, neighbourhood characteristics, and individual attitudes are associated with commuting mode choice and active travel behaviour in Windhoek, Namibia. As a medium-sized, rapidly urbanising African city often overlooked in empirical transport studies, Windhoek provides a contextually rich setting for exploring the complex relationships between mobility and spatial inequality. The originality of this research lies in its integrated, multi-scalar approach that combines detailed survey data, GIS-derived urban form indicators, and regression modelling to simultaneously analyse both commuting mode choice and neighbourhood-level active travel. Crucially, this study distinguishes between commuting and neighbourhood travel behaviours, revealing how different sets of factors operate at different spatial scales. Unlike existing studies that often treat active travel as homogeneous behaviour, this study disaggregates walking and cycling and demonstrates that their determinants vary significantly between long-distance commuting and localised neighbourhood movements. By incorporating socioeconomic status, spatial structure, infrastructure perception, and cultural constraints into a unified analytical framework, this study delivers an empirically rich and methodologically innovative contribution that advances both theory and practice in the field of African urban mobility. Furthermore, it provides empirical evidence on how socio-spatial inequality, the built environment, and perceived mobility constraints shape transport behaviour, thereby informing both local and regional policy debates. In doing so, it highlights the need to shift urban transport planning away from mobility-centric frameworks toward accessibility-based, equity-driven approaches that are aligned with national development priorities and broader continental agendas.

The remainder of this paper is organised as follows. Section 2 reviews the relevant literature on commuting mode choices and active travel. Section 3 outlines the study’s methodological approach, including data sources, variables, and analysis methods. Section 4 presents the results of the empirical analysis, followed by a discussion in Section 5, which interprets the findings by considering broader urban mobility challenges. Finally, Section 6 concludes with policy recommendations for promoting more inclusive and sustainable mobility in Windhoek.

2. Literature Review

Understanding mode choice between walking, cycling, and motorised transport necessitates engaging with the interrelation of psychological intention, structural conditions, and urban form. The theory of planned behaviour (TPB) [23], a widely utilised framework in transport research, emphasises that behavioural intention is shaped by attitudes, subjective norms, and perceived behavioural control. While TPB provides a valuable conceptual foundation, its explanatory capacity depends significantly on the environmental conditions that mediate intention–behaviour translation [23,24]. In contexts marked by infrastructure deficits and spatial exclusion, such as Windhoek, the theory’s individual-centred lens must be situated within broader socio-spatial realities.

Empirical research demonstrates that objective features of the built environment—such as sidewalk quality, intersection design, lighting, and land-use mix—shape individuals’ perceived control over mobility and influence the likelihood of walking or cycling [24,25,26]. In European cities, cycling is supported by coherent infrastructure, embedded cultural norms, and early cycling competence-building [27,28]. However, in cities where active travel is shaped by necessity rather than preference, such as in many Sub-Saharan African (SSA) settings, these features manifest differently [29]. Studies from Ghana [30], Kenya [31,32], Nigeria [33], and South Africa [34] underscore the structural constraints influencing mobility in low- and middle-income urban settings. In these contexts, walking and cycling are largely practised by low-income groups, not out of preference but due to limited access to alternative transportation options. These mobility patterns unfold in environments characterised by inadequate infrastructure, limited safety, and insufficient institutional support [35]. A parallel pattern emerges in Latin American urban contexts such as Bogotá, where compact urban form and spatial connectivity provide favourable conditions for walking [36,37]. However, widespread uptake of active transportation remains limited. Evidence from an eight-country study indicates that walking is more prevalent among lower-education and non-Caucasian populations, while cycling remains rare, with 94% of adults reporting no cycling activity. Gender disparities are notable, with men reporting higher levels of active transport than women. Structural constraints, safety concerns, and social perceptions continue to impede active travel, particularly among women and higher-income groups, underscoring the role of necessity over preference in shaping mobility choice [38]. Thus, across both regions, structural and sociocultural barriers significantly mediate the extent to which built environment advantages translate into increased active travel.

Mode choice determinants fall into four interlinked domains: (1) socioeconomic characteristics (e.g., income, gender, education, car access); (2) built environment factors (e.g., density, accessibility, proximity to amenities); (3) trip characteristics (e.g., distance, purpose); and (4) psychosocial variables (e.g., perceived safety, competence, norms). In high-income contexts, these factors tend to be well integrated into transport systems that offer genuine choice. In cities across the Netherlands, as well as in Amsterdam, Warsaw, and Copenhagen, car ownership does not preclude walking or cycling, owing to robust institutional commitments to integrated, multimodal transport systems [27,39,40,41].

By contrast, in many African cities, a paradox emerges wherein walking and cycling are widespread yet remain critically under-supported by formal infrastructure. These modes are often undertaken out of necessity rather than preference, within fragmented, car-oriented urban environments shaped by inequality and spatial neglect [42,43]. For instance, in cities such as Lagos, Nairobi, and Cape Town, high levels of walking are observed despite the absence of basic pedestrian infrastructure, reflecting a context of transport poverty rather than policy-driven modal shift [31,34,44]. Similarly, Ref. [29] highlights that in cities like Accra and Harare, urban sprawl does not diminish walking rates; instead, it coexists with necessity-driven mobility where public and private motorised alternatives are largely inaccessible. Eldeeb, G., et al. [45] demonstrates that in Hamilton, Canada, walking and cycling are significantly shaped by the built environment, particularly the density of sidewalks and bike lanes, and are most common in areas where these features are well developed. Here, active travel reflects a discretionary choice rather than a lack of alternatives, with usage patterns declining as distance from the urban core increases. These divergent cases underscore that while walking may be common across diverse contexts, the forces that sustain it, structural exclusion versus infrastructural enablement, are fundamentally distinct.

Moreover, active travel behaviours are deeply embedded within sociocultural and economic contexts, with marked contrasts between high- and low-income urban environments. In many cities across the Global South, walking and cycling are primarily undertaken out of necessity and often carry social stigma, being perceived as indicators of poverty and constrained choice [28,46]. In these settings, limited cycling competence serves both as a functional barrier and a symbolic marker of exclusion from more privileged modes of transport. By contrast, in high-income cities such as Amsterdam, Copenhagen, and Berlin, walking and cycling are widely accepted and adopted across social groups and are often regarded as sustainable and socially desirable forms of mobility [25,27]. These contexts benefit from early exposure to active travel, comprehensive infrastructure, and institutional support, contributing to their broad cultural acceptance. Collectively, these findings highlight that infrastructure alone is insufficient to drive modal shift; rather, mobility practices are shaped by intersecting features of class, culture, competence, and the perceived social meaning of mobility practices.

The application of the theory of planned behaviour (TPB) within African urban contexts necessitates a critical reconceptualisation that incorporates structural barriers, local mobility cultures, and historically embedded spatial injustices. For example, in Accra and Kumasi, evidence suggests that perceived constraints, such as safety concerns, inadequate infrastructure, and economic limitations, exert a stronger influence on active travel intentions than attitudinal predispositions [47,48]. Similarly, in Bogotá, walking behaviour has been shown to result from a complex interplay between spatial accessibility and socially embedded norms [49,50]. These findings highlight that travel mode decisions are not solely governed by individual cognition or preference but are fundamentally shaped by broader environmental, institutional, and sociocultural conditions.

Despite a growing body of international research on travel behaviour, significant gaps persist in understanding how objective environmental factors, such as infrastructure quality, commuting distance, and land-use patterns, interact with subjective dimensions, including attitudes, social norms, and perceived behavioural control. As [49] and [50] highlight, much of the behavioural mobility literature is shaped by empirical insights from the Global North, where urban systems are characterised by multimodal integration, institutional coherence, and high-quality non-motorised infrastructure. By contrast, medium-sized African cities such as Windhoek remain underrepresented in empirical research, despite facing acute and distinct challenges: sprawling low-density development, transport informality, socio-spatial inequality, and fragmented governance.

These conditions constrain not only physical mobility but also the formation of behavioural intentions, particularly among vulnerable populations. Existing studies tend to isolate either environmental determinants or individual attributes, leaving a limited understanding of how they jointly shape mobility behaviour in resource-constrained, exclusionary urban environments. Furthermore, structural constraints, such as gendered mobility norms, limited cycling competence, and safety fears, are rarely incorporated into formal transport models. This underscores the need for an integrated analytical approach that moves beyond infrastructure determinism to account for the co-production of travel behaviour by spatial, institutional, and psychosocial variables. This study responds to these critical knowledge gaps by developing a contextually grounded framework that synthesises built environment indicators with subjective constructs derived from the theory of planned behaviour (TPB). In doing so, it advances a more holistic understanding of active travel behaviour in under-researched, rapidly urbanising African contexts.

