Structural inequality shapes the daily rhythm of stationarity and mobility Authors: Zihao Tian, Chao Zhang, Weiyu Li, Lixin Tian, Michael M Danziger, Wenshan Liu, Qi R Wang Published: 2026 ABSTRACT AbstractHuman mobility exhibits rhythmic patterns that shape economic activity, social interactions, and access to essential services. Yet, while prior studies focus on spatial movement, the temporal structure of people’s stationarity—the act of dwelling—remains underexplored. Here, we introduce the Stationarity Sinusoidal Stability (STASIS) model to quantify the daily pulse of physical presence. Using three parameters—range, phase angle, and baseline—we show that majority-Black census block groups (CBGs) have smaller daily stationarity ranges, later activity peaks, and lower average home-presence baselines than majority-White and majority-Hispanic CBGs. The Rhythm Stability Index further indicates greater variability in activity timing among CBGs with higher proportions of Black residents, and this variability is associated with CBG-level economic precarity. Using COVID-19 as a stress test, our findings show that temporal stability in daily mobility rhythms differs across CBGs with different racial and Hispanic-origin compositions.Subject terms: Sociology, Applied physics Introduction Human mobility is dialectical. It operates within a fundamental daily pulse: the pervasive oscillation between activity and rest1–4. In this diurnal cycle, movement is structurally dependent on essential moorings—stable points in space and time that allow social life to function5,6. Governed by biological needs and social constraints7,8, home serves as the primary anchor for this rhythm. However, being home is not merely a static location or a lack of movement; it represents a state of fixity and stationarity, i.e. an active, necessary counterweight to travel9–11. Consequently, the timing and regularity of home presence create a measurable daily rhythm that is just as fundamental to the structure of everyday life as movement itself. Despite this interdependence, contemporary mobility research has prioritised the statistical mechanics of movement, focusing almost exclusively on spatial distributions, trajectory evolution and aggregate flows12–15. By contrast, the rhythmic structure of residential stationarity, specifically when people dwell at home and the regularity of this timing, has received comparatively less attention. This distinction is important because stationarity is not simply the absence of mobility; it is a manifestation of motility, i.e. the capacity to appropriate space and control one’s own time16,17. While spatial mismatch theories have long established how location shapes access to opportunity18,19, there remains a lack of unified and interpretable frameworks to understand the temporal architectures of daily life. Current models effectively capture where people go, but they do not explicitly quantify the capacity for fixity, leaving the structural rhythms of rest and recovery less well characterised. Rather than implying a fundamental shift beyond existing spatial frameworks, we argue that understanding these temporal architectures can serve as a complement to existing activity-space and multicontextual segregation research. Furthermore, the temporal stability of these rhythms is an important dimension of social inequality, building on the concept of temporal justice20,21. Mobility justice scholars argue that the ‘right to rest’ and the ability to maintain predictable routines are often less available to precariously employed or marginalised populations22,23. Stable daily rhythms generally rely on secure employment and consistent access to services. Conversely, temporal instability, which often appears as fragmented rest patterns or irregular home presence, is associated with shift-based work, economic precarity and difficulty aligning with dominant institutional schedules24,25. These rhythmic instabilities can reinforce disadvantage26,27. To capture this dimension, we move beyond measuring the duration of stationarity to measuring its predictability, examining how volatility in daily rhythms is associated with structural inequality. Here, we introduce the stationarity sinusoidal stability (STASIS) model. This framework pivots the analytical focus from spatial flows to daily patterns of presence, treating the daily oscillation between dwelling and movement as a measurable signal of social structure10. Whereas spatial models describe where people move, STASIS captures when daily transitions occur and, crucially, the consistency of that timing. The model yields three interpretable parameters: range (2A, the spatial extent of the daily pulse), phase angle (ϕ, the timing of peak activity) and baseline (B, the volume of stationarity). These parameters enable the derivation of the Rhythm Stability Index (RSIϕ), a metric that quantifies variability in daily activity timing. Applying this framework to more than 200,000 census block groups (CBGs) in the United States of America (USA) and using the COVID-19 pandemic as a structural stress test, we examine how temporal regularity in daily stationarity varies with CBG-level racial and Hispanic-origin composition. Results Stationarity sinusoidal stability (STASIS) model The daily mobility cycle can be represented as a regular shift between higher nighttime home presence and lower daytime home presence. In Fig. 1a, larger circles indicate higher proportions of devices observed at home at night; in Fig. 1b, daytime stationarity declines as devices are observed away from home. This cyclical pattern is not confined to a single location but is consistently observed across diverse urban regions, as illustrated