An Exploratory Review of Regional Perspectives on Social Capital and Occupational Studies

Abstract

Social capital is one of the most influential yet fragmented concepts in the social sciences. To gain insight into its substantive use within specific domains, this review explores how social capital (SC) has been applied in occupational studies, with particular attention to regional perspectives. Building on Adams and Fitch’s bibliometric mapping, it applies abstract-based topic modeling with citation data to identify thematic clusters and theoretical foundations within a corpus spanning over four decades. The results show that SC remains widely used across diverse themes. Citation patterns vary sharply across topics, with few sharing unified theoretical anchors. A closer look at studies on the topic of regional perspective reveals that SC is more employed as a background concept rather than through theorization or explicit operationalization. These findings refine Adams and Fitch’s conclusions on the fragmentation of SC research and highlight research opportunities for connecting SC mechanisms to regional perspectives.

1. Introduction

Are rural areas less occupationally segregated than large cities? Do the wider opportunities for social contact in urban environments foster greater inter-occupational mixing or, conversely, lead to fewer mixed contacts, as individuals can more easily form ties with occupationally similar others? This study takes these questions as its point of departure and asks to what extent regional perspectives have been studied in the research of social capital and occupation. Rather than revisiting broad debates over what social capital is or how it is operated, this study focuses on its practice in occupational studies and reads explicitly from the regional lens, intending to move the conversation from generic invocations of social capital to region-sensitive explanations.

It is well acknowledged that the concept of social capital (SC) is one of the most widely adopted yet conceptually diverse ideas in the social sciences (Erices-Ocampo et al. 2025). Since its theoretical formulation by Bourdieu (1985), Coleman (1988), and Putnam (1995), SC has been applied to examine and explain a broad array of social phenomena, including education, employment, health, and civic engagement. This theoretical flexibility has made it a popular umbrella term. Across fields, scholars have emphasized different components of SC, such as trust, reciprocity, social norms, or network structures, resulting in various interpretations and empirical applications. For instance, within sociology, researchers applying Putnam’s framework have alternately measured social capital as civic participation, trust, or community cohesion, which generated divergent results about its association with social integration (Paxton 1999). This has raised ongoing questions about how SC is actually used within and across different domains of research. To find an answer, many reviews have synthesized the expanding SC literature (e.g., Portes 1998; Quibria 2003; Koniordos 2008; Poder 2011; Carrillo-Alvarez et al. 2018; Ehsan et al. 2019; Alan and Köker 2021). Collectively, these works highlight the ambiguous nature of the concept: the wide range of ideas of social capital poses challenges for empirical research. Moreover, as the concept is expanded to explain a growing number of social phenomena, its heuristic precision and theoretical coherence are increasingly compromised.

Recent SC reviews have sought to use scientometrics to map the intellectual structure of SC scholarship. The latest research by Adams and Fitch (2023) is particularly noteworthy for introducing a citation network analysis that revealed the fragmented structure of SC research. Their study identified six broad clusters, including management, education, health, civic engagement, migration, and governance, each anchored in a distinct subset of foundational works, such as the cluster of management, which heavily relies on Burt (1992) and Coleman (1988). While this analysis offered insight into the empirical applicability of SC theories in a general sense, it provided little information about the substantive content of these clusters or how core theories are represented in specific clusters.

The present study takes an exploratory approach to a review of social capital in the subfield of occupational studies. An extended Latent Dirichlet Allocation (LDA) topic model is employed using both abstracts and citations. Unlike co-citation analyses limited to within-corpus citations, this approach considers cited references extended outside the corpus, allowing a more complete picture of how theoretical traditions circulate across topics. In addition to identifying the topics and the citations they include through the LDA approach, this study dives deeper into the regional perspectives of social capital by identifying the main keywords and citations in the designated topic.

This paper is organized as follows. Section 2 reviews the classic SC theories, summarizes existing systematic reviews of SC, and positions the present study. Section 3 describes the data and methods, including data collection, text processing of abstracts and references, and the extended LDA topic model with its evaluation. Section 4 reports the results by examining publishing and citation patterns across topics, and the role of regional perspectives in SC and occupational studies. Section 5 concludes with a summary and implications for future research.

2. Literature Review

To position the current study, this section first reviews the foundational theoretical traditions of SC and explains the two primary approaches of its application in research. It then summarizes existing systematic and bibliometric reviews, highlighting the fragmentation in definitions, particularly Adams and Fitch’s (2023) review, through a citation network approach.

2.1. Fundamental Theories of Social Capital

Social capital was first conceptualized in an explicitly sociological manner by Pierre Bourdieu in 1985. He introduced it alongside his ideas of economic and cultural capital as part of a broader framework for understanding different forms of “capital”. Building upon the idea of human capital from economics, the sociologist Bourdieu extended the logic of capital accumulation into the social domain and defined SC as “the aggregate of actual or potential resources linked to possession of a durable network of institutionalized relationships of mutual acquaintance and recognition” (p. 248). In this account, SC is generated and reproduced through investment in social relations and interaction that can maintain group membership and mutual recognition. In Koniordos’ (2008) review, Bourdieu’s SC is explicitly relational and class-conscious. That means SC is concentrated among the socially dominant class and serves to reproduce the existing hierarchies of privilege and class-based inequalities.

The next major step came from the sociologist Coleman (1988, 1990), who reinterpreted the concept within a rational choice framework. He defined social capital in functional terms as any aspect of social structures that facilitates certain actions of individuals or groups within that structure. Notably, social structure refers to normative and institutional arrangements, including trust, norms, obligations, and expectations that enable interaction and cooperation and achieve shared goals. He also suggested that SC differs from other forms of capital since it functions partly as a public good, which can benefit others beyond the actors who generate it and is present in family and community contexts.

A second line of theorizing is through the lens of social networks. M. Granovetter’s (1985) notion of embeddedness underscored the role of social ties in structuring economic and social action, distinguishing between the strength of weak and strong ties as different sources of information support. Building on Coleman’s conception, Lin (1999) also developed a network theory of social capital that formalized SC as a social resource embedded in social relations and network structures. His formulation keeps Coleman’s insight that social relations facilitate action but shifts the emphasis from normative cohesion to the instrumental mobilization of social contacts. Burt (1992) further emphasized the structural advantage of actors who bridge “structural holes”, gaps between otherwise disconnected actors, arguing that individuals who bridge such gaps gain informational and control benefits.