3. Materials and Methods

3.1. Research Questions, Approach, and Strategy

This study was guided by three interrelated research questions aimed at understanding the factors that shape active and non-motorised travel behaviour in Windhoek. First, how are socioeconomic factors, urban form, and accessibility attributes associated with residents’ commuting choices between motorised and non-motorised travel modes? Second, to what extent do socioeconomic conditions, urban form characteristics, neighbourhood features, and individual attitudes influence residents’ decisions to cycle within their neighbourhoods? Third, how do socioeconomic status, urban form, neighbourhood context, and personal attitudes shape residents’ decisions to walk in their neighbourhoods?

To answer the research questions, this study was conducted within a positivist paradigm, employing a quantitative, cross-sectional survey design. The survey methodology followed established practices in travel behaviour research, including those used by [51], integrating spatially stratified sampling with demographic alignment. This approach enabled the systematic collection and statistical analysis of numerical data, allowing for the identification of patterns and correlations essential to understanding mobility choices in Windhoek. The methodological design aligns with studies such as [51], which used large-scale survey data to explore how socioeconomic differences shape urban travel behaviour, demonstrating the effectiveness of survey-based methods for capturing urban mobility trends.

A total of 1000 respondents were selected from nine constituencies across Windhoek between October and December 2024, using a street-intersection proximity method to ensure spatial dispersion across diverse urban forms. Within each constituency, respondents were proportionally and randomly sampled according to local population densities and settlement patterns.

To further enhance representativeness across Windhoek’s socio-spatial spectrum, a stratified random sampling strategy was employed to capture variation between high-income neighbourhoods and middle- to low-income areas. Accordingly, 500 questionnaires were administered within each socioeconomic strata, ensuring balanced coverage of both spatial and socioeconomic profiles.

In the absence of up-to-date constituency-level income data, the sampling strategy was informed by poverty headcount rates from the Namibia Poverty Mapping Report of 2015 [52], derived from the 2011 Population and Housing Census. These poverty rates, defined as the proportion of residents living below the lower-bound poverty line, served as proxies for socioeconomic conditions. Constituencies such as John Pandeni, Katutura Central, Katutura East, Khomasdal, Samora Machel, Tobias Hainyeko, and Moses ǁGaroëb were characterised by relatively higher poverty levels (0.5–3.6%), whereas Windhoek East and Windhoek West recorded poverty rates below 0.1%, consistent with their higher levels of infrastructure provision.

3.2. Case Study

Windhoek, the capital of Namibia, is a rapidly urbanising, medium-sized African city that presents a compelling context for examining how urban form, neighbourhood characteristics, and socioeconomic conditions shape commuting mode choice and active travel. With a population of approximately 483,926 and an urbanisation rate of 4% per year [16], Windhoek spans a vast 5133 km² but maintains a low average population density of just 13.4 persons per km², typical of many sprawling cities in Sub-Saharan Africa.

This study focuses on nine constituencies: John Pandeni, Katutura Central, Katutura East, Khomasdal, Samora Machel, Tobias Hainyeko, Moses ǁGaroëb, Windhoek West, and Windhoek East, which together represent a cross-section of the city’s diverse built environments and socioeconomic disparities. As shown in Figure 1, these areas range from high-density, low-income neighbourhoods with limited infrastructure to low-density, affluent suburbs with better connectivity and services.

The spatial configuration of Windhoek reflects the enduring legacy of apartheid urban planning. Constituencies such as Katutura Central, Tobias Hainyeko, and Moses ǁGaroëb remain densely populated and characterised by informal settlement growth, poor infrastructure, and limited access to formal transportation. In contrast, Windhoek West and Windhoek East, though part of the urban core, are relatively affluent, well-serviced, and enjoy proximity to economic and administrative centres. This fragmented and unequal urban form continues to influence access to jobs, services, and transport, reinforcing exclusionary mobility patterns.

Windhoek is particularly well-suited as a case study for this research because it shares key structural features with many other small- and medium-sized cities across Sub-Saharan Africa. These include rapid but uneven urbanisation, spatial inequality rooted in historical planning, a high reliance on informal transport systems, and emerging pressures on infrastructure. Furthermore, the coexistence of formal and informal development in Windhoek, along with its transport and mobility disparities, reflects the broader set of urban challenges faced across the region.

In this context, understanding how neighbourhood-level factors shape commuting and active travel behaviours in Windhoek can provide valuable insights for promoting inclusive, sustainable mobility in other similarly structured African cities. The city’s heterogeneous urban fabric, ranging from informal settlements and township areas to affluent suburbs, provides a basis for comparative analysis that is both contextually grounded and broadly applicable to the dynamics of urbanisation in the Sub-Saharan African region.

3.3. Data and Variables

Data collection was carried out through in-person interview surveys conducted by trained enumerators who were stationed at selected street intersections across Windhoek. The final distribution of responses across the nine sampled constituencies was as follows: John Pandeni (n = 62), Katutura Central (n = 75), Katutura East (n = 82), Khomasdal (n = 70), Samora Machel (n = 80), Tobias Hainyeko (n = 76), Moses ǁGaroëb (n = 55), Windhoek West (n = 255), and Windhoek East (n = 245). This distribution reflects both the proportionality and the socio-spatial balance embedded in the sampling design, with adequate representation from both high-income and low-income urban areas. To further illustrate the territorial coverage achieved during data collection, Figure 2 presents the approximate geographic distribution of respondents’ home locations across Windhoek.

To ensure the validity and generalisability of the sample, key demographic variables, namely age group, gender, education level, and income, were cross-tabulated against official benchmarks provided by the 2023 Namibian Population and Housing Census. This comparative assessment demonstrated that the sample closely aligns with the demographic structure of Windhoek’s urban population. Such demographic comparability enhances the external validity of the findings and situates the study within established best practices in urban mobility research. Specifically, it reflects approaches that prioritise neighbourhood-based sampling frameworks and demographic fidelity to ensure robust behavioural analysis in spatially heterogeneous urban environments [51].

The questionnaire contained 105 questions across four sections: (1) neighbourhood characteristics (e.g., neighbourhood entertainment preference, security, safety), (2) motorised transport modes (e.g., mini-bus, taxi, public transit usage), (3) non-motorised travel (e.g., walking and cycling behaviour, perceptions of safety and convenience, NMT infrastructure), and (4) individual and household characteristics (e.g., age, gender, income, car ownership, and driving licence).

Table 1. Data and variables.

Variable	Variable Type	Description
Subjective Variables
Education	Dummy	0 = lower education (primary and secondary) 1 = higher education (college and university)
Income	Dummy	0 ≤ NAD 5000, 1 ≥ NAD 5000
Age group	Dummy	0 = under 35 years 1 = 35 years and above
Gender	Binary	Female = 1; male = 0
Driving licence	Continuous	0 = yes, 1 = no
Car ownership	Continuous	Number of cars owned in the respondent’s household
Number of adults in the household	Continuous	Number of adults in the respondent’s household
Number of household children	Continuous	Number of children in the respondent’s household
The primary mode of transport	Dummy	0 = motorised, 1 = non-motorised
Perceived cycling propensity	Continuous	Agreement with cycling to local destinations; measured on a continuous scale from 0 (strongly disagree) to 100 (strongly agree)
Perceived walking propensity	Continuous	Agreement with walking to local destinations; measured on a continuous scale from 0 (strongly disagree) to 100 (strongly agree)
Lack of NMT infrastructure	Continuous	Perception of absence of sidewalks or bike lanes; rated from 0 (strongly disagree) to 100 (strongly agree)
Neighbourhood entertainment preference	Continuous	Preference for local vs. distant entertainment locations; rated from 0 to 100
Perceived security in the neighbourhood	Continuous	Agreement with the statement “There is little security in the neighbourhood”; 0 = strongly disagree, 100 = strongly agree
Perceived safety when walking in the neighbourhood	Continuous	How safe the respondent feels walking in the neighbourhood due to traffic; rated from 0 to 100
Sense of belonging to my neighbourhood	Continuous	Agreement with the statement “I feel a sense of belonging to my neighbourhood”; rated from 0 to 100
Lack of cycling competence	Continuous	Agreement with the statement “I do not know how to cycle”; rated from 0 to 100
Social cultural barrier to walking	Continuous	Agreement with the statement “There are social and cultural barriers near my living place”; rated from 0 to 100
Neighbourhood shopping facilities	Continuous	Agreement with the statement “There are attractive shops or shopping centres in my neighbourhood”; rated from 0 to 100
Age disability constraint	Continuous	Agreement with the statement “I am too old/disabled to cycle”; rated from 0 to 100
Proximity of the house to work/study place	Continuous	Agreement with the statement “The house is near to my workplace/school”; rated from 0 to 100
Proximity to e-hailing stop	Continuous	Respondents rated how close the nearest e-hailing pickup point is; 0 = very far, 100 = very close
Commuting time	Continuous	Self-reported average one-way commute time in minutes
Objective Variables
Population density	Continuous	The average number of people per square kilometre (extracted from the Namibian 2023 Population and Housing Census Report)
Urban sprawl around home	Continuous	This variable was not self-reported but derived using geographic information systems (GISs) to quantify urban sprawl at the zonal level. Each respondent was assigned a corresponding Shannon entropy value based on their residential location
Commuting distance	Continuous	This variable was not self-reported but derived using geographic information systems (GISs). The commuting distance was calculated as the street-network distance between the nearest intersections to each respondent’s home and workplace
Intersection density around the workplace or university	Continuous	Number of street intersections per km2 around the respondent’s work or study location; derived from GIS data

