Social support and reference group: the dual action mechanism of the social network on subjective poverty

The study addresses how social networks influence subjective poverty in China and what role objective (relative) poverty plays in this relationship. Against the backdrop of China’s rising incomes, eradication of absolute poverty, and policy emphasis on relative poverty, the paper argues that subjective poverty—grounded in individuals’ well-being perceptions—is an essential complement to objective measures. The authors posit that social networks affect subjective poverty via two mechanisms: (1) social support (emotional and instrumental) that can directly improve subjective experiences and alleviate objective deprivation, and (2) reference group comparisons within networks that shape relative status perceptions. They further hypothesize that objective poverty mediates the social support–subjective poverty link and moderates the reference-group–subjective poverty link, weakening comparison effects among the objectively poor. The study’s purpose is to integrate these mechanisms in a unified framework to better understand drivers of subjective poverty and inform poverty governance in the era of common prosperity.

The review contrasts objective and subjective poverty, noting shifts from absolute to relative and from unidimensional to multidimensional objective measures, alongside critiques of rigid lines and limited attention to emotions and well-being. Subjective poverty emphasizes individuals’ perceived well-being within social reference frames, empowering those experiencing poverty and aligning with governance that respects subjectivity. Three main explanations for subjective poverty are discussed: (1) the relative income hypothesis (comparisons to peers and past selves shape perceptions), (2) the social inequality hypothesis (macro inequality structures affect subjective poverty, with mixed evidence and conditional effects via mobility expectations), and (3) multifactorial synthesis (income, deprivation, exclusion, demographics, education, employment, and networks), though often lacking theoretical integration. Gaps identified include insufficient attention to social networks as structural factors in subjective poverty, limited analysis of networks as reference groups (beyond social support), and inadequate exploration of interactions among objective poverty, networks, and subjective poverty.

Data: China Labor Dynamics Survey (CLDS) waves 2014, 2016, 2018 (2012 excluded due to missing variables), covering 29 provincial-level regions (excluding Hong Kong, Macao, Taiwan, Tibet, Hainan). The study merges household- and individual-level data and retains household heads’ individual questionnaires. Final sample N = 32,257 after removing missing values. Dependent variable: Subjective poverty measured as subjective socioeconomic status using Cantril’s ladder (1–10, reversed so higher values indicate higher subjective poverty). Robustness: a binary indicator where original responses 1–2 = poor (1) and 3–10 = non-poor (0). In the full sample, subjective poverty rate = 14.9%; mean degree = 6.586 (6.074 non-poor; 9.52 poor). Independent variables (social networks): - Social support (functional perspective): - ENSC (emotional support): number of close local people one can talk to about problems (logged). - INSC (instrumental support): number of close local people one can borrow 5,000 yuan from (logged). Resource-generator logic is applied to quantify amounts of specific support types. Descriptive means: INSC full 1.041 (non-poor 1.082; poor 0.809); ENSC full 1.156 (non-poor 1.192; poor 0.949). - Reference groups (comparisons): - HONC (homogeneous comparison): average of three items comparing living standards to relatives, neighbors, and classmates with the same education (1–5 from much lower to much higher; higher = higher status). - HENC (heterogeneous comparison): comparison to other members in the respondent’s municipal district (1–5; higher = higher status). Means: HONC full 2.737 (non-poor 2.811; poor 2.311). HENC full 2.469 (non-poor 2.549; poor 2.009). Mediator and moderator: Objective poverty (relative): binary indicator where household per capita income < 40% of the median per capita disposable income, calculated separately for urban vs. rural areas of current residence. Full-sample objective poverty rate = 29.1% (26.9% among subjective non-poor; 41.3% among subjective poor). Controls: gender, age, marital status, education (years), annual income, urban/rural, political status (party), region (eastern/central/western), survey year, migration experience, social insurance coverage, sense of fairness. Models: Baseline OLS for continuous subjective poverty (1). Mediation assessed via logit for objective poverty (2) and OLS including both social support and objective poverty (3). Moderation tested by including interactions between objective poverty and comparison variables (4). Endogeneity addressed via instrumental variables and 2SLS; IVProbit used for binary subjective poverty robustness; KHB method used to decompose mediation effects. Instruments: For social support, village-level mean social network size (excluding the individual) for both INSC and ENSC; Mandarin fluency (1–5) for INSC; off-duty language (Mandarin vs. dialect) initially for ENSC (found invalid), so village-level mean used. For comparisons, village-level mean HONC (excluding self) instruments HONC; province-level mean HENC (excluding self) instruments HENC. Instrument validity tested via first-stage F, Cragg-Donald and Kleibergen-Paap statistics, Hausman endogeneity tests, overidentification tests (Sargan/Hansen) where applicable. Hypotheses: H1.1/H1.2 on ENSC and INSC direct effects (with ENSC expected stronger), H1.3/H1.4 on HONC/HENC direct effects (HONC expected stronger), H2.1–H2.3 on mediation of objective poverty (with stronger mediation for INSC), and H3.1/H3.2 on moderation by objective poverty (weaker comparison effects among objectively poor).