Within this expanding field, Putnam extended SC from a property of individuals or networks to a collective attribute of communities and nations in his books Making Democracy Work (Putnam et al. 1993) and Bowling Alone (Putnam 1995). He defined SC as “features of social organization—networks, norms, and trust—that facilitate coordinated action” (p. 67), thereby foregrounding collective, community-level outcomes rather than individual advantage. While his wording resembles Coleman’s formulation, Putnam argues that the concept should be operationalized at the level of collective entities, such as regions, cities, or countries, linking it to civic participation, institutional performance, and even public health. His analyses thus shifted the focus from interpersonal relations to aggregate social indicators. Putnam also proposed the concepts of bonding capital (strong ties among similar people) and bridging capital (weak ties among people from diverse groups), both of which are crucial for a cohesive society. Unlike Coleman, who locates social capital within specific relational contexts that generate trust and obligations among actors, Putnam proposed community cohesion itself as both the source and manifestation of social capital, transforming SC into a macro-level societal resource.

These theoretical traditions have given rise to two primary approaches in subsequent research. One, following Bourdieu and Lin, considers SC as an individually held resource embedded in social relations. In this view, SC can function as a zero-sum good; that is, individuals with initial network advantages can maintain and reproduce these advantages by leveraging their network positions. The other, associated with Coleman and Putnam, conceptualizes SC as a collective good that generates positive-sum outcomes, with trust and cooperation benefiting the broader community. As will be shown in the analysis, the persistence of these divergent perspectives has produced a diverse and fragmented body of scholarship, where the concept of SC is widely applied but not consistently defined.

2.2. Systematic Reviews of Social Capital Research

The distinctive theoretical perspectives of SC have prompted efforts to synthesize and systematize the literature. A number of systematic reviews and bibliometric studies have mapped the evolution, applications, and disciplinary spread of the concept (e.g., Quibria 2003; Koniordos 2008; Poder 2011; Erices-Ocampo et al. 2025). Quibria (2003) shows that divergent uses of the concept underlie much of the definitional ambiguity, which has serious ramifications for empirical analysis; Koniordos (2008) specifies basic ways of conceptualizing SC and traces how the concept is adopted by international organizations and national governments in its ideological use; Poder (2011) highlights persistent measurement problems of the concept and the context-dependence of indicators; and Erices-Ocampo et al. (2025) identifies three primary theoretical dimensions of scholars’ conceptualizations of SC: (a) where beneficial resources reside within individuals; (b) beneficial network structure, differentiating closure from brokerage arrangements; and (c) the level to which rewards accrue, distinguishing individual from collective benefits. Combining these dimensions produces a unifying perspective that fosters reintegrating SC’s disconnected conceptualizations.

One notable social capital literature review is the study by Adams and Fitch (2023), which adopts a citation network approach. Whereas most earlier reviews reflect on the ambiguity and diverse use of SC as a concept, Adams and Fitch reconfirm multiple SC concepts in the literature, which are differentially drawn by researchers in disconnected communities, by examining the citation pattern.

Their corpus consisted of 21,160 publications that explicitly included “social capital” in the title, keywords, or abstract, drawn from the Web of Science Core Collection and published between 1 January 1974 and 31 October 2020. They constructed a citation network in which each paper “received” a tie from the papers it cites and applied community detection to estimate the network’s modular structure. Communities represent subsegments of the literature that are more likely to cite within themselves and less likely to cite across clusters. Within this corpus, citation counts ranged from 0 to 4095, from which the authors identified 12 “key papers” with more than 1000 in-corpus citations. The 12 key papers are listed in

Table 1. Top-cited papers in Adams and Fitch’s corpus.

No.	Authors and Year	Title	Type	Outlet (Journal/Book)	Volume (Issue)	Pages
1	(Adler and Kwon 2002)	Social Capital: Prospects for a New Concept	Journal article	Academy of Management Review	27(1)	17–40
2	(Bourdieu 1986)	The Forms of Capital	Book chapter	In J. Richardson (Ed.), Handbook of Theory and Research for the Sociology of Education		241–58
3	(Burt 1992)	Structural Holes: The Social Structure of Competition	Book	Harvard University Press
4	(Coleman 1990)	Foundations of Social Theory	Book	Harvard University Press
5	(Coleman 1988)	Social Capital in the Creation of Human Capital	Journal article	American Journal of Sociology	94(S1)	S95–S120
6	(M. S. Granovetter 1973)	The Strength of Weak Ties	Journal article	American Journal of Sociology	78(6)	1360–80
7	(Lin 2001)	Social Capital: A Theory of Social Structure and Action	Book	Cambridge University Press
8	(Nahapiet and Ghoshal 1998)	Social Capital, Intellectual Capital, and the Organizational Advantage	Journal article	Academy of Management Review	23(2)	242–66
9	(Portes 1998)	Social Capital: Its Origins and Applications in Modern Sociology	Journal article	Annual Review of Sociology	24(1)	1–24
10	(Putnam 2000)	Bowling Alone: The Collapse and Revival of American Community	Book	Simon & Schuster
11	(Putnam 1995)	Bowling Alone: America’s Declining Social Capital	Journal article	Journal of Democracy	6(1)	65–78
12	(Putnam et al. 1993)	Making Democracy Work: Civic Traditions in Modern Italy	Book	Princeton University Press

.

They further examined how each community engages with the 12 key papers, both by tracing the distribution of citations across communities and by analyzing community-level co-citation patterns. They revealed six communities in the citation network: A—management, B—governance, C—health-related, D—education and childhood, E—civic engagement, and F—migration and community cohesions. Some communities drew on the key papers in similar ways, while others exhibited markedly different patterns. For example, the management-focused community A relied heavily on Adler and Kwon (2002), Burt (1992), Coleman (1988), and especially Nahapiet and Ghoshal (1998), whereas other clusters were more singularly anchored. Coleman (1988) dominated cluster D, oriented toward education and childhood, while Putnam (2000) was the central citation in another cluster associated with civic engagement (cluster E). Overall, these patterns indicate the absence of a unified conceptualization of social capital. Instead, multiple and partly disconnected conceptualizations circulate in the literature, each sustained by different subfields.

