The data used in this study are drawn from the China Family Panel Studies(CFPS). The CFPS is conducted by Peking University’s China Social Science Survey Centre. Individual, household, and community data are collected from 25 provinces/municipalities/autonomous regions. In the survey, residents are asked about their family relationships, living environment, and economic situation via a questionnaire. In addition, to comprehensively assess residents’ health status, the survey evaluates their education, occupation, lifestyle, pension, and medical care . The research findings contained in this database are reliable because the survey process is scientific and standardized, and the data quality is high . The database has been updated every two years since 2010 and is now up to date through 2020.
The data used in this study are drawn from the 2020 CFPS individual pool, which contains a total of 28,590 records. First, we selected individuals aged 16 years or older (age > = 16), and then we selected individuals who received a self-response questionnaire (selfrpt = yes) and completed the interview (iinterv = has completed). Finally, we excluded cases that reported “missing”, “inapplicable”, “invalid responses”, “refusal to answer”, or “don’t know” for any of the variables used, resulting in a sample of 17,475 observations for the final analysis.
The health indicator we used is “general health”, which refers to self-rated health; this measure has been shown to be a reasonable indicator of an individual’s true level of health , taking into account both the respondent’s physical and mental health. It corresponds to the following question: “How would you rate your health status?”[QP201]. The response options are “1 excellent, 2 very good, 3 good, 4 fair, and 5 poor”. Based on the theoretical summary provided above, we used the indicators of cognitive social capital and structural social capital to measure social capital. With regard to structural social capital, studies utilizing CFPS data have frequently used the term “gift spending” to reflect social networks [34,35,36]; however, this indicator is household-based and difficult to specify at the individual level. The CFPS 2020 individual questionnaire does not include questions pertaining to informal social connections; thus, we selected “interpersonal relationships”  as an indicator of structural social capital at the individual level. The question for “interpersonal relationships” is “Do you think you are popular?” (PM2011). The respondents were asked to rate this statement on a scale ranging from 0 to 10, with 0 being the lowest and 10 being the highest. With respect to cognitive social capital, we selected the commonly used indicator “general trust”. This indicator has been used more frequently in other empirical studies [38,39,40,41,42]. The question for “general trust” is “In general, do you think that most people are trustworthy, or it is better to take greater caution when getting along with other people?” (PN1001). The response options were categorized as “Most people are trustworthy” and “The more caution, the better”.
Other conditions we selected were “age”, “gender”, “education”, “marriage” and “income”, which have been included as control variables in the overwhelming majority of social capital and health studies and play a broad and important role in the pathway by which social capital impacts health [39, 43,44,45,46]. The condition of “age” indicates the age of the respondent in the year the survey was conducted and corresponds to the code “age”, while “gender” corresponds to the code “gender”. “Education” corresponds to the code “cfps2020edu”, which indicates the highest level of education the respondent had completed in 2020 on a scale featuring 8 options, which were listed in the following order: “Illiterate or semiliterate, Primary school, Junior high school, High school/technical secondary school/technical school/vocational high school, Junior college, College degree, Master’s degree, and Doctorate”. We chose “marriage-last” to represent the respondent’s most recent marital status, which is categorized as “Married (with a spouse), Divorced, Unmarried, Widowed, or Cohabiting”. Given the reality of low self-reported income amounts, for the income profile, we selected job income satisfaction as a subjective indicator of income ; this measure corresponds to the question “How satisfied are you with your income from this job?” (QG401) This item was scored on a 5-point Likert scale ranging from very dissatisfied to very satisfied, with corresponding scores of 1–5.
We used fuzzy-set qualitative comparative analysis (fsQCA) to examine the sufficient and necessary conditions that lead to good general health as well as their various configurations. This study used the corresponding software fsQCA 4.1 for data analysis, following the form of presentation of results proposed by Ragin and Fiss  to organize the analysis results. QCA is a set-theoretic grouping analysis method based on Boolean algebra. Unlike traditional quantitative analysis, which is based on variance and the null hypothesis significance test, QCA treats cases as groups of variables and analyses the “necessary” or “sufficient” conditions for obtaining the desired outcome based on an identification of the particular outcome and variables to be explained, thereby holistically exploring “how” multiple concurrent causes and effects can generate complex problems . QCA exhibits two crucial characteristics: equivalence and asymmetry. Equivalence suggests that “multiple combinations of antecedent conditions are equally effective”, while asymmetry is interpreted as a situation in which “a condition (or a combination of conditions) that explains the presence of an outcome can be different from the conditions that lead to the absence of the same outcome” . Based on these two properties, QCA can analyse different configurations of necessary and sufficient conditions for attaining the target outcome , which is not possible using a linear regression model.