summarises the variables included in the analysis. These variables collectively capture the complex interrelations among the built environment, socioeconomic status, and mobility behaviour, providing a robust foundation for analysing commuting mode choice and active travel.

The questionnaire design was guided by the theory of planned behaviour (TPB), which informed the measurement of attitudes, subjective norms, and perceived behavioural control related to active travel. Attitudes were operationalised through indicators such as walking and cycling propensity, neighbourhood satisfaction, and perceived access to local amenities. Subjective norms were measured via perceived social and cultural barriers to walking, as well as self-reported cycling competence. Perceived behavioural control encompassed infrastructure availability, safety concerns, physical ability, and spatial proximity to destinations. These constructs enabled the contextual application of TPB within Windhoek’s urban environment. In addition, socio-demographic variables, including education, income, age, gender, car ownership, and household composition, were included to account for baseline differences in transport behaviour and constraints.

To complement these subjective measures, the study incorporated objective environmental variables. These included population density from the 2023 Namibian Census [16], urban sprawl (quantified using Shannon entropy via GIS), commuting distance (calculated through street-network analysis), and intersection density near workplaces or study locations (derived from OpenStreetMap data). These spatial metrics captured structural conditions that may affect mobility, offering a broader analytical lens beyond individual intention.

Collectively, these variables allowed the study to test TPB-consistent hypotheses: that more favourable attitudes, stronger perceived social support, and greater behavioural control are associated with increased likelihood of engaging in non-motorised travel. By integrating psychosocial and spatial dimensions, the study advances a more comprehensive model of active travel suited to the socio-spatial features of Windhoek.

Urban sprawl, commuting distance, and intersection density around respondents’ homes and workplaces were analysed using ArcGIS Pro v3.3.1. Land-use classification was conducted using satellite imagery sourced from the USGS Earth Explorer, which was processed to delineate land-use categories within the study area. Built-up areas were extracted from the classified data and converted into polygon features for further spatial analysis. The study area shapefiles were acquired and systematically divided into 400-metre grid zones, resulting in spatial units of 160,000 m² (0.16 km²) each. A total of 3323 uniform zones were created to structure the spatial analysis.

To maintain participant privacy, respondents were asked to indicate the street names of the closest intersection to their home and work/study location rather than to provide exact coordinates. Using the OpenStreetMap (OSM) road dataset, these intersection points were geocoded manually to enhance locational accuracy. Once plotted, these intersections were overlaid onto the built-up area map to generate 600-metre catchment buffers around the home and work/study zones. These catchment areas were then intersected with the pre-established grid zones to create individualised spatial units for each respondent, which were used to compute urban form variables such as sprawl, commuting distance, and intersection density.

Urban sprawl was measured using the Shannon entropy index, which quantifies the spatial dispersion of built-up areas across the zones. The following formula was used:

$$H = - \sum _{i=1}^{n}pᵢ·\mathit{ln}\left(pᵢ\right)$$

In the entropy formula, H represents the Shannon entropy for the respondent’s residential zone, n is the total number of zones, pᵢ is the proportion of built-up area in the iᵗʰ zone, and ln is the natural logarithm.

The proportion pᵢ is calculated using this formula:

$$p_i= {\displaystyle \frac{B_i}{\sum _{i=1}^n{B}_i}}$$

where Bᵢ is the built-up area in the: iᵗʰ zone, and ∑ Bᵢ is the total built-up area across all zones. This ensured that pᵢ reflected the relative share of built-up land in each zone.

To facilitate interpretation, the entropy values were normalised between 0 and 1, where 0 indicates a fully compact form, and 1 signifies a fully sprawled area. The Shannon entropy approach is well suited for urban sprawl analysis, as it captures not just the extent of built-up land but its spatial distribution, which is a key aspect of sprawl [53]. Its use in this study allows for an understanding of how compact or dispersed residential areas are within Windhoek’s urban fabric.

Intersection density was computed to evaluate the geographic connectivity of the street network across each zone. This statistic indicates the density of street intersections in a specific area and reflects the degree of connectedness (street connectivity) and accessibility in urban settings. This metric reflects the frequency of connections within a road network and is widely used in urban form and walkability assessments [54,55]. The computation was performed using GIS tools in ArcGIS Pro, employing the road network dataset acquired from OpenStreetMap (OSM). The following formula was used:

$$\mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{D}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y} = {\displaystyle \frac{\mathrm{N}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r} \mathrm{o}\mathrm{f} \mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}}{\mathrm{Z}\mathrm{o}\mathrm{n}\mathrm{e} \mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}\left({\mathrm{k}\mathrm{m}}^2\right)}}$$

Furthermore, the commuting distance was calculated using ArcGIS Pro and a network-based approach that measures the travel distance along the actual road network rather than relying on the straight-line (Euclidean) distance. The respondents’ home and workplace locations were geocoded using the nearest street intersections, and the origin-destination (OD) cost matrix tool was applied to compute the shortest travel path between each origin and destination. Distances were calculated in metres, allowing for high-resolution measurements and comparability across zones. This method accurately reflects real-world urban mobility conditions and enhances the precision of commuting distance estimates [56].

3.4. Characteristics of the Sample Size

Table 2. Descriptive statistics for categorical variables by neighbourhood type.

Variable	Category	Low-Income Area (%)	High-Income Area (%)	Total (%)
Gender	Female	52.2%	49.8%	51.0%
	Male	47.8%	50.2%	49.0%
Age group	Under 35 years	58.2%	55.0%	56.6%
	35 years and above	41.8%	45.0%	43.4%
Education level	Lower (primary and secondary)	57.0%	24.6%	40.8%
	Higher (college and university)	43.0%	75.4%	59.2%
Driving licence ownership	Yes	39.8%	63.0%	51.4%
	No	60.2%	37.0%	48.6%
Monthly household income	≤NAD 5000	54.8%	19.6%	37.2%
	>NAD 5000	45.2%	80.4%	62.8%
Commute mode to work/study	Motorised	88.6%	88.0%	88.3%
	Non-motorised (walking/cycling)	11.4%	12.0%	11.7%

and

Table 3. Descriptive statistics for continuous variables by neighbourhood type.

Variable	Neighbourhood Type	N	Min	Max	Mean	Std. Deviation
Population density	Low-income Area	500	3.2	4.0	3.693	0.2136
	High-income Area	500	2.8	3.3	3.042	0.2501
Number of adults in household	Low-income Area	500	1	6	2.69	1.177
	High-income Area	500	1	6	2.55	1.226
Number of children in the household	Low-income Area	500	0	7	1.57	1.348
	High-income Area	500	0	5	1.48	1.213
Number of cars in the household	Low-income Area	500	0	3	0.51	0.653
	High-income Area	500	0	3	0.93	0.851
Perceived cycling propensity (0–100)	Low-income Area	500	0	90	7.78	18.307
	High-income Area	500	0	90	7.92	18.525
Perceived walking propensity (0–100)	Low-income Area	500	10	100	59.00	23.582
	High-income Area	500	0	90	30.00	18.695

outline the demographic, socioeconomic, and mobility-related characteristics of survey respondents, disaggregated by neighbourhood income level. The analysis draws on both categorical and continuous variables to examine socio-spatial differences between Windhoek’s high-income and low-income urban areas. These descriptive statistics provide important context for interpreting patterns in travel behaviour across contrasting socioeconomic settings.