by the weekly rhythms in New York, Los Angeles, and Chicago (Fig. 1c). Stationarity Sinusoidal Stability (STASIS) model. Schematic illustration of the STASIS model, using Manhattan block groups from United States Census Bureau TIGER/Line Shapefiles (2019) as an example: a at night, as device-based home presence increases in Manhattan block groups, stationarity peaks; b during the day, stationarity declines as device-based home presence decreases. c Cyclical patterns of weekly rhythms in New York, Los Angeles and Chicago. d The STASIS model parameters: 2A represents the range (twice the amplitude), quantifying the span of daily variation; ϕ (phase angle) captures the timing of the transition (valley); and B (baseline) represents the average level of stationarity. e Fitted curves for devices at home per hour: grey lines represent individual working days in 2019 (n = 251), while the thick red line represents the mean fit. f Goodness-of-fit evaluated by Residual Sum of Squares (RSS) and R2. Scatter points represent individual working days in 2019 (n = 251). The solid line represents the linear regression fit (measure of centre), and the shaded band indicates the 95% confidence interval (error band). Spatial distribution of county-level stationarity across the United States (map outlines from United States Census Bureau TIGER/Line Shapefiles (2019)), showing the g peak (nighttime) and h valley (daytime). i Stationarity rhythms at the state level: blue curve = empirical data; red dashed curve = fitted model; shaded region = 95% confidence interval. Our STASIS model formalises human mobility and stationary patterns using sinusoidal waveforms. Thus, the model takes the following form: In Eq. (1), xt represents the radian value corresponding to time t, dividing 2π evenly into 24 h, with 0 corresponding to a zero radian value, i.e. midnight, and R(t) represents percentage of devices at home per hour. Three parameters in the model determine the key characteristics of the stationary rhythm. A is the amplitude, which quantifies the maximum range of variation in mobility. Since A represents half the difference between the peak and trough values, we use 2A to represent the full range (see Fig. 1d), which thus captures the full scale between activity and stationarity. The phase angle, ϕ, shifts the sinusoid along the time axis, reflecting the timing of peak mobility within the typical 24-h cycle. This shift helps pinpoint when the highest mobility or activity occurs, providing insights into daily patterns of human movement (see Fig. 1d). B represents the baseline, i.e. the central line around which the sinusoid oscillates. It, therefore, corresponds to the average level of stationarity, or the typical proportion of devices observed at home (see Fig. 1d). This parameter is essential for understanding the general tendency of device-based home presence. We test and evaluate our model based on three different mobility metrics from SafeGraph (see Data description). The first one, discussed in the main text, is the percentage of devices at home. It is calculated by R(t) = nij(t)/Nij, where nij(t) represents the number of devices at home at hour t of date j in CBG i, and Nij represents the number of devices at the date j in CBG i. We tested the model on multiple levels, including national, regional and city levels. The other two mobility metrics are: Rhome=nhome,ij/Nij, where nhome represents the number of devices completely at home daily and Rwork=nwork,ij/Nij, where nwork represents the number of devices observed at work. Here, the types of work are further broken down into full-time, part-time and delivery work. Note that the temporal resolution of R(t) is hourly and daily for Rhome and Rwork, with detailed results provided in Supplementary Fig. 25. To address potential demographic sampling biases, we implemented a weighting mechanism based on American Community Survey penetration rates28; detailed methodologies and comparative analyses confirming the consistency of our findings between weighted and unweighted datasets are provided in the Supplementary Information. The result at the national level for 2019 on the weighted datasets is presented in Fig. 1e, with each grey line representing the result of each working day and the red average across the year. The goodness of fit for each working day in 2019 is shown in Fig. 1f. Both R2 and Residual Sum of Squares (RSS) demonstrate robust performance. The spatial distribution of county-level stationarity during the peak (nighttime Fig. 1g) and valley (daytime, Fig. 1h) of the daily rhythm also align with intuition. At night (Fig. 1g), stationarity is uniformly high across the country, with most counties showing elevated R(t) values as devices are observed at home, while during the day (Fig. 1h), mobility increases nationally. The results on other scales and mobility metrics can be found in Supplementary Fig. 26. The generalisability of our model was further evaluated using the Rhome and Rwork metrics derived from the SafeGraph dataset, as shown in Supplementary Fig. 25. To test the robustness of the STASIS across different contexts, we also applied the model to human mobility data from Beijing (Supplementary Fig. 13) and Nanjing (Supplementary Fig. 12) in China, London in the United Kingdom (UK) (Supplementary Fig. 11), and Boston in the USA (Supplementary Fig. 10), as well as an anonymous metropolis in Japan (Supplementary Fig. 9). Across all datasets, the STASIS model consistently reproduced the characteristic daily rhythms of human activity, confirming its broad applicability and interpretability. Our model also captures stationary rhythms at the state level, as illustrated in Fig. 1i, where each circular