Dead Ends, Lost Time: Rural Mobility Constraints

Dead Ends, Lost Time: Rural Mobility Constraints | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Dead Ends, Lost Time: Rural Mobility Constraints Jiahang Liu, Pengjun Zhao, Zhengying Liu, Shixiong Jiang, Hao Wang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6871842/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Despite substantial global investments in rural road construction, a critical infrastructural deficiency—the prevalence of dead-end roads—continues to undermine mobility, exacerbates social inequities, and impedes sustainable development in rural communities worldwide. This study establishes the causal relationship between dead-end road density and travel outcomes, revealing how network topology fundamentally shapes rural accessibility patterns beyond simple distance metrics. Applying Double/debiased Machine Learning to data from 5,415 rural residents across China, we demonstrate that each unit increase in dead-end road density significantly extends travel time for discretionary activities by 26–40 minutes—effectively doubling average journey durations—while non-discretionary trips remain largely unaffected. Our causal analysis further reveals striking demographic heterogeneity, with elderly, male, higher-income, and self-employed individuals experiencing disproportionately greater mobility burdens. These findings challenge conventional rural development approaches that prioritise road quantity over network quality and connectivity. By identifying how specific network deficiencies systematically disadvantage certain demographic groups and activity types, we provide critical insights for policymakers seeking to enhance rural accessibility, promote inclusive development, and reduce infrastructure-induced inequalities. Strategic interventions targeting dead-end road reduction represent a high-leverage opportunity to transform rural infrastructure from a constraint into a catalyst for sustainable development. Earth and environmental sciences/Environmental social sciences/Socioeconomic scenarios Scientific community and society/Social sciences Figures Figure 1 Figure 2 Figure 3 Figure 4 Main While global attention celebrates the expansion of rural road coverage 1 , 2 , approximately 3 billion people worldwide reside in rural areas 3 , facing a profound yet largely invisible infrastructure challenge: not the absence of roads, but their inadequate connectivity. In rural China, residents often travel twice the theoretical distance to reach essential services—not due to geographic remoteness, but because of fragmented network topology characterised by dead-end roads. This fundamental qualitative deficiency in infrastructure design remains masked by conventional development metrics that emphasise road kilometres constructed rather than network functionality, creating a significant barrier to achieving multiple UN Sustainable Development Goals. Despite remarkable global investments in rural road construction, the prevalence of dead-end roads—pathways with limited connection to broader networks 4 —generates a paradoxical development trajectory where infrastructure expansion coexists with persistent accessibility constraints 5 . These structural inefficiencies force significant detours that disproportionately burden vulnerable populations, manifesting what we term infrastructure-induced inequality 7 , 8 . Unlike urban environments where road networks typically form interconnected grids 9 , rural transport systems often resemble fractured webs with critical missing links that dramatically inflate actual versus theoretical travel times 6 . The consequences extend beyond increased physical mobility, systematically constraining economic opportunities, limiting social inclusion, and restricting access to essential services 10 —transforming a seemingly technical infrastructure characteristic into a profound determinant of rural development outcomes. Our research presents groundbreaking causal evidence of this phenomenon, employing Double/debiased Machine Learning (DML) methodology to quantify precisely how dead-end road density affects travel patterns across different trip purposes and demographic groups. Analysing data from 5,415 rural residents across three geographically and economically diverse Chinese provinces, we demonstrate that each unit increase in dead-end road density dramatically increases travel time for discretionary activities by 26–40 minutes—effectively doubling the average journey time. These effects vary significantly by demographic characteristics, with elderly residents, higher-income individuals, and the self-employed bearing disproportionately heavier mobility burdens. This work advances beyond correlation-based studies by establishing a robust causal relationship between network structure and mobility outcomes. While previous research has documented rural-urban mobility disparities 11 , our findings illuminate how the specific topological features of transport networks—not merely distance—fundamentally shape accessibility patterns. This distinction matters tremendously for policy, suggesting that strategic interventions targeting network connectivity could yield greater benefits than simply adding more road kilometres. The distinction between non-discretionary trips (relatively unaffected) and discretionary journeys (severely impacted) further reveals how infrastructure deficiencies subtly reshape social participation and opportunity landscapes. As urbanisation accelerates globally, the risk of rural areas becoming mobility deserts intensifies, potentially deepening rural-urban divides and undermining sustainable development efforts. While urban planners increasingly embrace concepts like 15-minute cities, rural residents often navigate 60-minute villages—not because destinations are inherently distant, but because fragmented infrastructure imposes invisible detours and barriers. Our findings challenge policymakers to reimagine rural infrastructure development beyond conventional metrics, focusing on network functionality rather than mere coverage statistics. By addressing the dead-end road crisis, we can transform rural mobility into a catalyst rather than a constraint for sustainable and equitable rural development. Results Travel time characteristics of rural residents In this study, 5,415 rural residents were surveyed, with participants typically spending between 14.0 and 18.5 mins on five basic trips. Over half of the participants reported spending more than 10 mins on these trips (see Fig. 1 and Supplementary Table 3). Among these, subsistence trips took the most time, with a mean (SD/median) of 18.5 (21.1/12.0) mins, ranging from 1.0 to 300.0 mins. Shopping trips to stores took a mean (SD/median) of 18.1 (16.9/15.0) mins (range: 1.0–180.0 mins), while trips to rural periodic fairs took a mean (SD/median) of 15.0 (13.7/10.0) mins (range: 1.0–150.0 mins). Family visit trips required a mean (SD/median) of 21.8 (25.5/15.0) mins (range: 0.5–300.0 mins). Entertainment trips had the smallest mean (SD/median) of 14.0 (13.2/10.0) mins (range: 1.0–180.0 mins). Causal effects of dead-end road density on travel time Our analysis indicates that dead-end road density has a disproportionate effect on certain types of trips, particularly those involving shopping and entertainment, where time costs are more sensitive to detours. This suggests that rural road infrastructure, particularly the density of dead-end roads, plays a critical role in shaping activity-travel behaviour in rural areas. We found that increased dead-end road density significantly increased travel time for shopping and entertainment trips, but did not impact subsistence or family visit trips after adjusting for control variables (see Fig. 2 and Supplementary Table 5). Specifically, each additional unit of dead-end road density was associated with an increase of 21.5 mins (95% CI: 15.2–27.7) for shopping trips to stores, 19.0 mins (95% CI: 10.6–27.3) for shopping trips to fairs, and 15.2 mins (95% CI: 8.2–22.1) for entertainment trips. However, no significant effects were found for subsistence trips ( p = 0.96) and family visit trips ( p = 0.68). These findings remained consistent across all sensitivity analyses, reinforcing the robustness of our results. The relationships between dead-end road density and travel time remained unchanged regardless of variations in hyperparameters, sample deletions, outlier adjustments, and additions of township fixed effects. To address potential endogeneity, we used lagged dead-end road density as an instrumental variable in our analysis (see Fig. 3 and Supplementary Table 6). The results showed that the effects of dead-end road density on travel time remained significant. Specifically, each additional unit of dead-end road density resulted in an increase of 25.9 mins (95% CI: 0.30–51.4) for shopping trips to stores, 31.2 mins (95% CI: 18.8–41.6) for shopping trips to rural fairs, and 39.5 mins (95% CI: 23.8–55.1) for entertainment trips. Heterogeneity analysis The heterogeneity analysis highlights that the effects of road density are not uniform across all demographic groups. Population groups, such as the elderly, male, the high-income, and self-employed villagers, are particularly affected by the inefficiencies introduced by dead-end roads. Firstly, gender and age played crucial roles in how dead-end roads impacted travel time (see Fig. 4 a, 4 b and Supplementary Table 7, 8). Specifically, male villagers and elderly participants were more affected by the increased travel time associated with dead-end road density. Male villagers, who typically play a larger role in household subsistence activities, experienced longer travel times for subsistence trips. Similarly, elderly villagers, who tend to have lower mobility and fewer travel options, were significantly more affected by the detours caused by dead-end roads. The elderly group, in particular, experienced an average increase of 21.3 minutes in their subsistence trips (95% CI: 7.9–34.6), a considerable