Assimilation or Segregation? Evolutionary Trajectories and Driving Forces of Chinese Immigrant Residential Concentration in Seoul, South Korea

Abstract

The spatial distribution of immigrants and associated patterns of residential segregation and integration can manifest not only at the metropolitan scale but also at finer micro-spatial resolutions, reflecting the interaction between path dependence and structural reconfiguration. This article examines the micro-spatial residential patterns of Chinese immigrants in Seoul under institutional and market constraints. Using a Spatial Durbin Model and Multiscale Geographically Weighted Regression, it shows that from 2011 to 2025, immigrant settlements shifted from a monocentric pattern to a polycentric, functionally differentiated, and networked structure. While overall spatial embeddedness is high and segregation remains low, traditional cores such as Guro–Daerim persist. Selective clustering is shaped by path-dependent migrant networks, urban redevelopment policies, and intra-group differentiation, while infrastructure homogenization renders transportation accessibility a background condition. The findings support segmented assimilation theory in high-density East Asian cities and underscore the importance of incorporating immigrant needs into urban policy to promote inclusive integration.

1. Introduction

Against the backdrop of accelerating globalization and intensifying cross-border population mobility, urban migrant structures worldwide are undergoing profound restructuring [1]. According to the United Nations, the global international migrant population increased from 154 million in 1990 to 304 million in 2024, accounting for approximately 3.7% of the world’s total population [2]. The presence of migrant populations is inherently a spatial phenomenon, characterized by pronounced geographical disparities across different territorial contexts [3]. Migration has thus become a key force shaping contemporary urban demographic change. This process, however, does not unfold evenly across urban space. Differentiated housing markets, segmented labor structures, and uneven institutional contexts give rise to highly heterogeneous residential patterns both between and within cities [4,5,6,7].

At the macro level, global economic and political dynamics primarily shape the scale and destinations of population flows, whereas at the local level, societies influence migrants’ residential location choices and the degree of their spatial integration or segregation through the formulation of integration regimes and governance rules [8]. Examining residential assimilation and segregation from a migrant perspective is therefore not only central to understanding migrants’ living conditions but also provides a critical lens for interpreting urban socio-spatial structures and their underlying mechanisms of change [9].

Residential segregation of migrants is commonly defined as the extent to which migrants are spatially separated from [6,10], or co-reside with, other migrant groups or native residents at the neighborhood level, and its formation is jointly influenced by ethnic attributes, cultural distance, socioeconomic status, and institutional environments [3]. In contrast, spatial assimilation theory posits that ethnic or racial segregation in cities follows a staged trajectory: as migrants’ cultural adaptation deepens and their socioeconomic status improves, they gradually relocate to higher-quality neighborhoods and increasingly share urban space with mainstream populations [10,11,12]. From this perspective, assimilation should not be understood merely as the outcome of individual migrant behavior or cultural adaptation [13], but rather as a process embedded within national governance structures, functioning as an important mechanism through which population distribution is regulated via housing policies, urban planning, and institutional arrangements [14,15,16]. While these perspectives provide complementary analytical frameworks, empirical research consistently demonstrates that residential trajectories and their determinants vary substantially across migrant groups, housing regimes, and institutional contexts [15,16].

Since the early 2000s, South Korea has rapidly transformed from a traditional migrant-sending country into an emerging migrant-receiving destination. Within this process, Seoul has become the primary migrant concentration area due to its dense employment opportunities, public resources, and social networks [17,18,19,20]. Official statistics indicate that the foreign population in South Korea increased from approximately 65,000 in 1992 to about 1.49 million in 2024, representing nearly 6% of the national population. Among Seoul’s foreign residents, Chinese migrants constitute the largest and most internally differentiated group. This population includes ethnic Korean Chinese, who share linguistic and cultural affinities with Korean society, as well as non-Korean Chinese migrants from mainland China [21,22]. In recent years, the proportion of ethnic Korean Chinese has declined, while the size of the non-Korean Chinese population has remained relatively stable, indicating an internal restructuring of the Chinese migrant population with potential implications for residential demand and spatial behavior [21].

Early spatial assimilation models proposed that ethnic residential clustering would gradually weaken over time [23,24]. Subsequent research, however, has demonstrated that this trajectory is neither universal nor linear, especially under conditions of housing market stratification, institutional constraints, and persistent cultural differentiation [25,26,27,28,29]. In highly differentiated housing markets of global cities, processes such as urban redevelopment and gentrification further complicate residential adjustment, thereby limiting the explanatory power of a single assimilation framework [11,25,30,31,32]. Despite methodological advances, empirical evidence remains heavily concentrated in traditional immigrant-receiving regions such as North America, Europe, and Australia [33,34,35,36,37,38], whereas systematic studies focusing on emerging migrant destinations in East Asia remain relatively scarce [19,22,39].

In South Korea, the widespread prevalence of the Jeonse rental system and the relatively high cost of apartment housing compared to older detached housing appear to encourage immigrants’ residential concentration in aging urban neighborhoods. Consequently, assimilation patterns in apartment-dominated areas are weak, calling into question the applicability of suburbanization-based assimilation pathways commonly identified in Western urban research. Drawing on theories of residential assimilation and spatial segregation, this study systematically examines the formation mechanisms and evolutionary characteristics of migrant spatial clustering in East Asian cities, thereby addressing the Western-centric bias and the relative neglect of East Asian institutional contexts in existing research. Using Seoul, South Korea, as a case study, the paper focuses on the spatial distribution, temporal evolution, and institutional drivers of residential patterns among Chinese migrants.

The contributions of this study are threefold. First, it extends research on migrant residential assimilation and spatial segregation beyond its predominant focus on Western countries to East Asian urban contexts, providing new regional empirical evidence for existing theories. Second, by incorporating residential environments, public service provision, socioeconomic conditions, and transport accessibility into the analysis of migrant spatial clustering, the study reveals mechanisms shaping migrant residential patterns that are distinctive to East Asia. Third, within South Korea’s specific institutional setting, the study empirically examines the manifestations of spatial assimilation and spatial segregation in the residential patterns of Chinese migrants, thereby deepening understanding of the mechanisms underlying migrant socio-spatial differentiation across diverse institutional contexts.

In response to these theoretical debates and empirical gaps, this study examines the evolution of Chinese migrant residential patterns in Seoul and the structural factors shaping their spatial configuration. Specifically, it investigates whether Chinese migrant settlement patterns exhibit increasing dispersion consistent with spatial assimilation or continue to display sustained spatial concentration. The study addresses the following questions: How have Chinese migrant residential distributions in Seoul changed over time? To what extent do residential environments, public service provision, socioeconomic conditions, and transport accessibility influence these patterns? What forms of spatial assimilation and spatial segregation characterize the spatial clustering of Chinese migrants? Methodologically, the analysis integrates GIS-based spatial techniques with spatial econometric models to capture temporal and spatial variation in migrant settlement at the neighborhood (dong) level, while accounting for South Korea’s distinctive housing system and residential structure [40]. By empirically examining spatial assimilation arguments within an East Asian urban context, this study contributes new evidence to the understanding of migrant residential restructuring in emerging migrant-receiving cities.

2. Literature Review

2.1. Residential Segregation Theory

Residential segregation refers to the persistent spatial separation of social groups within systems of social stratification and urban space [41,42]. Its formation is commonly understood as the outcome of interacting institutional arrangements, housing and labor market mechanisms, and group-based residential preferences [43,44]. From a structural perspective, access conditions in housing markets, labor market segmentation, and the allocation of public resources shape the likelihood that different social groups can enter residential locations [27,34,39].

At the same time, immigrants actively participate in residential decision-making, often selecting locations based on cultural affinity, language use, and access to co-ethnic social networks [31,43]. Consequently, observed segregation patterns reflect the combined effects of economic resources, institutional constraints, and ethnic or social identities operating within urban space [45,46].

Immigrant residential segregation thus represents not only a spatial expression of social differentiation but also an important mechanism through which social inequality is reproduced in cities. For this reason, segregation has been widely employed as an indicator of spatial justice and social inclusion in urban studies [41,46,47]. In the context of accelerating transnational migration, ethnic enclaves are often interpreted as initial entry points and transitional spaces that facilitate immigrants’ incorporation into urban societies [47,48]. When facing linguistic barriers and limited access to employment information, newly arrived immigrants frequently rely on ethnic communities to secure jobs and social support, enabling the accumulation of early social capital [25,47,49]. However, when ethnic concentration becomes spatially entrenched and persists over time, it may restrict everyday interaction with the mainstream population and contribute to the long-term reproduction of residential inequality [48,49,50].

Comparative research further demonstrates that the mechanisms producing residential segregation are highly context-dependent. The concept of self-segregation emphasizes the active role of ethnic identity and cultural preferences in sustaining enclave formation [51]. Empirical evidence from migrant labor-oriented cities shows that occupational stratification closely overlaps with residential patterns, leading to the spatial concentration of low-skilled migrants [52]. Studies from Spain and Germany similarly indicate that regional economic conditions and labor market structures strongly influence immigrants’ initial residential choices, while subsequent mobility is increasingly shaped by levels of ethnic concentration and accumulated community social capital [53,54]. Overall, residential segregation emerges from the sustained interaction of institutional settings, market environments, and group-based residential strategies within specific spatial and temporal contexts [27].

2.2. Spatial Assimilation Theory

Spatial assimilation theory constitutes a foundational framework for understanding immigrant residential mobility and social incorporation. Rooted in the Chicago School tradition, it conceptualizes residential relocation as a spatial manifestation of upward socioeconomic mobility [26,55,56]. As immigrants spend more time in the host society, improvements in economic resources, language proficiency, and social networks are expected to facilitate residential convergence toward mainstream populations [26,57,58,59]. Empirical evidence from traditional immigrant-receiving countries suggests that this process is often associated with declining levels of ethnic segregation, increased intergroup contact, and enhanced social integration [60,61,62,63,64].

More recent studies, however, point to important limitations and variations in spatial assimilation trajectories. Persistent intergenerational neighborhood inequality among immigrant minorities has been documented even in welfare-state contexts, suggesting that spatial convergence may stall [26]. Housing tenure regimes have also been shown to play a decisive role in shaping migrant residential mobility pathways in highly regulated housing markets [32]. As migration streams become increasingly diverse and urban housing systems more segmented, the assumption of a single linear path of spatial assimilation has been widely questioned [58,65].