panel shows the diurnal movement pattern for one state in the USA, with time running around the circle (midnight at the top, noon at the bottom) and the radial axis indicating the percentage of devices observed at home. The blue curve is the empirical data, the red dashed curve is the fitted model, and the shaded region indicates the 95% confidence interval (CI). Overall, the figure demonstrates that a single sinusoidal model captures the daily rhythm well across diverse states, yet subtle variations in peak timing or amplitude reflect regional differences in daily routines (e.g. work schedules, time zones, or cultural norms). The results establish STASIS as a framework for quantifying the daily oscillation between stationarity and activity as measurable, comparable signals. Model parameters reveal CBG-level differences The STASIS model formalises the diurnal cycle of stationarity and transforms it into measurable, comparable signals. This interpretability allows us to move beyond aggregate flow statistics to examine the internal temporal structure of daily dwelling. By decomposing these rhythms into range, phase and baseline, the framework allows comparison of CBG-level stationarity patterns across racial and Hispanic-origin composition categories. To analyse CBG-level differences by racial and Hispanic-origin composition, we categorise CBGs based on a 50% threshold28,29 into majority non-Hispanic White, majority non-Hispanic Black, and majority Hispanic groups. Figure 2a illustrates the daily stationarity rhythms for the three CBG racial/Hispanic-origin categories. Across all categories, the general daily pattern is preserved, with nighttime stationarity peaking around midnight and a clear trough during the day. However, subtle but consistent differences emerge in the shape of the curves. The pattern for majority-White CBGs appears more stretched, indicating a sharper contrast between nighttime and daytime immobility. Meanwhile, the curves for majority-Hispanic and especially majority-Black CBGs appear more compressed, suggesting a more constrained daily mobility rhythm. While these differences align with prior research on disparities in social media usage and mobile phone communications25,28,30, here we quantify analogous rhythmic differences using device-based human mobility measures. Differences in stationarity range, timing and baseline. a Daily stationarity rhythms for three census block group (CBG) racial/Hispanic-origin categories: majority-White, majority-Hispanic and majority-Black CBGs. b–d Distributions of STASIS model parameters across 251 working days in 2019. Data are presented as split violin plots comparing weighted (blue) and unweighted (red) datasets across majority-White, majority-Hispanic and majority-Black CBGs. Within each violin, the central dashed line indicates the median and the thickened region represents the interquartile range (IQR). b Range (2A): The full amplitude of the daily rhythm, representing twice the model amplitude A and capturing the total variation between peak and trough levels of the signal. c Phase angle (ϕ): The phase shift of the sinusoidal function, indicating the timing of peak mobility within the 24-h cycle. d Baseline (B): The baseline level of the sinusoidal function, corresponding to the average proportion of devices observed at home. All parameters are estimated from the STASIS model (see ‘Methods’ for model specification). e State-level performance of the three parameters. Missing colours indicate insufficient data for the corresponding CBG category. The disparities hinted at in the overall shapes of the daily stationarity curves become clearer when we examine the underlying model parameters. Fig. 2b–d show how differences across CBG categories are captured by the three parameters of our STASIS model. 2A was the smallest in the majority-Black CBGs (Fig. 2b (red), 0.203, 95% CI: 0.199–0.206), while majority-White CBGs exhibited the highest range (blue, 0.292, 95% CI: 0.286–0.298), followed by the majority-Hispanic CBGs (orange, 0.260, 95% CI: 0.255–0.265). The comparison indicates smaller daily stationarity ranges in majority-Black CBGs than in majority-White and majority-Hispanic CBGs. The phase angle (ϕ) also differed substantially across CBG categories. Majority-Black CBGs had a later ϕ at 1:32 PM (95% CI: 1:28–1:36 PM), almost 45 min later than the values we observed from majority-White and majority-Hispanic CBGs with (ϕ)’s at 12:49 PM (95% CI: 12:45–12:52 PM) and 12:52 PM (95% CI: 12:48–12:55 PM), respectively (Fig. 2c). Similarly, the baseline (B), which indicates the typical level of stationarity or the proportion of devices observed at home, was lower in majority-Black CBGs (Fig. 2d, 0.428 (95% CI: 0.427–0.430, majority-Black) vs. 0.461 (95% CI: 0.458–0.463, majority-White), 0.459 (95% CI: 0.457–0.461, majority-Hispanic). Figure 2 e breaks down the three model parameters across states in the USA. Each coloured dot represents the average parameter value for majority-White (blue), majority-Hispanic (orange) and majority-Black (red) CBGs within a given state; missing colours indicate that data for the corresponding CBG category is unavailable due to insufficient representation. The figure reveals a consistent pattern nationwide. With few exceptions, majority-White CBGs exhibit larger range values, earlier peak mobility times (i.e. higher ϕ), and higher stationarity baselines compared to majority-Black and majority-Hispanic CBGs. Conversely, majority-Black CBGs generally show a smaller range, later peak mobility and lower baseline, with majority-Hispanic CBGs generally falling between the two. These patterns are also observed across regional and city-level summaries, as