The descriptive profile in Table 2 reveals marked socio-spatial contrasts between Windhoek’s high-income and low-income neighbourhoods. Although gender and age group distributions are relatively balanced across the two contexts, significant differences emerge in terms of educational attainment, household income, and mobility resources. In high-income neighbourhoods, 75.4% of respondents reported higher education compared to 43.0% in low-income areas. Similarly, 80.4% of high-income area respondents reported household incomes exceeding NAD 5000, in contrast to 45.2% in the low-income areas. Car ownership and access to a driving licence follow similar patterns, with higher levels observed in high-income neighbourhoods. Despite these socioeconomic differences, reliance on motorised transport is high in both areas, at approximately 88%, suggesting widespread dependence on motorised mobility, albeit likely through differing modalities: private car use in high-income areas and shared or informal transport in low-income areas. These findings highlight the enduring structural inequalities shaping urban mobility in Windhoek and support the imperative for planning interventions that respond to the city’s socio-spatial diversity.

Descriptive statistics for continuous variables in Table 3 demonstrate marked socio-spatial disparities between Windhoek’s high-income and low-income neighbourhoods. Population density was higher in low-income areas (M = 3.69) relative to high-income areas (M = 3.04), consistent with patterns of densification in informally developed or under-serviced settlements. Household composition also differed, with low-income neighbourhoods reporting marginally higher numbers of adults and children per household, indicative of larger household sizes and potential overcrowding. Car ownership levels were substantially lower in low-income areas (M = 0.51) compared to high-income areas (M = 0.93), reflecting limited access to private vehicles and increased reliance on alternative modes. Walking levels were significantly higher in low-income neighbourhoods (M = 59.0) than in high-income ones (M = 30.0), likely driven by necessity rather than modal preference given the restricted availability of formal transport services. In contrast, cycling levels remained low across both neighbourhood types, suggesting systemic barriers to active travel, such as infrastructural deficits, safety concerns, or cultural perceptions. These findings highlight the structural inequalities that shape transport behaviours in Windhoek and emphasise the importance of context-sensitive planning approaches that account for divergent urban forms, income levels, and mobility constraints.

3.5. Analytical Methods

To answer the first research question, a binary logistic regression model was estimated to examine how urban form, socioeconomic characteristics, and accessibility attributes influence residents’ commuting mode choice in Windhoek. The dependent variable was commuting mode choice, coded as a binary outcome:

$$Y = \left\{\begin{array}{c}1 if respodent uses non-motorised transport (walking/cycling)\\ 0 if respodent uses motorised transport \end{array}\right.$$

The probability of choosing a non-motorised commuting mode is expressed as:

$$\mathrm{P}\left(\mathrm{Y}\mathrm{ᵢ}=1\right)={\displaystyle \frac{1}{\left(1+{\mathrm{e}}^{-\mathrm{Z}\mathrm{ᵢ}}\right)}}$$

where Zi is the linear predictor defined by:

$$\begin{gathered} \mathrm{Z}\mathrm{ᵢ}={\mathsf{\beta }}^0+{\mathsf{\beta }}^1\left(\mathrm{C}\mathrm{o}\mathrm{m}\mathrm{m}\mathrm{u}\mathrm{t}\mathrm{e}\mathrm{D}\mathrm{i}\mathrm{s}\mathrm{t}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^2\left(\mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{D}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^3\left(\mathrm{U}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}\mathrm{S}\mathrm{p}\mathrm{r}\mathrm{a}\mathrm{w}\mathrm{l} \mathrm{a}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{d} \mathrm{w}\mathrm{o}\mathrm{r}\mathrm{k}\right)\mathrm{ᵢ}+ \\ {\mathsf{\beta }}^4\left(\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{x}\mathrm{E}-\mathrm{H}\mathrm{a}\mathrm{i}\mathrm{l}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^5·\mathrm{C}\mathrm{o}\mathrm{m}\mathrm{m}\mathrm{u}\mathrm{t}\mathrm{e}\mathrm{T}\mathrm{i}\mathrm{m}\mathrm{e}\mathrm{ᵢ}+{\mathsf{\beta }}^6·\mathrm{E}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{ᵢ}+{\mathsf{\beta }}^7\left(\mathrm{G}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{r}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^8\left(\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}\right)\mathrm{ᵢ}+ \\ {\mathsf{\beta }}^9\left(\mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{L}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{e}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^{10}\left(\mathrm{C}\mathrm{a}\mathrm{r}\mathrm{O}\mathrm{w}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{h}\mathrm{i}\mathrm{p}\right)\mathrm{ᵢ}+{\mathsf{\beta }}^{11}\left(\mathrm{N}\mathrm{u}\mathrm{m}\mathrm{C}\mathrm{h}\mathrm{i}\mathrm{l}\mathrm{d}\mathrm{r}\mathrm{e}\mathrm{n}\right)\mathrm{ᵢ}+\mathsf{\epsilon }\mathrm{ᵢ} \end{gathered}$$

(1)

The initial models included commuting trip frequencies, population density, and urban sprawl around home; however, during the model refinement process, predictors that did not meet the threshold for statistical significance (p < 0.05) were excluded. This approach improved the model efficacy and focused on the analysis of variables with meaningful explanatory power. The model was estimated using SPSS Version 27 with non-motorised commuting as the event category.

A multiple linear regression (MLR) model was applied to assess the extent to which socioeconomic characteristics, urban form, neighbourhood features, and individual attitudes determine residents’ decisions to cycle within their neighbourhoods. Initially, a broad set of independent variables was included in the model. However, variables with high p-values were removed because of statistical insignificance, and only those with statistically significant effects were retained in the final model. Explanatory variables were grouped into four conceptual categories: urban form (e.g., population density, urban sprawl around home, cycling infrastructure), neighbourhood characteristics (e.g., security, sense of belonging, access to neighbourhood shopping facilities, neighbourhood entertainment preferences), individual attitudes (e.g., perceived competence, cultural barriers, and age-related constraints), and socioeconomic attributes (e.g., age, education, driving licence status).

The estimated model can be represented as follows:

$$\begin{array}{cc}\mathrm{C}\mathrm{y}\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{q}\mathrm{ᵢ}=& {\mathsf{\beta }}^0+{\mathsf{\beta }}^1\mathrm{P}\mathrm{o}\mathrm{p}\mathrm{D}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{ᵢ}+{\mathsf{\beta }}^2\mathrm{U}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}\mathrm{S}\mathrm{p}\mathrm{r}\mathrm{a}\mathrm{w}\mathrm{l}\mathrm{ᵢ}+{\mathsf{\beta }}^3\mathrm{L}\mathrm{a}\mathrm{c}\mathrm{k}\mathrm{C}\mathrm{y}\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{S}\mathrm{k}\mathrm{i}\mathrm{l}\mathrm{l}+{\mathsf{\beta }}^4\mathrm{N}\mathrm{M}\mathrm{T}\mathrm{I}\mathrm{n}\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{ᵢ}\hfill \\ & +{\mathsf{\beta }}^5\mathrm{N}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{b}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{h}\mathrm{o}\mathrm{o}\mathrm{d}\mathrm{S}\mathrm{e}\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{ᵢ}+{\mathsf{\beta }}^6\mathrm{B}\mathrm{e}\mathrm{l}\mathrm{o}\mathrm{n}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{ᵢ}+{\mathsf{\beta }}^7\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{i}\mathrm{o}\mathrm{C}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{l}\mathrm{B}\mathrm{a}\mathrm{r}\mathrm{r}\mathrm{i}\mathrm{e}\mathrm{r}\mathrm{ᵢ}\hfill \\ & +{\mathsf{\beta }}^8\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{ᵢ}+{\mathsf{\beta }}^9\mathrm{S}\mathrm{h}\mathrm{o}\mathrm{p}\mathrm{p}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{A}\mathrm{c}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{ᵢ}+{\mathsf{\beta }}^{10}\mathrm{A}\mathrm{g}\mathrm{e}\mathrm{D}\mathrm{i}\mathrm{s}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{ᵢ}+{\mathsf{\beta }}^{11}\mathrm{A}\mathrm{g}\mathrm{e}\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{p}\mathrm{ᵢ}\hfill \\ & +{\mathsf{\beta }}^{12}\mathrm{E}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{ᵢ}+{\mathsf{\beta }}^{13}\mathrm{D}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{L}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{e}\mathrm{ᵢ}+\mathsf{\epsilon }\mathrm{ᵢ}\hfill \end{array}$$

(2)

In this formulation, each β coefficient represents the marginal change in neighbourhood perceived cycling propensity associated with a one-unit change in the respective predictor, holding all other variables constant. The term εi is the error term, which captures unobserved influences on perceived cycling propensity and is assumed to be normally distributed with constant variance.