2.3. The Present Study

The systematic reviews and bibliometric review confirmed the conceptual ambiguity of social capital, empirical issues, and the expansion of SC across disciplines. In Adams and Fitch’s citation network approach, the labeling of communities was heuristic, inferred from journal categories and keywords rather than algorithmically derived from the textual content of the papers themselves. The present study contributes two new perspectives to the earlier studies by narrowing the focus to occupational studies and by applying a different topic modeling approach based on both abstracts and citations.

The application of topic modeling in bibliometrics allows us to go beyond the results of conventional bibliometric analyses, disclosing influence and co-citation structure, to obtain further information on the articles’ contents of a steadily expanding corpus. Topic modeling approaches have been applied to systematic reviews in various fields (Maier et al. 2018; Bai et al. 2021; Guillén-Pacho et al. 2024), but not yet to the review of SC research. Moreover, existing systematic reviews and topic-modeling studies have relied largely on abstracts and keywords, which capture the topical focus of research but not the theoretical foundations on which studies are built.

The occupational dimension of SC is relevant because of its close connection to questions of labor, work, and access to resources through social networks and links directly to stratification, opportunity structures, and labor market outcomes. Furthermore, this study places particular emphasis on regional perspectives, given that cultural, institutional, and occupational structures vary geographically and shape both the meaning and application of SC in practice.

3. Data and Method

This section describes the collection of 2408 articles from the Web of Science core collection and text preprocessing, including the preparation of abstract and cited reference data for each of the articles. This section then introduces the extended Latent Dirichlet Allocation topic model that jointly incorporates abstracts and references and explains the model evaluation procedures. Together, the methods provide a way for identifying thematic topics and their associated theories within the literature.

3.1. Data Collection

The source data were retrieved from the Web of Science (WoS) core collection on 13 March 2025. WoS was selected because it is widely recognized for applying a more selective and transparent journal curation process, which indexes high-quality, peer-reviewed publications.

Considering that the concept of SC began to be systematically theorized and empirically studied in the academic literature starting in the mid-1980s, the current study spans the literature from 1985 to 2025. The foundational works of Bourdieu (1985), Coleman (1988), and Robert Putnam (1995) introduced more formal definitions and operationalizations of the concept, which allowed for cumulative empirical research and gained prominence as a recognized field of inquiry.

Although the concept of SC is theorized, measured, and applied differently across disciplinary contexts, there are many publications from other domains, such as the natural sciences, biology, arts, and humanities, and the selection of articles was limited to those indexed in the Social Sciences Citation Index (SSCI) to ensure that the included studies are situated within the disciplinary scope of the social sciences.

To identify relevant studies that incorporate social capital theory in occupational contexts, specific words and phrases were used in the search, occurring in titles and abstracts. This generated a much smaller corpus than the one by Adams and Fitch (2023), who included articles from 1974 onwards and did not restrict themselves to specific topics related to occupation. Thus, for example, works related to civic engagement were not included in this study’s corpus. Details on the search and filter process are listed by steps in

Table 2. Search queries and refinement procedure.

Step	Results	Refinement
1	4589	TOPIC: ((TI = (“social capital” AND (occupation* OR job* OR “labor market*” OR “labour market*”)) OR TI = (“social network*” AND (occupation* OR job* OR “labor market*” OR “labour market*”))) OR (AB = (“social capital” AND (occupation* OR job* OR “labor market*” OR “labour market*”)) OR AB = (“social network*” AND (occupation* OR job* OR “labor market*” OR “labour market*”)))) TIME SPAN: 1 January 1985 to 1 January 2025
2	4261	Refined by LANGUAGES: (ENGLISH)
3	3579	Refined by DOCUMENT TYPES: (ARTICLE)
4	2408	Refine by Social Sciences Citation Index (SSCI) *

* Concretely, this excludes 1171 articles in journals indexed only in SCIE (e.g., Nature, The Lancet), AHCI (e.g., Critical Inquiry), or ESCI (e.g., Cogent Social Sciences). For journals cross-indexed in SSCI and another index, such as the Journal of Mixed Methods Research and American Journal of Public Health, they are included in the corpus.

.

As a result, 2408 records were retrieved. The retrieved articles were saved as research information system (.ris) documents, including full records1 (e.g., metadata—authors, title, abstract, keywords, cited reference, year of publication, source; times cited in the WoS core collection; cited reference count; and so on).

Only one article was removed from the dataset due to the objective absence of an abstract (Flap and de Graaf 1986). In total, 2407 documents were retained for analysis.

Figure 1 presents the distribution of collected articles over time (from 1985 up to and including the full year of 2024). As expected, there is a notable increase in the number of publications in this field over the past 40 years. Specifically, in the late 1980s to early 1990s, there are just a few publications per year. From the mid-1990s onwards, the number began to grow slowly and steadily. From 2005 to 2010, there is an increase in growth. After 2010, the trends present a continuing sharp increase and peaks around 2020 and 2022, where the annual publications reach about 160 articles. After 2022, the count declines slightly, though it remains high above 150. Part of the increase mirrors the overall expansion of global academic output, and the slight decrease after 2022 could reflect the impact of the COVID-19 pandemic rather than a true decline in research interest, when many research projects were delayed or disrupted during 2020 and 2021. Although a third-order polynomial might capture the non-linear increase and recent decrease in the overall publication trend, the limited number of recent observations does not allow firm conclusions about a structural decline in the number of publications after 2022.

3.2. Text Processing

3.2.1. Abstract Data

First, the abstracts of the 2407 academic articles were subsetted to build a corpus. Next, they were tokenized into 216,790 term components, following the instructions suggested by Maier et al. (2018). Specifically, the punctuation and numbers were removed, and the words were lowercased, which were then tokenized into single grams and their part-of-speech tags. Accordingly, the words were lemmatized into their base or dictionary form (lemma), for example, “studies” was transformed into “study”. Stemming, (i.e., chopping off prefixes or suffixes to reduce a word to its stem or root form, such as “studi” in the former case, or “immigration” being transformed to “immigrat”), was not used because stems are less interpretable in the later stage to identify latent topics in the literature. Then, basic stopwords, such as “the”, “is”, and “at”, that do not contain meaningful information for this review were removed (see Supplement Material S1 for the stopwords).

In addition, a semantic map was created to group related terms under one term to reduce redundancy and repetition, such as from “immigration” to “migrant”, from “young” to “age”, and from “women” to “gender” (See Supplement Material S2 for the semantic abstraction). Terms in over 80% of the documents and less than 0.5% of the documents were then removed to prune the corpus, as proposed by Maier et al. (2018) and Guillén-Pacho et al. (2024). This resulted in terms that are frequent and exhibit a unique power of representation in the documents.