impact on their daily routines. In terms of household income, the analysis revealed that higher-income villagers were more affected by dead-end road density, particularly in the context of shopping trips (see Fig. 4 c and Supplementary Table 9). Villagers with upper-middle income showed a significant increase of 16.6 minutes in subsistence trips (95% CI: 4.6–28.5), reflecting a higher reliance on more distant, often urban, destinations for purchasing goods. These findings suggest that wealthier individuals tend to engage in shopping trips that involve traveling longer distances, and dead-end roads increase their travel burden. Surprisingly, high-income villagers demonstrated a decrease in time spent on family visit trips, with a reduction of 32.7 minutes (95% CI: -60.3 to -5.0). This reduction may reflect a shift in behaviour, with wealthier individuals opting for more convenient, less time-consuming alternatives to traditional family visits, such as meeting at third-party venues. Employment type was another key factor influencing travel time (see Fig. 4 d and Supplementary Table 10). Self-employed villagers, who typically have more flexible schedules and travel to a wider range of destinations, were the most affected by dead-end roads, particularly for entertainment trips. These individuals experienced a significant increase in travel time of 35.8 minutes (95% CI: 22.1–49.4). In contrast, farmers, whose daily activities are concentrated within a small geographic area, experienced less of an impact from dead-end roads, with only a modest increase of 5.4 minutes (95% CI: -0.1 to 10.9). This suggests that the effect of road infrastructure on travel behaviour is more pronounced for those with broader mobility patterns. Full-time workers also showed an increase in time spent on family visit trips (24.2 minutes, 95% CI: -2.0 to 50.5), but no significant effect on entertainment trips. Their fixed working hours may take away relatively limited leisure time, so they are less sensitive to dead-end roads for entertainment trips. Discussion This study provides a comprehensive analysis of the impact of dead-end road density on rural residents’ activity-travel behaviour, drawing from a large sample of 5,415 participants across three major geographical regions in China. The findings reveal significant variations in how dead-end road density affects different types of activities, with notable consequences for mobility, social inclusion, and rural development. Our results highlight that dead-end road density increases travel time primarily for shopping and entertainment trips, while subsistence and family visit trips remain largely unaffected. This distinction is important because it underscores the nature of discretionary versus non-discretionary travel. Non-discretionary trips, such as subsistence and family visits, are generally more structured, often involving fixed destinations and driven by social or cultural obligations 12 . In contrast, shopping and entertainment trips are discretionary, often involving more flexible destinations that require more complex and circuitous routes, especially when road networks are fragmented 13 . This finding aligns with previous studies showing that built environment factors have a greater impact on non-work trips compared to work-related travel, where destinations are more predictable and fixed 14 . Our study contributes to this literature by providing new insights into how infrastructure challenges disproportionately affect discretionary travel, which is often associated with higher mobility demands and broader social participation. Moreover, our use of causal inference techniques, particularly DML, provides robust evidence of the causal relationship between dead-end road density and travel time, addressing the limitations of previous studies that have relied primarily on correlation-based methods 15 . By applying this method to a partially linear instrumental variables model, we ensure that the estimates are unbiased, thus strengthening the validity of our findings. The estimated increases in travel time—ranging from 26 to 40 mins for shopping and entertainment trips—are more than doubling the travel time mean of 14–18 mins. These effects are substantial, especially considering that many urban areas operate within a 15-minute community for daily needs 16 , 17 . This finding challenges the assumption that rural mobility is solely constrained by long geographical distances to essential services, highlighting the additional burdens created by poor road infrastructure. The heterogeneity analysis reveals that the effects of dead-end road density vary significantly across different demographic groups. For instance, male and older villagers experience greater increases in time of subsistence trips, with older participants particularly vulnerable due to their reduced mobility and limited travel options 18 , 19 . These findings suggest that dead-end roads exacerbate transport-related social exclusion for specific groups, particularly the elderly, who face greater challenges in maintaining their social and economic activities. Similarly, higher-income villagers, who often travel longer distances for higher-quality goods, are more affected by dead-end road density, particularly for shopping trips. Interestingly, high-income individuals show a decrease in time of family visit trips, likely as a result of changing social dynamics where these individuals may prefer meeting family members at more accessible venues, thus reducing the traditional need for long visits. This shift in behaviour suggests that while road infrastructure influences mobility, it also interacts with socioeconomic factors to shape activity-travel patterns 20 . Furthermore, self-employed villagers—who tend to have more dispersed travel destinations—were found to be the most affected by dead-end road density, experiencing significantly longer travel time for entertainment trips. On the other hand, farmers, whose daily activities are confined to their immediate rural surroundings, were less impacted by dead-end roads, suggesting that road infrastructure challenges have less of an effect on their relatively localised mobility patterns 21 . These findings underscore the need for targeted interventions that consider the diverse travel needs of different groups, particularly vulnerable populations who are more likely to face increased travel burdens due to fragmented road networks. Our study makes a significant contribution to the understanding of rural transport disadvantages, offering valuable insights for planners and policymakers. By addressing the inefficiencies created by dead-end roads and improving connectivity, we can reduce the time costs associated with travel, which in turn can enhance the quality of life for rural residents and reduce social exclusion. In particular, policies aimed at improving access to essential services, such as shopping centres and entertainment venues, should prioritise addressing the infrastructure gaps that disproportionately affect specific groups, such as the elderly and self-employed. Furthermore, the findings suggest that high-income villagers may benefit from improved access to higher-quality goods and services, highlighting the role of road infrastructure in shaping economic opportunities and social mobility, further promoting health and well-being 22 , 23 . There are some limitations that must be considered. First, the reliance on self-reported travel time data introduces the possibility of recall bias, which may affect the accuracy of the reported travel time. Additionally, seasonal variations may influence travel patterns, particularly for activities with cyclical destinations. Second, while OSM provides comprehensive road data, there may still be gaps in coverage, particularly in remote rural areas. Finally, our results are limited to rural residents in China aged 18 and older, and further research is needed to examine whether these findings can be generalised to other populations or countries. Nevertheless, our study provides strong evidence of the impact of dead-end road density on rural mobility and highlights the importance of addressing infrastructure deficiencies to promote more equitable and sustainable rural development. Methods Study design and population The primary data were derived from a large-scale survey conducted in 2024 targeting permanent residents living in villages across South China, North China, and Northwest China. Using a stratified multi-stage sampling design, we first selected three representative provinces—Guangdong, Henan, and Gansu—that exhibit substantial variation in spatial locations, geographical conditions, and socioeconomic indicators across the three regions. Within each province, rural townships (the most basic administrative units in China’s rural governance structure) were systematically sampled. The sampling procedure continued with the selection of 3–6 villages per township, stratified based on their distances from the township centre. Within each selected village, approximately 15 households were randomly sampled. Trained investigators conducted face-to-face interviews with participants, obtaining written informed consent from all respondents. This research protocol received ethical approval from the Peking University. Of the 5,980 rural residents who initially participated, 5,645 completed the questionnaire (94.4% response rate). Data cleaning procedures excluded respondents who were under 18 years of age, those with invalid responses, and participants whose survey completion time was less than 5 minutes (indicating potentially unreliable responses). After further excluding cases with missing data on travel patterns and key socioeconomic variables, the final analytical sample comprised 5,415 participants from 277 villages (see Supplementary Fig. 1 and Table 1). Key variables Participants reported travel time from home to destinations in minutes. We categorised all trips into four types based on purpose: subsistence, shopping, family visit, and