Positive and Negative Effects of Finance-based Social Capital on Incident Functional Disability and Mortality: An 8-year Prospective Study of Elderly Japanese

Background Rotating savings and credit associations (ROSCAs) involve group financial self-help activities. These voluntary financial cooperative associations—mujin in Japanese—are found in some rural areas of Japan. Cross-sectional evidence suggests that active participation in mujin correlates with rich social capital and better functional capacities among older adults. However, the effect of mujin on subsequent health outcomes is unknown. Methods In 2003, we conducted a baseline interview survey of 583 functionally independent adults randomly selected from Yamanashi Prefecture residents aged 65 years or older. They were followed up until 2011. We used proportional hazards models, and factor analysis of 8 mujin-related questions identified 2 components: the “intensity and attitude” and “financing” aspects of mujin. Results The hazard ratios (HRs) for incident functional disability—identified by using the public long-term care insurance database—per 1-SD increase in factor scores were 0.82 (95% CI: 0.68–0.99) for the intensity and attitude score and 1.21 (1.07–1.38) for financing score. Adjustments for age, sex, marital status, household composition, physical health, education, income, and other factor scores only slightly attenuated these HRs. The results for mortality models were very similar to those for incident functional disability. Conclusions ROSCA-type activities in Japan could have beneficial effects on the health of older adults if used primarily for the purpose of friendship. Mujin for aggressively financial purposes might be somewhat harmful, as such activities might reflect the “dark side” of social capital, ie, overly demanding expectations of group conformity.


INTRODUCTION
Accumulating evidence suggests social capital is an important social determinant of individual health. 1 According to a seminal work of Putnam, 2 social capital refers to "features of social organization, such as trusts, norms, and networks, that can improve the efficiency of society by facilitating coordinated actions."Putnam introduced the concept of the financial cooperative association, called a rotating savings and credit association (ROSCA), which is a form of informal local microfinance that builds strong structural social capital. 2,3In a ROSCA, members regularly make a fixed deposit at scheduled meetings, after which a different assigned member takes the aggregate deposit.This practice continues until each member has taken an aggregate deposit.Therefore, to be successful, ROSCA members must trust one another implicitly. 4cial capital should be developed within the intrinsic historical and cultural contexts of each community, to overcome shared challenges. 2 As in many developing countries today, ROSCAs were very popular in Japan until the early postwar era.Japanese ROSCAs are called mujin in eastern Japan, tanomoshi-ko in western Japan, and moai in Okinawa Prefecture.Hereafter, for simplicity we will refer to all such organizations as mujin throughout this article. 5,6lthough mujin have largely disappeared in Japan, they remain active in some rural and remote regions, including Yamanashi and Okinawa, where many communities are bound by strong cultural ties that allow them to share socioeconomic challenges. 7For example, in the Yamanashi area, 66% of older (age 65+ years) adults were currently or formerly engaged in mujin.The figure was 40% for the entire adult population.Among the older group, 70% have multiple mujin memberships, and 79% have continued to their mujin for longer than 10 years.Most (93%) mujin are held monthly; the mean deposit is 5703 Japanese yen and the mean party fee is 2311 yen.Payment and financing policies vary; in every round, a deposit taker is selected by auction or in a predetermined order.The taker may pay a fixed fee or a proportional amount of the bidding fee, which is used as interest for those who take deposits in a later round.However, this interest system has largely disappeared in many extant mujin.
Nevertheless, unlike traditional ROSCAs, many extant mujin focus less on the financial needs of their members than on the need to promote friendship and social bonding among members. 8,9For most (95%) participants, mujin are enjoyable, an important source of information, and essential to daily living and health.
In 2003, the results of a survey in the Yamanashi area showed a positive cross-sectional association between mujin participation and functional capacity among elderly Japanese (≥65 years). 9The same study proposed that, as a highly homogeneous group, mujin were a form of structural social capital that fostered strong membership bonds.However, this evidence was obtained from a cross-sectional analysis that was subject to reverse causation (ie, healthier individuals might be likelier to participate in mujin).Thus, the prospective association between mujin participation and subsequent health needs to be studied.In addition, because mujin activities can have various functional aspects, there is a need to evaluate whether different aspects of mujin have independent effects on the health of elderly Japanese.For example, Kawachi et al suggested the possibility of negative effects of social capital on health if group membership requires unfavorable conformity. 10n the present study, we used 8-year prospective data from the Yamanashi Healthy-Active Life Expectancy (Y-HALE) study to investigate the independent effects of multiple aspects of mujin on onset of functional disability and mortality in Yamanashi Prefecture, Japan.We studied elderly Japanese (age ≥65 years) because social capital is particularly important for that age group because of their close bonds with community members and because the probability of failing health increases with age. 10 Because society is aging at an unprecedented pace in Japan and many other countries, 11 postponement of long-term care (LTC) is a critical public health concern worldwide.
Our hypotheses were as follows: (1) because mujin strongly foster bonding among members, intensive participation in mujin is beneficial to the health of elders in terms of maintaining functional capacity and longevity independent of sociodemographic characteristics, (2) the beneficial health effects of mujin are explained by differences in the financial condition and health behavior of participants and nonparticipants, and (3) the health effects of mujin differ according to their components, ie, some components may have detrimental effects.