In the final stage of text processing, a triple-blind human review of all terms was conducted by the first three authors, and the final 928 terms with a frequency of 2 or higher were selected. This removed 971 terms that were too general or too domain-specific (See Supplement Material S3 for all removed terms).

Figure 2 is a word cloud presentation of the most frequently used terms in the corpus, with 928 terms. The word cloud is dominated by terms like gender, migrant, work, support, education, and information. Terms relating to the SC mechanism, such as group, community, tie, resources, and structure, are mostly mid-sized. Regional or place-related words, such as urban, rural, community, local, area, country, and residence, also appear, though at smaller sizes, indicating their lower frequency. Other words, such as satisfaction, performance, career, and stress, representing typical occupational outcomes, are also less frequent.

3.2.2. Reference Data

To bring cited references into the corpus, the cited references were extracted from the metadata, which were coded in WoS format. It was not possible to obtain this information for 32 of the cited references of all 2407 articles. The references for the 32 articles were manually searched and coded into WoS format. The references were then standardized using their DOI numbers, which produced a set of 56,988 references. Initially, the references were tokenized into a simplified format consisting of the first author, year, and journal abbreviation, resulting in 145,417 entries. However, this approach proved unreliable, as author names in WoS format were coded in inconsistent ways, and equivalent terms were not recognized as identical.

To refine the citation data further, Crossref API, a public service that provides access to Crossref’s registry of DOIs and bibliographic metadata, was queried to match each reference’s author names and publication year to a unique DOI and then filter records by citation frequency. Specifically, the DOIs were pruned by applying the same thresholds of 80% and 0.5% occurrence rates, retaining only the references cited more than 13 times, the lowest frequency of selected terms, resulting in 314 references.2 It was found that applying the upper threshold (80%) had no additional effect: no single reference was cited in over 80% of the documents, which also confirms that the corpus contains no near-universal “top references” and that citations are relatively dispersed. This pattern underscores the dispersed nature of citations within the field, where many references appear only once or twice. Following the approach of Eklund and Nelhans (2017), the retained references were transformed into the form of “Author_Year_DOI” and integrated with the terms extracted from the abstracts, which completed the textual dataset for analysis. The top 20 cited references with DOIs are listed in

Table 3. Top 20 references with DOIs cited by the articles in the sample.

DOI	Frequency	First Author	Year	Ref.
10.1086/225469	396	Granovetter	1973	(M. S. Granovetter 1973)
10.1086/228943	278	Coleman	1988	(Coleman 1988)
10.2307/202051	154	Granovetter	1983	(M. Granovetter 1983)
10.1146/annurev.soc.24.1.1	152	Portes	1998	(Portes 1998)
10.2307/259373	137	Nahapiet	1998	(Nahapiet and Ghoshal 1998)
10.1146/annurev.soc.27.1.415	133	McPherson	2001	(McPherson et al. 2001)
10.1146/annurev.soc.25.1.467	120	Lin	1999	(Lin 1999)
10.2307/1519749	108	Mouw	2003	(Mouw 2003)
10.1145/358916.361990	103	Putnam	2000	(Putnam 2000)
10.1007/978-1-4615-1225-7_19	102	Marsden	2001	(Marsden and Gorman 2001)
10.5465/amr.2002.5922314	97	Adler	2002	(Adler and Kwon 2002)
10.2307/2095260	97	Lin	1981	(Lin et al. 1981)
10.1017/CBO9780511815447	96	Lin	2001	(Lin 2001)
10.1037/0021-9010.88.5.879	89	Podsakoff	2003	(Podsakoff et al. 2003)
10.1016/0378-8733(86)90003-1	83	Lin	1986	(Lin and Dumin 1986)
10.5465/3069452	83	Seibert	2001	(Seibert et al. 2001)
10.1257/0002828041464542	82	Calvó-Armengol	2004	(Calvó-Armengol and Jackson 2004)
10.1086/228311	81	Granovetter	1985	(M. Granovetter 1985)
10.1257/0022051043004595	77	Ioannides	2004	(Ioannides and Loury 2004)
10.2307/2657311	76	Bian	1997	(Bian 1997)

.

Note that the DOI-based extraction omits books and book chapters that commonly do not have a DOI number3. To supplement them, the six prominent books and book chapters identified by Adams and Fitch (2023) in Table 1 were included in a separate search. In this search, for each book, the references were matched only when the cited reference string contained the author name, publication year, and a distinctive title substring uniquely associated with the book or book chapter (e.g., Structural Holes, Foundations of Social Theory, Social Capital). This procedure ensured that same-year journal articles by the same authors were not classified as book citations. The resulting matches were recorded using “author, year” identifiers. Their global frequency in the corpus is displayed in

Table 4. Frequency of key books and book chapters cited by the articles in the sample.

Author, Year	Book or Book Chapter	Frequency
(Lin 2001)	Social Capital: A Theory of Social Structure and Action	241
(Burt 1992)	Structural Holes: The Social Structure of Competition	143
(Coleman 1990)	Foundations of Social Theory	137
(Putnam 2000)	Bowling Alone: The Collapse and Revival of American Community	109
(Putnam et al. 1993)	Making Democracy Work: Civic Traditions in Modern Italy	95
(Bourdieu 1986)	The Forms of Capital	28

below.

3.3. Extended Latent Dirichlet Allocation Topic Model

3.3.1. The LDA Model for Abstract and References

The Latent Dirichlet Allocation (LDA) topic model is used to identify latent themes within text (Blei 2003). The model assumes that each document can be classified as a probabilistic mixture of multiple topics, while a distinct distribution of words characterizes each topic. LDA estimates both the topic probability within documents and the word distributions within topics. Thus, LDA allows for reducing high-dimensional textual data into a lower-dimensional representation of topics, making it applicable in fields such as information retrieval and bibliometrics. Within the context of scientometric analysis, LDA has been applied as a practical approach for identifying the thematic structure of large text collections in different disciplines (Maier et al. 2018; Bai et al. 2021; Guillén-Pacho et al. 2024). More information on the LDA model is given in Appendix A.1.