To answer the third research question, a multiple linear regression (MLR) model was employed to assess the extent to which socioeconomic factors, urban form, neighbourhood characteristics, and individual attitudes were associated with walking behaviour in Windhoek. This approach is appropriate for modelling the relationship between a continuous dependent variable and a set of theoretically grounded predictors. The independent variables include four conceptual categories: urban form (e.g., population density, urban sprawl, perceived NMT infrastructure), neighbourhood characteristics (e.g., perceived safety and security, sense of belonging), individual attitudes, and socioeconomic attributes (e.g., income, education, and car ownership). While a wide range of predictors were initially included in the model, variables with high p-values were removed because of statistical insignificance, and only those with statistically significant effects were retained in the final model. Their removal improved the model’s precision while maintaining conceptual coherence.

The estimated model can be expressed as:

$$\begin{array}{cc}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{q}\mathrm{ᵢ}=& {\mathsf{\beta }}^0+{\mathsf{\beta }}^1\mathrm{P}\mathrm{o}\mathrm{p}\mathrm{D}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{ᵢ}+{\mathsf{\beta }}^2\mathrm{U}\mathrm{r}\mathrm{b}\mathrm{a}\mathrm{n}\mathrm{S}\mathrm{p}\mathrm{r}\mathrm{a}\mathrm{w}\mathrm{l}\mathrm{ᵢ}+{\mathsf{\beta }}^3\mathrm{N}\mathrm{M}\mathrm{T}\mathrm{I}\mathrm{n}\mathrm{f}\mathrm{r}\mathrm{a}\mathrm{ᵢ}+{\mathsf{\beta }}^4\mathrm{N}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{b}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{h}\mathrm{o}\mathrm{o}\mathrm{d}\mathrm{S}\mathrm{e}\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{ᵢ}\hfill \\ & +{\mathsf{\beta }}^5\mathrm{S}\mathrm{a}\mathrm{f}\mathrm{e}\mathrm{t}\mathrm{y}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{ᵢ}+{\mathsf{\beta }}^6\mathrm{B}\mathrm{e}\mathrm{l}\mathrm{o}\mathrm{n}\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{ᵢ}+{\mathsf{\beta }}^7\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{ᵢ}+{\mathsf{\beta }}^8\mathrm{E}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{ᵢ}\hfill \\ & +{\mathsf{\beta }}^9\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{ᵢ}+{\mathsf{\beta }}^{10}\mathrm{C}\mathrm{a}\mathrm{r}\mathrm{O}\mathrm{w}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{h}\mathrm{i}\mathrm{p}\mathrm{ᵢ}+{\mathsf{\beta }}^{11}\mathrm{N}\mathrm{u}\mathrm{m}\mathrm{A}\mathrm{d}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{s}\mathrm{ᵢ}+\mathsf{\epsilon }\mathrm{ᵢ}\hfill \end{array}$$

(3)

In this formulation, each β coefficient represents the estimated change in perceived walking propensity for a one-unit change in the corresponding predictor, holding all other variables constant. The error term εi accounts for the residual variation in perceived walking propensity not explained by the included predictors and is assumed to be normally distributed with constant variance.

4. Results

4.1. Estimated Model: Determinants of Commuting Mode Choice in Windhoek

The binary logistic regression model explored the determinants of commuting mode choice in Windhoek using IBM SPSS Statistics (version 28). As shown in

Table 4. Association between socioeconomic factors, urban form, and accessibility attributes and commuting mode choice in Windhoek.

Variables Group	Variables	B	S.E.	Wald	p	Exp (B)
Urban form objective	Commuting distance	−0.000	0.000	4.515	0.034	1.000
	Intersection density around workplace or university	0.001	0.000	3.869	0.049	1.001
	Urban sprawl around home	−0.029	0.003	70.348	<0.001	0.971
Urban form subjective	Proximity to e-hailing stop	−0.013	0.005	8.058	0.005	0.987
Mobility pattern	Commuting time in minutes	−0.048	0.013	14.196	<0.001	1.049
Control variables	Education	0.466	0.289	2.614	0.106	1.594
	Driving licence	−0.601	0.289	4.331	0.037	0.548
	Gender	−0.524	0.241	4.740	0.029	0.592
	Car ownership	−1.645	0.297	30.580	<0.001	5.180
	Household children	0.134	0.098	1.852	0.174	1.143
	Income	−0.699	0.283	6.097	0.014	0.497
	Constant	2.469	0.676	13.333	<0.001	11.812
Specification Tests
	Nagelkerke R2	Omnibus Test
	0.40		Chi-square	df	p
			230.797	11	<0.001
	Observations		Hosmer and Lemeshow Test
	N = 1000		Chi-square	df	p
			7.527	8	0.481

Notes: dependent variable: commuting mode choice (0 = motorised, 1 = non-motorised).

, out of the 11 predictors included in the model, 9 were statistically significant at p < 0.05. Among these, four variables are highly significant (p < 0.01): urban sprawl, commuting time, proximity to e-hailing stop, and car ownership. Five other variables were moderately significant (0.01 < p < 0.05): commuting distance, intersection density, driving licence, gender, and income.

The strongest predictor of commuting mode was car ownership, with individuals who owned a car being over five times less likely to walk or cycle to work (B = −1.645, p < 0.001, Exp(B) = 5.180). This finding reflects the high dependency on private vehicles when available and the structural limitations faced by those without a car. Similarly, longer commuting times significantly reduced the likelihood of using non-motorised modes (B = −0.048, p < 0.001, Exp(B) = 1.049), indicating that each additional minute of commuting decreased the odds of walking or cycling by approximately 4.8%. The built environment also plays a notable role. Urban sprawl around the home was negatively associated with non-motorised commuting (B = −0.029, p < 0.001, Exp(B) = 0.971), suggesting that residents in more sprawled and disconnected neighbourhoods are less likely to walk or cycle when going to work or study.

In contrast, intersection density around the workplace was positively associated with non-motorised mode use (B = 0.001, p = 0.049, Exp(B) = 1.001); however, this association occurred with a small effect size, indicating that more connected urban networks near employment or study centres may slightly support active commuting.

Interestingly, proximity to e-hailing stops was also a significant predictor (B = −0.013, p = 0.005, Exp(B) = 0.987) but in the direction suggesting that greater perceived access to ride-hailing services reduced the likelihood of non-motorised travel. This may reflect a substitution effect in which easier access to convenient motorised options discourages walking or cycling. Sociodemographic characteristics were also found to be influential. Income was negatively associated with active commuting (B = −0.699, p = 0.014, Exp(B) = 0.497), indicating that higher-income individuals were less likely to use non-motorised modes, possibly because of greater vehicle access or residential location choices. Possession of a driving licence similarly reduced the likelihood of walking or cycling (B = −0.601, p = 0.037, Exp(B) = 0.548). Gender was a significant predictor (B = −0.524, p = 0.029, Exp(B) = 0.592), with females being less likely than males to engage in non-motorised commuting, possibly because of safety concerns, caregiving responsibilities, or differing perceptions of urban walkability.

The binary logistic regression model demonstrates a strong overall fit and explanatory power in predicting commuting mode choice in Windhoek. The Nagelkerke R² value of 0.40 indicates that 40% of the variance in active versus motorised commuting is explained by the model, a substantial proportion for behavioural models in urban transport research. Furthermore, the omnibus test of model coefficients was highly significant (χ² = 230.797, p < 0.001), confirming that the set of included predictors collectively enhances the model’s predictive capability. Additionally, the Hosmer–Lemeshow test (χ² = 7.527, p = 0.481) yielded a non-significant result, indicating a good fit between the predicted and observed values. These diagnostic indicators suggest that the model is both statistically sound and practically useful for understanding the determinants of mode choice in Windhoek.