entertainment. Subsistence trips encompass travel to work, agricultural activities, or school—activities essential for sustaining livelihood. Shopping trips involve travel to purchase necessary goods or services, with two primary destinations identified: physical stores and rural periodic fairs. Family visit trips constitute travel to visit relatives, while entertainment trips represent discretionary travel for social or leisure purposes. Dead-end roads were defined as roads with only one or no point of access to other roads. Rural road data was extracted from the OpenStreetMap (OSM) dataset (2024) via the Geofabrik Download Server ( https://download.geofabrik.de/ ). This dataset, despite limitations, represents the most comprehensive open-source rural road database available, with accuracy substantially improved in recent years 24 , 25 . Using the township as our basic analytical unit based on villagers’ typical activity ranges, we extracted road networks within township administrative boundaries and planarised all roads at intersection points. We identified dangle points (points without connection to other roads) using the Feature Vertices to Points tool. Dead-end roads were characterised as roads with one or two dangle points, with length measured as the distance from the dangle point to either an intersection point or another dangle point. Our primary metric, dead-end road density, was calculated by dividing the total length of dead-end roads by township area. All spatial analyses were performed using ArcGIS 10.8. Control variables We controlled for two groups of variables: socioeconomic characteristics and township-level variables. Socioeconomic characteristics included age, gender, education years, employment type, self-reported health, household size, structure, income, and vehicle ownership (see Supplementary Table 2). Self-reported health was measured on a four-point scale (1 “strongly disagree” to 4 “strongly agree” in response to “I think my health is poor”), with scores reversed so higher values indicated better health. Household size represented the number of cohabiting members, household structure indicated the presence of children and elderly, and vehicle ownership encompassed cars, bicycles, motorcycles, and tricycles. Township-level variables comprised built and natural environmental factors. Built environmental variables included population density (residents per township area from 2023 LandScan dataset at 1-km resolution), road density (road length per township area from OSM data), proximity measures (shortest distance to city/county centres and highways/local roads calculated using “geodist” package in Stata 18.0), and infrastructure accessibility (number of healthcare, educational, and shopping facilities, plus bus stations from Amap 2024). Natural environmental variables encompassed slope (from Copernicus DEM dataset at 30m × 30m resolution), temperature, and precipitation (both at 1km × 1km resolution from the National Tibetan Plateau Data Centre). All geographic data were matched to township-level administrative boundaries and processed using raster calculation tools to derive township-level metrics (see Supplementary Table 4). Statistical analysis With the increasing focus on causal inference in social sciences and economics, machine learning (ML) has emerged as a powerful approach for identifying causal effects. In this study, we employed Double/debiased Machine Learning (DML) to examine the causal relationship between dead-end roads and villagers’ travel time. DML relaxes assumptions about intrinsic data correlations, allowing for more flexible functional forms and improved stability. This approach combines the advantages of machine learning in processing high-dimensional data and model recognition while maintaining the rigor of econometric models in causal inference 26 . We began by constructing a partial linear model: $$\:\begin{array}{c}Y={\theta\:}_{0}D+{f}_{0}\left(\varvec{X}\right)+U,\:\:E\left[U|\varvec{X},\:D\right]=0\#\left(1\right)\end{array}$$ $$\:\begin{array}{c}D={m}_{0}\left(\varvec{X}\right)+V,\:\:E\left[V|\varvec{X}\right]=0\#\left(2\right)\end{array}$$ where Y is the travel time, D is the number or length of dead-end roads, X is a vector consists of control variables, U and V are the stochastic errors, \(\:{f}_{0}\left(\varvec{X}\right)\) and \(\:{m}_{0}\left(\varvec{X}\right)\) are unknown nonparametric functions. We are only interested in the parameter \(\:{\theta\:}_{0}\) , which represented the average treatment effect of D on Y . All DML analyses were performed using the “ddml” package in Stata 18.0 26 . To ensure model reliability, we followed Chernozhukov et al. 27 recommendations by implementing a fivefold cross-fitting procedure. Since DML randomly splits samples during cross-fitting, we fitted the same model with five different random splits and reported the median estimator to account for randomness effects. Given the flexibility in ML method selection within the DML framework, we employed a stacking approach to reduce misspecification risk. This method constructs combinations of diverse base machine learners, assigning nonzero weights based on constrained least squares, with the final result being a weighted average of predictions 28 . We selected six mainstream base learners: linear regression (LR), lasso with cross-validated penalty (LASSO), support vector machines (SVM), feed-forward neural nets (FNN), random forests (RF), and gradient boosting trees (GBT). To address the nested data structure (villagers within townships) potentially violating DML’s independently and identically distributed samples assumption, we used cluster robust standard errors to prevent statistical significance overstatement 29 . Additionally, we achieved independent sample-splitting by randomly assigning folds by cluster, ensuring hold-out samples contained no observations from clusters used in training. We conducted multiple sensitivity analyses to validate benchmark results, including: setting cross-fitting folds to 10; removing subdistrict ( jiedao ) samples which might exhibit different travel behaviours than rural townships; winsorising continuous variables at the 1st and 99th percentiles to mitigate outlier effects; and including township dummy variables to control for unobserved characteristics across townships. To identify causal effects, we employed an instrumental variable approach within DML to address potential endogeneity. Following previous literature 30 , 31 , we used historical road networks as instruments since current roads build upon old roads while previous roads remain unassociated with current individual activities. We adopted 2021 dead-end road density as an instrument, avoiding premature OSM data quality issues. Weak instrument tests confirmed its validity, with first-stage F-statistics exceeding the recommended value of 10 in all significant models 32 . For heterogeneity analysis, we estimated dead-end road density impacts across demographic groups including sex, age, household income, and employment types, implementing 10-fold cross-fitting to ensure reliability given smaller subgroup sample sizes 26 . Further analyses examined other dead-end road indicators (number, length, various ratios), revealing negligible effects on travel time and highlighting the importance of dead-end road density specifically (Supplementary Table 11). Declarations Author Contribution Conceptualization: Q.L.; data collection: P.Z., Z.L., S.J., H.W., H.Z., Z.L., Y.T., Q.L.; analytical tools: J.L. and Q.L.; results analysis: J.L.; manuscript draft and reviewing: J.L. and Q.L.; funding acquisition: P.Z. and Q.L.; all authors read and approved the manuscript. Data Availability The raw CoVIDA and HBS data are protected and are not available due to data privacy laws. References Alessandretti, L., Aslak, U. & Lehmann, S. The scales of human mobility. Nature 587 , 402-407 (2020). Laurance, W. F. et al. A global strategy for road building. Nature 513 , 229-232 (2014). United Nations Department of Economic and Social Affairs. World Social Report 2021: Reconsidering Rural Development (UN DESA, 2021). Schlossberg, M., Greene, J., Phillips, P. P., Johnson, B. & Parker, B. School Trips: Effects of Urban Form and Distance on Travel Mode. Journal of the American Planning Association 72 , 337-346 (2006). Hanig, L. et al. Finding gaps in the national electric vehicle charging station coverage of the United States. Nature Communications 16 , 561 (2025). Yuan, D., Zhao, P., Yu, Z. & Liu, Q. Villagers' travel burden and the built environment in rural China: Evidence from a national level survey. Journal of Transport Geography 113 , 103717 (2023). Mackett, R. L. The health implications of inequalities in travel. Journal of Transport & Health 1 , 202-209 (2014). Weiss, D. J. et al. Global maps of travel time to healthcare facilities. Nature Medicine 26 , 1835-1838 (2020). Leite, D. & De Bacco, C. Similarity and economy of scale in urban transportation networks and optimal transport-based infrastructures. Nature Communications 15 , 7981 (2024). Asher, S. & Novosad, P. Rural Roads and Local Economic Development. American Economic Review 110 , 797-823 (2020). Schläpfer, M. et al. The universal visitation law of human mobility. Nature 593 , 522-527 (2021). Chen, C. & Mokhtarian, P. L. Tradeoffs between Time Allocations to Maintenance Activities/Travel and Discretionary Activities/Travel. Transportation 33 , 223-240 (2006). Dharmowijoyo, D. B. E., Susilo, Y. O. & Karlström, A. On complexity and variability of individuals’ discretionary activities. Transportation 45 , 177-204 (2018). Lee, M., Barbosa, H., Youn, H., Holme, P. & Ghoshal, G. Morphology of travel routes and the organization of cities. Nature Communications 8 , 2229 (2017). Jiang, N. et al. Traffic related activity pattern of Chinese adults: a nation-wide population based survey. Journal of Exposure Science & Environmental Epidemiology 33 , 482-489 (2023). Abbiasov, T. et al. The 15-minute city quantified using human mobility data. Nature Human Behaviour 8 , 445-455 (2024). Bruno, M., Monteiro Melo, H. P., Campanelli, B. & Loreto, V. A universal framework for inclusive 15-minute cities. Nature Cities 1 , 633-641 (2024). Ford, J. A., Wong, G., Jones, A. P. & Steel, N. Access to primary care for socioeconomically disadvantaged older people in rural areas: a realist review. BMJ Open 6 , e010652 (2016). Zhao, P. & Yu, Z. Investigating mobility in rural areas of China: Features, equity, and factors. Transport Policy 94 , 66-77 (2020). Chen, Z. et al. Contrasting social and non-social sources of predictability in human mobility. Nature Communications 13 , 1922 (2022). González, M. C., Hidalgo, C. A. & Barabási, A.