Recent developments in bibliometric and scientometric research have extended the application of LDA beyond abstracts (or full texts) to the analysis of citation patterns. Eklund and Nelhans (2017), for example, applied LDA to citation data to map patterns of citations. As such, the approach in this study complements existing text-based and citation-based techniques for clustering research that might bridge the two approaches. Specifically, using the data from both abstract and cited references, the analysis incorporates citations alongside abstracts into LDA models, enabling a comprehensive mapping of how key theoretical contributions are situated within specific thematic domains.4 This method enriches textual data with citation elements and responds to calls also from the citation analysis literature to combine citation analysis with content information (Liu et al. 2013; Zou et al. 2021).

3.3.2. Model Setup and Evaluation

One common challenge in LDA is the determination of the optimal number of topics (K) that can be interpreted substantively in terms of topics. This study examines four complementary quantitative criteria to determine this number: CV and UMASS coherence score, topic diversity, and held-out perplexity, which jointly capture semantic interpretability and statistical fit, following previous works (see Maier et al. 2018 for more information on these criteria). Model estimation was carried out using the LdaModel implementation in the Gensim library (Python 3.11.12). The models were estimated with fifteen passes through the corpus and a fixed random seed to guarantee replicability, as the variational inference procedure of LDA relies on stochastic initialization and randomized updates whose outcomes can vary across runs, particularly when multiple passes over the corpus are employed. The model performance was evaluated over a range of candidate models with the number of topics (K) set from 1 to 30. The lower bound K = 1 serves as a reference model with only one topic against which gains in the statistical fits of models with more than one topic can be assessed. The upper bound (K = 30) was selected to remain within the commonly used 10–30 range in bibliometric studies while avoiding topic fragmentation and overly fine-grained topics that would make a larger number of topics difficult to interpret in a corpus of 2407 documents and would increase computational cost.

Figure 3 displays the distribution of four metrics over models with K from 1 to 30. A solution with 17 topics is likely to provide the appropriate balance, as perplexity exhibits the lowest curve, the CV coherence score achieves the second local maximum, the UMASS coherence score also improves after a drop-down with fewer topics, and topic diversity remains relatively high (over 70%), showing that the top 25 words are largely non-overlapping across topics.

Beyond statistical fit, K = 17 is adopted because it preserves a stable rural–urban topic across adjacent K, which indicates a clear theme on the “regional” perspective, and at the same time, it identifies citation-centric topics that are absent at lower K.

Table 5. Topic evolution: top terms by number of topics (K = 2–17).

K	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
1	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant	migrant
2	work	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender	gender
3		organization	support	support	work	child/family	child/education	child/family	child/education	child/family	child/entrepreneurship	child/family	child/education	child/user	child/family	child/family
4			tie	organization	organization	organization	organization	organization	organization	organization/user	organization/user	organization/work	occupation/class	organization/satisfaction	organization/satisfaction	organization/satisfaction
5				community	group/community	group/community	group/support/community	group/community	group/community	group/support	group/community	group/community	group/community	community/group	community/group	community/group
6					tie	professional	work	work	tie/search	tie	tie	search/tie	tie/search	search/wage	search/tie	tie/contact
7						search/tie	information/professional	professional/information	performance/information	performance/organization	performance/organization	performance/knowledge	performance/support	performance/organization	performance/work	search/wage
8							tie	risk	risk	risk	risk	teacher/risk	mobility/partner	teacher/resource	teacher/resource	mobility/education
9								search/tie	rural/teacher/	professional/online	online/information	information/	communicationn/skill	career/professional	professional/	skill/organization
									community			professional			information
10									skill/support	school/teacher	education	satisfaction/work	satisfaction/care	skill/mobility	employer/mobility	mobility/education
11										work/satisfaction	work/satisfaction	occupation/covid	work/user	occupation/structure	risk/class	risk/stress
12											rural/urban/community	rural/urban/community	rural/urban/community	urban/rural/area	urban/rural/community	urban/rural/community
13												tie/wage	career/networking	work/support	support/work	support/work
14													tie/(M. S. Granovetter 1973)	care/patient	skill/participation	participation/activity
15														activity/entrepreneurship	public/policy	economic/public
16															care/patient	care/patient
17																class/occupation

Note: K denotes the total number of topics specified in the model and does not indicate the ordinal position of topics. In subsequent analyses and in the main text, topic identifiers follow the original LDA output, where topic numbering may differ from the reordered presentation used in this table to illustrate topic evolution over time.

presents the evolution of the top word for each topic with increasing K. As expected, as K increases, broad themes fracture into more specific ones. A “rural” topic becomes identifiable at K = 10, yet this topic still blends with education (e.g., “teacher” ranks second). From K = 12 onward, the rural–urban theme stabilizes as its top terms remain largely unchanged across K = 11–17, while education-specific terms migrate to other topics. Regarding the citation terms, these are sparse with lower K and are absent from the top 30 topics. At K = 12, only Topic #6 (“tie”) carries noticeable citation terms, and this lineage is consistent with Topic #11(“tie”) at K = 17 after alignment. Two other distinct citation topics emerge at K = 17: #8 (“search”) and #14 (organization). The correspondence between topics identified in the abstract-only and abstract–reference models at the selected K is further documented in Appendix A.2.

4. Results

As mentioned in Section 3.3.2, the preferred LDA model estimates 17 latent topics. For each article in the sample, a dominant topic was assigned by selecting the topic with the highest probability in the article’s topic distribution5. Section 4.1 describes the frequency distribution of publications across topics over time and how the rural–urban topic changes over time. Section 4.2 presents how citations vary based on the identified research themes, and particularly the prominent citations in the rural–urban topic. Section 4.2, with a focus on specialization within the rural–urban topic, explores how the sample, methods, framing of social capital theory, and regional perspectives are covered in this niche and hidden theme.

4.1. Publishing Pattern of the Regional Perspective on Social Capital and Occupation Study over 40 Years

Figure 4 shows the proportion of publications per year for each of the 17 topics. In the limited number of publications from the years 1985 to 1995, there are a few topics, among which the topic leadership–risk–workplace and topic support–work–satisfaction dominate. With the rising number of publications since 1990, the number of topics increased as well, with varying “peaks”. For instance, in the years 2000 and 2002, the topic of community–group–residence dominated, whereas between 2007 and 2016, the migration topic was most important. The share of the urban–rural topic remained low and irregular across the entire period, with occasional isolated peaks in 1994 and 2003, but no sustained upward trend. This pattern suggests that while the urban–rural theme does surface in the literature, it does so episodically rather than forming a mainstream line of inquiry.

