PaperlayerView exact revision on arXiv

Source-linked AI summary

Social Network Structure, Wealth, and Wealth Inequality Across Cultures

Eleanor A. Power, Monique Borgerhoff Mulder, Samuel Bowles, Matthew O. Jackson, Jeremy Koster, Daniel Redhead, Thomas Rutter, Sahana Subramanyam, Justin Weltz, Nurul Alam, Sarah Alami, Alexandra Alvergne, Curtis Atkisson, Michele Barnes, Bret Beheim, Christine M. Beitl, Madeline Brown, Mark Caudell, Wendy Chávez-Páez, Komal Chauhan, Joshua Cinner, Siobhán Cully, Augusto Dalla Ragione, Angelina L. DeMarco, Ivan Deschenaux, Federico Fernandez, Juan Pablo Ferreiro, Drew Gerkey, Matthew Gervais, Christopher Golden, Gianluca Grimalda, Werner Hertzog, Paul L. Hooper, Karen Kramer, Geoff Kushnick, Banrida Langstieh, Rodrigo Lazo, Sheina Lew-Levy, Shane Macfarlan, Emmanuel Maliti, Karl J. Mertens, Madalena Monteban, Rafael Morais Chiaravalloti, Daniel Murphy, Kathryn Oths, Alejandro Pérez Velilla, Emily Post, Sean Prall, Cody Ross, Anirudh Sankar, Brooke Scelza, Michael Schnegg, Edmond Seabright, Mary K. Shenk, Kathrine E. Starkweather, Chun-Yi Sum, Bram Tucker, Bapu Vaitla, Vivek Venkataraman, John P. Ziker

arXiv:2608.25488v1cs.SIecon.GN

TL;DR

Research has lacked broad cross-cultural evidence on how social-network structure relates to wealth and inequality. Using household wealth and multiplex support-network data from 46 communities across 28 countries, the study finds that wealthier households are better connected and that poorer households are less connected to wealthier ones in more unequal communities.

  • Problem

    Evidence on how wealth and social connections relate across diverse human livelihoods remains limited.

  • Method

    The study combines household asset inventories with directed, weighted support networks collected across 46 diverse communities in 28 countries.

  • Results

    Wealthier households are better connected, provide and access more support, and tend to connect with wealthier households; poorer households interact less with wealthier ones in more unequal communities.

  • Takeaways & Limitations

    The findings show that economic connectedness is consistently associated with wealth and wealth inequality across highly diverse communities.

  • Takeaways & Limitations

    The analysis focuses on within-community connections and does not capture resources and opportunities available through ties beyond the local community.

Abstract

from arXiv · show

Despite theory tying wealth inequality to social structure, empirical evidence has been limited to a few studies based on online social media data. This study uses a very different type of data, expands the global coverage to very different types of societies, and investigates new questions. In particular, we collect data from ~3500 sharing units (households) in 46 communities across the globe, representing considerable human social and cultural diversity. In each, we analyze the relationship between people's material wealth and the structure of social networks: borrowing money, sharing food, working together, socializing, etc. In almost all communities, a sharing unit's material wealth is positively associated with the number of other sharing units it both helps and is helped by. A sharing unit's wealth is also associated with the relative wealth of the sharing units to which it is linked---a form of economic homophily. Notably, communities with greater wealth inequality are also characterized by a network structure in which poorer sharing units are less well connected to wealthier ones. We augment our unique cross-cultural data with other community-level environmental, institutional, and economic attributes, opening new avenues for future research into the co-determination of wealth and social networks.

1 Introduction

This study examines how household material wealth relates to social-network connections within 46 diverse communities across 28 countries, and how community network structure relates to wealth inequality. Its contribution is a cross-cultural dataset combining detailed household possessions with multiplex social relationships.