4.2. Estimated Model: Determinants of Perceived Cycling Propensity in Windhoek

The Multiple linear regression model in

Table 5. Determinants of neighbourhood walking behaviour: socioeconomic factors, urban form, neighbourhood characteristics, and individual attitudes in Windhoek.

Variables Group			Variables				B		Beta		t		p	VIF
			(Constant)				30.794				4.649		<0.001
Urban form objective			Population density				−2.908		−0.063		−1.817		0.070	1.588
			Urban sprawl around home				−0.039		−0.072		−2.276		0.023	1.302
Urban form subjective			Lack of cycling competence				−0.161		−0.388		−13.863		<0.001	1.030
			Lack of NMT infrastructure				−0.051		−0.062		−2.170		0.030	1.078
			Perceived security in the neighbourhood				0.625		0.180		6.253		<0.001	1.082
			Sense of belonging to my neighbourhood				0.035		0.058		1.982		0.048	1.128
			Sociocultural barrier				−0.039		−0.072		−2.276		0.023	1.302
			Neighbourhood entertainment preference				0.039		0.070		2.382		0.017	1.126
			Neighbourhood shopping facilities				0.055		0.100		3.163		0.002	1.306
			Age/disability constraint				−0.037		−0.065		−2.136		0.033	1.215
Socioeconomic			Age group				2.753		0.074		2.504		0.012	1.150
			Education				−2.006		−0.054		−1.681		0.093	1.333
			Driving licence				−4.266		−0.116		−3.627		<0.001	1.339
Model Summary
R	R Square			Adjusted R Square				Std. Error of the Estimate
0.498	0.248			0.238				16.065
ANOVA F-Test
		Sum of Squares			df	Mean Square				F		p
Regression		84,007.151			13	6462.089				25.039		<0.001
Residual		254,470.349			986	258.084

presents the determinants of perceived cycling propensity in the neighbourhood of Windhoek.

Among the 13 predictors included, four variables were highly significant (p < 0.01): lack of cycling competence, perceived neighbourhood security, driving licence possession, and availability of shopping facilities. Six variables were moderately significant (0.01 < p < 0.05): urban sprawl, social and cultural barriers, sense of neighbourhood belonging, neighbourhood entertainment preference, lack of NMT infrastructure, and age group. An additional two variables were marginally significant (0.05 < p < 0.1): population density and education. These findings highlight the complex nature of cycling behaviour shaped by psychological, environmental, and socio-demographic factors.

The most influential barrier was the lack of cycling competence, which reduced the perceived cycling propensity by nearly 39% based on its standardised effect size (beta = –0.388). Perceived neighbourhood security increased cycling by 18%, making it the most influential positive factor. Access to shopping facilities (+10%) and entertainment opportunities (+7%) also promotes cycling, highlighting the importance of local amenities. However, having a driving licence reduced perceived cycling propensity by 12%. Urban sprawl around homes, which is one of the urban form objective factors, decreased cycling by 7%. while social norms and age/disability constraints decreased cycling by approximately 6–7%. A stronger sense of neighbourhood belonging increased cycling modestly (+6%), whereas older age groups were slightly more likely to cycle (+7.4%). This may indicate a shift in age-related trends, potentially driven by lifestyle changes, increased leisure time, or health-oriented behaviours among older adults.

The MLR model in Table 5 highlights that while access to infrastructure and services is important, perceptions of safety, cultural norms, and individual competence are equally, if not more strongly, associated with perceived cycling propensity. Therefore, effective interventions should combine infrastructure investment with capacity building, safety enhancement, and community-level engagement to foster inclusive and sustainable cycling habits.

The multiple linear regression model offers a statistically meaningful explanation of perceived cycling propensity in Windhoek, accounting for approximately 25% of the variance (R² = 0.248). The adjusted R² of 0.238 accounts for the number of predictors and suggests a similar degree of fit. The model’s F-statistic (F = 25.039, p < 0.001) confirms that the overall set of predictors significantly improves prediction accuracy over a baseline model. Moreover, the standard error of the estimate (16.065) provides a reasonable margin of prediction error, acceptable within behavioural modelling contexts. Importantly, no multicollinearity issues were detected, as all VIF values remain below 1.34, ensuring that the estimated effects are statistically reliable and not distorted by overlapping predictor influences. These diagnostics collectively affirm the model’s robustness and its suitability for interpreting the complex relationship of psychological, environmental, and demographic factors influencing cycling behaviour.

4.3. Estimated Model: Determinants of Perceived Walking Propensity in Windhoek

The MLR model in

Table 6. Determinants of walking behaviour: socioeconomic factors, urban form, neighbourhood characteristics, and individual attitudes in Windhoek.

Variables Group			Variables			B	Beta			t	p	VIF
			(Constant)			−13.097				−1.551	0.121
Urban form objective			Population density			15.043	0.234			6.945	<0.001	1.607
			Urban sprawl around home			0.142	0.188			5.892	<0.001	1.446
Urban form subjective			Lack of NMT infrastructure			−0.130	−0.096			−3.512	<0.001	1.056
			Neighbourhood entertainment preference			3.163	0.061			2.264	0.024	1.041
			Perceived security in the neighbourhood			0.101	0.101			3.687	<0.001	1.057
			Perceived safety when walking in the neighbourhood			0.085	0.075			2.724	0.007	1.081
			Sense of belonging to my neighbourhood			0.053	0.058			2.038	0.042	1.159
Socioeconomic			Education			−0.046	−0.063			−2.302	0.022	1.055
			Income			−0.067	−0.083			−2.757	0.006	1.297
			Car ownership			−2.331	−0.071			−2.554	0.011	1.105
			Number of adults in the household			0.460	0.022			0.784	0.433	1.067
Model Summary
R	R Square		Adjusted R Square		Std. Error of the Estimate
0.55	0.304		0.295		21.614
ANOVA
		Sum of Squares		df	Mean Square			F	Sig.
Regression		201,073.205		12	16,756.100			35.869	<0.001
Residual		461,076.795		987	467.150

presents the determinants of perceived walking propensity in the Windhoek neighbourhood. The model included 11 predictors, of which six were highly significant (p < 0.01): population density, urban sprawl around home, lack of NMT infrastructure, perceived neighbourhood security, perceived safety when walking, and income. Four variables were moderately significant (0.01 < p < 0.05): neighbourhood entertainment preference, sense of neighbourhood belonging, education, and car ownership.

The most associated predictor was population density, which increased walking propensity by approximately 23% for every standard deviation increase in density (β = 0.234, p < 0.001). This suggests that denser neighbourhoods in Windhoek are more conducive to walking, likely because of limited access to public transport and transport service affordability. Urban sprawl, which is typically associated with car dependency in the Global North context, was positively associated with walking propensity (18.8%, beta = 0.188, p < 0.001). This finding, while different from that in the Global South, reflects the spatial structure of an emerging African city, where the urban periphery tends to be more densely populated than the affluent, car-dominated urban core. In these peripheral settlements, limited access to motorised transport compels residents to walk out of necessity.

The regression model further revealed that several subjective social–environmental variables were significantly associated with the perceived walking propensity f. Perceived security had a strong and statistically significant effect (+10.1%, p < 0.001), indicating that for every percentage increase in how secure individuals feel from crime-related threats (e.g., theft, harassment), walking propensity increased by approximately 10%. Similarly, a percentage increase in perceived safety while walking, which reflects confidence in pedestrian infrastructure and protection from traffic risks, was associated with a 7.5% increase in perceived walking propensity (p = 0.007).

Beyond safety and security, the social and experiential qualities of neighbourhoods are also significantly correlated with walking propensity. A percentage increase in preference for neighbourhood entertainment (e.g., valuing local parks, cafés, or recreational spaces) was associated with a 6% increase in walking propensity (p = 0.024), whereas a sense of belonging to the neighbourhood was linked to a 5.8% increase in walking (p = 0.042). This indicates that individuals who feel connected to their community, neighbours, and local surroundings are more likely to walk regularly. A sense of belonging can foster social interactions, increase comfort in public spaces, and enhance perceptions of overall neighbourhood liveability, all of which contribute to higher walking levels.