- Social support (direct effects): Both INSC and ENSC reduce subjective poverty. OLS with both supports: INSC −0.119***; ENSC −0.093*** per unit. Accounting for endogeneity (2SLS), effects are larger: with single endogenous regressor, INSC −0.483*** (SE 0.038) and ENSC −0.213*** (SE 0.023). With both instrumented jointly (Model 6), INSC −0.259*** and ENSC −0.289***, indicating ENSC’s stronger alleviating effect, confirming H1.2 after correcting endogeneity (H1.1 also supported). - Reference group comparisons (direct effects): OLS with both comparisons: HONC −0.721*** (SE 0.018) and HENC −0.268*** (SE 0.013) per unit; HONC’s effect is stronger (H1.3–H1.4 supported). With IV, single-variable models show larger magnitudes: HONC −1.844*** (SE 0.154); HENC −0.739*** (SE 0.054). In the joint IV model, HONC remains large and significant (−1.901***, SE 0.101) while HENC becomes non-significant (−0.032), reinforcing HONC’s dominance. - Mediation by objective poverty: INSC lowers the odds of objective poverty by 45.77% (exp(−0.612)−1), and ENSC by 19.02% (exp(−0.211)−1). In outcome models adding objective poverty, objective poverty increases subjective poverty by 0.316–0.346 units. Indirect effect shares: INSC path ≈ 12.42% (direct ≈ 87.58%); ENSC path ≈ 4.77% (direct ≈ 95.23%). H2.1 and H2.2 supported; mediation is modest, with stronger mediation for INSC than ENSC, supporting H2.3. - Moderation by objective poverty: Interaction terms are positive and significant, indicating that being objectively poor weakens the alleviating effects of comparisons: OLS interactions HENC×Objective poverty 0.064***; HONC×Objective poverty 0.081***. With IV, interactions are larger: HENC×Objective poverty 0.586***; HONC×Objective poverty 0.761***. H3.1–H3.2 supported. - Robustness: IVProbit using binary subjective poverty confirms main findings: ENSC remains significant; INSC becomes non-significant when both included; HONC remains significant while HENC becomes non-significant. KHB decomposition shows consistent mediation: for INSC, total −0.346***, direct −0.249***, indirect −0.098***; for ENSC, total −0.343***, direct −0.257***, indirect −0.086***.

The findings demonstrate dual mechanisms through which social networks shape subjective poverty. Social support—especially emotional support—directly enhances subjective well-being and reduces perceived poverty, with some additional alleviation via improvements in objective living standards. Reference-group comparisons within social networks strongly influence perceived poverty; higher relative status within homogeneous networks (relatives, neighbors, similarly educated classmates) substantially lowers subjective poverty, and these effects surpass those from heterogeneous comparisons. Objective poverty plays two roles: it partially mediates the link from social support to subjective poverty (more so for instrumental support that directly affects material conditions) and moderates comparison effects by dampening their influence among the objectively poor, consistent with need-priority and conditional relative income hypotheses. These results refine theories of subjective poverty by integrating subjective social comparison within network contexts and by showing conditionality on objective deprivation. For policy, alleviating objective poverty remains foundational, while activating social networks’ informal insurance and emotional support can further reduce subjective poverty.

The paper contributes by integrating social networks’ social support and reference group functions within a unified framework to explain subjective poverty. Empirically, both instrumental and emotional supports reduce subjective poverty, with emotional support having a stronger direct effect after addressing endogeneity. Network-based comparisons also matter, especially within homogeneous networks, where higher relative status sharply lowers subjective poverty. Objective poverty mediates the support–subjective poverty relationship (more through instrumental support) and moderates comparison effects, weakening them among the objectively poor. These insights underscore the importance of consolidating gains in objective living conditions and leveraging social networks’ informal insurance and emotional support to alleviate subjective poverty. Future research should refine measurement of reference groups (allow self-selected referents), improve identification of heterogeneous network members, align support and comparison measures to the same network members, and incorporate objective indicators of network members’ socioeconomic status to construct more accurate social comparison metrics.

- Reference group measurement was researcher-provided (relatives, neighbors, similarly educated classmates) rather than self-selected, potentially misaligning with respondents’ actual salient referents. - Classification of network members, especially for heterogeneous comparisons, could be improved; current HENC is indirect and may mix relationship strengths. - Social network measures did not allow aligning the same network members across support and reference-group roles, limiting the ideal integrated framework. - Data constraints limit direct measurement of objective attributes of network members (e.g., their income/SES) to form objective comparison variables. - As with survey-based network measures, potential endogeneity and measurement error remain despite IV strategies.