detailed in Supplementary Figs. 28–35 and Supplementary Figs. 38–39. Taken together, these analyses indicate that CBG-level differences in daily mobility rhythms are observed across multiple geographic aggregations, while some states show deviations or incomplete data31,32. The smaller mobility range (2A) observed in majority-Black CBGs is consistent with a compressed daily stationarity rhythm. This pattern may reflect capability constraints discussed in the spatial mismatch literature, including weaker transit service33,34 and exclusionary housing policies18,35. The delayed phase angle (ϕ) indicates later timing of peak activity in majority-Black CBGs, potentially related to low-wage and shift-based employment patterns36,37. The lower baseline (B) indicates lower average home presence and may reflect irregular hours and complex commutes that reduce opportunities for sustained residential stationarity22,38. Rhythm stability index While range (2A) and baseline (B) capture the magnitude of stationarity, the phase angle (ϕ) captures its timing. Variability in this timing provides one way to examine the structural conditions of daily life. Drawing on the distinction between constrained immobility and restorative fixity9,10, we define the Rhythm Stability Index (RSIϕ = Var(ϕ)) to measure the temporal integrity of this daily anchor. Crucially, the measured RSIϕ provides an empirical proxy that is consistent with the theoretical concept of motility, which is the capacity to control one’s own schedule16. A low RSIϕ indicates more stable activity timing, whereas high values indicate greater variability in the timing of daily mobility. This measure provides an empirical way to assess how predictable daily stationarity rhythms vary across CBGs and how those variations relate to CBG-level socioeconomic conditions. Figure 3a, d illustrate differences in rhythm stability by CBG racial and Hispanic-origin composition. In Fig. 3a, majority-Black CBGs (red) exhibit higher variance (RSIϕ) than majority-Hispanic (orange) and majority-White (blue) CBGs. The distributions for majority-White and majority-Hispanic CBGs are more tightly clustered, whereas majority-Black CBGs show a broader spread in daily activity timing. Figure 3d maps the timing of the stationarity trough (i.e. peak activity), showing that majority-Black CBGs have both later and more dispersed timing than majority-White and majority-Hispanic CBGs. Rhythm Stability Index (RSIϕ) reveals instability. a Distribution of RSIϕ and variance of ϕ within individual CBGs in 2019 (red: majority-Black; orange: majority-Hispanic; blue: majority-White). b RSIϕ across CBG racial/Hispanic-origin population-share thresholds. Scatter plot and linear regression showing the relationship between RSIϕ and CBG population-share thresholds, ranging from 50% to 100%. The shaded bands represent the 95% confidence intervals around the mean RSIϕ. Statistics are derived from n = 193,024 independent CBGs (majority-White: n = 131,712; majority-Hispanic: n = 20,455; majority-Black: n = 15,163). Replicates are independent spatial units, defined as unique CBGs. Each data point represents the mean RSIϕ for all CBGs at a specific integer percentage of population concentration. No technical replicates were used in this analysis. c Stratification by data density (quartiles of device counts). Data are presented as box plots. The centre line represents the median, white diamonds indicate the mean, the box limits correspond to the first and third quartiles (IQR), and the whiskers extend to 1.5 × IQR. Outliers are not shown for visual clarity. Sample sizes (n) for each group (majority-White, majority-Hispanic and majority-Black CBGs, respectively) are: Low (n = 33,246, 4419, 4994), Medium-Low (n = 32,162, 5773, 4097), Medium-High (n = 32,146, 5912, 3383) and High (n = 34,158, 4351, 2689). d Valley time corresponding to ϕ for each CBG at the 50% threshold (10,000 sampled points per group; dashed lines indicate group means). e, f Spatial distribution of RSIϕ in Atlanta and St. Louis. g Ordinary least squares (OLS) regression results on determinants of temporal dispersion in mobility, as detailed in Supplementary Note 9. Figure 3b shows how RSIϕ evolves with increasing proportions of each racial or Hispanic-origin category within CBGs. As the proportion of each population moves from 50 to 100%, we progressively restrict the analysis to CBGs with higher concentrations of a single racial or Hispanic-origin category. The results show that as the proportion of Black residents increases, RSIϕ also rises, indicating greater variability in the timing of daily mobility. In contrast, majority-White and majority-Hispanic CBGs show lower and more stable RSIϕ values across thresholds. This pattern indicates increasing fragmentation of activity timing in CBGs with higher proportions of Black residents. To assess whether the observed differences in rhythm stability reflect data sparsity in CBGs with different racial and Hispanic-origin compositions, we conducted a robustness check by stratifying CBGs into four quartiles based on device counts (Fig. 3c). This analysis shows that differences in RSIϕ between CBG categories are observed across all levels of data density. Specifically, the mean RSIϕ values for majority-Black versus majority-White CBGs remain distinct across the four quartiles: Q1 (low: 0.283, 95% CI: 0.279–0.287 vs. 0.197, 95% CI: 0.195–0.198), Q2 (0.182, 95% CI: 0.17–0.185 vs. 0.118, 95% CI: 0.117–0.119), Q3 (0.124, 95% CI: 0.121–0.127 vs. 0.083, 95% CI: 0.082–0.084) and Q4 (high: 0.114, 95% CI: 0.109–0.119 vs. 0.080, 95% CI: 0.079–0.082) (see Supplementary Note 5 for details). These results confirm that while lower sample sizes generally