  • Study scope: The study analyzes economic, demographic, and social-support network data from 46 mostly rural communities across 28 countries.The communities span different stages of market integration and diverse livelihoods.
  • Research questions: The analysis investigates whether sharing-unit wealth relates to network connections and whether community network structure relates to wealth inequality.It focuses on within-community household relationships and across-community associations between network structure and inequality.
  • Main contribution: The study reports strong and generally consistent relationships between material wealth and network position across highly diverse communities.It also considers whether differences in these associations relate to community attributes such as wage labor and private-property-rights inclusiveness.
  • Conceptual framework: The paper distinguishes support solicitation, support provision, and economic connectedness as three network-based forms of social capital.Economic connectedness concerns access to relatively wealthy others.
  • Data contribution: The dataset combines household inventories of possessions with multiplex relationships involving food sharing, information exchange, and money lending.These relationships are presented as important to daily lives and livelihoods.

2 Fieldsites and Data

The study draws on diverse data from 46 communities worldwide, using sharing units as the resource-pooling unit. Researchers combine detailed wealth inventories with locally meaningful social-support nominations to construct network measures of social capital and connected-unit wealth.

  • 46 communities worldwide span substantial variation in geography, infrastructure, services, population density, and subsistence practices.Most communities are rural and remote, but some lie in high-density areas near highways and government services.
  • Sharing units average four to five individuals, with about 60 sharing units per community, and represent the units in which resources are pooled.Their membership varies across communities, from nuclear families to other household configurations.
  • The cross-cultural design contrasts with prior inequality studies and enables locally meaningful measures of social connections across multiple interaction types.The study constructs measures of access to support, provisioning of support, and Average Alter Wealth among support connections.
  • Detailed demographic, economic, and social data were collected from virtually all sharing units in each community.Researchers enumerated property and assets to derive material wealth and recorded nominations of people consulted for different types of support.

3 Within-Site Results: Sharing Units’ Wealth and Social Capital

Within communities, wealthier sharing units generally have greater support access and provisioning, and their social-capital ties are positively associated with wealth. Wealth is also related to the wealth of connected sharing units, with these associations remaining robust across alternative measures and network constructions.

  • Within-Site Results: Sharing Units’ Wealth and Social Capital: Across the majority of communities, wealthier sharing units have greater support access and are more often called on to provision support.These relationships are assessed using percentile-ranked material wealth and support access or provisioning per capita.
  • Within-Site Results: Sharing Units’ Wealth and Social Capital: Wealth is associated with the material wealth of connected sharing units, capturing a form of economic or wealth homophily.The study measures economic connectedness using the average wealth of a sharing unit’s alters.
  • Within-Site Results: Sharing Units’ Wealth and Social Capital: For all four network measures of social capital, positive correlations occur across communities more often than expected, with positive and statistically significant cross-community average effects.The cross-community averages are estimated using random-effects meta-analyses to account for variation in relationship strength.
  • Within-Site Results: Sharing Units’ Wealth and Social Capital: The associations remain robust to alternative wealth accountings, network constructions, and versions of economic connectedness.Alternative wealth measures include unweighted absolute measures, adult-weighted measures, and controls for sharing-unit size.

4 Cross-Site Results: Communities’ Wealth Inequality, Network Patterns, and Other Characteristics

Across communities, wealth inequality is not consistently associated with inequality in support access or provisioning, but it is strongly negatively associated with poorer sharing units’ economic connectedness. Other community and national characteristics show some positive associations with wealth inequality, although the limited correlations preclude confident conclusions and causal inference.

  • Support inequality: Wealth inequality has no consistent association with inequality in sharing units’ support access or support provisioning.The support-access correlation is positive and borderline significant under one wealth accounting but is not consistent across alternative accountings.
  • Economic connectedness: Higher community wealth Gini coefficients are strongly and significantly associated with lower economic connectedness among poorer sharing units.Relative Average Alter Wealth measures below-median sharing units’ supporters’ wealth relative to above-median sharing units’ supporters; a value of 1 indicates equal alter wealth.
  • Economic connectedness: Robustness analyses using alternative wealth-class splits and multilevel models suggest that connections betw
  • Other correlates: Three variables show significant positive associations with wealth inequality: wage or salaried labor time, a national private property rights index, and marginally, the Human Influence Index.The reported relationships concern communities with more wage and salaried work, countries with more widely available private property rights, and communities with higher Human Influence Index values.
  • Interpretation and limitations: Because the analysis includes many variables and limited correlations, relationships between other characteristics and inequality cannot be confidently identified or ruled out.The authors present Lasso and multivariate regressions as supplementary analyses and reserve deeper causal investigation for future longitudinal data.