A growing body of research further indicates that socioeconomic advancement does not necessarily translate into improved residential outcomes [66,67,68,69]. Housing supply systems and urban planning institutions function as gatekeeping mechanisms that structure neighborhood access and may reinforce residential inequality [70,71,72]. Building on these insights, segmented assimilation theory proposes that immigrants may follow divergent pathways—including upward mobility, persistent marginalization, or selective integration that preserves ethnic identity—depending on institutional contexts, ethnic stratification, and available social capital [30,72,73]. From this perspective, spatial assimilation is not an inevitable outcome, but a contingent process shaped by local housing regimes and institutional conditions.

2.3. Applicability of Theory in High-Density East Asian Urban Contexts

Most theories of immigrant residential choice and spatial differentiation were developed in the context of low-density, automobile-oriented cities in Europe and North America. In contrast, high-density East Asian megacities such as Seoul are characterized by compact built environments, housing systems dominated by ownership, distinctive rental institutions, and strong state involvement in urban redevelopment [40,73,74]. These structural conditions fundamentally reshape residential opportunities and constrain the direct applicability of Western-based theoretical frameworks.

Housing supply structures and rental systems exert a particularly strong influence on immigrant residential choices. In South Korea, the Jeonse rental system requires substantial lump-sum deposits, creating high entry barriers for low-income households [75]. At the same time, the prevalence of large-scale branded apartment complexes further elevates housing prices and deposit requirements [76,77]. Under these constraints, many immigrants are excluded from mainstream housing markets and instead concentrate in older single-family houses and low-rise multifamily dwellings, forming dense immigrant residential clusters with relatively limited living conditions [73,78].

Urban redevelopment practices have also exerted sustained effects on immigrant residential patterns. Since the 1960s and 1970s, large-scale reconstruction and regeneration projects in South Korea have contributed to improvements in the physical environment and the expansion of housing supply [74]. However, rapid increases in land and housing prices during redevelopment frequently exceed the affordability thresholds of low-income residents, resulting in displacement. Compared with Western cities, immigrant residential clusters in high-density East Asian cities are more prone to spatial relocation under redevelopment pressure, prompting immigrants to seek lower-cost alternatives within the urban core or to relocate toward metropolitan peripheries [73,79,80]. These dynamics indicate that residential segregation and spatial assimilation in East Asian cities are deeply shaped by housing institutions and redevelopment regimes, underscoring the need for contextualized theoretical adaptation rather than direct application of Western models.

These dynamic processes indicate that residential segregation and spatial assimilation in high-density East Asian cities are not driven solely by socio-cultural factors but are instead jointly shaped by housing system arrangements and housing market structures. Accordingly, analyses of migrant spatial clustering and socio-spatial differentiation should systematically incorporate housing-related variables, such as housing types and rent levels, into the analytical framework. Moreover, it is necessary to examine the formation mechanisms and evolutionary trajectories of residential clustering among specific nationality groups at finer spatial scales, in order to more accurately capture the dynamics of migrant settlement patterns.

2.4. Immigrant Residential Concentration in Seoul

Since the mid-1990s, the continuous growth of the foreign-born population has accelerated South Korea’s transition toward a multicultural society [80,81,82]. In this context, the spatial distribution of foreign residents has gradually emerged as a key issue in urban geography and population studies. Existing research generally indicates that the level of residential segregation among immigrants in South Korea has reached, and in some cases exceeded, that observed in several developed countries, with evidence pointing to a continued intensification over time [82]. Early studies consistently show that approximately two-thirds of the foreign population is concentrated in the Seoul Capital Area, underscoring the close relationship between foreign population growth, diversification of migrant types, and the evolution of urban spatial structure [83].

Building on this line of inquiry, an increasing number of studies have focused on the city of Seoul to examine the residential distribution of foreign populations. Drawing on analytical frameworks of residential segregation and socio-spatial differentiation, this body of literature documents spatial patterns across different nationalities and visa categories [80,84]. Quantitative indicators, most notably the Dissimilarity Index, are widely applied to measure nationality-based residential segregation. Empirical findings consistently demonstrate that foreign populations in Seoul exhibit highly differentiated spatial distributions across nationality groups [84]. These patterns are further shaped by cross-national differences in chain migration dynamics, spatial diffusion trajectories, and processes of residential differentiation over time. In addition, existing evidence suggests that levels of residential segregation are systematically associated with the economic development of migrants’ countries of origin [80].

Regarding spatial scale and analytical design, several studies have emphasized the importance of adopting finer spatial units, such as dong, to capture the morphology and formation mechanisms of foreign population concentrations more accurately [81]. Nevertheless, most empirical analyses continue to rely on districts as the primary unit of observation [85]. At this scale, studies typically incorporate a set of socioeconomic and spatial indicators, including total population size, foreign population counts, visa categories, nationality composition, gross regional domestic product, and the spatial distribution of industrial complexes, to explain variations in foreign population distribution [12,85]. Findings from this line of research indicate that residential patterns of foreign populations are shaped not only by nationality and migration purpose, but also by local economic opportunity structures and housing-related conditions [19,85,86,87].

Empirical research further reveals clear patterns of social stratification in the residential distribution of foreign populations in Seoul. When foreign residents are classified by nationality, level of economic development in their countries of origin, or socioeconomic status, substantial differences emerge in residential location, degree of segregation, and spatial stability [12,86]. Foreign residents from economically developed countries tend to concentrate in spatially limited areas characterized by better housing conditions and higher residential quality [81,84]. By contrast, migrants from less developed or lower- and middle-income countries are more widely dispersed and form multiple nationality-based residential clusters within the city [12,82]. These stratified patterns highlight the combined influence of economic conditions, housing structures, and institutional arrangements on foreign population clustering [82].

Overall, while existing studies have provided robust empirical evidence on the general spatial distribution and social stratification of foreign populations in Seoul at macro- and meso-spatial scales, research focusing on a single nationality group and examining dong-level clustering patterns and their underlying mechanisms remains limited.

3. Data and Methodology

3.1. Data Sources and Processing

3.1.1. Data Sources and Population Definitions

This study integrates four complementary data domains: demographic characteristics, built environment attributes, spatial distribution of public service facilities, and transportation network features. Population data and core geospatial layers—including administrative boundaries, road networks, and building footprint vectors—were obtained from the Seoul Open Data Plaza (https://data.seoul.go.kr/, accessed on 27 August 2025) and the National Spatial Information Platform of South Korea (https://www.vworld.kr/, accessed on 27 August 2025). These datasets are compiled and released by governmental agencies and exhibit high spatial precision and temporal consistency, providing a robust empirical basis for dong-level spatial analysis.

Chinese population data for Seoul were assembled from two primary sources: administrative population registers for 2011 and 2016, and ambient population estimates derived from anonymized mobile phone data for 2021 and 2025. Administrative registers reflect legally declared residential addresses, whereas ambient population estimates capture observed spatial presence and activity patterns. Integrating these sources implies a shift in the underlying population concept—from de jure residence to de facto spatial presence—which may introduce measurement inconsistencies, particularly for individuals whose registered residence differs from their primary activity locations [88]. Such definitional differences may affect the interpretation of absolute population magnitudes; therefore, the analysis focuses primarily on relative spatial patterns and spatial concentration measures, which are less sensitive to definitional discrepancies than absolute counts.

Long-term Chinese residents were defined as individuals residing in South Korea for more than 90 days, consistent with official statistical definitions and the Seoul living population estimation framework. In the mobile phone-based datasets, foreign residents were identified using anonymized foreign subscriber identifiers associated with holders of alien registration cards. Individuals classified as long-term residents had valid registration status and sustained presence exceeding 90 days, as inferred from longitudinal activity records. Although this classification approach aligns with administrative definitions, some misclassification between short-term and long-term residents may occur due to delayed registration, temporary departures, or anonymization constraints. These errors are expected to be limited and spatially random, and therefore unlikely to systematically bias neighborhood-level spatial patterns.

To approximate residential population distribution, a single temporal snapshot at 03:00 a.m. on Mondays was used. Prior mobility research consistently demonstrates that early morning hours most closely correspond to residential locations, as commuting and discretionary activities are minimal during this period [89,90]. Sensitivity analyses using alternative time points (02:00 and 04:00 a.m.) and midweek observations produced highly similar spatial distributions and model estimates, supporting the robustness of this temporal choice [89,90].

For years in which Chinese population data were available only at the gu-level, a population-weighted areal interpolation approach was employed to disaggregate district-level counts to the dong-level. Dong-level ambient population served as the weighting variable, under the assumption that the spatial distribution of long-term Chinese residents broadly follows overall residential population patterns within each district. This assumption may underestimate localized ethnic clustering or enclave formation, potentially attenuating fine-scale concentration estimates. Nevertheless, population-based weighting preserves underlying residential structure more effectively than uniform area-based allocation and reduces artificial spatial smoothing.

3.1.2. Spatial Units, Boundary Harmonization, and Data Processing

Administrative dong boundaries underwent several adjustments during the study period due to jurisdictional restructuring and boundary realignment. To ensure temporal comparability, all datasets were harmonized to the 2025 boundary system using spatial overlay and area-weighted interpolation. This boundary harmonization procedure may introduce spatial smoothing and minor allocation errors, particularly in areas experiencing substantial administrative reconfiguration; however, such effects are expected to be limited at the neighborhood scale and unlikely to affect analyses of relative spatial patterns.

To further mitigate biases associated with heterogeneous administrative unit sizes and the modifiable areal unit problem (MAUP), a regular 250 m × 250 m grid was constructed for complementary spatial analysis. This resolution follows the official Seoul grid-based living population estimation framework and represents a trade-off between capturing fine-grained spatial heterogeneity in Seoul’s high-density urban environment and minimizing random noise associated with excessively fine grids (e.g., 100 m). Sensitivity analyses using coarser grids (e.g., 500 m) produced consistent clustering patterns, indicating that the main findings are robust to grid resolution.

Spatial disaggregation was conducted using population-based weighting derived from dong-level ambient population estimates, rather than area- or housing-based proxies. This approach assumes that ambient population density is an appropriate proxy for residential distribution, particularly in high-rise, high-density urban environments where land area or building counts may poorly represent actual population distribution. Integer constraints were imposed using the largest remainder method to ensure consistency between disaggregated dong-level counts and original gu-level totals. This procedure can introduce minor rounding errors that propagate spatially; however, such errors are numerically small, spatially diffuse, and unlikely to materially affect spatial concentration measures or econometric estimates. All spatial datasets were projected into the KGD 2002 Central Belt 2010 coordinate reference system. Prior to model estimation, all continuous variables were standardized using Z-score normalization to remove scale effects and facilitate comparison of estimated coefficients.