4.2. Citation Patterns Across the Themes of Social Capital and Occupation Study

This section examines the association between the citations and the topics for each of the 17 topics. A display of the probability distribution of top terms for all topics can be found in Appendix A.4.

Most topics do not have predominant citations among the top terms (Appendix A.4.1 and Appendix A.4.2). This means that referencing practices are uneven across the topics, and only a small subset of the identified topics is shaped by relatively unified bibliographies6. Moreover, the co-occurrence of cited references and the abstract’s words in a topic’s top terms is likely to reflect that those references are the anchor points or theoretical landmarks in that topic. In other words, the top references among the top terms indicate that this particular theme of study leans heavily on that key literature.

Figure 5a presents the distribution of the 20 top words of the selected topic, with the terms rural and urban having the highest probabilities. This suggests a focus on regional contexts, particularly comparisons or distinctions between urban and rural settings. Terms such as class, area, economic, household, city, and local follow closely and point to discussions that link location with socioeconomic conditions. Their relatively high weights indicate that analyses of local economic context, household circumstances, and class-related elements are common within this topic. The presence of income, informal, poverty, access, and land at mid-range probabilities suggests attention to material and infrastructural conditions that vary across regions. Some articles on this topic may relate to regional disparities in livelihoods and access to resources. Lower-probability terms like community, production, family, and migrant indicate that the regional focus possibly overlaps with discussions of community structures, family arrangements, and population movement.

Just like for other topics, in the rural–urban topic, no individual citation appears among the top twenty terms. The five most frequent citations, as Figure 5b shows, are Coleman (1988), Portes (1998), Woolcock and Narayan (2000), Massey (1990), and Massey et al. (1994) (see Appendix A.4.3 for the distribution of citations of the remaining topics). These works carry very low probabilities and function as a theoretical signpost for a very small subset of articles on this topic, which nevertheless deserve further exploration. Coleman (1988) emphasizes family- and community-level structures, which could be a flexible framework for understanding how social capital operates differently across local settings, applicable to studies comparing the social organization of rural and urban areas. Portes (1998) has a widely referenced overview of the SC concept, summarizing its sources and consequences. The presence of these two citations should be read as evidence that a small portion of the publications cite foundational theories of social capital, not as an indication that these works are central to urban–rural research specifically. The citations of Woolcock and Narayan’s (2000) work on social capital and development, and Massey’s (1990) and Massey et al.’s (1994) works on migration networks and cumulative causation mark the application of their theories on migration and community research. The relatively low probabilities of the corresponding terms (i.e., migrant, community) in the main term distribution in Figure 5a suggest that both of these works occupy a niche within the rural–urban topic.

Regarding the six prominent books and book chapters listed in Table 3, they were cited in 24 out of 139 articles under the rural–urban topic.

Table 6. Frequency of key books and book chapters cited by the articles in Topic 6 (rural–urban-community).

Author, Year	Book or Book Chapter	Frequency
(Lin 2001)	Social Capital: A Theory of Social Structure and Action	8
(Burt 1992)	Structural Holes: The Social Structure of Competition	3
(Coleman 1990)	Foundations of Social Theory	9
(Putnam 2000)	Bowling Alone: The Collapse and Revival of American Community	4
(Putnam et al. 1993)	Making Democracy Work: Civic Traditions in Modern Italy	5
(Bourdieu 1986)	The Forms of Capital	4

summarizes the distribution of citations to six foundational books and book chapters within articles assigned to the rural–urban topic. Coleman’s (1990) Foundations of Social Theory book and Lin’s (2001) Social Capital: A Theory of Social Structure and Action are the most cited books in this topic. This suggests that articles in the rural–urban topic may emphasize micro–macro linkage, rational action, and resource-based conceptions of SC when analyzing rural–urban and/or community processes. Citations to Putnam et al.’s (1993) works, Making Democracy Work and Bowling Alone (Putnam 2000), are present but more modest. This indicates that civic engagement and collective action frameworks are relevant to the topic but not dominant. The comparatively lower presence of Burt’s (1992) Structural Holes and Bourdieu’s (1986) The Forms of Capital suggests that network brokerage and class-based conceptions of capital may play a role in a few articles. These works are likely used to examine bridging ties, inequalities, or symbolic resources in rural–urban interactions.

4.3. The Regional Perspective on Social Capital and Occupation Study

This section examines the articles in the topic of urban and rural to understand how social capital is theorized, which occupational factors have been studied, and the role of regional perspectives in these studies.

Table 7. Top 19 articles in the distribution of topic probability over Topic 6.

Title	Author and Year	Journal
The limits to cumulative causation: International migration from Mexican urban areas	(Fussell and Massey 2004)	Demography
Growing food to grow cities? The potential of agriculture for economic and community development in the urban United States	(Vitiello and Wolf-Powers 2014)	Community Development Journal
Religious Diversity, Islam, and Integration in Western Europe-Dissecting Symbolic, Social, and Institutional Boundary Dynamics	(Koenig 2023)	Cologne Journal of Sociology and Social Psychology
Urban orchard in a megacity: formality and informality in China	(MacLachlan et al. 2026)	Eurasian Geography and Economics
Essential multiple functions of farms in rural communities and landscapes	(Milestad et al. 2011)	Renewable Agriculture and Food Systems
Neighborhoods and their impacts on the informal food economy of Bengaluru	(Anand and Jagadeesh 2022)	Cities
Cultural economy and the creative field of the city	(Scott 2010)	Geografiska Annaler, Series B: Human Geography
Structural Model of Community Social Capital for Enhancing Rural Communities Adaptation against the COVID-19 Pandemic: Empirical Evidence from Pujon Kidul Tourism Village, Malang Regency, Indonesia	(Prayitno et al. 2022)	Sustainability
Couples’ places of meeting in late 20th century Britain: Class, continuity and change	(Lampard 2007)	European Sociological Review
Touring as labour: mobilities and reconsideration of tour guiding in everyday life	(Chen and Chang 2020)	Tourism Geographies
Achieving inclusive urbanization through county-led industrial specialization: from the perspective of rural labor supply	(Hu et al. 2024)	China Agricultural Economic Review
Implications of Migration on Employment and Occupational Transitions in Tanzania	(Mueller et al. 2019)	International Regional Science Review
Small industrial towns in Moravia: a comparison of the production and post-productive eras	(Vaishar et al. 2023)	European Planning Studies
Thinking global but acting local: The middle classes in the city	(Butler 2002)	Sociological Research Online
From London to Los Angeles: a comparison of local labour market processes in the US and UK film industries	(Blair et al. 2003)	The International Journal of Human Resource Management
Layers, flows and intersections: Jeronymo Jose de Mello and artisan life in Rio de Janeiro, 1840s-1880s	(Frank 2007)	Journal of Social History
Migration, Informal Labour and (Trans) Local Productions of Urban Space—The Case of Dhaka’s Street Food Vendors	(Etzold 2016)	Population, Space and Place
Impacts of Chinese Urbanization on Farmers’ Social Networks: Evidence from the Urbanization Led by Farmland Requisition in Shanghai	(Xu et al. 2016)	Journal of Urban Planning and Development
Do rural migrants benefit from labor market agglomeration economies? Evidence from Chinese cities	(Yang et al. 2020)	Growth and Change