increase variance, higher RSIϕ in majority-Black CBGs remains evident across device-penetration strata. Figure 3e, f map the spatial distribution of RSIϕ in Atlanta and St. Louis, with each circle representing a CBG. The colour again indicates the CBG majority-group category while the size of each circle reflects the magnitude of RSIϕ, i.e. the variability in daily mobility timing. A clear spatial pattern emerges in both cities: majority-Black CBGs (red) consistently exhibit larger circles, indicating higher variance in mobility timing. These CBGs are spatially clustered, especially in north St. Louis and south and west Atlanta. In contrast, majority-White CBGs (blue) are associated with smaller circles, indicating more stable daily rhythms, and tend to cluster in different parts of each city. This visual reinforces earlier findings by showing that higher RSIϕ values are spatially clustered and often align with CBGs that have higher proportions of Black residents. To further examine the structural roots of the temporal instability revealed by our model, we conducted a series of regression analyses using a comprehensive set of socio-economic and demographic variables (Supplementary Tables 12–13). Figure 3g summarises the results across five progressively enriched models. A consistent finding is that the proportion of Black residents remains positively and significantly associated with higher RSIϕ, even after controlling for built environment and economic factors. This association suggests that CBG-level racial composition remains related to rhythm variability beyond the included income and built-environment measures. The positive associations with poverty, unemployment and low educational attainment indicate that CBG-level economic conditions are also associated with less stable daily activity timing. Together, these results show that higher RSIϕ is associated with intersecting CBG-level socioeconomic conditions rather than individual preference alone. Pandemic stress test While our analysis has thus far characterised the stable rhythms of 2019, extreme events serve as critical stress tests for the model’s generalisability. We validated the STASIS framework against multiple acute crises, including winter storms, hurricanes and wildfires (see Supplementary Note 8), confirming its ability to capture sudden ruptures in daily life. Among these crises, the COVID-19 pandemic provides a prolonged period of widespread public health recommendations and policy restrictions that substantially reduced mobility in many settings. This period changed the usual relationship between mobility and stationarity: staying at home became a recommended or required public health behaviour in many jurisdictions. We use this period to examine whether changes in daily mobility rhythms differed across CBGs with different racial and Hispanic-origin compositions. The aggregate impact of the crisis is visible in the transformation of the daily pulse (Fig. 4a). Following the national emergency declaration in the USA in March 2020, the rhythm between mobility and stationarity was disrupted significantly. The post-emergency curve (red) illustrates a pronounced increase in daytime stationarity, reflecting a broad reduction in discretionary movement. This contraction is further contextualised in the temporal heatmap (Fig. 4b), which captures the immediate change following March 13, 2020. The transition from low stationarity (green) to high stationarity (deep pink) across weekdays signals a sudden increase in home presence following the emergency declaration and related public health restrictions. The gradual return toward pre-pandemic levels by September and October motivates analysis of how these changes varied across CBG categories. Changes in stationarity rhythms during COVID-19 in 2020. a Fitted weekday curves for devices at home per hour: grey lines represent individual working days (n = 253 in total), with blue and red lines indicating the mean fits before (n = 49) and after (n = 204) the national emergency, respectively. b Daily mean R(t) over 24 h in selected months of 2020. c, e, g Values and changes in 2A, ϕ and B during weekdays by CBG racial/Hispanic-origin category. d, f, h Violin plots comparing parameter changes between 2019 (blue) and 2020 (red). Disaggregating these shifts by racial and Hispanic-origin composition (Fig. 4c–h) shows different trajectories across CBG categories. First, Fig. 4c, d show a decline in range (2A) immediately following the declaration. This contraction was largest in majority-Black CBGs. Majority-Black CBGs, which already had the lowest mobility range pre-pandemic, saw their mean value decline to 0.136 (95% CI: 0.133–0.139), compared to 0.190 (95% CI: 0.185–0.196) for majority-White CBGs and 0.179 (95% CI: 0.175–0.183) for majority-Hispanic CBGs. This indicates a larger reduction in the observed daily activity range39,40. Following the onset of the pandemic, the daily rhythm shifted later for all groups, with a larger delay in majority-Black CBGs, where the average phase angle shifted by 25.5 (95% CI: 25.3–25.7) minutes. The median delay was 30.5 min (Fig. 4f), higher than in the other CBG categories. This larger phase shift is consistent with higher RSIϕ values and may reflect irregular essential-work schedules during the pandemic period. The baseline stationarity (B) also changed unevenly across CBG categories. Although stay-at-home recommendations and restrictions were widespread, Fig. 4g, h shows that majority-Black CBGs experienced the smallest increase in device-based home presence. Pre-to-post pandemic distributions show an upward shift in stationarity for majority-White CBGs (0.461 (95% CI: 0.458–0.464) to 0.495 (95% CI: 