5 Discussion

Across diverse communities, wealth is closely tied to social capital and economic connectedness, while wealth inequality is associated with weaker links between poorer and wealthier sharing units. The discussion highlights plausible mechanisms, broader contributions, and priorities for longitudinal and network research.

  • Core findings: Wealthier sharing units are better connected, receive and provide more support, and tend to link with wealthier sharing units.These patterns indicate that material wealth and social capital are closely tied within communities.
  • Contribution: The study broadens evidence on economic connectedness across geographic and social contexts using relationships spanning social and economic exchange.Its findings are consistent with prior evidence linking economic connectedness to upward income mobility.
  • Wealth homophily: Wealth homophily varies across contexts, with caste segmentation, state policies, charity, and egalitarianism offering plausible explanations for cross-wealth connection patterns.The authors argue that new theoretical frameworks are needed to explain the range and patterning of wealth homophily and its association with inequality.
  • Wealth inequality: Stronger private property rights and greater participation in wage or salaried labor are positively associated with wealth inequality.The discussion suggests possible links to wealth monopolization and a dual economy, while emphasizing that causal pathways require further investigation.
  • Future research: Future research should distinguish support types and examine which cross-wealth connections produce dependency, exploitation, mutuality, or other consequences.The current analysis aggregates multiple support relationships, limiting assessment of how particular ties bridge wealth disparities.
  • Future research: Future work should examine extra-community ties and intra-household dynamics as resources and mechanisms shaping wealth accumulation and distribution.The authors identify longitudinal data, external connections, household composition, and unequal sharing of benefits as important open questions.

6 Methods

The study combines near-complete community surveys with asset-based wealth inventories and social-network data aggregated at the sharing-unit level. It derives unit- and community-level measures and analyzes within-site and cross-site relationships using correlations and complementary models, while withholding the underlying data because of privacy constraints.

  • Data collection: Researchers surveyed effectively all sharing units in each community, seeking at least one adult representative per unit where possible.
  • Measurement: Material wealth aggregates each sharing unit’s property and assets from extensive inventories, with estimated values converted into US Dollars for comparison.Inventories include land, livestock, household goods, and tools.
  • Network construction: Researchers construct weighted, directed sharing-unit networks from reported social connections spanning tangible aid, behavioral assistance, socializing, and information sharing.The composite network aggregates individual responses to the survey questions.
  • Network measures: Support access is the out-degree of the network, whereas support provisioning is the in-degree, because ties point toward requested support and resources flow oppositely.
  • Community measures: Community measures include wealth and support Ginis, plus Relative Average Alter Wealth, which normalizes poor sharing units’ average alter wealth by that of above-median units.
  • Analysis: Within-site analyses use percentile-rank correlations and pooled random-effects meta-analysis, while cross-site analyses primarily use bivariate correlations with complementary multilevel and regression models.
  • Data availability: The underlying data cannot currently be shared because community-specific consent processes include varied assurances of privacy and access.Site-level aggregate measures are provided in the Supplementary Information.

Author Affiliations · A Supplementary Information · A.1 Data Collection

The ENDOW project assembled cross-cultural data on sharing units, wealth, and social networks through collaborating researchers working in diverse communities. Data collection occurred across sites from 2017 to 2025, using broad community coverage and multiple dimensions of wealth.