-L. Understanding individual human mobility patterns. Nature 453 , 779-782 (2008). Liu, Y. & Li, Y. Revitalize the world’s countryside. Nature 548 , 275-277 (2017). Pearsall, H. et al. Advancing equitable health and well-being across urban–rural sustainable infrastructure systems. npj Urban Sustainability 1 , 26 (2021). Herfort, B., Lautenbach, S., Porto de Albuquerque, J., Anderson, J. & Zipf, A. A spatio-temporal analysis investigating completeness and inequalities of global urban building data in OpenStreetMap. Nature Communications 14 , 3985 (2023). Koks, E. E. et al. A global multi-hazard risk analysis of road and railway infrastructure assets. Nature Communications 10 , 2677 (2019). Ahrens, A., Hansen, C. B., Schaffer, M. E. & Wiemann, T. ddml: Double/debiased machine learning in Stata. The Stata Journal 24 , 3-45 (2024). Chernozhukov, V. et al. Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal 21 , C1-C68 (2018). Ahrens, A., Hansen, C. B. & Schaffer, M. E. pystacked: Stacking generalization and machine learning in Stata. The Stata Journal 23 , 909-931 (2023). Abadie, A., Athey, S., Imbens, G. W. & Wooldridge, J. M. When Should You Adjust Standard Errors for Clustering? The Quarterly Journal of Economics 138 , 1-35 (2023). Baum-Snow, N. Did Highways Cause Suburbanization? The Quarterly Journal of Economics 122 , 775-805 (2007). Duranton, G. & Turner, M. A. The Fundamental Law of Road Congestion: Evidence from US Cities. American Economic Review 101 , 2616-2652 (2011). Alejo, J., Galvao, A. F. & Montes-Rojas, G. A first-stage representation for instrumental variables quantile regression. The Econometrics Journal 26 , 350-377 (2023). Additional Declarations No competing interests reported. Supplementary Files supplementaryinformation.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6871842","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":513334364,"identity":"587bac24-308a-46ec-b297-30a581f7e6d2","order_by":0,"name":"Jiahang Liu","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Jiahang","middleName":"","lastName":"Liu","suffix":""},{"id":513334366,"identity":"20dad22a-320f-4664-bab7-ea337f3e1d8a","order_by":1,"name":"Pengjun Zhao","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Pengjun","middleName":"","lastName":"Zhao","suffix":""},{"id":513334369,"identity":"add6a3b3-be74-445b-b6e4-04b898c2e444","order_by":2,"name":"Zhengying Liu","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Zhengying","middleName":"","lastName":"Liu","suffix":""},{"id":513334370,"identity":"461d4277-f0f7-4580-948f-6b8d346c8ba0","order_by":3,"name":"Shixiong Jiang","email":"","orcid":"","institution":"Shenzhen Technology University","correspondingAuthor":false,"prefix":"","firstName":"Shixiong","middleName":"","lastName":"Jiang","suffix":""},{"id":513334373,"identity":"921556c3-3f70-4763-a874-c745a2629599","order_by":4,"name":"Hao Wang","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Wang","suffix":""},{"id":513334375,"identity":"01fc00db-d242-4bf1-bc0f-f946fe221bbc","order_by":5,"name":"Hongjian Zhao","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Hongjian","middleName":"","lastName":"Zhao","suffix":""},{"id":513334377,"identity":"de8c05d9-dfe9-4192-8ccd-4323ef6ba46f","order_by":6,"name":"Zhaoxiang Li","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Zhaoxiang","middleName":"","lastName":"Li","suffix":""},{"id":513334378,"identity":"3dbd7546-43af-4f25-82e5-9ba2423665ee","order_by":7,"name":"Yushun Tang","email":"","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":false,"prefix":"","firstName":"Yushun","middleName":"","lastName":"Tang","suffix":""},{"id":513334379,"identity":"6a0dc8ce-4f16-4927-befc-e87e17ce7663","order_by":8,"name":"Qiyang Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAtUlEQVRIiWNgGAWjYHACxgMfGBJADAPi1PMA8cEZJGs5zEOSFnv+wxsO2+5IS2xgb94mwVBzhwhbJNIKDueeyUls4DlWJsFw7BkxWngMDue2VSQ2SOSYSTA2HCZCC/8Zg8OWIC3yb4jVwpBjcJixDegwCR5itdxIKzjYeybNuI0nrdgi4RgRWtj7D2988HNHsmw/++GNNz7UEKGFARQdjA0MDGwgZgJRGmBaRsEoGAWjYBTgBABggjlXVmktPAAAAABJRU5ErkJggg==","orcid":"","institution":"Peking University Shenzhen Graduate School","correspondingAuthor":true,"prefix":"","firstName":"Qiyang","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-06-11 12:23:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6871842/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6871842/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91415165,"identity":"3cd1399f-8533-4c64-8316-8ca42d9f1057","added_by":"auto","created_at":"2025-09-16 09:15:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":154670,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTravel time characteristics of study participants.\u003c/strong\u003e In the box plots, thick bars represent the median value, box shapes represent the range between the first and quartile third quartiles. The scatter and violin plots show the data distributions. N represents sample sizes.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/1b4a2a38960ca0d0790fee61.png"},{"id":91415162,"identity":"571d7385-680d-41dc-b5fd-d893f9e7e067","added_by":"auto","created_at":"2025-09-16 09:15:31","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":191268,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe effects of dead-end road density on travel time.\u003c/strong\u003e BA and SA represent benchmark and sensitivity analysis, respectively. In SA, we (1) change the hyperparameter of DML, (2) delete particular samples, (3) modify outliers and (4) add township fixed effects. Points represent the estimated parameters, and numbers indicate the significant average treatment effects. Error bars represent 95% confidence intervals. N represents the sample sizes. Asterisks indicate significance levels: \u003csup\u003e*\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.1, \u003csup\u003e**\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05, \u003csup\u003e***\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/58bd45037073a02ea6a9d628.png"},{"id":91415163,"identity":"c35116df-9f05-4c52-b1b6-4cd0f966124f","added_by":"auto","created_at":"2025-09-16 09:15:31","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":67122,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResults of instrumental variable analysis.\u003c/strong\u003e IV represents instrumental variable analysis. Points represent the estimated parameters, and numbers indicate the significant average treatment effects. Error bars represent 95% confidence intervals. N represents the sample sizes. Asterisks indicate significance levels: \u003csup\u003e*\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.1, \u003csup\u003e**\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05, \u003csup\u003e***\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/c90a38c9c0b0f69b2357c750.png"},{"id":91416366,"identity":"cc9dca43-ca78-4c93-a0e9-61d950999b98","added_by":"auto","created_at":"2025-09-16 09:31:32","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":730805,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResults of heterogeneity analysis.\u003c/strong\u003e We consider four types of heterogeneity, including \u003cstrong\u003e(a)\u003c/strong\u003e gender: males and females; \u003cstrong\u003e(b)\u003c/strong\u003e age: young adults (18 to 44 years), the middle-aged (45 to 59 years), and the elderly (60 years and over); \u003cstrong\u003e(c)\u003c/strong\u003e household income: low (less than 29999 CNY), low-middle (30000 to 59999 CNY), middle (60000 to 89999 CNY), and high income (more than 90000 CNY); \u003cstrong\u003e(d)\u003c/strong\u003e employment types: full-time workers, temporary workers, self-employed workers, farmers, and others (e.g., the unemployed, the retired, and students). Points represent the estimated parameters, and numbers indicate the significant average treatment effects. Error bars represent 95% confidence intervals. N represents the sample sizes. Asterisks indicate significance levels: \u003csup\u003e*\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.1, \u003csup\u003e**\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05, \u003csup\u003e***\u003c/sup\u003e \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/ba033e363ce5a1afa18dd116.jpeg"},{"id":93812030,"identity":"f61398ef-d2b6-4c00-8ad0-ec02b500a5d5","added_by":"auto","created_at":"2025-10-17 20:31:30","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1753331,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/bcb9ef01-fdf5-484c-9991-fa2c4b91fa59.pdf"},{"id":91415167,"identity":"bf76077e-b195-4e90-a985-8c6dc60985d5","added_by":"auto","created_at":"2025-09-16 09:15:32","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":353425,"visible":true,"origin":"","legend":"","description":"","filename":"supplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6871842/v1/8c8ea4e29e9d0107dc0350fa.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Dead Ends, Lost Time: Rural Mobility Constraints","fulltext":[{"header":"Main","content":"\u003cp\u003eWhile global attention celebrates the expansion of rural road coverage\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, approximately 3\u0026nbsp;billion people worldwide reside in rural areas\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, facing a profound yet largely invisible infrastructure challenge: not the absence of roads, but their inadequate connectivity. In rural China, residents often travel twice the theoretical distance to reach essential services\u0026mdash;not due to geographic remoteness, but because of fragmented network topology characterised by dead-end roads. This fundamental qualitative deficiency in infrastructure design remains masked by conventional development metrics that emphasise road kilometres constructed rather than network functionality, creating a significant barrier to achieving multiple UN Sustainable Development Goals.