However, several constraints were found to significantly reduce walking propensity. Lack of non-motorised transport (NMT) infrastructure was associated with a 9.6% decrease in walking propensity (p < 0.001), indicating that inadequate sidewalks, crossings, or pedestrian connectivity remain key barriers to walking in Windhoek. Among the socioeconomic variables, income had a significant negative association (–8.3%, p = 0.006), suggesting that higher-income individuals walk less, likely due to greater reliance on motorised modes. Similarly, higher levels of education and car ownership were associated with a 6% and 7% decrease, respectively (p = 0.022 and p = 0.011), further reinforcing the patterns seen in other urban settings where socioeconomic advantage correlates with reduced active travel.

The multiple linear regression model predicting perceived walking propensity in Windhoek demonstrates a strong and statistically significant model fit. With an R² of 0.304 and an adjusted R² of 0.295, the model explains nearly 30% of the variation in walking propensity, a substantial figure for behavioural studies in urban mobility. As noted by [57], R² values of 0.10 or higher are acceptable when the explanatory variables are statistically significant, particularly in studies of social phenomena. The model’s F-statistic (F = 35.869, p < 0.001) confirms its overall statistical significance, indicating that the included predictors significantly enhance explanatory power over a baseline model. The standard error of 21.614 reflects an acceptable degree of prediction error given the outcome’s complexity. Moreover, multicollinearity is not a concern, as all VIF values are comfortably low (below 1.6), suggesting that each variable contributes distinct information to the model. Together, these diagnostics affirm the model’s robustness in identifying key socio-spatial and attitudinal determinants of walking in Windhoek.

5. Discussion

This study explored how socioeconomic factors, urban form, neighbourhood characteristics, and individual attitudes are associated with commuting mode choice and neighbourhood-level active travel behaviour (cycling and walking) in Windhoek, Namibia. The findings across the three models revealed a complex relationship between structural conditions, personal circumstances, and environmental perceptions that collectively shape how residents move through their cities. In doing so, the study confirms several patterns observed in the global urban mobility literature while revealing unique characteristics tied to an emerging African city’s socio-spatial and infrastructural realities.

5.1. Commuting Mode Choice: Socioeconomic and Urban Form Determinants of Motorised vs. Non-Motorised Travel

Car ownership emerged as the strongest predictor of commuting modes in Windhoek, with car-owning individuals being over five times less likely to walk or cycle. This finding is consistent with [30] in Accra and [33] in Nigerian cities, where access to private vehicles significantly decreased the likelihood of engaging in active travel, even for short distances. These patterns suggest not only attitudinal preferences for motorised transport but also prevailing subjective norms that frame walking and cycling as indicators of socioeconomic disadvantage. In the context of Windhoek, such preferences are further reinforced by low perceived behavioural control, as infrastructural fragmentation and extended travel distances undermine the practicality of non-motorised commuting, regardless of individual intention.

Commuting time was another significant factor, with each additional minute of travel time reducing the likelihood of choosing non-motorised travel by nearly 5%. This echoes findings from Kenya [32], where commuters facing longer travel distances were significantly more inclined to use motorised transport due to the practical limitations of active travel. Unlike compact cities in Latin America (e.g., Bogotá), where land-use integration moderately supports walkability [36], Windhoek’s spatial structure imposes significant constraints on perceived and actual agency. These limitations, when viewed through the TPB lens, lower perceived behavioural control and thus suppress the translation of positive attitudes into action. Urban sprawl around the home is negatively associated with walking and cycling for commuting trips. While in high-income contexts such as Hamilton, Canada, sprawl typically corresponds with low-density, car-dependent neighbourhoods [45], in African cities like Lagos and Johannesburg [33,34], it often coincides with high-density informal settlements that remain spatially disconnected from formal transport infrastructure. This disconnection limits access to non-motorised modes despite the potential benefits of higher population density. In Windhoek, such environments function as both physical and sociocultural barriers, where inadequate infrastructure is compounded by social norms that stigmatise walking and cycling, particularly among more affluent groups. A notable pattern was observed with proximity to e-hailing stops, which exhibited a negative association with walking and cycling, indicating a potential substitution effect. This suggests a substitution effect like that observed in U.S. cities where ride-hailing has drawn users away from active and public transport. In African contexts, informal ride-hailing often functions as a critical stopgap, but without integration with non-motorised infrastructure, it risks further eroding the perceived utility and legitimacy of walking and cycling.

Driving licence possession further decreased the likelihood of active commuting. This aligns with findings in Nigeria [33], where driving credentials are associated not only with mobility privilege but also with perceived autonomy, making non-motorised options appear undesirable or unfeasible. Unlike in the Netherlands [27], where high cycling infrastructure offsets the impact of licence ownership, the lack of viable alternatives in Windhoek exacerbates the reliance on cars.

A set of moderately significant factors further reveals the structural and psychosocial dimensions of commuting mode choice. Gender disparities, as documented in Nairobi and South African cities, demonstrate how concerns around safety and entrenched social norms constrain women’s participation in active travel [31,34]. Driving licence possession, strongly associated with mobility privilege in Nigerian contexts [33], similarly reduced the uptake of non-motorised modes, particularly in environments where viable alternatives are limited. Income-related differences further accentuate this divide: in Accra and Kumasi, wealthier individuals exhibited greater reliance on private vehicles, while low-income populations walked primarily out of necessity [47,48]. The presence of e-hailing services, though addressing transport gaps, also appeared to substitute short-distance walking and cycling, consistent with trends observed in Nairobi. Intersection density showed a modest positive association with active travel, as found in Bogotá [58], yet its limited influence in Windhoek suggests that mere connectivity is insufficient without accompanying pedestrian infrastructure and safety measures.

Overall, these patterns demonstrate that attitudes, norms, and control beliefs, as conceptualised in the TPB, are deeply embedded in Windhoek’s socio-spatial structure. Unlike in high-income contexts, active travel here is shaped less by preference and more by constrained agency, infrastructural exclusion, and urban form.

5.2. Neighbourhood Cycling Behaviour: Perceptions, Competence, and Social Norms

The analysis of perceived cycling propensity reveals that psychosocial and structural constraints are strongly associated with the likelihood of engaging in cycling within the neighbourhood. The most significant barrier identified was a lack of cycling competence, echoing findings across African cities, including Accra and Nairobi [30,31], where limited exposure to cycling, poor training opportunities, and fear of traffic discourage uptake. This reflects a low sense of control over cycling as a viable mode of travel. Unlike in contexts such as the Netherlands or Copenhagen, where early socialisation and embedded infrastructure cultivate near-universal cycling competence [27], Windhoek exhibits a structural absence of normative and institutional reinforcement, making cycling behaviourally and symbolically inaccessible to many. Perceived neighbourhood security also emerged as a strong positive predictor of cycling propensity, particularly for women. Studies from Nairobi and Johannesburg show that safety concerns, especially related to crime and harassment, constrain women’s mobility and reduce their confidence in choosing active modes [31,34]. These safety concerns constitute a critical dimension of perceived behavioural control and intersect with social norms that limit women’s physical autonomy. In contrast, cities like Berlin or Warsaw benefit from integrated safety planning, well-lit paths, active street fronts, and routine policing that enhance perceived security and normalise cycling [25,40].

Possession of a driver’s licence was negatively associated with cycling, echoing findings from Lagos [44], where driving credentials are closely linked to elevated social status and access to motorised mobility. Within the framework of the theory of planned behaviour, this pattern reflects attitudinal inclinations favouring car use and entrenched normative perceptions that frame cycling as a mode of last resort. By contrast, in the Netherlands, car ownership does not significantly diminish cycling uptake, largely due to robust cycling infrastructure and deeply embedded cultural norms that support multimodal mobility practices [41]. The availability of shopping facilities and neighbourhood entertainment was significantly positively correlated with cycling, aligning with evidence from Bogotá and Lima, where land-use mix supports utility cycling [38]. This highlights how perceptions of convenience and efficiency shape positive attitudes towards cycling, particularly when trip purposes are varied and destinations are proximate. While Windhoek lacks the formal compactness of these cities, targeted improvements in mixed-use planning could help foster attitudes conducive to active travel. By enhancing local access to essential services over short distances, urban planning can simultaneously promote utility-based cycling and reduce dependence on private cars for everyday trips.