  • A.1 Data Collection: ENDOW began in 2017 to support cross-cultural comparisons of inequality by involving researchers contributing data from their field communities.Because the sample depended on willing researchers with active fieldsites, inferences primarily concern a wide range of communities rather than a representative global sample.
  • A.1 Data Collection: Researchers reported each community’s primary and secondary subsistence modes as part of the cross-cultural dataset.These characteristics are summarized in Figure A1 across the represented communities.
  • A.1 Data Collection: The sharing unit is the fundamental unit of analysis, representing the unit to which resources are principally pooled and shared.It often, but not always, corresponds to a household, and sharing units average four or five members across ENDOW communities.
  • A.1 Data Collection: Each community’s sampling and sharing-unit characteristics were summarized alongside material wealth measured in U.S. dollars around the time of data collection.The summaries include sharing-unit membership, network-question sampling, and wealth distributions across communities.
  • A.1 Data Collection: Researchers gathered information on virtually all sharing units, achieving average community coverage of 96%.Measures included material wealth, embodied wealth, and relational wealth, encompassing resources, skills, and access to social support networks.
  • A.1 Data Collection: Data collection was conducted between 2017 and 2025, with timing determined by community-member availability and researchers’ scheduling constraints.Seasonal constraints affected when data could be collected, and communities joined the project at different dates.
  • A.1 Data Collection: Each community received relevant ethical approval, with an overarching review conducted by the University of Cincinnati Institutional Review Board.The overarching review was identified as Study ID 2016-4691.
  • A.1 Data Collection: Submitted data files underwent completeness and error checks, followed by researcher clarification of ambiguities or missing information.Checks included identifying implausible age differences between mothers and children.

A.2 Material Wealth Data · A.3 Social Network Data

Material wealth is measured from valued household assets, while relational wealth is captured through ten name-generator questions covering economic exchange and support relationships. Wealth distributions are examined using total, per-capita, and per-adult measures, with imputation used for incomplete enumerations.

  • A.2 Material Wealth Data: Researchers exhaustively enumerated economically productive property and material assets for each sharing unit, excluding items such as clothing from detailed enumeration.Relevant items were identified through resident conversations and visits to homes in each community.
  • A.2 Material Wealth Data: Researchers assigned current market or replacement values to enumerated items, including rebuilding-cost valuations for houses when market prices were unavailable.Values were determined through resident conversations and visits to local markets.
  • A.2 Material Wealth Data: Total sharing-unit wealth sums each enumerated quantity multiplied by its associated item value, wi = Pk qik ·pk.The primary measure is the total value of enumerated assets, including baseline items.
  • A.2 Material Wealth Data: Incomplete item enumerations were handled with multivariate imputation via chained equations, producing 50 datasets whose median supplied the paper’s baseline value.The paper evaluates robustness to this imputation approach in Section D.2.
  • A.2 Material Wealth Data: Cross-site inequality patterns persist whether wealth is measured at the sharing-unit, per-capita, or per-adult level.Lorenz curves compare sharing-unit-, adult-, and person-level wealth distributions across sites.
  • A.3 Social Network Data: Relational wealth is measured with ten name-generator questions, combining standardized and community-tailored wording presented in a fixed order.Further methodological detail is provided in a forthcoming companion paper.
  • A.3 Social Network Data: The first four network questions covered borrowing and lending a week’s wages plus asking for and giving common goods in sharing-unit exchanges.The wage amount was set in local currency, and goods questions were site-specific and double sampled.

A.3.1 Network Construction

The main analysis constructs a weighted, directed composite network of sharing units by aggregating person-level support nominations across question layers. Because sampling and sharing-unit size affect nomination opportunities, the study accounts for aggregation choices, tests proportional and raw alternatives, and excludes double-sampled prompts.

  • Aggregation considerations: Sampling strategy and sharing-unit size create unequal opportunities to name support partners, requiring choices about aggregating person-level responses into sharing-unit networks.The aggregation must account for questions asked to specific people and variation in who was sampled within sharing units.
  • Composite network: The main network is weighted and directed, with sharing units as nodes and layer weights normalized by the number of surveyed individuals in the nominating unit before summing across layers.For example, two respondents naming three and two people in another unit produce a borrowing-money layer weight of (3 + 2)/2 = 2.5.
  • Alternative networks: The proportional composite network replaces nomination counts with the proportion of nominating individuals in sharing unit i who nominate any member of sharing unit j.For each respondent in sharing unit i, the presence of nominations to sharing unit j is represented as a binary value of 0 or 1.
  • Alternative networks: The raw composite network retains directed, weighted sharing-unit nodes but assigns each layer a weight equal to the total number of nominations from sharing unit i to sharing unit j.The study reports that results are similar when using the proportional composite and raw composite networks.
  • Prompt selection: The analysis omits double-sampled name generators because of weighting complexities, restricting the network to prompts asking whom respondents would make requests of.The excluded prompts asked whom respondents would lend a week’s wages or household items to; incorporating them is planned for future work.