3.1.3. Data Limitations

Several limitations should be acknowledged. First, administrative registers and mobile phone-based ambient population data are grounded in different population concepts, which may introduce measurement inconsistencies, particularly for highly mobile individuals or those with mismatched registered and activity locations. Second, the reliance on a single early-morning temporal snapshot may not capture atypical nocturnal activity patterns, although sensitivity analyses indicate limited influence on spatial outcomes. Third, population-weighted areal interpolation assumes that the spatial distribution of long-term Chinese residents broadly follows overall residential population patterns within each district, which may underestimate localized ethnic clustering and potentially attenuate fine-scale segregation measures. Fourth, rounding errors introduced by integer constraints may propagate spatially, although these errors are numerically small and spatially diffuse. Collectively, these limitations are unlikely to affect the identification of relative spatial patterns or the inference of dong-level drivers, but they should be considered when interpreting absolute population magnitudes.

3.1.4. Research Framework

This study develops an integrated analytical framework that combines spatial pattern identification, driving mechanism analysis, and assessments of spatial heterogeneity and assimilation, with the aim of characterizing the evolution of Chinese immigrant residential patterns in Seoul and elucidating their underlying formation mechanisms (Figure 1).

First, a dong-level spatial database was constructed using Chinese immigrant population data for four time points (2011, 2016, 2021, and 2025). Kernel density estimation, location quotient analysis, and standard deviational ellipse techniques were employed to identify clustering cores, degrees of concentration, and directional shifts over time, thereby providing an overall depiction of spatial pattern formation and evolution.

Second, based on the identified spatial configurations, a system of explanatory variables capturing residential environment, public services, economic structure, and accessibility was developed. All variables were standardized and tested for multicollinearity. Spatial dependence was assessed using Moran’s I, followed by the spatial autoregressive model (SAR), spatial error model (SEM), and Spatial Durbin model (SDM). Lagrange Multiplier (LM), Likelihood Ratio (LR), and Wald tests were jointly applied to select the optimal model, and model fit was further compared using the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and log-likelihood values. Finally, spatial effects were decomposed into direct and spillover effects using the partial derivatives method.

Finally, a multiscale geographically weighted regression (MGWR) model was employed to capture spatial variations in the effects of driving factors across multiple spatial scales. In addition, the Dissimilarity Index, Isolation Index, and Theil’s H Index were used to evaluate temporal changes in the degree of spatial assimilation, and robustness analysis was conducted.

3.2. Construction of the Variable System

3.2.1. Dependent Variable

Immigrant residential concentration intensity was measured using the proportion of the Chinese population relative to the total population within each spatial unit ($Y_i$), calculated as follows:

$$Y_i={\displaystyle \frac{C_i}{P_i}}\times 100\%$$

(1)

where $C_i$ denotes the number of Chinese immigrants within administrative dong i, and $P_i$ represents the total population of the same unit. This indicator directly captures the relative concentration and visibility of immigrants in urban social space and has been widely used in empirical studies of ethnic residential segregation.

3.2.2. Explanatory Variables

To systematically examine the mechanisms underlying the formation of residential patterns among Chinese immigrants in Seoul, this study draws on residential location choice theory and the spatial assimilation framework. Explanatory variables were constructed across four dimensions: residential environment, public services, economic sectors, and accessibility. In total, 15 indicators were selected to capture the multidimensional determinants of immigrant residential concentration at the dong-level (Table 1).

(1)

Residential environment

Immigrants’ residential choices exhibit strong path dependence [91]. Previous studies indicate that established immigrant concentrations shape subsequent settlement decisions by reducing information costs, providing social support, and reinforcing place-based identity [92]. Such path dependence reflects both long-term accumulations of ethnic settlement and short-term spatial adjustments driven by recent migration flows. Accordingly, it is essential to capture immigrant settlement patterns from both historical and dynamic perspectives when examining the role of residential environments in shaping location choices.

To operationalize these mechanisms, two indicators were included: immigrant stock and immigrant flow [91]. Immigrant stock represents the size of the immigrant population within a given spatial unit at a specific point in time and reflects the long-term settlement foundation; specifically, the number of Chinese immigrants in each dong in 2011 was used as the baseline stock indicator. Immigrant flow captures recent changes in settlement patterns and reflects the capacity of residential environments to attract or repel new immigrants beyond existing concentrations; it was defined as the net change in the Chinese immigrant population between 2021 and 2025.

Beyond immigrant networks, housing market characteristics constitute key structural constraints on immigrants’ residential choices [14,93,94]. While higher housing prices are often associated with better accessibility and living conditions, they may impose affordability constraints on immigrants with limited economic resources. Within the Korean housing system, a higher proportion of apartments is typically associated with higher housing costs and middle-class-dominated communities, which tend to generate screening effects for newly arrived immigrants [94,95,96]. In contrast, older residential areas characterized by lower housing prices and more flexible entry conditions often function as initial settlement spaces for immigrant populations [94]. Housing prices, rental levels, housing age, and the proportion of apartments jointly reflect dong-level economic entry thresholds and social class composition.

To ensure comparability across rental contracts with heterogeneous deposit–rent structures, this study employed the Monthly Equivalent Rent Burden per Unit Area as the primary housing cost indicator. Tenant-paid security deposits were treated as opportunity costs rather than sunk payments. Accordingly, deposits were annualized using the prevailing Seoul housing market capitalization rate of 5.5% (2025) and converted into a monthly equivalent cost. This imputed cost was then added to the nominal monthly rent to derive the total effective monthly housing expenditure. The aggregated cost was normalized by leased floor area (m²), converted into pyeong using a factor of 3.3, and adjusted by a monetary scaling coefficient of 10,000. This approach effectively mitigates structural biases associated with high-deposit/low-rent and low-deposit/high-rent contracts, yielding a standardized measure of true rental prices per unit area.

(2)

Public service

The spatial distribution of public service resources reflects immigrants’ trade-offs between daily convenience and institutional support in residential decision-making. Public services may also indirectly influence immigrant settlement patterns by shaping housing prices and neighborhood upgrading processes [94,97,98]. Accordingly, three indicators were included: education facility density, medical facility density, and cultural facility density.

Education and medical facilities represent levels of human capital accumulation and basic living security [80,94]. High densities of such facilities are often associated with improved urban living conditions and increased housing values, which may generate affordability pressures for immigrants with limited economic resources [94,98,99]. In contrast, cultural facilities indicate the openness and diversity of public spaces and provide important venues for intercultural interaction and social integration [94,98].

(3)

Economic sectors

The spatial configuration of employment opportunities constitutes a fundamental determinant of immigrant residential location choices. From a jobs–housing balance perspective, industrial structure influences immigrant clustering patterns by shaping employment accessibility for labor forces with different skill levels and labor market positions.

Labor-intensive industry density and service industry density were used to capture the spatial distribution of employment opportunities characterized by relatively low entry barriers. Specifically, labor-intensive industry density refers to the concentration of factories, hazardous material storage and processing facilities, warehousing facilities, and transportation-related infrastructure. These sectors are typically associated with physically intensive work, low skill requirements, and strong employment orientation, thereby generating locational pull effects for low-skilled and employment-driven immigrant populations [95,96,97,98]. Service industry density encompasses commercial and consumer-oriented facilities, including retail and business establishments, entertainment venues, accommodation facilities, automobile-related services, and tourism and leisure facilities. Such sectors are characterized by flexible employment arrangements, relatively low language and credential requirements, and high labor demand, providing accessible job opportunities for immigrants facing linguistic and skill constraints. Together, labor-intensive and service-oriented industries constitute critical entry points into urban labor markets for newly arrived immigrants and are closely linked to the spatial concentration of ethnic economic activities [95,96,97,98].

In contrast, office facility density reflects the spatial concentration of high-end business activities and white-collar employment, which typically require higher levels of human capital, formal credentials, and language proficiency. This indicator allows for the examination of potential spatial associations between high-skill employment centers and immigrant residential patterns, particularly with respect to differentiated settlement strategies across immigrant skill groups [95,96,97,98].

(4)

Accessibility

Transportation conditions play a fundamental role in shaping immigrant residential decisions and daily spatial behaviors by influencing commuting costs, access to employment opportunities, and the spatial extent of daily activity spaces [99,100]. Public transit provision is a key determinant of accessibility for immigrant populations, who often exhibit lower levels of private car ownership. Well-developed transit systems reduce mobility constraints, enhance access to employment and public services, and support residential stability and neighborhood attachment [99,100]. Road network infrastructure further influences urban accessibility by shaping regional connectivity and facilitating spatial mobility. Higher road network density is commonly associated with expanded commuting ranges and urban spatial expansion, which may indirectly affect immigrant residential patterns through changes in accessibility and land-use structures [99,100].

Given that this study examines Chinese residential concentration at the dong-level, a 15 min walking threshold was adopted to delineate local activity spaces and approximate community-level daily mobility. This threshold has been widely applied in neighborhood-based activity space research and aligns with contemporary urban planning frameworks emphasizing walkable community life circles, such as the “15 min city” concept promoted in Seoul. To ensure robustness, sensitivity analyses using alternative walking-time thresholds were conducted. Based on average walking speed, a 15 min walk corresponds to an effective service area with a buffer radius of approximately 800 m. Accordingly, circular buffers with an 800 m radius were constructed to represent pedestrian-accessible areas. Public transit accessibility was operationalized as the number of subway stations and bus stops reachable within a 15 min walking distance, capturing the spatial availability of public transportation opportunities within residents’ daily activity spaces [101].

Road network accessibility was defined as spatial accessibility within a 30 min travel time, calculated using road design speed as the impedance factor. Travel speeds for different road classes were specified in accordance with the Regulations on the Structural and Facility Standards of Roads issued by the Korean Ministry of Land, Infrastructure and Transport (December 2021): 100 km·h⁻¹ for urban expressways, 80 km·h⁻¹ for arterial roads, 60 km·h⁻¹ for secondary arterial roads, 50 km·h⁻¹ for collector roads, and 40 km·h⁻¹ for local roads. For bicycle-exclusive lanes, a travel speed of 30 km·h⁻¹ was adopted based on the Regulations on the Structural and Facility Standards for Bicycle Facilities enacted in February 2017.