gives an overview of the first 19 articles assigned to Topic 6 with a probability over 0.67. Most articles were published in recent years, across various journals in sociology, geography, and labor economics. Among these articles, four of the top five citations associated with this topic, as shown in Figure 5b, are cited by Fussell and Massey (2004) and Vitiello and Wolf-Powers (2014), respectively. The six prominent books and book chapters in Table 3 are not cited by any of the top 19 articles.

Second, we consider that the top-ranked papers contain the most unambiguous thematic expression of the topic. The first 10 papers are selected for a qualitative review, which includes six qualitative and three quantitative studies, as well as one mixed study.

Table 8. Systematic review of the top 10 rural and urban articles.

Author (Year)	Sample	Method	Social Capital/Network	Occupational Outcome/Factor	Regional/Spatial Perspective	Core Theories Used in the Paper
Fussell and Massey (2004)	Mexican Migration Project (households of Mexican communities)	Event-history models	Community- and family-level social capital (social ties to rural-based migrants)	None (outcome is migration likelihood; occupation as a personal economic indicator)	US immigration with Mexican origins (urban and rural origin communities)	The theory of cumulative causation (Myrdal 1957)
Vitiello and Wolf-Powers (2014)	Field research across six U.S. cities (municipalities, nonprofits, residents)	Qualitative case studies/fieldwork	None (only as an implication of agricultural development on social capital)	Job creation, workforce integration pathways	United States (urban)	Theory of consumption-driven urban economic development (Markusen and Schrock 2009)
Koenig (2023)	Muslim minorities	Conceptual synthesis (theory/review)	Religion-based segregation in social networks	Religion-based labor market disadvantages contribute to social boundaries	Western Europe	Theories related to the boundary paradigm (Loveman and Muniz 2007)
MacLachlan et al. (2024)	Single U-pick lychee orchard case in Pearl River Delta, China	Qualitative observations and interviews	Informal social network in urban agriculture	Informal economic activity and formal occupations coexist	Densely populated urban environment, China	None
Milestad et al. (2011)	Four Swedish rural communities (village action groups and farmers)	Qualitative semi-structured interviews	Rural social capital via village action groups supporting agriculture and community functions	Job creation and the local economy	Sweden (rural)	Theory of multifunctional agriculture (Wilson 2007)
Anand and Jagadeesh (2022)	Informal food businesses in different neighborhoods	Mixed (survey and interviews)	Social network creating vending opportunities	Occupational mobility of vendors	India (Bengaluru); land typologies	Theory on informal enterprises (e.g., Hosier 1987)
Scott (2010)	None	Theoretical synthesis	Social networks that bind workers together in urban space	Local creative industry labor market	US cities, urban economy	Theories related to the creative field of the city (e.g., Landry and Bianchini 1995)
Prayitno et al. (2022)	Residents of Pujon Kidul Tourism Village	Confirmatory factor analysis and Structural Equation Modeling (AMOS)	Measures trust, norms, and networks (i.e., willingness to build cooperation, participation in religious activities, etc.); analyzes effects on resilience	Indirect: community resilience as the outcome sustaining village tourism/agritourism jobs	Indonesia (Malang Regency)	Theories related to community resilience (e.g., Nugraha et al. 2021)
Lampard (2007)	Great Britain couples	Quantitative analysis of survey data	Social network as a context for meeting	Occupational class homogamy patterns	Great Britain	Theory on places of meeting (Bozon and Héran 1987)
Chen and Chang (2020)	Tour guides	Qualitative interviews	Tour guides’ relational ties in their work	Tour guiding labor conditions and everyday employment realities	China (south)	Theories of tourism mobilities (e.g., Sheller and Urry 2006)

presents a summary of the sample, method, framework of social capital/network, occupational outcome, and regional perspectives. It can be seen that these articles cover a wide range of methodologies, from qualitative studies (e.g., field work, observation, and interviews) to quantitative analysis (e.g., national surveys) to mixed studies with both interviews and surveys. The sample, or unit of analysis, is also diverse, varying from individual to household and city. The quantitative studies (including the only mix study) touch on (1) direct labor market outcomes, such as employment, wage, and stability, (2) micro- or self-enterprise, including street vending (Anand and Jagadeesh 2022), smallholder/urban agriculture (MacLachlan et al. 2026), and (3) homogeneity that is based occupational class positions, such as homogamy (Lampard 2007). Qualitatively, researchers focus on (4) occupational experience and processes. For example, Anand and Jagadeesh (2022) study the occupational mobility of vendors from small businesses. Chen and Chang (2020) focus on working conditions and the interpersonal experience of a specific occupation.

From the regional perspective, there are three lenses in the articles. First, some are region-specific, which highlights analysis as a single locale or a particular geographical unit, such as the UK (Lampard 2007), Western Europe (Koenig 2023), a village (Milestad et al. 2011; Prayitno et al. 2022), or a particular city (Vitiello and Wolf-Powers 2014). By this design, the regional dimension comprises the geographical basis of the case study. Second, regions in some studies are considered as institutional, industrial, or spatial contexts beyond geographical units. For example, MacLachlan et al. (2026) examine the influence of formal regulation by local authorities, which is particularly pronounced in urban land of China, and Anand and Jagadeesh (2022) explore the effects of land typology in different neighborhoods. Third, there is a comparative lens in terms of the mechanisms, effects, or contexts between rural and urban regions. One example is Fussell and Massey’s (2004) work on comparing the migration likelihood of people with rural and urban origins.