A.4 Measures Constructed from Network and Wealth Data … B.1 Different Network Constructions

The paper constructs wealth-sensitive network measures, assembles community-level environmental, institutional, economic, and ethnographic attributes, and tests whether results persist across alternative network constructions. Across proportional and raw composite networks, the reported results are similar to the main findings.

  • A.4 Measures Constructed from Network and Wealth Data: The analysis uses node wealth attributes and edge-weighted network ties to construct measures of alters’ wealth and economic connectedness.Wealth may be measured as absolute wealth, wealth per adult, wealth per capita, or size-adjusted wealth.
  • A.4 Measures Constructed from Network and Wealth Data: Average Alter Wealth is calculated separately for poorer and richer sharing units, with relative average alter wealth normalizing the poorer group by the richer group.The measures can also be constructed separately for supporter and supportee tie directions.
  • A.5 Community-Level Information: Community-level analyses treat production processes, environmental factors, and work arrangements mainly as potential confounders of inequality–network relationships.These attributes are supplemented by national, local geospatial, and researcher-provided information.
  • A.5.1 National Attributes: National attributes include GDP per capita, the national Gini coefficient, adult literacy, governance indices, and state authority characteristics.Governance measures cover corruption, socioeconomic exclusion, private property rights, and freedom of domestic movement.
  • A.5.2 Local Geospatial Attributes: Local geospatial measures include population density, nightlight radiance, human influence, travel time to dense population, temperature-related measures, and precipitation.Measures are raster-based, centered on community coordinates, and aggregated within buffers where specified.
  • A.5.3 Researcher-Provided Attributes: Researcher-provided attributes cover subsistence, exchange, labor, livelihoods, shocks, and norms governing work and access to property or production.The questionnaire distinguishes within- and outside-community activities and asks about climate, disease, violence, and market shocks.
  • B Robustness of Within-Community Results: Robustness analyses compare the main composite network with proportional composite and raw composite constructions.The robustness section evaluates whether the within-community results depend on how network edges are constructed.
  • B.1 Different Network Constructions: Results from both alternative networks are similar to the main results in Figure 3.Figures A12 and A13 report proportional-composite results, while Figures A14 and A15 report raw-composite results.

B.2 Different Accountings for Sharing Unit Composition … C.7 Quartile Splits

Robustness analyses show that the paper’s core wealth–network relationships generally persist across alternative measures, models, inequality metrics, and network constructions, though support access is sensitive to some size adjustments. Cross-community wealth inequality is also linked to weaker economic connectedness, while simulations rule out a simple wealth-band mechanism.