Table 1. Selection and quantitative measures of explanatory variables.

Category	Variable	Calculation Method	Source
Residential Environment	Migrant Stock	Total stock of Chinese immigrants (2011)	[91]
	Migrant Flow	Inflow of Chinese immigrants (2021–2025)	[91]
	Housing Price	Average listed housing price (KRW/m2)	[93,94,95]
	Rent Price	Monthly equivalent rent burden per m2 (KRW/m2)	[14,93,102]
	Housing Age	Average year of housing construction	[94,95]
	Apartment Ratio	Proportion of apartment units in total housing stock (%)	[94,95]
Public Services	Education Facility Density	Number of education facilities per km2	[94,98,99]
	Medical Facility Density	Number of medical facilities per km2	[94,99]
	Cultural Facility Density	Number of cultural facilities per km2	[94,98,99]
Economic Sectors	Labor-Intensive Industry Density	Number of labor-intensive enterprises per km2	[95,97,98]
	Service Industry Density	Number of service industry facilities per km2	[95,97,98]
	Office Facility Density	Number of office facilities per km2	[95,97,98]
Accessibility	Subway Station Accessibility	Spatial accessibility within a 15 min walking time, calculated as the number of subway stations reachable using a pedestrian road network, with walking time as the impedance.	[99,101,102]
	Bus Stop Accessibility	Spatial accessibility within a 15 min walking time, calculated as the number of bus stops reachable using a pedestrian road network, with walking time as the impedance.	[99,101,102]
	Road Network Accessibility	Spatial accessibility within a 30 min travel time, calculated using a road network, with road design speed specified by road class as the impedance.	[99,101,102]

Table 1. Selection and quantitative measures of explanatory variables.

Category	Variable	Calculation Method	Source
Residential Environment	Migrant Stock	Total stock of Chinese immigrants (2011)	[91]
	Migrant Flow	Inflow of Chinese immigrants (2021–2025)	[91]
	Housing Price	Average listed housing price (KRW/m2)	[93,94,95]
	Rent Price	Monthly equivalent rent burden per m2 (KRW/m2)	[14,93,102]
	Housing Age	Average year of housing construction	[94,95]
	Apartment Ratio	Proportion of apartment units in total housing stock (%)	[94,95]
Public Services	Education Facility Density	Number of education facilities per km2	[94,98,99]
	Medical Facility Density	Number of medical facilities per km2	[94,99]
	Cultural Facility Density	Number of cultural facilities per km2	[94,98,99]
Economic Sectors	Labor-Intensive Industry Density	Number of labor-intensive enterprises per km2	[95,97,98]
	Service Industry Density	Number of service industry facilities per km2	[95,97,98]
	Office Facility Density	Number of office facilities per km2	[95,97,98]
Accessibility	Subway Station Accessibility	Spatial accessibility within a 15 min walking time, calculated as the number of subway stations reachable using a pedestrian road network, with walking time as the impedance.	[99,101,102]
	Bus Stop Accessibility	Spatial accessibility within a 15 min walking time, calculated as the number of bus stops reachable using a pedestrian road network, with walking time as the impedance.	[99,101,102]
	Road Network Accessibility	Spatial accessibility within a 30 min travel time, calculated using a road network, with road design speed specified by road class as the impedance.	[99,101,102]

3.3. Model Specification

3.3.1. Spatial Pattern Identification

(1)

Kernel density estimation

To capture the overall spatial configuration and temporal dynamics of Chinese immigrant residential distribution, kernel density estimation (KDE) was applied to residential location data for the years 2011, 2016, 2021, and 2025. KDE generates a continuous density surface by smoothing point-based population data, thereby facilitating the identification of core settlement areas and their expansion or relocation over time. The kernel density function is defined as:

$$f(x)={\displaystyle \frac{1}{nh}}\sum _{i=1}^{n}K\left({\displaystyle \frac{x-x_i}{h}}\right)$$

(2)

where $f\left(x\right)$ denotes the estimated density at location $x$; $K(·)$ represents the Gaussian kernel function; $h$ is the bandwidth parameter; $n$ denotes the number of residential points; and $x_i$ indicates the location of the $i$-th residence.

In kernel density estimation, the choice of bandwidth critically influences the resulting density surface and the identification of spatial clustering patterns. Although adaptive bandwidth can better reveal local details in heterogeneous point distributions, to maintain temporal comparability across different study years, a fixed bandwidth determined using Silverman’s rule of thumb was ultimately adopted. Comparisons of kernel density surfaces across multiple years allow for a systematic examination of the emergence, diffusion, and spatial relocation of Chinese immigrant residential clusters within the Seoul metropolitan area.

(2)

Location quotient

The LQ was employed to measure the relative concentration of Chinese immigrants in each administrative unit compared with the citywide average. It is calculated as:

$$L{Q}_i={\displaystyle \frac{\left(M_i/P_i\right)}{\left(M/P\right)}}$$

(3)

where $M_i$ denotes the number of Chinese immigrants in district $i$, $P_i$ represents the total population of district $i$, and $M$ and $P$ indicate the total Chinese immigrant population and total population of Seoul, respectively. An $L{Q}_i$ value greater than 1 indicates that the concentration of Chinese immigrants in district $i$ exceeds the citywide average.

(3)

Standard deviation ellipse

Standard deviation ellipses were used to characterize the directional tendency and spatial dispersion of Chinese immigrant residential patterns. Shifts in the centroid of the ellipse indicate dominant migration directions, while changes in the lengths of the major and minor axes and in the ellipse area reflect variations in the degree of spatial concentration over time.

3.3.2. Determinants of Residential Distribution

(1)

Multicollinearity Diagnostics

Based on the system of 15 explanatory variables constructed in the previous section, street-level datasets were generated through GIS-based spatial overlay procedures. All variables were standardized using Z-score normalization to eliminate scale effects and enhance the comparability of regression coefficients. VIF diagnostics were conducted to detect multicollinearity; variables with VIF values exceeding 5 were removed or combined to ensure robust model estimation.

(2)

Spatial autocorrelation testing

Prior to model estimation, spatial dependence in immigrant residential distribution was assessed using Moran’s I statistic:

$$I={\displaystyle \frac{N}{W}}{\displaystyle \frac{{\sum }_i{\sum }_j{w}_{ij}(x_i-\stackrel{-}{x})(x_j-\stackrel{-}{x})}{{\sum }_i(x_i-\stackrel{-}{x}{)}^2}}$$

(4)

where $N$ denotes the number of dongs in Seoul; $x_i$ and $x_j$ represent the Chinese immigrant population in dong $i$ and $j$; $\stackrel{-}{x}$ is the mean Chinese immigrant population across all dongs; $w_{ij}$ denotes the spatial weight between units $i$ and $j$; and $W={\sum }_i{\sum }_j{w}_{ij}$. An inverse-distance spatial weight matrix was adopted in this study.

(3)

Specification and selection of spatial econometric models

To account for spatial dependence, SAR, SEM, and SDM were estimated and compared:

SAR

$$Y_i=\rho W{Y}_i+X_i\beta +{\epsilon }_i$$

(5)

SEM

$$Y_i=X_i\beta +{\mu }_i,{\mu }_i=\lambda W{\mu }_i+{\epsilon }_i$$

(6)

SDM

$$Y_i=\rho W{Y}_i+X_i\beta +W{X}_i\theta +{\epsilon }_i$$

(7)

where $Y_i$ denotes the proportion of Chinese immigrants at the street level; $X_i$ represents the matrix of explanatory variables; $W$ is the spatial weight matrix; $\rho$ is the spatial autoregressive coefficient; $\theta$ captures spatial spillover effects; and ${\epsilon }_i$ is the random error term.

Model selection proceeded in three steps. First, ordinary least squares (OLS) regression was estimated, followed perform LM test. When LM-Lag and LM-Error tests were significant, the SDM was chosen. Then, Wald and LR tests were applied to assess whether the SDM could be simplified to an SAR or SEM. If the null hypothesis of simplification was rejected at the 5% significance level, the SDM was retained as the optimal specification.

3.3.3. Spatial Heterogeneity and Assimilation Testing

(1)

MGWR

To further capture spatial heterogeneity in the effects of explanatory variables, a MGWR model was employed:

$$y_i={\beta }_0(u_i,v_i)+\sum _{k=1}^K{\beta }_{b{w}_k}(u_i,v_i)x_{ik}+{\epsilon }_i$$

(8)

where $\left(u_i,v_i\right)$ denotes the centroid coordinates of street $i$; $b{w}_k$ represents the optimal bandwidth for the $k$-th explanatory variable, determined by minimizing the corrected Akaike Information Criterion (AICc). MGWR allows different explanatory variables to operate at distinct spatial scales, thereby providing a nuanced understanding of spatial heterogeneity in immigrant residential choice.

(2)

Spatial assimilation testing

Residential segregation indices quantify spatial inequalities in population distribution and the structural conditions shaping potential social interactions. Beyond descriptive mapping, segregation metrics provide empirical evidence for identifying socio-economic, institutional, and policy-driven processes that generate and reproduce spatial stratification [103]. Index selection was guided by research objectives, spatial scale, and analytical units [104]. This study employs three complementary indices: the D, II, and H. D captures spatial evenness [105], II measures intra-group exposure [106,107], and H quantifies entropy-based compositional divergence across spatial units [108]. These indices approximate potential spatial contact patterns rather than realized social interactions and are interpreted as structural indicators of spatial opportunity and constraint [106,107]. Results may be subject to the MAUP; robustness checks across spatial scales were conducted.

(a)

Dissimilarity Index

The D measures the degree to which two population groups are unevenly distributed across spatial units and represents the minimum proportion of one group that would need to relocate to achieve spatial parity [25,64,105,106]. Owing to its interpretability and cross-study comparability, D is widely used as a benchmark measure in segregation research [109]. The index is defined as:

$$D={\displaystyle \frac{1}{2}}\sum _{i=1}^n\mid {\displaystyle \frac{m_i}{M}}-{\displaystyle \frac{p_i}{P}}\mid$$

(9)

where $n$ is the number of dong units, $m_i$ is the number of Chinese immigrants in dong $i$, $M$ is the total Chinese immigrant population in Seoul, $p_i$ is the total population in dong $i$, and $P$ is the total population of Seoul. D ranges from 0, indicating complete integration, to 1, indicating complete segregation, and reflects the proportion of the minority population that would need to relocate to replicate the metropolitan population composition [25].