0.491–0.498)) and majority-Hispanic CBGs (0.458 (95% CI: 0.456–0.461) to 0.491 (95% CI: 0.488–0.494)), whereas majority-Black CBGs showed a smaller increase (0.426 (95% CI: 0.424–0.428) to 0.444 (95% CI: 0.441–0.448)). This attenuated response indicates lower observed increases in home presence in majority-Black CBGs during the pandemic period. These findings suggest that pandemic-period changes in mobility rhythms were associated with pre-existing structural conditions. These divergent rhythms are likely associated with occupational composition and lack of telework feasibility in CBGs with higher shares of minority populations36. Majority-White CBGs showed larger increases in baseline (B) and maintained greater rhythm stability, whereas majority-Black CBGs showed smaller increases in protective stationarity, higher temporal instability (ϕ) and greater spatial constriction (2A) during the study period37,41. Consequently, the pandemic clarified pre-existing differences in the distribution of motility. The lower increase in protective fixity in majority-Black CBGs suggests that the ability to increase home presence during a public health crisis depends partly on structural conditions such as work requirements, housing conditions and transport access42,43. The persistence of these gaps late into 2020 suggests that they reflect longer-term structural constraints rather than only short-term disruption. Discussion Our study, while effectively operationalising the dialectic of mobility and stationarity to reveal hidden temporal disparities, shares certain limitations common to research relying on large-scale passive data collection44–47. First, although our primary dataset represents a significant cross-section of the population in the USA, it remains a proxy derived from mobile devices and is subject to sampling biases. We addressed this by incorporating demographic re-weighting and, crucially, by validating the STASIS model across five diverse metropolitan areas in four countries. This cross-regional robustness suggests that the core parameters—range, phase and baseline—capture broad properties of urban life rather than artefacts of a single dataset. Second, our analysis operates at the spatial resolution of CBGs in the USA without considering the option of virtual mobility. While this scale effectively captures CBG-level structural inequalities, it inevitably smooths over within-CBG heterogeneity, potentially obscuring how individual-level constraints (such as gendered caregiving duties or specific employment types) manifest within the same spatial unit. Third, this reliance on aggregate data precludes a full disaggregation of non-employed statuses beyond formal unemployment—such as retirement, homemaking, or disability. As a substantial body of transport research demonstrates, these statuses strongly shape distinct temporal activity structures and spatial participation across the life course48–51. Consequently, part of the residual racial disparity observed in our models may reflect this unobserved heterogeneity in CBG-level activity compositions, with race partially capturing underlying differences in activity profiles. However, the persistence of a statistically significant race effect after controlling for unemployment indicates that these rhythmic disparities cannot be reduced entirely to labour-market participation. Rather, they reflect broader, systematic inequalities in mobility opportunities and transport accessibility documented across racialized groups52,53. Finally, while the STASIS model successfully captures the signal of temporal instability (RSIϕ), the current analysis remains primarily descriptive and diagnostic. Our framework identifies where and when rhythm ruptures occur, but specific causal mechanisms—such as gig-economy labour laws, transit scheduling, or housing eviction policies—cannot be directly inferred from the current cross-sectional design. Future research integrating individual-level surveys, longitudinal data or quasi-experimental designs is required to definitively disentangle these structural constraints from individual activity composition. Furthermore, it is important to distinguish the empirical measurements produced by the STASIS framework from their broader theoretical interpretation. Empirically, the model provides a parsimonious representation of daily stationarity rhythms through three parameters—amplitude, phase angle and baseline—together with the RSIϕ, enabling systematic and scalable comparison of rhythmic patterns across CBGs and urban systems. In this sense, the framework functions primarily as a descriptive and comparative measurement tool for characterising temporal patterns of presence and absence in urban space. Concepts such as motility, mobility inequality, or temporal injustice therefore operate as interpretative lenses through which these measured disparities can be situated within wider debates on time-space constraints, differential access to mobility resources and socially structured rhythms of everyday life6,12,16,54. From this perspective, the present analysis identifies where and when systematic rhythm disruptions occur across the urban system, while the specific causal mechanisms underlying these patterns—such as labour-market institutions, transport service provision, residential segregation, or land-use constraints—cannot be directly inferred from the current cross-sectional design. Despite these limitations, our analyses contribute to the measurement of temporal patterns in human mobility. First, we introduce a unified framework that complements existing research on activity spaces and multicontextual segregation by linking theoretical insights from the new mobilities