  • B.2 Different Accountings for Sharing Unit Composition: Alternative accounting for sharing unit composition leaves the broad wealth–network pattern comparable, although out-degree–wealth associations are somewhat weaker without size weighting.Per-adult weighting produces a similar broad pattern, while residualizing on log(size) or size fixed effects preserves provisioning but not access relationships.
  • B.3 Different Measures of Economic Connectedness: Alternative economic-connectedness measures remain positively associated with sharing unit wealth across most communities.This holds for Unit Economic Connectedness using supporters or supportees and for both absolute wealth and wealth per capita; median and summed alter wealth show consistent positive associations.
  • B.4 Multilevel Models of Degree, AAW, and Wealth: Pooled and multilevel analyses find positive average associations between wealth and support access, support provisioning, and Average Alter Wealth across communities.Multilevel average slopes are almost identical to pooled coefficients, while site-specific slopes vary; except for Panel B, negative slopes lie within 2τ of the pooled slope.
  • C Robustness of Across-Community Results; C.1 Different Accountings for Sharing Unit Composition: Across-community relationships between wealth Ginis and social-capital measures are examined under per capita, absolute, per-adult, and size-adjusted accountings.The measures include support provisioning, support access, Relative Average Alter Wealth, Economic Connectedness, and Relative Connectedness.
  • C.2 Using Incoming Nominations: Incoming-nomination measures provide equivalent cross-site tests of wealth inequality and economic connectedness using supportees rather than supporters.The analyses cover Relative Average Alter Wealth, Economic Connectedness, and Relative Connectedness.
  • C.3 Using Absolute Wealth: Absolute-wealth versions compare below-median and above-median sharing units using ratios of summed or median alter wealth.These community-level measures are constructed analogously to Relative Average Alter Wealth.
  • C.4 Using Different Measures of Inequality: Alternative inequality measures align positively with the wealth-per-capita Gini, supporting robustness to outliers with extreme wealth.The alternatives are the 90th-to-10th percentile and 80th-to-20th percentile wealth-per-capita ratios.
  • C.5 Ruling Out Wealth-Band Tie Formation: The observed negative association between wealth inequality and economic connectedness is not explained by absolute wealth-band tie formation.Alter-wealth dispersion correlates .85 with community wealth dispersion, or .89 after omitting the top 5 percent; simulated correlations lie far from the observed value, and excess RAAW remains negatively correlated with the wealth-per-capita Gini.

C.8 Multilevel Model for Economic Connectedness … D.1 Noise Perturbation of Item Values

The paper models economic connectedness across wealth quartiles, finds that stronger poorest–wealthiest ties accompany lower inequality, and tests whether wealth-network findings withstand measurement-error perturbations. It also defines wealth modularity to assess wealth-based network segregation.

  • C.8 Multilevel Model for Economic Connectedness: A multilevel generative model estimates economic connectedness from wealth-class positions, ego survey size, site-level confounders, and community wealth inequality.The model uses random effects and posterior inference to quantify quartile-pair connectedness.
  • C.8 Multilevel Model for Economic Connectedness: Communities where the wealthiest quartile is better connected to the poorest quartile have lower wealth inequality.This result uses wealth per capita and is shown through model posteriors across quartile combinations.
  • C.8 Multilevel Model for Economic Connectedness: Across all 16 fits, no divergent transitions occurred, R̂ ≤1.004 for reported parameters, and α1 effective sample sizes exceeded 1,100 bulk and 1,300 tail.Some fits reached maximum treedepth around 25%, indicating reduced sampling efficiency without corresponding divergences or poor convergence diagnostics.
  • C.9 Wealth Modularity: Wealth modularity measures how strongly network edges correspond to wealth-class divisions, maximizing modularity over every possible two-class division of each community’s wealth distribution.The maximum is compared with a null distribution generated by permuting sharing-unit wealth.
  • D Consequences of Wealth Mis-measurement: Wealth estimates may omit assets or require nonmarket valuation judgments, and cross-validation against true wealth is infeasible.These limitations motivate sensitivity analyses of downstream inequality and connectedness statistics.
  • D Consequences of Wealth Mis-measurement: The wealth Gini is not related across communities to mean wealth, using either absolute or per capita wealth.This lack of association reduces concern that measurement error systematically drives wealth-inequality estimates.
  • D Consequences of Wealth Mis-measurement: Relative Average Alter Wealth is similarly not related to community median wealth, further reducing concern about measurement error.The result is shown using wealth per capita.
  • D.1 Noise Perturbation of Item Values: 50 perturbed wealth datasets per community yielded fairly invariant wealth Gini coefficients and very robust within-site material–social wealth relationships.Each item value was multiplied by 0.8 or 1.2 with equal probability before recomputing downstream statistics.

D.2 Multiple Imputation of Missing Item Data

For communities with incomplete item inventories, the study uses multiple imputation to create 50 datasets per site. Imputed wealth Gini coefficients are fairly invariant across stochastic imputations, and affected sites do not systematically differ from other sites.

  • Incomplete item inventories affect a subset of sharing units in some communities.
  • 50 imputed datasets per site are constructed using the mice package in R.
  • Wealth Gini coefficients remain fairly invariant across stochastic imputation realizations.Figure A51 summarizes the median and an approximately 95% confidence interval across 50 trials.
  • Sites requiring item-inventory imputation do not systematically differ from sites where wealth imputation is unnecessary.