(b)

Isolation Index (II)

The II captures the exposure dimension of segregation by estimating the probability that a group member encounters a co-ethnic resident within the same spatial unit [25,110]. Unlike D, which reflects distributional evenness, II reflects within-group spatial contact intensity and the experiential dimension of segregation [110,111]. The index is defined as:

$$II=\sum _{i=1}^n\left({\displaystyle \frac{m_i}{M}}\right)\left({\displaystyle \frac{m_i}{p_i}}\right)$$

(10)

where $p_i$ is the total population in unit $i$. II ranges from 0 to 1, with higher values indicating stronger intra-group residential exposure [112]. Because II is mechanically influenced by minority group size, it is interpreted jointly with D and temporal trends rather than as an absolute segregation indicator.

(c)

Theil’s H Index (H)

H provides an entropy-based measure of deviations between local and metropolitan population compositions and allows decomposition across nested spatial scales [108]. Entropy for each spatial unit is defined as:

$$E_i=P_i\mathrm{l}\mathrm{n}\left({\displaystyle \frac{1}{P_i}}\right)+(1-P_i)\mathrm{l}\mathrm{n}\left({\displaystyle \frac{1}{1-P_i}}\right)$$

(11)

where $P_i$ denotes the proportion of Chinese immigrants in dong $i$. Entropy equals zero when a single group dominates and is maximized when groups are evenly represented.

Theil’s H for Seoul is computed as:

$$H=\sum _{i=1}^N{\displaystyle \frac{T_i}{T_m}}\left({\displaystyle \frac{E_m-E_i}{E_m}}\right)$$

(12)

where $N$ is the number of dong units, $T_i$ and $E_i$ denote the population size and entropy for dong $i$, and $T_m$ and $E_m$ represent metropolitan totals. H ranges from 0, indicating no segregation, to 1, indicating complete segregation, capturing population-weighted deviations from the metropolitan composition [108].

(d)

Robustness Check Using Spatially Explicit Measures

To bolster methodological rigor, we conducted a spatially explicit robustness check by supplementing aspatial segregation indices with alternative spatial configurations [113]. Aligning with the study’s multi-scalar framework, all core analyses were replicated using a 250 m × 250 m grid system. This resolution serves as the standardized micro-spatial unit for Seoul’s big data analytics, enabling a more granular representation of spatial proximity and socioeconomic interaction.

Population counts were disaggregated from administrative boundaries (Dong) to grid cells using area-weighted interpolation. To ensure the structural integrity of this spatial transformation, we implemented a dual-stage validation protocol:

First, Global Conservation Testing: The mass balance was verified by confirming that the aggregate population of the source units remained congruent with the integrated sum of the target grid cells.

Second, Distributional Fidelity Assessment: The statistical consistency of the downscaling process was validated via KDE. This step ensured that the grid-level data accurately preserved the spatial heterogeneity and statistical moments (e.g., skewness and peak intensity) inherent in the original administrative distribution.

The results derived from this spatially explicit specification were systematically com-pared with the Dong-level findings. This comparative analysis allowed us to evaluate the sensitivity of observed patterns to the MAUP and confirm the consistency of our conclusions across divergent spatial scales.

4. Results

4.1. Evolutionary Characteristics of the Spatial Pattern of Chinese Immigrant Settlements

4.1.1. Spatiotemporal Evolution of Residential Agglomeration

At the citywide scale, the spatial configuration of Chinese immigrant settlements in Seoul between 2011 and 2025 exhibits a dual dynamic characterized by strong path dependence alongside phased structural adjustment (Figure 2). By 2011, the Guro–Daerim cluster in the southwestern part of the city had already consolidated its position as the dominant settlement core. This dominance persisted through 2016, 2021, and 2025, demonstrating a high degree of spatial stability. Cross-sectional KDE further indicates that while the primary agglomeration remained firmly anchored in Guro-gu and Yeongdeungpo-gu, Gwangjin-gu gradually emerged as a distinct secondary growth pole after 2016, giving rise to an increasingly polycentric settlement structure.

At the Gu-level, Guro-gu and Yeongdeungpo-gu consistently exhibited exceptionally high LQ values (LQ > 14) across all four observation years, constituting the unequivocal core areas of Chinese immigrant concentration. Following 2016, Gwangjin-gu progressively consolidated its role as a secondary core. Although other districts, such as Mapo-gu and Dongdaemun-gu, experienced localized fluctuations in clustering intensity, the industrial clusters in the southwest and the university-oriented zones in the eastern part of Seoul consistently functioned as the principal spatial carriers of immigrant expansion.

At the administrative dong-level, LQ analysis reveals pronounced internal restructuring within these core areas (Figure 3). Garibong-dong, representing the traditional settlement nucleus, reached a peak LQ of 21.6 in 2016, substantially exceeding that of other neighborhoods. However, its LQ remained at a comparable level in 2025 and declined to 14.4 in 2021, indicating a relative weakening of agglomeration dominance within the traditional core. In contrast, emerging enclaves such as Jayang 4-dong experienced a marked increase in LQ, reaching 9.8 by 2025, which signals a significant intensification of clustering in non-traditional settlement areas.

A longitudinal comparison of agglomeration intensity across different regional typologies (Figure 4) further demonstrates that, except for a few peripheral districts, traditional core areas (e.g., Guro-gu) consistently maintained substantially higher concentration levels than emerging residential zones (e.g., Jung-gu and Seodaemun-gu). Emerging areas characterized by university and commercial functions exhibited clustering densities markedly lower than those of the southwestern core during the 2011–2016 period. Moreover, temporal trajectories of LQ values within traditional cores indicate a continuous intensification of agglomeration between 2011 and 2016, followed by stabilization or a slight decline after 2021. This pattern suggests the presence of a “saturation effect,” whereby the spatial carrying capacity of traditional settlement cores has approached its upper limit.

4.1.2. Directionality and Scale Effects of the Residential Spatial Pattern

At the macro scale, standard deviational ellipse (SDE) analysis reveals a strong correspondence between the directional structure of Chinese immigrant residential space and the overall population distribution of Seoul (Figure 5). Administrative dongs with larger population sizes and higher levels of immigrant concentration exerted a decisive influence on the geometric configuration of the ellipses. In the initial phase (2011), the centroid of the SDE was firmly anchored in the southwestern part of the city, corresponding to the Guro–Daerim traditional core. However, between 2021 and 2025, a notable spatial shift occurred, with the centroid moving northeastward to approximately 126°57′27.47″ E, 37°31′38.64″ N, indicating a structural reorganization of internal settlement patterns.

Changes in SDE geometric parameters (

Table 2. Standard Deviation Ellipse Related Parameters.

Year	Coordinate	Major Axis (m)	Minor Axis (m)	Rotation (°)	Area (km2)	Oblateness
2011	126°57′15.48″ E, 37°31′15.96″ N	10,866.74	5296.02	61°42′	180.79	0.5126
2016	126°57′7.56″ E, 37°31′14.16″ N	10,660.72	5093.76	60°35′	170.59	0.5222
2021	126°57′7.56″ E, 37°31′20.64″ N	10,838.97	5231.43	59°34′	178.13	0.5173
2025	126°57′27.47″ E, 37°31′38.64″ N	10,897.31	5285.02	58°2′	180.92	0.5150

and Figure 5) further illustrate a phased process characterized by “agglomeration intensification–boundary contraction–subsequent expansion.” During the first stage (2011–2016), the intensification of high-density settlement areas led to a contraction of the ellipse area from 180.79 km² to 170.59 km², reflecting spatial compaction primarily driven by infill development within the Guro–Daerim core. In the subsequent stage (2021–2025), as dispersal tendencies re-emerged, the residential boundary expanded toward suburban areas and eastern education–commercial centers, with the ellipse area increasing to 180.92 km².

Despite this outward diffusion, the overall directional orientation of Chinese immigrant residential space remained highly stable throughout the study period, with the azimuth consistently ranging between 58° and 62°. This orientation closely aligns with Seoul’s principal urban development axis and major transit corridors, particularly Subway Lines 2 and 7. These findings indicate that the spatial expansion of Chinese immigrant settlements is not random but remains strongly embedded within the existing urban spatial structure and transportation framework, reflecting a pronounced scale-constrained pattern.

4.2. Spatial Dependence and Determinants of Chinese Immigrant Residential Patterns

4.2.1. Diagnostic Tests for Spatial Autocorrelation and Multicollinearity

(1)

Multicollinearity Diagnostics

Prior to estimating spatial econometric models, it is essential to examine the presence of multicollinearity among explanatory variables, as excessive collinearity may undermine the stability and precision of coefficient estimates. Accordingly, VIF tests were conducted for 15 explanatory variables, including Migrant Stock, Migrant Flow, Housing Price, and Rent Price.

As shown in Figure 6, VIF values for all explanatory variables range from 1.115 to 2.532, remaining well below the conventional threshold of 10 and beneath the more conservative cutoff value of 5 commonly applied in empirical research. Among the covariates, Housing Price (VIF = 2.532) and Service Industry Density (VIF = 2.284) exhibit relatively higher collinearity levels; however, these values remain within acceptable limits. In contrast, variables such as Migrant Stock and Cultural Facilities display VIF values close to unity, indicating a high degree of independence from other explanatory variables.

Overall, the diagnostic results suggest that the explanatory variable set is structurally stable and free from serious multicollinearity, thereby satisfying the methodological requirements for subsequent spatial regression analysis.

(2)

Global Spatial Autocorrelation Test (Moran’s I)

Following the confirmation of limited multicollinearity, the spatial dependence of migrant population distribution in China in 2025 was examined using the Global Moran’s I statistic. The results indicate that Moran’s I equals 0.0655, with an associated p-value of 0.000000. This finding rejects the null hypothesis of spatial randomness and provides strong evidence of positive spatial autocorrelation in migrant distribution. Specifically, regions with relatively high migrant population shares tend to be spatially clustered, while areas with lower migrant shares exhibit similar geographic concentration patterns.