paradigm5,6 with computational mobility analysis. By operationalising the dialectic of movement and moorings through the lens of statistical physics and big data, we connect sociological concepts of movement and moorings with measurable temporal mobility signals. While traditional frameworks have excelled at describing spatial flows (e.g. gravity models, origin-destination matrices)55,56 or stochastic individual trajectories (e.g. random walks, Lévy flights)57,58, they often treat stationarity as a zero-value residual—a mere absence of movement. By contrast, our STASIS model treats the temporal structure of fixity as a primary, structured signal, aligning with time-geographic perspectives that view daily life as a rhythmic oscillation constrained by physiological needs and institutional authority2,7. By formalising these daily cycles through a continuous sinusoidal function, our framework shifts the analytical focus from where people go to when and how consistently they return, offering a scalable method to examine temporal architectures of inequality that aggregate flow models obscure. The three parameters of the STASIS model show how CBG-level structural conditions are associated with temporal patterns of stationarity. The smaller mobility range (2A) observed in majority-Black CBGs provides an additional temporal measure relevant to theories of spatial mismatch59,60. Rather than indicating a limitation of physical distance alone, this compressed rhythm suggests that barriers to transportation and accessibility61–63 may be associated with reduced differentiation between daytime activity and nighttime home presence. The lower baseline (B) indicates lower average home presence in CBGs with higher shares of minority populations, which may be related to multiple jobs, extended labour hours, or other constraints on sustained residential stationarity. We interpret the phase angle (ϕ) and Rhythm Stability Index (RSIϕ) as descriptive measures of activity timing and its variability. A later phase angle indicates later timing of peak activity relative to standard institutional schedules24. A higher RSIϕ indicates greater variability in this timing. Together, ϕ and RSIϕ provide a CBG-level measure of temporal instability that can be compared with CBG-level socioeconomic conditions and used to examine temporal precarity as a structural pattern rather than an individual time-management outcome. Finally, while we validated the STASIS model across multiple acute extreme events to ensure generalisability, the COVID-19 pandemic provides a prolonged systemic shock that changed daily mobility rhythms. The analysis indicates that changes in range (2A), phase (ϕ) and baseline (B) differed across CBGs with different racial and Hispanic-origin compositions. In majority-Black CBGs, simultaneous spatial constriction (2A), temporal destabilisation (ϕ) and a smaller rise in protective fixity (B) identify a pattern consistent with reduced capacity to increase home presence during public health restrictions. These findings move beyond general observations of disproportionate impact37,41 by identifying specific mobility-rhythm parameters associated with those differences. These findings suggest that urban policy can incorporate temporal as well as spatial dimensions of mobility. Transport equity could address delayed phase angle (ϕ) by expanding high-frequency service during off-peak hours for workers with non-traditional schedules. Housing policy could support residential stability through anti-displacement measures that help maintain baseline stationarity (B). During public health crises or natural disasters, resource allocation could account for CBGs with high RSIϕ, where fragmented routines may increase vulnerability to sudden disruptions. Labour protections such as fair-workweek laws may also help stabilise shift scheduling64. Methods Data This study utilises SafeGraph mobility data to analyze human stationarity dynamics. We leverage several key fields from the SafeGraph dataset. origin_census_block_group: records the geographic location using the 12-digit Federal Information Processing Standards code for CBGs. date_range: specifies the temporal coverage of each record. device_count: denotes the number of devices observed within their home CBG on a given day. at_home_by_each_hour: a 24-element array indicating the number of devices at home during each hour of the day (local time). The first element corresponds to 00:00–1:00 AM. In our analysis, we use data corresponding to each hour from 00:00 to 23:00. completely_home_device_count: represents the number of devices that did not leave their home location throughout the entire day. We also calculate the total number of work_device_count by summing the following three variables: full_time_work_behavior_devices, part_time_work_behavior_devices, and delivery_behavior_devices. delivery_behavior_devices identifies devices that stopped at three or more locations for over 20 min each within a single day. In addition, demographic and socioeconomic characteristics at the census block group level are obtained from the 2015–2019 American Community Survey (ACS), provided by the United States Census Bureau (2021). Key variables include pct_NH_White_alone_ACS_15_19, pct_Hispanic_ACS_15_19, pct_NH_Blk_alone_ACS_15_19, among others. To ensure consistency in daily mobility patterns, we excluded data corresponding to weekends and federal holidays in the USA, including New Year’s Day, Martin Luther King Jr. Day, Presidents’ Day, Memorial Day, Independence Day, Labor Day, Columbus Day, Veterans Day, Thanksgiving Day and Christmas Day. Our analysis is therefore based exclusively on weekdays throughout the calendar year, to minimise