E Kinship · F The Relationship of Other Community Characteristics with Wealth Inequality and Economic Connectedness

Kinship ties provide baseline support across communities but do not primarily explain the study’s wealth–network patterns. The paper also examines how additional community characteristics relate to wealth inequality and economic connectedness.

  • E Kinship: Relative Average Alter Wealth from kin connections and non-kin connections is uncorrelated across communities.The comparison uses Relative Average Alter Wealth derived separately from primary-kin and non-primary-kin connections.
  • E Kinship: There is no consistent pattern for whether kin-based Relative Average Alter Wealth is higher or lower than non-kin-based Relative Average Alter Wealth.This comparison is shown for communities using the overall network, primary-kin connections, and non-primary-kin connections.
  • E Kinship: Kinship provides a baseline of support in all communities but is not the main explanation of the study’s results.This conclusion synthesizes the kin-versus-non-kin analyses.
  • E Kinship: Kinship-based and non-kinship-based network measures are compared in relation to material wealth across communities.The analysis restricts the composite network to primary-kin or non-primary-kin connections and examines support access, support provisioning, and Average Alter Wealth.
  • F The Relationship of Other Community Characteristics with Wealth Inequality and Economic Connectedness: The paper explores how other community characteristics relate to wealth inequality and economic connectedness.The analysis considers variables discussed in Section 4 and presented in Table A5.
  • F The Relationship of Other Community Characteristics with Wealth Inequality and Economic Connectedness: Table A5 provides a basic description of each additional community variable, including its information source and measurement scale.Exact site-level values are reported in Table A16 alongside other site-level variables in Section G.

F.1 Correlations of Community Characteristics with Wealth Inequality · F.2 Correlations of Community Characteristics with Economic Connectedness · G Site-Level Variables

Community wealth inequality is most strongly associated with economic connectedness, private property rights, and wage or salaried labor, while robust predictors of economic connectedness remain unidentified. The paper also documents the site-level variables and alternative measurement conventions used across analyses.

  • F.1 Correlations of Community Characteristics with Wealth Inequality: Table A6 reports bivariate correlations between community characteristics and the Gini of wealth per capita.These correlations correspond to the relationships shown in the main text’s Figure 5.
  • F.1 Correlations of Community Characteristics with Wealth Inequality: Relative Average Alter Wealth, private property rights, and time in wage or salaried labor are the leading LASSO predictors of the Gini of wealth per capita.These variables also show significant bivariate associations with the Gini; results remain similar after a Puffer transform.
  • F.1 Correlations of Community Characteristics with Wealth Inequality: Relative Average Alter Wealth is significantly and strongly negatively correlated with inequality across regression specifications.Communities where poorer residents are better connected to wealthier residents have lower inequality.
  • F.1 Correlations of Community Characteristics with Wealth Inequality: Time spent in wage or salaried labor shows a consistent positive effect on the Gini of wealth per capita.Communities whose residents spend more time in wage or salaried work have greater inequality.
  • F.2 Correlations of Community Characteristics with Economic Connectedness: No robust predictors of Relative Average Alter Wealth were found among the community variables considered.Time spent in wage or salaried labor is near significance and negatively associated with Relative Average Alter Wealth.
  • F.2 Correlations of Community Characteristics with Economic Connectedness: Bivariate correlations with Relative Average Alter Wealth are reported using wealth per capita, with 95 percent confidence intervals based on Fisher’s z-transformation.Bolded rows identify intervals excluding zero.
  • G Site-Level Variables: Site-level variables cover wealth inequality, support, Relative Average Alter Wealth, economic and relative connectedness, normed wealth modularity, and other characteristics.Where appropriate, measures are provided using absolute, per-capita, per-adult, and size-adjusted sharing-unit composition, and for both supporters and supportees.
  • G Site-Level Variables: Tables A15 and A16 provide site-level measures of normed wealth modularity and other site-level variables.Table A16 also includes a continuation of the other site-level variables.
Loading 2608.25488v1…