\u003c/p\u003e\u003cp\u003eDespite remarkable global investments in rural road construction, the prevalence of dead-end roads\u0026mdash;pathways with limited connection to broader networks\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e\u0026mdash;generates a paradoxical development trajectory where infrastructure expansion coexists with persistent accessibility constraints\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. These structural inefficiencies force significant detours that disproportionately burden vulnerable populations, manifesting what we term infrastructure-induced inequality\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Unlike urban environments where road networks typically form interconnected grids\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, rural transport systems often resemble fractured webs with critical missing links that dramatically inflate actual versus theoretical travel times\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. The consequences extend beyond increased physical mobility, systematically constraining economic opportunities, limiting social inclusion, and restricting access to essential services\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e\u0026mdash;transforming a seemingly technical infrastructure characteristic into a profound determinant of rural development outcomes.\u003c/p\u003e\u003cp\u003eOur research presents groundbreaking causal evidence of this phenomenon, employing Double/debiased Machine Learning (DML) methodology to quantify precisely how dead-end road density affects travel patterns across different trip purposes and demographic groups. Analysing data from 5,415 rural residents across three geographically and economically diverse Chinese provinces, we demonstrate that each unit increase in dead-end road density dramatically increases travel time for discretionary activities by 26\u0026ndash;40 minutes\u0026mdash;effectively doubling the average journey time. These effects vary significantly by demographic characteristics, with elderly residents, higher-income individuals, and the self-employed bearing disproportionately heavier mobility burdens.\u003c/p\u003e\u003cp\u003eThis work advances beyond correlation-based studies by establishing a robust causal relationship between network structure and mobility outcomes. While previous research has documented rural-urban mobility disparities\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, our findings illuminate how the specific topological features of transport networks\u0026mdash;not merely distance\u0026mdash;fundamentally shape accessibility patterns. This distinction matters tremendously for policy, suggesting that strategic interventions targeting network connectivity could yield greater benefits than simply adding more road kilometres. The distinction between non-discretionary trips (relatively unaffected) and discretionary journeys (severely impacted) further reveals how infrastructure deficiencies subtly reshape social participation and opportunity landscapes.\u003c/p\u003e\u003cp\u003eAs urbanisation accelerates globally, the risk of rural areas becoming \u003cem\u003emobility deserts\u003c/em\u003e intensifies, potentially deepening rural-urban divides and undermining sustainable development efforts. While urban planners increasingly embrace concepts like 15-minute cities, rural residents often navigate 60-minute villages\u0026mdash;not because destinations are inherently distant, but because fragmented infrastructure imposes invisible detours and barriers. Our findings challenge policymakers to reimagine rural infrastructure development beyond conventional metrics, focusing on network functionality rather than mere coverage statistics. By addressing the dead-end road crisis, we can transform rural mobility into a catalyst rather than a constraint for sustainable and equitable rural development.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eTravel time characteristics of rural residents\u003c/h2\u003e\u003cp\u003eIn this study, 5,415 rural residents were surveyed, with participants typically spending between 14.0 and 18.5 mins on five basic trips. Over half of the participants reported spending more than 10 mins on these trips (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Supplementary Table\u0026nbsp;3). Among these, subsistence trips took the most time, with a mean (SD/median) of 18.5 (21.1/12.0) mins, ranging from 1.0 to 300.0 mins. Shopping trips to stores took a mean (SD/median) of 18.1 (16.9/15.0) mins (range: 1.0\u0026ndash;180.0 mins), while trips to rural periodic fairs took a mean (SD/median) of 15.0 (13.7/10.0) mins (range: 1.0\u0026ndash;150.0 mins). Family visit trips required a mean (SD/median) of 21.8 (25.5/15.0) mins (range: 0.5\u0026ndash;300.0 mins). Entertainment trips had the smallest mean (SD/median) of 14.0 (13.2/10.0) mins (range: 1.0\u0026ndash;180.0 mins).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eCausal effects of dead-end road density on travel time\u003c/h3\u003e\n\u003cp\u003eOur analysis indicates that dead-end road density has a disproportionate effect on certain types of trips, particularly those involving shopping and entertainment, where time costs are more sensitive to detours. This suggests that rural road infrastructure, particularly the density of dead-end roads, plays a critical role in shaping activity-travel behaviour in rural areas.\u003c/p\u003e\u003cp\u003eWe found that increased dead-end road density significantly increased travel time for shopping and entertainment trips, but did not impact subsistence or family visit trips after adjusting for control variables (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Supplementary Table\u0026nbsp;5). Specifically, each additional unit of dead-end road density was associated with an increase of 21.5 mins (95% CI: 15.2\u0026ndash;27.7) for shopping trips to stores, 19.0 mins (95% CI: 10.6\u0026ndash;27.3) for shopping trips to fairs, and 15.2 mins (95% CI: 8.2\u0026ndash;22.1) for entertainment trips. However, no significant effects were found for subsistence trips (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.96) and family visit trips (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.68).\u003c/p\u003e\u003cp\u003eThese findings remained consistent across all sensitivity analyses, reinforcing the robustness of our results. The relationships between dead-end road density and travel time remained unchanged regardless of variations in hyperparameters, sample deletions, outlier adjustments, and additions of township fixed effects.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo address potential endogeneity, we used lagged dead-end road density as an instrumental variable in our analysis (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Supplementary Table\u0026nbsp;6). The results showed that the effects of dead-end road density on travel time remained significant. Specifically, each additional unit of dead-end road density resulted in an increase of 25.9 mins (95% CI: 0.30\u0026ndash;51.4) for shopping trips to stores, 31.2 mins (95% CI: 18.8\u0026ndash;41.6) for shopping trips to rural fairs, and 39.5 mins (95% CI: 23.8\u0026ndash;55.1) for entertainment trips.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eHeterogeneity analysis\u003c/h3\u003e\n\u003cp\u003eThe heterogeneity analysis highlights that the effects of road density are not uniform across all demographic groups. Population groups, such as the elderly, male, the high-income, and self-employed villagers, are particularly affected by the inefficiencies introduced by dead-end roads.\u003c/p\u003e\u003cp\u003eFirstly, gender and age played crucial roles in how dead-end roads impacted travel time (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea, \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb and Supplementary Table\u0026nbsp;7, 8). Specifically, male villagers and elderly participants were more affected by the increased travel time associated with dead-end road density. Male villagers, who typically play a larger role in household subsistence activities, experienced longer travel times for subsistence trips. Similarly, elderly villagers, who tend to have lower mobility and fewer travel options, were significantly more affected by the detours caused by dead-end roads. The elderly group, in particular, experienced an average increase of 21.3 minutes in their subsistence trips (95% CI: 7.9\u0026ndash;34.6), a considerable impact on their daily routines.\u003c/p\u003e\u003cp\u003eIn terms of household income, the analysis revealed that higher-income villagers were more affected by dead-end road density, particularly in the context of shopping trips (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec and Supplementary Table\u0026nbsp;9). Villagers with upper-middle income showed a significant increase of 16.6 minutes in subsistence trips (95% CI: 4.6\u0026ndash;28.5), reflecting a higher reliance on more distant, often urban, destinations for purchasing goods. These findings suggest that wealthier individuals tend to engage in shopping trips that involve traveling longer distances, and dead-end roads increase their travel burden. Surprisingly, high-income villagers demonstrated a decrease in time spent on family visit trips, with a reduction of 32.7 minutes (95% CI: -60.3 to -5.0). This reduction may reflect a shift in behaviour, with wealthier individuals opting for more convenient, less time-consuming alternatives to traditional family visits, such as meeting at third-party venues.