The presence of pronounced spatial dependence implies the existence of spatial spillover effects and highlights the inadequacy of conventional OLS models for capturing spatial interactions. Consequently, the application of spatial econometric techniques is both methodologically justified and empirically necessary.

4.2.2. Spatial Econometric Model Specification and Estimation Results

(1)

Model Selection and Specification Tests

Spatial dependence diagnostics based on OLS residuals were conducted to determine the appropriate spatial econometric specification (Figure 7). The LM-Lag test was highly significant (12.523, p < 0.01), whereas the LM-Error test was not significant (p = 0.356). However, both Robust LM-Lag (16.947, p < 0.01) and Robust LM-Error (5.274, p < 0.05) tests were significant, indicating the coexistence of spatial lag dependence and spatial error correlation. These results suggest that a conventional OLS model is insufficient to capture the underlying spatial processes.

Accordingly, the SDM was adopted as the general specification. Likelihood Ratio (LR) and Wald tests further rejected the null hypotheses that the SDM could be simplified to a SAR (35.549, p < 0.01) or a Spatial Error Model (SEM) (LR = 44.381, p < 0.01). Consistently, the Wald statistic (34.51, p < 0.01) confirmed that such simplifications were inappropriate.

These findings indicate that the SDM effectively captures spatial dependence in the dependent variable as well as spatial externalities associated with explanatory variables, thereby reducing potential bias arising from omitted spatially correlated factors. The SDM is therefore employed to examine the effects of Migrant Stock, Migrant Flow, housing variables, and related covariates on migrant population distribution.

(2)

Estimation Results of the SDMs

The SDM was employed to investigate spatial dependence in migrant inflows. The estimation results reveal a strong negative spatial autoregressive parameter (ρ = −0.905), indicating that characteristics of neighboring regions exert competitive and suppressive spillover effects on local migration outcomes.

Compared with the SAR model, the SDM demonstrates superior model performance (log-likelihood: −414.98 vs. −432.75; AIC: 895.95 vs. 901.50), suggesting that incorporating both spatially lagged dependent and independent variables is necessary to adequately capture the spatial dynamics of migration.

(a)

Direct Effects

As shown in Figure 8, Migrant Stock exhibits the strongest positive direct effect on migrant inflows (β = 0.501, p < 0.01). Migrant Flow also shows a significant positive association (β = 0.163, p < 0.01), although with a smaller magnitude. These findings indicate that migration networks constitute the dominant determinants of migrant settlement patterns.

In contrast, contemporaneous local socioeconomic variables, including housing prices, rental prices, housing characteristics, facility densities, and transport accessibility—are not statistically significant after controlling for migration networks and spatial dependence. This suggests that local attributes alone do not independently explain migrant residential patterns once network effects and spatial interactions are accounted for.

(b)

Spatial Lag Effects

Spatial lag effects, representing the influence of neighboring regional characteristics, exhibit patterns distinct from contemporaneous local effects. The spatial lag of Migrant Stock is positive and statistically significant (β = 1.681, p = 0.038), indicating that migrant populations in adjacent regions increase local migrant inflows. Combined with the negative spatial autoregressive parameter, this finding suggests spatial clustering within regional systems accompanied by competitive interactions across spatial units (Figure 9).

Several spatially lagged socioeconomic variables show marginal statistical significance. Higher medical facility density in neighboring regions is positively associated with local migrant inflows (β = 1.664, p = 0.089). Similarly, the density of labor-intensive industries in adjacent areas is positively associated with migrant inflows (β = 2.542, p = 0.066). In contrast, neighboring housing prices exhibit a relatively large negative effect, although this does not reach conventional significance levels (β = −2.243, p = 0.110).

Overall, these results suggest that spillover effects from neighboring regions—particularly migration-related and industrial factors—are more influential than local attributes for the outcome variable examined in this study, and that interregional relationships are predominantly competitive rather than complementary.

4.2.3. Decomposition of Spatial Effects

To further interpret the underlying spatial processes, the coefficients of the SDM were decomposed into direct, indirect, and total effects, explicitly accounting for spatial feedback mechanisms (Figure 10).

(1)

Effects of Core Explanatory Variables

After decomposing the SDM effects, Migrant Stock emerged as the only core explanatory variable with statistically significant impacts, demonstrating strong explanatory power. The direct effect of Migrant Stock was 0.497 and highly significant at the 1 percent level (p < 0.01). This finding indicates that the accumulation of existing migrant populations exerts a substantial positive local influence on the dependent variable, suggesting that population agglomeration and stock-based network effects constitute a key driving force of local development.

In contrast, the indirect spillover effect of Migrant Stock was positive at 0.649 but did not reach statistical significance, implying that interregional linkage effects associated with migrant populations in neighboring areas are not clearly identifiable in the current model specification. Driven primarily by the strong direct effect, the total effect of Migrant Stock reached 1.145 and was significant at the 5% level, confirming from a system-wide spatial perspective that increasing migrant stock contributes positively to overall regional development.

(2)

Effects of Migrant Flows and Housing Characteristics

Unlike the stock effect, Migrant Flow exhibited neither significant direct effects at 0.163 nor indirect effects at 0.125. This contrast with Migrant Stock suggests that short-term and dynamic population movements have not yet generated stable or observable scale effects on the outcome variable, whereas the accumulation of stable resident migrant populations plays a more critical role.

With respect to housing market characteristics, most indicators were statistically insignificant but exhibited discernible directional tendencies. Housing prices with a direct effect of −0.061 and an indirect effect of −1.153, as well as housing age with a direct effect of −0.061, displayed negative coefficients, suggesting a potential suppressive influence of higher housing costs and housing aging on the dependent variable, although these effects did not reach conventional significance levels. Rental prices and the share of apartment housing showed weak positive tendencies, but these associations were likewise not statistically significant.

(3)

Spatial Feedback from Facility Density and Transport Accessibility

The examination of public service provision, industrial agglomeration, and transport accessibility indicates that most indicators did not exhibit statistically significant direct or indirect effects; however, the coefficient directions provide preliminary insights into spatial dynamics. Education and medical facility densities showed positive direct effect tendencies, suggesting that improved public service provision may serve as a potential supportive factor for local development.

In contrast, the densities of labor-intensive industries, service industries, and office facilities generally displayed negative direct effects but positive indirect spillover effects. This divergence in effect directions suggests complex spatial interactions, potentially reflecting local competition alongside regional complementarity in industrial distribution, although these interpretations remain tentative due to estimation uncertainty.

Finally, metro station density, bus stop density, and road network accessibility exhibited positive tendencies for local contributions, but neither their direct nor indirect spatial marginal effects were statistically significant in the current model.

4.3. Spatial Heterogeneity and Multi-Scalar Dynamics

4.3.1. Model Fit Comparison

This study applied both geographically weighted regression (GWR) and MGWR models. The results indicate that the MGWR model outperformed the GWR model across all evaluation metrics, demonstrating higher goodness of fit and greater model parsimony (Figure 11). Specifically, MGWR achieved an R² value of 0.666 and an adjusted R² of 0.616, indicating an improved ability to capture spatial variation in the data. Moreover, the lower AICc value suggests that MGWR provides a better model fit while avoiding overfitting, confirming its robustness as an analytical framework. Accordingly, subsequent analyses are primarily based on the MGWR results.

4.3.2. Spatial Heterogeneity Decomposition of Core Variables

Based on the MGWR model, the spatial scale and heterogeneity intensity of each explanatory variable were identified using bandwidth values and the proportion of significant coefficients. Smaller bandwidths indicate stronger spatial heterogeneity, whereas larger bandwidths indicate more spatially stable global effects (Figure 12).

(1)

Core Variables with Strong Spatial Heterogeneity

These variables exhibit highly spatially differentiated effects and constitute key drivers of geographic variation in the study phenomenon. Migrant Stock displayed the smallest bandwidth of 45, indicating the strongest spatial heterogeneity among all variables. Its coefficients were significantly positive in 73.05% of the sample locations, with values ranging from 0.385 to 4.237. This confirms that Migrant Stock is the primary driver of spatial differentiation, although the magnitude of its positive effect varies substantially across locations.

(2)

Important Variables with Moderate Spatial Heterogeneity

These variables exhibit spatially varying effects at broader regional scales and play a moderating role in shaping spatial outcomes. Migrant Flow had a bandwidth of 112, with significant coefficients in only 50.63% of the sample locations, indicating that short-term migration dynamics have not yet formed a stable or ubiquitous influence pattern. Regarding industrial structure, labor-intensive industry density with a bandwidth of 155 and a significant proportion of 44.33%, and office facility density with a bandwidth of 142 and a significant proportion of 58.69%, also exhibited spatially heterogeneous effects. Their coefficients displayed both positive and negative signs, with ranges from −0.140 to 0.165 and from −0.134 to 0.814, respectively. This suggests that these factors promote development in some areas while exerting suppressive effects in others.

(3)

Spatially Stable Background Variables

Most variables, including housing prices, facility densities, and transport accessibility indicators, reached the maximum bandwidth of 386, indicating relatively stable effects across the study area and limited explanatory power for spatial variation. These variables can be considered global background conditions. An exception is service industry density. Although its bandwidth of 297 suggests relatively stable spatial effects, its coefficients were significantly negative in 97.48% of the sample locations, with values ranging from −0.262 to −0.254. This indicates a pervasive and consistent suppressive effect throughout the study region.

4.3.3. Synthesis of Spatial Drivers and Regional Differentiation Mechanisms

Integrating the model comparison and spatial effect decomposition results, the MGWR model is better suited for analyzing the spatial structure of the study data and more effectively reveals spatial heterogeneity in driving factors (Figure 13).

Migrant Stock emerges as the dominant driver of spatial differentiation, exhibiting strong positive effects with pronounced location dependence. Service industry density represents a global suppressive factor with a stable negative influence across the entire study area. Migrant Flow and selected industrial factors, including labor-intensive and office sectors, function as localized moderating factors, with their effects varying in magnitude and direction across regions. In contrast, most housing characteristics, public service facilities, and transport accessibility indicators act as spatially stable background variables, forming the structural context of regional development rather than primary sources of spatial variation.

Overall, the observed geographic pattern is primarily shaped by localized accumulation of Migrant Stock, constrained by the global suppressive influence of service industry density, while other factors play secondary and regionally contingent roles.