By checking the citations, it is found that social capital and networks are more often used as implicit and descriptive concepts than as measured quantities. Across the 10 articles, there are only two articles that include explicit citations to the foundational social capital/network literature. First, Fussell and Massey (2004) cite Coleman (1988) and Bourdieu (1986) to formalize cumulative causation mechanisms, whereby within-community ties lower the costs of migration and shape subsequent labor market entry. Their work measures the quantities of SC as community-level social capital, which is measured as the migration prevalence ratio, and as family-level social capital, which is measured based on two ties: whether the respondent’s parents had prior U.S. experience and whether any sibling had ever been to the United States.

Second, in Vitiello and Wolf-Powers’ (2014) qualitative study, social capital is still used as a broad concept implicitly. They cite both Coleman (1988) and M. Granovetter (1985) in the context below:

“Economic security and workforce integration are often intertwined, aided by social capital—embedded norms and relationships that provide information, moral and material support (M. Granovetter 1985; Coleman 1988).”

(Vitiello and Wolf-Powers 2014, p. 8)

Clearly, their citation context does not further provide any justification in terms of the theoretical framework of their research. Their citation invokes social capital rather than building on or testing social capital theory. Social capital is used as a commonsense umbrella, where social networks can help people get information and other benefits, similar to the widely used concept of “human capital”.

While screening the other two quantitative research studies, Prayitno et al. (2022) operationalize community social capital based on trust and norms, such as willingness to build cooperation and participation in religious activities, without explicit citations of the classical social capital literature. Lampard (2007) does not operationalize social capital or network but mentions that social capital research indicates class differences in the opportunities to meet people through involvement in voluntary associations and less formal networks. These citations only use social capital to provide rhetorical legitimacy and background framing, which is not a formal theoretical framework for analysis. Even with the measurement, it is considered an aggregated quantity.

As for the other qualitative studies, they do not have explicit citations to the foundational social capital/network literature, despite the general use of SC as relational resources in place-specific contexts. For example, Koenig (2023) and Scott (2010), in their theoretical synthesis, refer to social capital as a network property that associates with (1) religion-based segregation that generates labor market disadvantages and (2) interpersonal contact at the workplace in urban space to maintain up-to-date job-related information. Across other interview studies, MacLachlan et al. (2026) show how informal networks in urban agriculture enable the coexistence of informal and formal work. Milestad et al. (2011) highlights collective rural social capital, organized through village action groups, in supporting agriculture and local job creation. Anand and Jagadeesh (2022) examine how social networks create vending opportunities and facilitate occupational mobility among informal food vendors in Bengaluru, India. Chen and Chang (2020) focus on relational ties among tour guides, showing how everyday network relations affect employment conditions in tourism work in South China.

Beyond the first 10 articles in the rural–urban topic, we further examined the citation patterns of the remaining 129 articles. Among these, 32 articles explicitly cite classical social capital theorists: 12 cite Coleman, four cite Putnam, five cite Bourdieu, three cite Lin, and 23 cite Granovetter. Of these 32 articles, only 16 explicitly apply a social capital framework in their empirical analyses, including five quantitative studies, 10 qualitative studies, and one mixed-method study. Among the quantitative contributions, three operationalize social capital using Lin’s network-based framework. For example, Lu et al. (2013) measure individual-level social capital through the indices of extensity, upper reachability, and range, while Song (2020) and Song and Pettis (2020) adopt Lin’s network approach to capture access to social resources. The other two studies draw on alternative theoretical perspectives: Shui et al. (2022) conceptualize social capital as access to broader societal institutions following Coleman’s framework, whereas Sunikka-Blank et al. (2019) operationalize three forms of social capital through norms or values, information channels, and collective expectations inspired by Putnam’s formulation. Overall, these patterns indicate that although classical SC theories are cited within the remaining articles in the rural–urban topic, only a minority of studies explicitly operationalize the concept within a coherent theoretical framework. Network-based perspectives, drawing on Lin, seem more present in the remaining minority of studies.

5. Summary and Conclusions

The current study explored 40 years with increasing numbers of publications on the application of the concept of social capital in occupational studies, with a special interest in the area of regional studies. In a corpus consisting of abstracts and citations of 2407 publications with DOIs, the publications were classified as belonging to 17 different topics, one of which was identified as regional. Because books without DOIs were not included in the corpus, their citations were separately counted. The inclusion of citations with DOIs in the corpus underscored the diversity of the topics, with limited references to classic social capital theory. The regional topic had rural and urban as the most frequent terms and occupies a modest position in the field. In a closer scrutiny of the design and theoretical foundation of the ten papers most strongly associated with the rural–urban topic, it was found that social capital was implicit in most papers (operationalized by social networks) and only referred to Coleman (1988), Bourdieu (1986), and M. Granovetter (1985) in two papers. In most papers, social capital or social networks were generally used as the instrumental, relational resources in place-specific contexts to motivate the study or derive hypotheses, without explicit citations.

These findings both complement and refine the insights of Adams and Fitch (2023). While their citation network analysis demonstrated that social capital research can be divided into six broad and relatively disconnected clusters, the present study uncovered a finer-grained thematic structure within one specific area, the occupation domain. Whereas Adams and Fitch’s heuristically labeled clusters, this study identified topics through a topic modeling approach based on textual content. The results confirm that the weak integration of theoretical anchors, as observed by Adams and Fitch, persists within a single applied domain of social capital research. Thus, social capital continues to serve as a general concept rather than as a specified theoretical or analytical framework.

Notwithstanding the perhaps underrepresented social capital citations, our findings suggest that the field has reached a stage of diversification without corresponding theoretical consolidation. In regionally oriented occupation research in particular, SC is frequently cited as a general relational resource, rather than specified as a mechanism linking spatial context to occupation factors relevant to the questions raised at the outset of this study: whether rural areas are less occupationally segregated than large cities, and whether wider opportunities for social contact in urban environments foster greater inter-occupational mixing or instead reinforce homophily.

Future research would therefore benefit from articulating region-specific mechanisms, such as spatial opportunity structures, cumulative causality, or institutional dependency, that capture how spatial context shapes social capital formation, network structures, or resource distribution. Advancing this agenda requires moving from generic use of social capital to region-sensitive and mechanism-based explanations. Future research can more directly address the questions of occupational segregation and inter-occupational mixing that motivated this study and contribute to a more theoretically integrated empirical research agenda.