\u003c/p\u003e\u003cp\u003eEmployment type was another key factor influencing travel time (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed and Supplementary Table\u0026nbsp;10). Self-employed villagers, who typically have more flexible schedules and travel to a wider range of destinations, were the most affected by dead-end roads, particularly for entertainment trips. These individuals experienced a significant increase in travel time of 35.8 minutes (95% CI: 22.1\u0026ndash;49.4). In contrast, farmers, whose daily activities are concentrated within a small geographic area, experienced less of an impact from dead-end roads, with only a modest increase of 5.4 minutes (95% CI: -0.1 to 10.9). This suggests that the effect of road infrastructure on travel behaviour is more pronounced for those with broader mobility patterns. Full-time workers also showed an increase in time spent on family visit trips (24.2 minutes, 95% CI: -2.0 to 50.5), but no significant effect on entertainment trips. Their fixed working hours may take away relatively limited leisure time, so they are less sensitive to dead-end roads for entertainment trips.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study provides a comprehensive analysis of the impact of dead-end road density on rural residents\u0026rsquo; activity-travel behaviour, drawing from a large sample of 5,415 participants across three major geographical regions in China. The findings reveal significant variations in how dead-end road density affects different types of activities, with notable consequences for mobility, social inclusion, and rural development.\u003c/p\u003e\u003cp\u003eOur results highlight that dead-end road density increases travel time primarily for shopping and entertainment trips, while subsistence and family visit trips remain largely unaffected. This distinction is important because it underscores the nature of discretionary versus non-discretionary travel. Non-discretionary trips, such as subsistence and family visits, are generally more structured, often involving fixed destinations and driven by social or cultural obligations\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. In contrast, shopping and entertainment trips are discretionary, often involving more flexible destinations that require more complex and circuitous routes, especially when road networks are fragmented\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. This finding aligns with previous studies showing that built environment factors have a greater impact on non-work trips compared to work-related travel, where destinations are more predictable and fixed\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Our study contributes to this literature by providing new insights into how infrastructure challenges disproportionately affect discretionary travel, which is often associated with higher mobility demands and broader social participation.\u003c/p\u003e\u003cp\u003eMoreover, our use of causal inference techniques, particularly DML, provides robust evidence of the causal relationship between dead-end road density and travel time, addressing the limitations of previous studies that have relied primarily on correlation-based methods\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. By applying this method to a partially linear instrumental variables model, we ensure that the estimates are unbiased, thus strengthening the validity of our findings. The estimated increases in travel time\u0026mdash;ranging from 26 to 40 mins for shopping and entertainment trips\u0026mdash;are more than doubling the travel time mean of 14\u0026ndash;18 mins. These effects are substantial, especially considering that many urban areas operate within a 15-minute community for daily needs\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. This finding challenges the assumption that rural mobility is solely constrained by long geographical distances to essential services, highlighting the additional burdens created by poor road infrastructure.\u003c/p\u003e\u003cp\u003eThe heterogeneity analysis reveals that the effects of dead-end road density vary significantly across different demographic groups. For instance, male and older villagers experience greater increases in time of subsistence trips, with older participants particularly vulnerable due to their reduced mobility and limited travel options\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. These findings suggest that dead-end roads exacerbate transport-related social exclusion for specific groups, particularly the elderly, who face greater challenges in maintaining their social and economic activities. Similarly, higher-income villagers, who often travel longer distances for higher-quality goods, are more affected by dead-end road density, particularly for shopping trips. Interestingly, high-income individuals show a decrease in time of family visit trips, likely as a result of changing social dynamics where these individuals may prefer meeting family members at more accessible venues, thus reducing the traditional need for long visits. This shift in behaviour suggests that while road infrastructure influences mobility, it also interacts with socioeconomic factors to shape activity-travel patterns\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eFurthermore, self-employed villagers\u0026mdash;who tend to have more dispersed travel destinations\u0026mdash;were found to be the most affected by dead-end road density, experiencing significantly longer travel time for entertainment trips. On the other hand, farmers, whose daily activities are confined to their immediate rural surroundings, were less impacted by dead-end roads, suggesting that road infrastructure challenges have less of an effect on their relatively localised mobility patterns\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. These findings underscore the need for targeted interventions that consider the diverse travel needs of different groups, particularly vulnerable populations who are more likely to face increased travel burdens due to fragmented road networks.\u003c/p\u003e\u003cp\u003eOur study makes a significant contribution to the understanding of rural transport disadvantages, offering valuable insights for planners and policymakers. By addressing the inefficiencies created by dead-end roads and improving connectivity, we can reduce the time costs associated with travel, which in turn can enhance the quality of life for rural residents and reduce social exclusion. In particular, policies aimed at improving access to essential services, such as shopping centres and entertainment venues, should prioritise addressing the infrastructure gaps that disproportionately affect specific groups, such as the elderly and self-employed. Furthermore, the findings suggest that high-income villagers may benefit from improved access to higher-quality goods and services, highlighting the role of road infrastructure in shaping economic opportunities and social mobility, further promoting health and well-being\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThere are some limitations that must be considered. First, the reliance on self-reported travel time data introduces the possibility of recall bias, which may affect the accuracy of the reported travel time. Additionally, seasonal variations may influence travel patterns, particularly for activities with cyclical destinations. Second, while OSM provides comprehensive road data, there may still be gaps in coverage, particularly in remote rural areas. Finally, our results are limited to rural residents in China aged 18 and older, and further research is needed to examine whether these findings can be generalised to other populations or countries.\u003c/p\u003e\u003cp\u003eNevertheless, our study provides strong evidence of the impact of dead-end road density on rural mobility and highlights the importance of addressing infrastructure deficiencies to promote more equitable and sustainable rural development.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eStudy design and population\u003c/h2\u003e\u003cp\u003eThe primary data were derived from a large-scale survey conducted in 2024 targeting permanent residents living in villages across South China, North China, and Northwest China. Using a stratified multi-stage sampling design, we first selected three representative provinces\u0026mdash;Guangdong, Henan, and Gansu\u0026mdash;that exhibit substantial variation in spatial locations, geographical conditions, and socioeconomic indicators across the three regions. Within each province, rural townships (the most basic administrative units in China\u0026rsquo;s rural governance structure) were systematically sampled.\u003c/p\u003e\u003cp\u003eThe sampling procedure continued with the selection of 3\u0026ndash;6 villages per township, stratified based on their distances from the township centre. Within each selected village, approximately 15 households were randomly sampled. Trained investigators conducted face-to-face interviews with participants, obtaining written informed consent from all respondents. This research protocol received ethical approval from the Peking University.\u003c/p\u003e\u003cp\u003eOf the 5,980 rural residents who initially participated, 5,645 completed the questionnaire (94.4% response rate). Data cleaning procedures excluded respondents who were under 18 years of age, those with invalid responses, and participants whose survey completion time was less than 5 minutes (indicating potentially unreliable responses). After further excluding cases with missing data on travel patterns and key socioeconomic variables, the final analytical sample comprised 5,415 participants from 277 villages (see Supplementary Fig.\u0026nbsp;1 and Table\u0026nbsp;1).