QCA can be divided into 3 basic categories based on the type of variables: crisp-set qualitative comparative analysis (csQCA), multi-value qualitative comparative analysis (mvQCA), and fuzzy-set qualitative comparative analysis (fsQCA). csQCA analyses can address only binary categorical variables. mvQCA analyses superior for multicategorical nominal variables, but they are only suited to deal with the kind problem; cases can only be assigned to one of the categories associated with the categorical variables . The emergence of fsQCA further enhances researchers’ ability to analyse fixed-distance and fixed-ratio variables, allowing QCA to address not only category problems but also degree-varying problems and partial membership, in which context cases have an affiliation score ranging between 0 (nonmembership) and 1 (full membership) . Therefore, fsQCA was chosen for this study to facilitate the analysis of multiple variable types. As a methodological innovation, QCA aims to identify causal relationships among variable groupings and outcomes using case-to-case comparisons, thereby answering the question, “What groupings of variables lead to the desired outcome? ”  QCA conceptualizes causality in terms of complex causation characterized by jointness, equivalence, and asymmetry. Combined with a group analysis approach, QCA researchers are able to extend the extant causal theoretical framework based on additivity and symmetry and revisit previous empirical findings and contradictory conclusions . Using QCA analysis, researchers can also identify variable groupings of states with equifinality, thus improving their understanding of the differentiated driving mechanisms that lead to outcomes in different case scenarios and facilitating further discussion of the fit and substitution relationships among conditions. In addition, researchers can further compare the groupings of variables that lead to the results in question and broaden the dimensions of their theoretical explanations with regard to specific research questions .
QCA identify sufficient or necessary conditions by reference to affiliation between sets. A configuration is necessary if it is a consistent superset of the outcome; similarly, if the configuration is a consistent subset of the outcome, then the sufficiency of the configuration is indicated . In crisp-set QCA, the boundaries are clearer, whereas in fuzzy-set QCA, the results must be judged using two metrics that vary between 0 and 1: consistency and coverage. For necessity analysis, the thresholds for the two indicators are typically 0.90 and 0.60, respectively , while the requirements for sufficiency analysis are relatively lenient, with consistency results greater than 0.80 being accepted .
Before performing fsQCA, the variables for the cause and effect conditions must be calibrated to fuzzy sets ranging from 0 and 1, with 0 denoting nonmembership and 1 denoting full membership. We used fsQCA software to complete the calibration process of the variables . As suggested by a relevant study , the variables must be transformed into a calibrated set using three substantively meaningful thresholds: full membership, full nonmembership, and a crossover point that reflects maximum ambiguity. The dichotomy of health is more common, and in this context, we referred to several empirical studies that have placed the option ‘fair’ in the ‘unhealthy’ category [62, 63]; thus, we defined “excellent, very good, and good” as indicating full membership and “fair and poor” as indicating nonmembership. Next, we chose “general trust” as an indicator of cognitive social capital, and the possible responses for the question were treated as dichotomous by assigning “yes” to 1 and “no” to 0; scores of 1 and 0 were then classified as indicating full membership and nonmembership, respectively, without calibration. The structural social capital indicator took a score ranging from 0–10, for which we used the 90th, 10th, and 50th percentiles of the original distribution to define the thresholds and intersection points, which were calculated to be 10, 7, and 5, respectively. In the other conditions, age was calibrated in the same way as the structural social capital indicator, and gender was assigned a value of 1 for females and 0 for males to emphasize the analysis of females given that females are more likely to report poor health. With regard to education, we classified “junior college and college degree” as one category and then assigned a score ranging from 1–7, with 6 indicating full set membership, 4 indicating intermediate membership and 2 indicating full set nonmembership according to the 7-point Likert scale . Income was calibrated on a 5-point Likert scale, with full set membership, intermediate membership, and full set nonmembership corresponding to scores of 5, 3, and 1, respectively. Further details can be found in Table 1.
The sufficiency analysis begins with a “truth table”, which includes all logically possible configurations of conditions and requires thresholds for case frequency and consistency level to be established manually. Typically, the frequency threshold is 1 or 2, but due to the large sample size of 17,475 observations, we set the frequency threshold for this analysis to 10 and the consistency level threshold to the commonly employed threshold of 0.8. The preconditions were then classified as either core or peripheral based on the following criteria: “Core conditions are those that are both parsimonious and intermediate solutions, while peripheral conditions are those that occur only in intermediate solutions” , and core conditions can be viewed as “decisive causal components” .