4.3.4. Spatial Measurement of Residential Assimilation and Segregation

(1)

Distributional Characteristics and Evolution of the Spatial D Index

The spatiotemporal analysis of the Local D index reveals a predominantly low-level yet structurally differentiated spatial pattern at the administrative dong-level, characterized by pronounced spatial heterogeneity (Figure 14). Overall, the magnitude of spatial isolation remains limited, indicating that the residential distribution of Chinese immigrants in Seoul is neither marked by strong citywide segregation nor by complete spatial assimilation. Instead, it reflects a hybrid regime that combines widespread spatial integration with localized concentrations.

In terms of overall distribution, D values for most neighborhoods consistently remained below 0.001 across all temporal cross-sections, indicating extremely low levels of residential segregation. This pervasive low magnitude suggests that Chinese immigrants are broadly embedded within the mainstream residential structure, with their spatial distribution largely aligned with the demographic and housing patterns of the host population. This pattern is particularly evident in middle- to high-income districts and peripheral new towns, where immigrant settlement appears to be constrained primarily by prevailing housing institutions and market mechanisms rather than driven by ethnic clustering.

In contrast, a limited number of neighborhoods persistently exhibited substantially higher D values, constituting the primary loci of spatial differentiation. Representative high-isolation areas in 2025 include Guro 2-dong (0.0207), Daerim 2-dong (0.0194), Jayang 4-dong (0.0110), Doksan 4-dong (0.0054), and Yeomchang-dong (0.0019). These neighborhoods maintained relatively elevated index values across multiple time periods, suggesting the presence of stable social or ethnic clustering processes. They can thus be interpreted as enduring cores of immigrant concentration embedded within otherwise integrated urban contexts.

From a longitudinal perspective, the evolution of the D index exhibits a dynamic pattern of localized transformation rather than monotonic convergence or divergence. For instance, Garibong-dong experienced a substantial decline in D values from 0.0185 in 2011 to 0.0089 in 2025, indicating a marked reduction in spatial isolation and possible processes of neighborhood restructuring or demographic mixing. Conversely, Daerim 3-dong recorded an increase from 0.0068 in 2011 to 0.0135 in 2025, suggesting a renewed intensification of spatial differentiation. Similarly, Hyehwa-dong displayed a sharp increase from 0.0008 to 0.0073 over the same period, implying significant socio-spatial reconfiguration, potentially associated with university-related populations and service-sector clustering.

Spatial trajectories also varied across urban subregions. High-status districts such as Gangnam, Seocho, and Songpa consistently exhibited low D values throughout the study period, indicating limited spatial isolation and highly dispersed settlement patterns. This suggests that although Chinese immigrants have increasingly gained access to high-end housing markets, their residential presence in these areas remains statistically diffuse and does not translate into meaningful spatial clustering.

(2)

Spatial Patterns and Phased Dynamics of the Spatial Isolation Index (P)

The Spatial Isolation Index (P) further substantiates the pronounced polarization embedded in Seoul’s residential landscape (Figure 15). Overall, the results reveal a highly skewed distribution, characterized by a pervasive background of extremely low isolation coexisting with a limited number of persistent, high-intensity isolation clusters.

At a general level, p-values for most administrative neighborhoods remained very small, with more than 90% of neighborhoods exhibiting values below 0.001 and many falling within the range of 10⁻⁶ to 10⁻⁴ throughout the study period. This pattern indicates that the target population constitutes an absolute numerical minority within Seoul and that most foreign residents experience a high degree of exposure to the native population in their daily lives. Such low isolation levels were particularly evident in high-income districts such as Gangnam and in newly developed residential areas, where p-values remained consistently negligible and showed no systematic upward trend over time.

By contrast, a small subset of neighborhoods consistently displayed markedly elevated P-values, indicating persistent spatial isolation and enclave-like settlement patterns. Traditional migrant settlement hubs located in Guro-gu and Yeongdeungpo-gu recorded isolation indices on the order of 10⁻² as early as 2011. For instance, Garibong-dong reached a p-value of 0.0143 in 2011, while Daerim 2-dong recorded 0.0118 in the same year. Although p-values in some of these areas declined over time—Garibong-dong decreased to 0.0040 by 2025—their absolute levels remained several orders of magnitude higher than those observed elsewhere, confirming their persistence as deeply entrenched zones of spatial isolation. In contrast, Daerim 2-dong exhibited an increase to 0.0128 by 2025, and Guro 2-dong rose from 0.0046 to 0.0095, suggesting an intensification of enclave formation and growing intra-group residential concentration. Other high-isolation neighborhoods in 2025 included Daerim 3-dong (0.0049), Guro 4-dong (0.0069), and Jayang 4-dong (0.0035), reflecting both labor-oriented and family-based migrant concentrations.

From a temporal perspective, the evolution of the P index followed a phased restructuring trajectory. During the 2011–2016 period, isolation levels in core enclave areas reached localized peaks, reflecting the consolidation of traditional migrant clusters. This was followed by a partial attenuation of isolation between 2016 and 2021, likely associated with urban redevelopment, population dispersal, and policy interventions. However, the post-2021 period did not exhibit a uniform decline; instead, a process of localized re-differentiation emerged. While original enclave cores maintained relatively high isolation levels, p-values increased substantially in neighborhoods associated with universities and older inner-city districts. For example, Hyehwa-dong increased from 0.0001 in 2011 to 0.0020 in 2025, and Sinchon-dong reached 0.0023 in 2025, indicating the emergence of new, institutionally anchored micro-clusters of international students and foreign residents.

(3)

Spatial Patterns and Phased Dynamics of the H

The longitudinal analysis of the H index and entropy ($E_i)$ from 2011 to 2025 reveals a clear three-phase evolution in the spatial distribution of Chinese immigrants across Seoul (Figure 16).

First, traditional immigrant enclaves in the southwestern region exhibit persistent and intensifying concentration patterns. In Daelim 2-dong, $E_i$ consistently exceeds the high threshold of 0.6, while the H index shows a significant downward trend in negative values, declining from −0.008 in 2011 to a projected −0.016 by 2025. This pattern quantitatively identifies Daelim 2-dong as a primary anchor of Chinese settlement in Seoul, where segregation is not only sustained but increasingly reinforced through growing demographic weight within the metropolitan system. Similarly, Guro 2-dong and Guro 4-dong maintain high entropy levels and strongly negative H values, with Guro 2-dong reaching a peak negative contribution of −0.0224 by 2025, indicating its dominant role in shaping regional segregation dynamics.

Second, new immigrant clusters are emerging in the northeastern and central districts, indicating spatial diffusion beyond traditional enclaves. In Zayang 4-dong (Gwangjin-gu), Eᵢ increases from 0.35 in 2011 to 0.40 in 2025, accompanied by a decline in the H index from −0.004 to −0.014. This shift reflects the consolidation of a new ethnic cluster proximate to Konkuk University, consistent with the development of an emerging Chinatown-type enclave. In addition, Hyehwa-dong is projected to experience an entropy increase to 0.366 by 2025, suggesting concentrated settlement patterns associated with academic migrants and younger demographic cohorts.

Third, high-income and highly educated districts exhibit persistently low levels of immigrant integration. In areas such as Gangnam-gu and Seocho-gu, including Daechi-dong and Seocho-dong, H values remain consistently positive but small, ranging from 0.001 to 0.003, while Eᵢ remains extremely low. These indicators suggest that Chinese immigrants are sparsely distributed in these affluent districts, maintaining a positive deviation from the metropolitan average, and effectively indicating their relative exclusion from these urban cores.

In this study, high-entropy areas such as Garibong-dong (0.51) and Daelim-dong (0.62) represent urban contexts characterized by either highly mixed ethnic compositions or dominance by the target immigrant group. A notable anomaly is observed in Yongsin-dong, where the projected 2025 Eᵢ reaches 0.692. As this value approaches the theoretical maximum, it serves as a predictive signal that the district is transitioning toward a hyper-dense immigrant enclave. Regarding the H index, absolute values for most administrative dongs remain below 0.005. From a metropolitan perspective, this indicates that although localized ethnic clustering is evident, Seoul has not yet reached the level of systemic racial segregation observed in some Western contexts where H exceeds 0.3. Chinese immigrants are increasingly expanding beyond the traditional southwestern concentration zone of Guro, Geumcheon, and Yeongdeungpo, and are moving toward economically dynamic districts such as Mapo and Gwangjin, where new micro-clusters are forming.

4.3.5. Results of Robustness Check via Spatially Explicit Measures

(1)

Conservation Law Verification

Robustness checks were conducted for four time points (2011, 2016, 2021, and 2025) to evaluate the consistency of the spatial downscaling results. Aggregate conservation tests show that discrepancies between administrative totals and grid-integrated values are negligible, with relative errors of 0.0062%, 0.0069%, 0.0059%, and 0.0061%, respectively, all well below the 0.1% tolerance threshold. Regression results further confirm global conservation, with coefficients of determination (R²) exceeding 0.99 and observations tightly aligned along the 1:1 reference line (Figure 17).

(2)

Distributional Consistency and Visualization Analysis

Distributional consistency was assessed using KDE and correlation diagnostics (Figure 18 and Figure 19). KDE curves exhibit near-complete overlap between administrative-level data and 250 m grid outputs across all years, indicating that key distributional properties, including skewness and peak intensity, are preserved. Scatter plots reveal a strong linear correspondence with no systematic outliers. Minor localized deviations (approximately −1.5%) observed in a small number of units are attributable to boundary edge effects and do not affect global consistency. Overall, the results demonstrate high robustness of the downscaling procedure in terms of both aggregate conservation and distributional fidelity.

5. Discussion

5.1. Spatiotemporal Evolution of Chinese Immigrant Enclaves at the Microscale

Immigrant clustering constitutes a fundamentally spatial process whose morphology and dynamics are jointly shaped by local embeddedness and macro-structural forces [3]. This study demonstrates that the spatial evolution of Chinese immigrant enclaves in Seoul is characterized by a path-dependent core structure accompanied by phased spatial reconfiguration. Consistent with patterns observed in global cities, strong spatial inertia effects are evident [46,47]. The southwestern cluster centered on Guro-gu and Daerim-dong has consolidated into a persistent and dominant ethnic core, reflecting lock-in mechanisms generated by early migrant networks, cumulative ethnic economies, and affordable industrial-era housing stock.