\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eKey variables\u003c/h3\u003e\n\u003cp\u003eParticipants reported travel time from home to destinations in minutes. We categorised all trips into four types based on purpose: subsistence, shopping, family visit, and entertainment. Subsistence trips encompass travel to work, agricultural activities, or school\u0026mdash;activities essential for sustaining livelihood. Shopping trips involve travel to purchase necessary goods or services, with two primary destinations identified: physical stores and rural periodic fairs. Family visit trips constitute travel to visit relatives, while entertainment trips represent discretionary travel for social or leisure purposes.\u003c/p\u003e\u003cp\u003eDead-end roads were defined as roads with only one or no point of access to other roads. Rural road data was extracted from the OpenStreetMap (OSM) dataset (2024) via the Geofabrik Download Server (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://download.geofabrik.de/\u003c/span\u003e\u003cspan address=\"https://download.geofabrik.de/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). This dataset, despite limitations, represents the most comprehensive open-source rural road database available, with accuracy substantially improved in recent years\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Using the township as our basic analytical unit based on villagers\u0026rsquo; typical activity ranges, we extracted road networks within township administrative boundaries and planarised all roads at intersection points. We identified dangle points (points without connection to other roads) using the Feature Vertices to Points tool. Dead-end roads were characterised as roads with one or two dangle points, with length measured as the distance from the dangle point to either an intersection point or another dangle point. Our primary metric, dead-end road density, was calculated by dividing the total length of dead-end roads by township area. All spatial analyses were performed using ArcGIS 10.8.\u003c/p\u003e\n\u003ch3\u003eControl variables\u003c/h3\u003e\n\u003cp\u003eWe controlled for two groups of variables: socioeconomic characteristics and township-level variables. Socioeconomic characteristics included age, gender, education years, employment type, self-reported health, household size, structure, income, and vehicle ownership (see Supplementary Table\u0026nbsp;2). Self-reported health was measured on a four-point scale (1 \u0026ldquo;strongly disagree\u0026rdquo; to 4 \u0026ldquo;strongly agree\u0026rdquo; in response to \u0026ldquo;I think my health is poor\u0026rdquo;), with scores reversed so higher values indicated better health. Household size represented the number of cohabiting members, household structure indicated the presence of children and elderly, and vehicle ownership encompassed cars, bicycles, motorcycles, and tricycles.\u003c/p\u003e\u003cp\u003eTownship-level variables comprised built and natural environmental factors. Built environmental variables included population density (residents per township area from 2023 LandScan dataset at 1-km resolution), road density (road length per township area from OSM data), proximity measures (shortest distance to city/county centres and highways/local roads calculated using \u0026ldquo;geodist\u0026rdquo; package in Stata 18.0), and infrastructure accessibility (number of healthcare, educational, and shopping facilities, plus bus stations from Amap 2024). Natural environmental variables encompassed slope (from Copernicus DEM dataset at 30m \u0026times; 30m resolution), temperature, and precipitation (both at 1km \u0026times; 1km resolution from the National Tibetan Plateau Data Centre). All geographic data were matched to township-level administrative boundaries and processed using raster calculation tools to derive township-level metrics (see Supplementary Table\u0026nbsp;4).\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eWith the increasing focus on causal inference in social sciences and economics, machine learning (ML) has emerged as a powerful approach for identifying causal effects. In this study, we employed Double/debiased Machine Learning (DML) to examine the causal relationship between dead-end roads and villagers\u0026rsquo; travel time. DML relaxes assumptions about intrinsic data correlations, allowing for more flexible functional forms and improved stability. This approach combines the advantages of machine learning in processing high-dimensional data and model recognition while maintaining the rigor of econometric models in causal inference\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. We began by constructing a partial linear model:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\begin{array}{c}Y={\\theta\\:}_{0}D+{f}_{0}\\left(\\varvec{X}\\right)+U,\\:\\:E\\left[U|\\varvec{X},\\:D\\right]=0\\#\\left(1\\right)\\end{array}$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\begin{array}{c}D={m}_{0}\\left(\\varvec{X}\\right)+V,\\:\\:E\\left[V|\\varvec{X}\\right]=0\\#\\left(2\\right)\\end{array}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cem\u003eY\u003c/em\u003e is the travel time, \u003cem\u003eD\u003c/em\u003e is the number or length of dead-end roads, \u003cb\u003eX\u003c/b\u003e is a vector consists of control variables, \u003cem\u003eU\u003c/em\u003e and \u003cem\u003eV\u003c/em\u003e are the stochastic errors, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{0}\\left(\\varvec{X}\\right)\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{m}_{0}\\left(\\varvec{X}\\right)\\)\u003c/span\u003e\u003c/span\u003e are unknown nonparametric functions. We are only interested in the parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\theta\\:}_{0}\\)\u003c/span\u003e\u003c/span\u003e, which represented the average treatment effect of \u003cem\u003eD\u003c/em\u003e on \u003cem\u003eY\u003c/em\u003e.\u003c/p\u003e\u003cp\u003eAll DML analyses were performed using the \u0026ldquo;ddml\u0026rdquo; package in Stata 18.0\u003csup\u003e26\u003c/sup\u003e. To ensure model reliability, we followed Chernozhukov et al.\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e recommendations by implementing a fivefold cross-fitting procedure. Since DML randomly splits samples during cross-fitting, we fitted the same model with five different random splits and reported the median estimator to account for randomness effects. Given the flexibility in ML method selection within the DML framework, we employed a stacking approach to reduce misspecification risk. This method constructs combinations of diverse base machine learners, assigning nonzero weights based on constrained least squares, with the final result being a weighted average of predictions\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. We selected six mainstream base learners: linear regression (LR), lasso with cross-validated penalty (LASSO), support vector machines (SVM), feed-forward neural nets (FNN), random forests (RF), and gradient boosting trees (GBT).\u003c/p\u003e\u003cp\u003eTo address the nested data structure (villagers within townships) potentially violating DML\u0026rsquo;s independently and identically distributed samples assumption, we used cluster robust standard errors to prevent statistical significance overstatement\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. Additionally, we achieved independent sample-splitting by randomly assigning folds by cluster, ensuring hold-out samples contained no observations from clusters used in training. We conducted multiple sensitivity analyses to validate benchmark results, including: setting cross-fitting folds to 10; removing subdistrict (\u003cem\u003ejiedao\u003c/em\u003e) samples which might exhibit different travel behaviours than rural townships; winsorising continuous variables at the 1st and 99th percentiles to mitigate outlier effects; and including township dummy variables to control for unobserved characteristics across townships.\u003c/p\u003e\u003cp\u003eTo identify causal effects, we employed an instrumental variable approach within DML to address potential endogeneity. Following previous literature\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e, we used historical road networks as instruments since current roads build upon old roads while previous roads remain unassociated with current individual activities. We adopted 2021 dead-end road density as an instrument, avoiding premature OSM data quality issues. Weak instrument tests confirmed its validity, with first-stage F-statistics exceeding the recommended value of 10 in all significant models\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. For heterogeneity analysis, we estimated dead-end road density impacts across demographic groups including sex, age, household income, and employment types, implementing 10-fold cross-fitting to ensure reliability given smaller subgroup sample sizes\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Further analyses examined other dead-end road indicators (number, length, various ratios), revealing negligible effects on travel time and highlighting the importance of dead-end road density specifically (Supplementary Table\u0026nbsp;11).\u003c/p\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eConceptualization: Q.L.; data collection: P.Z., Z.L., S.J., H.W., H.Z., Z.L., Y.T., Q.L.; analytical tools: J.L. and Q.L.; results analysis: J.L.; manuscript draft and reviewing: J.L. and Q.L.; funding acquisition: P.Z. and Q.L.; all authors read and approved the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe raw CoVIDA and HBS data are protected and are not available due to data privacy laws.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlessandretti, L., Aslak, U. \u0026amp; Lehmann, S. The scales of human mobility. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e587\u003c/strong\u003e, 402-407 (2020). \u003c/li\u003e\n\u003cli\u003eLaurance, W. F.\u003cem\u003e et al.\u003c/em\u003e A global strategy for road building. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e513\u003c/strong\u003e, 229-232 (2014). \u003c/li\u003e\n\u003cli\u003eUnited Nations Department of Economic and Social Affairs. \u003cem\u003eWorld Social Report 2021: Reconsidering Rural Development\u003c/em\u003e (UN DESA, 2021). \u003c/li\u003e\n\u003cli\u003eSchlossberg, M., Greene, J., Phillips, P. P., Johnson, B. \u0026amp; Parker, B. School Trips: Effects of Urban Form and Distance on Travel Mode. \u003cem\u003eJournal of the American Planning Association\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 337-346 (2006). \u003c/li\u003e\n\u003cli\u003eHanig, L.\u003cem\u003e et al.\u003c/em\u003e Finding gaps in the national electric vehicle charging station coverage of the United States. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 561 (2025). \u003c/li\u003e\n\u003cli\u003eYuan, D., Zhao, P., Yu, Z. \u0026amp; Liu, Q. Villagers\u0026apos; travel burden and the built environment in rural China: Evidence from a national level survey. \u003cem\u003eJournal of Transport Geography\u003c/em\u003e \u003cstrong\u003e113\u003c/strong\u003e, 103717 (2023). \u003c/li\u003e\n\u003cli\u003eMackett, R. L. The health implications of inequalities in travel. \u003cem\u003eJournal of Transport \u0026amp; Health\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 202-209 (2014). \u003c/li\u003e\n\u003cli\u003eWeiss, D. J.\u003cem\u003e et al.\u003c/em\u003e Global maps of travel time to healthcare facilities. \u003cem\u003eNature Medicine\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1835-1838 (2020). \u003c/li\u003e\n\u003cli\u003eLeite, D. \u0026amp; De Bacco, C. Similarity and economy of scale in urban transportation networks and optimal transport-based infrastructures. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 7981 (2024). \u003c/li\u003e\n\u003cli\u003eAsher, S. \u0026amp; Novosad, P. Rural Roads and Local Economic Development. \u003cem\u003eAmerican Economic Review\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 797-823 (2020). \u003c/li\u003e\n\u003cli\u003eSchl\u0026auml;pfer, M.\u003cem\u003e et al.\u003c/em\u003e The universal visitation law of human mobility. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e593\u003c/strong\u003e, 522-527 (2021). \u003c/li\u003e\n\u003cli\u003eChen, C. \u0026amp; Mokhtarian, P. L. Tradeoffs between Time Allocations to Maintenance Activities/Travel and Discretionary Activities/Travel. \u003cem\u003eTransportation\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 223-240 (2006). \u003c/li\u003e\n\u003cli\u003eDharmowijoyo, D. B. E., Susilo, Y. O. \u0026amp; Karlstr\u0026ouml;m, A. On complexity and variability of individuals\u0026rsquo; discretionary activities. \u003cem\u003eTransportation\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 177-204 (2018). \u003c/li\u003e\n\u003cli\u003eLee, M., Barbosa, H., Youn, H., Holme, P. \u0026amp; Ghoshal, G. Morphology of travel routes and the organization of cities. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 2229 (2017). \u003c/li\u003e\n\u003cli\u003eJiang, N.\u003cem\u003e et al.\u003c/em\u003e Traffic related activity pattern of Chinese adults: a nation-wide population based survey. \u003cem\u003eJournal of Exposure Science \u0026amp; Environmental Epidemiology\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 482-489 (2023). \u003c/li\u003e\n\u003cli\u003eAbbiasov, T.\u003cem\u003e et al.\u003c/em\u003e The 15-minute city quantified using human mobility data. \u003cem\u003eNature Human Behaviour\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 445-455 (2024). \u003c/li\u003e\n\u003cli\u003eBruno, M., Monteiro Melo, H. P., Campanelli, B. \u0026amp; Loreto, V. A universal framework for inclusive 15-minute cities. \u003cem\u003eNature Cities\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 633-641 (2024). \u003c/li\u003e\n\u003cli\u003eFord, J. A., Wong, G., Jones, A. P. \u0026amp; Steel, N. Access to primary care for socioeconomically disadvantaged older people in rural areas: a realist review. \u003cem\u003eBMJ Open\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e010652 (2016). \u003c/li\u003e\n\u003cli\u003eZhao, P. \u0026amp; Yu, Z. Investigating mobility in rural areas of China: Features, equity, and factors. \u003cem\u003eTransport Policy\u003c/em\u003e \u003cstrong\u003e94\u003c/strong\u003e, 66-77 (2020). \u003c/li\u003e\n\u003cli\u003eChen, Z.\u003cem\u003e et al.\u003c/em\u003e Contrasting social and non-social sources of predictability in human mobility. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1922 (2022). \u003c/li\u003e\n\u003cli\u003eGonz\u0026aacute;lez, M. C., Hidalgo, C. A. \u0026amp; Barab\u0026aacute;si, A.-L. Understanding individual human mobility patterns. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e453\u003c/strong\u003e, 779-782 (2008). \u003c/li\u003e\n\u003cli\u003eLiu, Y. \u0026amp; Li, Y. Revitalize the world\u0026rsquo;s countryside. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e548\u003c/strong\u003e, 275-277 (2017). \u003c/li\u003e\n\u003cli\u003ePearsall, H.\u003cem\u003e et al.\u003c/em\u003e Advancing equitable health and well-being across urban\u0026ndash;rural sustainable infrastructure systems. \u003cem\u003enpj Urban Sustainability\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 26 (2021). \u003c/li\u003e\n\u003cli\u003eHerfort, B., Lautenbach, S., Porto de Albuquerque, J., Anderson, J. \u0026amp; Zipf, A. A spatio-temporal analysis investigating completeness and inequalities of global urban building data in OpenStreetMap. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 3985 (2023). \u003c/li\u003e\n\u003cli\u003eKoks, E. E.\u003cem\u003e et al.\u003c/em\u003e A global multi-hazard risk analysis of road and railway infrastructure assets. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 2677 (2019). \u003c/li\u003e\n\u003cli\u003eAhrens, A., Hansen, C. B., Schaffer, M. E. \u0026amp; Wiemann, T. ddml: Double/debiased machine learning in Stata. \u003cem\u003eThe Stata Journal\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 3-45 (2024). \u003c/li\u003e\n\u003cli\u003eChernozhukov, V.\u003cem\u003e et al.\u003c/em\u003e Double/debiased machine learning for treatment and structural parameters. \u003cem\u003eThe Econometrics Journal\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, C1-C68 (2018). \u003c/li\u003e\n\u003cli\u003eAhrens, A., Hansen, C. B. \u0026amp; Schaffer, M. E. pystacked: Stacking generalization and machine learning in Stata. \u003cem\u003eThe Stata Journal\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 909-931 (2023). \u003c/li\u003e\n\u003cli\u003eAbadie, A., Athey, S., Imbens, G. W. \u0026amp; Wooldridge, J. M. When Should You Adjust Standard Errors for Clustering? \u003cem\u003eThe Quarterly Journal of Economics\u003c/em\u003e \u003cstrong\u003e138\u003c/strong\u003e, 1-35 (2023). \u003c/li\u003e\n\u003cli\u003eBaum-Snow, N. Did Highways Cause Suburbanization? \u003cem\u003eThe Quarterly Journal of Economics\u003c/em\u003e \u003cstrong\u003e122\u003c/strong\u003e, 775-805 (2007). \u003c/li\u003e\n\u003cli\u003eDuranton, G. \u0026amp; Turner, M. A. The Fundamental Law of Road Congestion: Evidence from US Cities. \u003cem\u003eAmerican Economic Review\u003c/em\u003e \u003cstrong\u003e101\u003c/strong\u003e, 2616-2652 (2011). \u003c/li\u003e\n\u003cli\u003eAlejo, J., Galvao, A. F. \u0026amp; Montes-Rojas, G. A first-stage representation for instrumental variables quantile regression. \u003cem\u003eThe Econometrics Journal\u003c/em\u003e\u003cstrong\u003e26\u003c/strong\u003e, 350-377 (2023). \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6871842/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6871842/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDespite substantial global investments in rural road construction, a critical infrastructural deficiency\u0026mdash;the prevalence of dead-end roads\u0026mdash;continues to undermine mobility, exacerbates social inequities, and impedes sustainable development in rural communities worldwide. This study establishes the causal relationship between dead-end road density and travel outcomes, revealing how network topology fundamentally shapes rural accessibility patterns beyond simple distance metrics. Applying Double/debiased Machine Learning to data from 5,415 rural residents across China, we demonstrate that each unit increase in dead-end road density significantly extends travel time for discretionary activities by 26\u0026ndash;40 minutes\u0026mdash;effectively doubling average journey durations\u0026mdash;while non-discretionary trips remain largely unaffected. Our causal analysis further reveals striking demographic heterogeneity, with elderly, male, higher-income, and self-employed individuals experiencing disproportionately greater mobility burdens. These findings challenge conventional rural development approaches that prioritise road quantity over network quality and connectivity. By identifying how specific network deficiencies systematically disadvantage certain demographic groups and activity types, we provide critical insights for policymakers seeking to enhance rural accessibility, promote inclusive development, and reduce infrastructure-induced inequalities. Strategic interventions targeting dead-end road reduction represent a high-leverage opportunity to transform rural infrastructure from a constraint into a catalyst for sustainable development.\u003c/p\u003e","manuscriptTitle":"Dead Ends, Lost Time: Rural Mobility Constraints","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-16 09:15:27","doi":"10.21203/rs.3.rs-6871842/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9afe14e1-ab76-43ea-b86f-ff6514da69bd","owner":[],"postedDate":"September 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54534234,"name":"Earth and environmental sciences/Environmental social sciences/Socioeconomic scenarios"},{"id":54534235,"name":"Scientific community and society/Social sciences"}],"tags":[],"updatedAt":"2025-10-17T20:23:18+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-16 09:15:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6871842","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6871842","identity":"rs-6871842","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}