the influence of atypical mobility behaviour observed during weekends and public holidays. STASIS model with angular frequency To further examine the temporal dynamics of human stationarity, we extend the main model by incorporating an angular frequency term to characterise rhythmic variations over time. The stationarity sinusoidal stability (STASIS) model with angular frequency is defined as: In Eq. (2), ω represents the angular frequency, which governs the rate of fluctuation in the mobility pattern, i.e. the time interval between the maximum and minimum values of the human mobility ratio. The angular frequency ω captures the periodicity of mobility changes. Specifically, ω < 1 corresponds to a cycle duration longer than 24 h, indicating slower variations in daily mobility, while ω > 1 corresponds to a cycle shorter than 24 h, reflecting more rapid shifts. This extension allows us to capture subtle intra-day mobility rhythms that are not evident in the baseline model. A detailed illustration of this extended model and its empirical fitting across selected study sites is provided in Supplementary Figs. 46–49. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Acknowledgements The authors thank Mario L. Small and Albert-László Barabási for insightful discussions. Author contributions Z.T., C.Z., L.T. and Q.R.W. conceived the study and designed the research. Z.T. and C.Z. developed the theoretical model and performed the numerical analysis. Z.T., C.Z., W.Y.L., M.M.D. and W.S.L. processed and analysed the human mobility data. Z.T., C.Z. and Q.R.W. wrote the initial draft of the manuscript. L.T. and Q.R.W. supervised the project. All authors contributed to the discussion of the results and reviewed and approved the final manuscript. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding Z.T. discloses support for the research of this work from the National Natural Science Foundation of China [grant number 72401127]. L.T. discloses support for the research of this work from the Major Program of the National Social Science Foundation of China [grant number 22&ZD136], the National Natural Science Foundation of China [grant number 72174091] and the National Key R&D Program of China [grant number 2020YFA0608601]. W.Y.L. discloses support for the research of this work from the National Natural Science Foundation of China [grant number 72504198], the Natural Science Foundation of Jiangsu Province [grant number BK20240987] and the Natural Science Foundation of the Jiangsu Higher Education Institutions [grant number 23KJB120012]. C.Z., M.M.D., W.S.L and Q.R.W. declare no relevant funding. Data availability The mobility data used in this study are proprietary and were obtained from SafeGraph (https://www.safegraph.com). These data are subject to licensing restrictions and are not publicly available. Access to the data can be requested through the SafeGraph data access program (https://www.safegraph.com) and requires a valid data use agreement. CBG-level socio-demographic variables, including race, age and education, were derived from the United States Census Bureau’s American Community Survey (ACS) 2015–2019 5-year estimates (https://www.census.gov/programs-surveys/acs). Built environment metrics were obtained from the United States Environmental Protection Agency’s Smart Location Database (Version 3.0) (https://www.epa.gov/smartgrowth/smart-location-mapping). The processed and aggregated datasets underlying all figures and results in this study, along with the source data for the main figures, are publicly available via GitHub at https://github.com/zc16423/stationarity_mobility_rhythm_analysis. Code availability The computer code used to perform the analyses and generate the results presented in this study has been deposited in the Zenodo repository with the Digital Object Identifier 10.5281/zenodo.1943805465. The development version of the code is also available on GitHub at https://github.com/zc16423/stationarity_mobility_rhythm_analysis. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Zihao Tian, Chao Zhang. Contributor Information Chao Zhang, Email: zhangchao16423@126.com. Lixin Tian, Email: tianlx@ujs.edu.cn. Qi R. Wang, Email: q.wang@northeastern.edu Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-75778-z. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Transparent Peer Review file Data Availability Statement The mobility data used in this study are proprietary and were obtained from SafeGraph (https://www.safegraph.com). These data are subject to licensing restrictions and are not publicly available. Access to the data can be requested through the SafeGraph data access program (https://www.safegraph.com) and requires a valid data use agreement. CBG-level socio-demographic variables, including race, age and education, were derived from the United States Census Bureau’s American Community Survey (ACS) 2015–2019 5-year estimates (https://www.census.gov/programs-surveys/acs). Built environment metrics were obtained from the United States Environmental Protection Agency’s Smart Location Database (Version 3.0) (https://www.epa.gov/smartgrowth/smart-location-mapping). The processed and aggregated datasets underlying all figures and results in this study, along with the source data for the main figures, are publicly available via GitHub at https://github.com/zc16423/stationarity_mobility_rhythm_analysis. The computer code used to perform the analyses and generate the results presented in this study has been deposited in the Zenodo repository with the Digital Object Identifier 10.5281/zenodo.1943805465. The development version of the code is also available on GitHub at https://github.com/zc16423/stationarity_mobility_rhythm_analysis.