However, this core structure is not static. The empirical results reveal pronounced phase-based restructuring. Declining segregation and isolation indices in traditional hubs such as Garibong-dong, combined with increasing clustering indices in Daerim 2-dong and Guro 2-dong, indicate internal reconfiguration rather than simple enclave decay. These changes coincide with major redevelopment initiatives, including Garibong-dong’s designation as a redevelopment promotion zone after 2008, which induced migrant displacement and relocation to adjacent neighborhoods such as Daerim-dong. This finding underscores the role of urban governance and redevelopment policies as exogenous drivers actively reshaping immigrant residential geographies.

Moreover, the emergence of secondary growth poles in northeastern districts such as Gwangjin-gu (e.g., Jayang 4-dong) reflects spatial diffusion beyond traditional industrial enclaves. The northeastward shift in the center of the standard deviational ellipse further corroborates a structural transition from labor-oriented industrial clusters toward education- and service-oriented residential patterns. The Jayang 4-dong cluster, located in proximity to Konkuk University and characterized by Chinese-oriented commercial amenities, exemplifies a post-industrial immigrant settlement pattern catering to students and new professional migrants. Taken together, these findings indicate that Chinese immigrant enclaves in Seoul are evolving from a monocentric enclave system toward a polycentric and functionally differentiated networked structure.

5.2. The Complex Landscape of Residential (De)Segregation

Immigration often inscribes spatial patterns of segregation within host cities, with segregation intensity typically increasing at finer analytical scales [3]. By examining local segregation indices at the administrative dong-level, this study reveals a highly nuanced residential landscape that challenges the binary narrative of segregation versus assimilation. At the metropolitan scale, both the Local D and P indices indicate overwhelmingly low levels of segregation and isolation across most neighborhoods. More than 90% of dongs exhibit D and p values below 0.001, suggesting that Chinese immigrants remain an absolute numerical minority and are broadly embedded within the mainstream residential structure.

This widespread low magnitude indicates strong exposure to the native population and quasi-assimilation in most parts of Seoul, particularly in middle- to high-income districts and newly developed suburban areas where settlement patterns are largely shaped by housing market institutions rather than ethnic clustering. Nevertheless, pronounced local heterogeneity persists. Persistent high-segregation and high-isolation clusters are concentrated in southwestern districts such as Guro-gu and Yeongdeungpo-gu, where Daerim 2-dong and Guro 2-dong exhibit consistently elevated D and p values and strongly negative contributions to Theil’s H index. These areas represent enduring cores of labor-oriented immigrant concentration embedded within otherwise integrated urban contexts.

Conversely, affluent districts such as Gangnam and Seocho display persistently low D, p, and entropy values, indicating that although Chinese immigrants have gained access to high-end housing markets, their residential presence remains statistically diffuse and does not translate into meaningful spatial clustering.

A particularly salient finding is the emergence of new micro-clusters structured by intra-group differentiation. University-adjacent neighborhoods such as Hyehwa-dong and Jayang 4-dong show increasing D, p, and entropy values, indicating concentrated settlement associated with international students and highly educated migrants. This suggests that contemporary residential segregation is increasingly stratified along education, occupation, and life-course dimensions, generating new spatial logics distinct from traditional labor-based enclaves. Importantly, overall H index values remain extremely low at the metropolitan scale, with absolute values for most neighborhoods below 0.005 and far below segregation thresholds reported in Western cities. This indicates that Seoul has not developed systemic racial segregation structures comparable to those observed in North American or European contexts.

5.3. Multiscale Determinants of Enclave Formation and Segregation

The factors shaping immigrant concentration and segregation exhibit inherent scale sensitivity and complexity [4,7,114]. By integrating results from the SDM and MGWR, this study reveals a multilevel and dynamic framework of determinants.

First, the central role of network effects. The SDM results confirm that Migrant Stock exerts the most robust and strongly positive direct effect, firmly demonstrating that immigrant networks and chain migration constitute the most fundamental drivers of residential location choice [4], creating a self-reinforcing positive feedback loop for enclave formation.

Second, the interactive dynamics of spatial competition. The significantly negative spatial autoregressive parameter indicates competitive spatial dependence among enclaves. Growth in immigrant concentration in one area may exert inhibitory spillover effects on neighboring areas, suggesting a quasi-zero-sum competition for migrant resources and attractiveness within the regional system, rather than simple synergistic agglomeration.

Third, scale heterogeneity in local driving factors. Local factors that are globally insignificant in the SDM (e.g., housing prices, facility density) exhibit strong spatial heterogeneity in the MGWR results, confirming that their impacts are highly context-dependent. For example, Office Facility Density shows an almost universally stable negative effect, indicating that Chinese immigrant enclaves tend to (actively or passively) avoid high-end modern service cores. This pattern is consistent with explanations related to ethnic economic preferences and socioeconomic barriers [47]. However, this contrasts with findings from cities such as Sydney [7], underscoring the city-specific logic of immigrant settlement: in Sydney, high-quality education and amenities may dominate, whereas in Seoul, existing networks and historical industrial pathways appear more influential.

Finally, reconsidering the role of transportation. Although the initial concentration of Chinese migrants in southwestern Seoul was associated with transportation accessibility, transportation variables in the model did not exhibit significant direct effects. A plausible inference is that in global cities like Seoul, the homogenization of high-level infrastructure reduces the explanatory power of transportation as a differentiating location factor, with its influence internalized as a background condition rather than an explicit determinant.

6. Conclusions

6.1. Conclusions and Implications

Against the backdrop of accelerating global transnational migration, the residential geography of immigrants provides an important lens through which to examine urban socio-spatial restructuring and processes of social integration. Building on the longstanding debate between assimilation and segregation, this study examined the spatiotemporal evolution and underlying mechanisms of Chinese immigrant settlement patterns in Seoul between 2011 and 2025. The primary objective was to assess whether these residential patterns tend toward spatial dispersion within the mainstream population or maintain localized concentration under institutional and market constraints.

The results indicate that the residential patterns of Chinese immigrants in Seoul do not conform to a model of pervasive segregation or linear spatial assimilation. Instead, they reflect a composite spatial configuration in which generally low levels of segregation and isolation at the metropolitan scale coexist with persistent and evolving localized clustering at the neighborhood scale. At the citywide level, Chinese immigrants remain a relatively small population group and are broadly embedded within Seoul’s mainstream residential structure. This overall pattern suggests a relatively high degree of exposure to the native population and a tendency toward spatial integration across most neighborhoods.

At the local scale, however, traditional settlement cores—particularly the Guro–Daerim cluster—continue to function as key loci of spatial differentiation, exhibiting comparatively elevated D, p, and entropy values as well as negative contributions to Theil’s H index. At the same time, the emergence of new micro-clusters in northeastern and central districts, such as Jayang 4-dong and Hyehwa-dong, indicates spatial diffusion beyond traditional industrial enclaves and highlights the growing importance of education- and service-oriented settlement processes. These patterns suggest that Chinese immigrants’ residential integration in Seoul is better characterized as selective and segmented assimilation shaped by institutional and structural constraints, rather than as a simple equilibrium diffusion process implied by classical spatial assimilation theory.

With respect to driving mechanisms, historical settlement patterns appear to exert a persistent influence on contemporary spatial configurations, indicating a path-dependent process in which established immigrant communities shape subsequent residential choices through housing availability, social networks, and informational channels. In addition, the spatial distribution of Chinese immigrants is influenced by the interaction of migrant network dynamics, urban redevelopment and regeneration policies, and cross-scale contextual factors that vary across neighborhoods, including industrial structure, educational institutions, and service environments. The resulting pattern reflects neither uniform assimilation nor systemic segregation, but rather a dynamic, polycentric, and locally clustered spatial configuration emerging from the interplay of global mobility, urban institutions, and intra-group differentiation.

From a theoretical perspective, this study provides empirical evidence for reassessing the applicability of spatial assimilation theory in high-density East Asian metropolitan contexts. In contrast to assumptions derived largely from low-density Western urban settings, the Seoul case suggests that residential assimilation is closely linked to housing institutions, redevelopment trajectories, and broader processes of socioeconomic stratification, rather than being a straightforward function of length of residence. The findings are consistent with the segmented assimilation framework, indicating that different immigrant subgroups—including labor migrants, students, and highly educated professionals—may follow differentiated spatial incorporation pathways within the same urban environment.

From a practical standpoint, the results offer important implications for urban governance and social inclusion in emerging immigrant destinations. The formation and persistence of Chinese immigrant clusters in Seoul are associated not only with cultural preferences or ethnic self-selection but also with institutional housing constraints, redevelopment-related displacement, and spatially uneven access to urban resources. Accordingly, efforts to mitigate spatial separation and promote inclusive integration may benefit from policy interventions that enhance immigrant access to affordable and stable housing through housing supply systems, rental regulations, and urban renewal strategies, while paying particular attention to minimizing displacement during redevelopment processes.

In sum, this study demonstrates that in a megacity such as Seoul, the residential patterns of Chinese immigrants cannot be adequately captured by a binary distinction between assimilation and segregation. Instead, they are best understood as a spatially heterogeneous intermediate condition characterized by broad integration, persistent enclave cores, and emerging functionally differentiated micro-clusters. This pattern contributes to the empirical literature on migration in East Asia and refines the boundary conditions of spatial assimilation theory, offering insights into immigrant spatial incorporation in high-density and institutionally structured urban contexts.

6.2. Limitations and Future Directions

Despite the systematic findings obtained through spatial econometric modeling and multiscale analysis, several limitations of this study should be acknowledged. First, data on the Chinese population in Seoul were derived from administrative registration records for 2011 and 2016 and from dynamic population estimates based on anonymized mobile phone signaling data for 2021 and 2025. Differences in statistical definitions and methodological approaches across these data sources may have affected the continuity of longitudinal comparisons and, to some extent, the consistency of the research findings. Second, reliance on aggregated administrative neighborhood-level data limits the direct examination of individual migration behaviors and residential decision-making processes. Third, due to data availability constraints, heterogeneity within the Chinese immigrant population—particularly in terms of legal status, length of residence, and socioeconomic characteristics—could not be fully differentiated.

Future research should integrate longitudinal micro-level data with qualitative research approaches to capture the dynamic mechanisms shaping immigrant residential trajectories more comprehensively. In addition, extending comparative analyses to other major metropolitan regions in East Asia would further enhance understanding of residential assimilation and spatial segregation patterns in emerging immigrant-receiving societies.