The moderate to marginal association of factors such as urban sprawl, educational attainment, and infrastructure provision further suggests that perceived cycling propensity in Windhoek is shaped as much by socio-spatial narratives as by physical conditions. While sprawl constrains infrastructure delivery, echoing challenges in Johannesburg and Lagos, it also erodes perceived behavioural control by fragmenting safe, connected cycling routes. The negative association with higher education levels reinforces the attitudinal and normative dimensions of the TPB, as cycling continues to be framed as a mode of necessity, deterring uptake among more affluent or educated populations, as seen in cities like Accra and Kumasi.

5.3. Neighbourhood Walking Behaviour: Density, Perceptions, and Spatial Necessity

Walking propensity in Windhoek is driven primarily by structural conditions and perceived environmental constraints rather than voluntary mode preference. High population density was strongly associated with increased walking; however, unlike high-income contexts, where density coincides with service proximity and multimodal integration [41], Windhoek’s dense areas are often informal, peripheral, and underserved. Similarly, urban sprawl was positively associated with walking; this finding is seemingly counterintuitive yet consistent with findings from African cities such as Lusaka, Lagos, and Nairobi, where walking is driven by economic compulsion amid infrastructure deficits [31,59]. Within the TPB framework, these trends suggest that while individual attitudes toward walking may be neutral or positive, the decision to walk is primarily shaped by constrained behavioural control. Perceived pedestrian safety and neighbourhood security emerged as distinct yet equally critical determinants of walking in Windhoek. In this context, the study highlights that safety concerns stem primarily from traffic-related risks, while security pertains to fears of crime and social vulnerability, especially pronounced among women and youth in peripheral areas. Both factors significantly shape perceived behavioural control, a core component of the theory of planned behaviour. Similar dynamics have been observed in South African cities, where the convergence of infrastructural deficits and social insecurity deters walking even where it remains the most viable transport option [35]. Conversely, the lack of non-motorised transport (NMT) infrastructure was negatively associated with walking propensity. Despite walking being prevalent in Windhoek, the absence of sidewalks, crossings, lighting, and pedestrian continuity undermines both the actual and perceived feasibility of walking. These infrastructural shortcomings highlight exclusionary mobility patterns and reduce individual agency, limiting the translation of intention into action. This observation corroborates evidence from Latin American and SSA cities, where fragmented or poorly maintained infrastructure diminishes the utility and desirability of walking [38]). Income and education were both negatively associated with walking, consistent with research from Accra and Kumasi, where higher socioeconomic groups opt for private transport to avoid the stigma and discomfort of walking [43,48]. Here, subjective norms work against active travel, reinforcing class-based mobility patterns that undermine modal equity.

Overall, walking in Windhoek reflects a complex relationship between spatial compulsion and perceptual barriers. TPB remains a useful interpretive framework, yet its behavioural assumptions must be grounded in the lived realities of exclusionary urbanism. In contexts such as Windhoek, perceived behavioural control is not a function of personal efficacy alone but of institutional failure, infrastructural neglect, and social risk.

5.4. Study Limitations

While this study provides important insights into the correlations of commuting mode choice and active travel behaviour in Windhoek, several limitations should be acknowledged. First, the analysis was based on cross-sectional survey data, which limits the ability to draw causal inferences. While statistical models employing binary logistics and multiple linear regression are appropriate for identifying robust associations, future longitudinal or panel data could provide deeper insights into how commuting and active travel behaviours evolve. Second, while the spatial indicators used (e.g., population density, intersection density, sprawl) are grounded in the established urban form literature, they may not fully capture the qualitative nuances of mobility constraints experienced by residents in informal settlements, especially in contexts with uneven service provision and infrastructure quality. Nonetheless, integrating high-resolution spatial data with survey-based perceptions represents a methodological advancement in Sub-Saharan mobility research. Finally, although comparisons are drawn with high-income cities to contextualise the findings, the unique socio-spatial and cultural context of Windhoek may limit the generalisability of results across regions. Rather than being a weakness, this highlights the value of context-specific empirical research in underrepresented urban settings and offers a solid foundation for comparative urban studies and policy translation.

6. Conclusions

This study examined how urban form, neighbourhood characteristics, and socioeconomic factors are associated with commuting mode choice and neighbourhood-level active travel (walking and cycling) in Windhoek, Namibia. By analysing commuting decisions for work and study trips alongside everyday walking and cycling behaviour within neighbourhoods, this study provides a multidimensional understanding of mobility in the rapidly urbanising African context. Commuting behaviour in Windhoek is shaped by a combination of socioeconomic status and spatial form, with car ownership, income level, driving licence possession, commuting time, and urban sprawl significantly associated with the use of motorised modes over non-motorised modes. These patterns align with trends observed in other African and South Asian cities such as Lagos, Accra, and Colombo, where vehicle ownership serves not only functional needs but also conveys social status and security within fragmented and unreliable transport systems. However, in Windhoek, the relationship between mobility and urban form is further complicated by low institutional capacity, peripheral densification without service integration, and limited modal alternatives. These conditions create a mobility environment in which active travel for commuting is largely constrained to those without access to private vehicles, not as a matter of choice but of necessity. The findings suggest that structural inequalities, rather than behavioural attitudes alone, continue to dictate mobility outcomes across diverse Global South contexts, with Windhoek offering a unique case where spatial marginality and transport poverty converge in shaping commuting decisions.

In contrast, the analysis of walking and cycling behaviour focused on neighbourhood-level movements rather than commuting. Walking, though prevalent, is largely driven by necessity rather than deliberate urban planning. Its positive association with both density and sprawl reflects settlement patterns in peripheral areas marked by infrastructural neglect and limited transport alternatives, rather than the presence of pedestrian-oriented environments.

Cycling, by contrast, remains a marginal mode of transport. Although its propensity increases in neighbourhoods with access to shops, recreational amenities, and higher levels of perceived security, overall uptake is constrained by inadequate infrastructure, limited cycling competence, and enduring social stigma. These constraints echo patterns observed in cities such as Accra, Lagos, and Nairobi, where cycling is often perceived less as a viable mode of transport and more as a sign of socioeconomic disadvantage. In Windhoek, perceived behavioural control is undermined by concerns over traffic safety and personal security, particularly among women and youth, despite generally positive attitudes toward active mobility.

These findings illustrate a marked divergence between walking and cycling; while walking is sustained through structural exclusion, cycling is constrained by both symbolic and practical barriers. This highlights the need to move beyond infrastructure-focused determinants of active travel. In Windhoek and in similar cities across the Global South, mobility behaviour is co-produced through the interrelations of spatial form, social stratification, and psychosocial perception. The application of the theory of planned behaviour reveals that attitudes, subjective norms, and perceived behavioural control are not fixed determinants but are shaped by historically embedded inequalities and structurally uneven urban environments.

Together, these results illustrate that while walking and cycling are prevalent in Windhoek, they are often undertaken under conditions of constraint rather than choice. Unlike cities with supportive infrastructure and active travel cultures, active mobility in Windhoek is deeply shaped by transport poverty, infrastructural gaps, and spatial exclusion. This study highlights the need to differentiate between commuting and neighbourhood travel behaviour when designing policy responses, as each is influenced by distinct but overlapping factors.

This study recommends a multi-pronged, equity-focused approach to transport planning in Windhoek that prioritises safe, inclusive, and accessible infrastructure for walking and cycling, particularly in high-density, low-income neighbourhoods where active travel is already a necessity. Investments in sidewalks, protected bike lanes, and well-lit pedestrian environments should be complemented by better integration with public and informal transport systems to support multimodal travel. Policies must also address behavioural and cultural barriers, including cycling competence and gendered safety concerns, through education, community programs, and gender-sensitive urban design. Finally, urban development should promote compact, mixed-use neighbourhoods to reduce travel distances and support everyday active travel, reinforcing walking and cycling as viable, dignified modes of transport.

Future research should expand on this study’s cross-sectional insights by adopting longitudinal and mixed-method approaches to explore how travel behaviours change over time and in response to interventions. Comparative studies between medium-sized African cities would help to determine whether the patterns observed in Windhoek reflect broader regional trends or local specificities. Additionally, qualitative studies focusing on gender, culture, and safety perceptions would enrich the understanding of behavioural barriers to active travel. Finally, experimental or pilot studies that evaluate the effects of specific infrastructure upgrades, such as new sidewalks, protected bike lanes, or lighting improvements, could offer valuable, context-specific evidence for policymakers and urban designers.