Study participants
A full description of the sampling procedure for the Y-HALE study has been published elsewhere. 9,12In short, baseline data were collected from a 2-stage probability sample of 1800 older adults (≥65 years) who resided in Yamanashi Prefecture in 2002.In 2003, more-detailed questions were asked to 583 randomly selected participants who agreed to participate in this second survey.They were physically and cognitively independent in undertaking activities of daily living (ADL), ie, they were not qualified to receive public LTC insurance benefits.The municipal government determines applicant eligibility for LTC insurance benefits based on physical and cognitive functions, using a standardized procedure that includes a physician examination, computer-based provisional judgment, and a final decision made by a local insurance committee. 13Therefore, at baseline, we excluded residents who qualified for insurance benefits, as they already had ADL disabilities.Trained investigators visited the participants and gathered information on sociodemographic characteristics, social activities, diet, psychosocial factors, social capital, and health status.The response rate was 98%.After excluding participants with missing data for key measures, including age, sex, and number of household members, we analyzed data from 562 respondents (96%) to the 2003 survey.This study was approved by the Ethics Board of the School of Medicine, University of Yamanashi, and conforms to the principles of the Declaration of Helsinki.

Follow-up
We conducted yearly follow-up consultations by mail or, when requested, by telephone interview.Participants reported health status, including recent qualifications for LTC insurance benefits (dates of qualification and levels of benefits for which they qualified).For participants who died, we asked their family members to report the date of death and cause of death.We confirmed responses concerning qualification for benefits by phoning participants' officially assigned care-plan managers.To remain apprised of deaths, we also used alternate sources, such as death announcements in local newspapers, which covered 90% of deaths reported by participants' family members.If we discovered or suspected that a participant had died, we mailed or phoned the person's family to confirm the accuracy of the information.Participants who withdrew from the study because they did not want to participate, and those who could not be reached for any reason (such as departure from the study area), were treated as censored.

Onset of functional disability
Japanese public LTC insurance has a 7-level authorization system that is based on the severity of physical and cognitive disabilities.This system is used to determine maximum usage of LTC services.In this study, we defined onset of functional disability as the point when a participant was certified as having a level 3 or greater LTC need.This level was selected because there is a clear boundary in disability severity between levels 2 and 3: people at level 2 or lower require only partial help to fulfill their basic ADL, such as toileting and bathing, whereas those at level 3 or greater are completely dependent on assistance for many ADL (Table 1).Also, using level 3 as a cutoff minimized the potential for bias from participants who did not apply for LTC care services because they already had sufficient informal support for their care needs.Participants with level 3 certification require full ADL support, and it is thus far less likely that they would lack a level 3 or higher certification, because once people start using LTC care services, they are likely to obtain the necessary certification levels that meet actual care needs so as to maximize the available monthly cost limit for services covered by LTC insurance.

Scoring mujin domains
Participants were asked about their history of participation in mujin, number of current mujin memberships, current average frequency of participation in mujin meetings per month, maximum number of mujin memberships held at 1 time, and duration, frequency, type, financing deposit, and party fee for their primary mujin group.Complete information on the questionnaire regarding mujin has been published elsewhere. 9e conducted a factor analysis of these questions by using principal component methods.We used promax rotation to address correlations between factors and decided the number of factors on the basis of standard criteria, namely, by referring to the eigenvalue and evaluating the internal consistency of primary questions for each factor.To maximize data available for factor analysis, we imputed the mean value of each variable for missing observations.Although we gathered additional information on mujin characteristics, such as major topics of conversation in mujin and demographic characteristics of members, we did not use these data in our factor analysis because this additional information represented categorical variables that either could not be dealt with using a conventional factor analysis technique or did not compose a valid factor that could be reasonably characterized.
Rather than analyzing the indirect contextual social capital effects of mujin group involvement, we evaluated the direct effects of mujin on participant health because the contextual effect of mujin should be evaluated by using the mujin as the unit rather than by the geographical or administrative community unit, though we did not have any relevant information on the mujin groups in which the respondents participated.However, the importance of determining the effect of individual social capital rather than that of aggregated social capital is evident, as social capital measured at the individual level collectively constitutes social capital at the community level. 14 15socioeconomic status (educational attainment and individual income), levels of social activity, and health behaviors (smoking habit, alcohol consumption, and exercise habits).Social activity was measured using the question, "How frequently do you participate in the following community activities (community festivals, residents' associations, senior citizens' clubs, hobby activities, volunteer activities, and activities based on special/traditional skills)?".This Japanese social activity scale for older people was validated by Ohno 16 and allowed us to determine whether a participant was socially inactive, normal, active, or very active.

Statistical analysis
Using the statistical analysis software package SAS version 9.2, we conducted Cox proportional hazards regression and modeled the principal component scores for mujin with covariates.We visually and statistically confirmed a proportional hazard assumption.We created 2 separate models, with onset of functional disability (or being certified for level 3 or greater LTC need) and mortality as outcomes.In the disability models, all deceased cases were treated as censored.

RESULTS
In total, we observed 4055 person-years.During 8 years of follow-up, 126 participants developed a level 3 or greater LTC need and another 95 died, including 11 who had received level 3 or greater LTC certification before death.There were 56 dropouts (ie, individuals who refused to participate in further studies and became nonresponsive).The incidence rates of onset of functional disability and mortality varied with respect to participant socioeconomic status and social capital (Tables 2 and 3).The variables used for factor analysis were correlated with each, and correlation coefficients ranged between 0.49 and 0.94 (Table 4), excepting deposit for group financing, which was more weakly correlated with the other variables (correlation coefficient, 0.33-0.43).Factor analysis identified 1 component from the 8 questions on mujin (Table 5).We called this "intensity and attitude" toward mujin, which was reflected in all questions except those querying the amount of the deposit for group financing.
Moreover, although the eigenvalue was not large (0.762), we adopted another factor to which the amount of the deposit for group financing made a significant contribution.We accepted this second factor because it may reflect the financing aspect of mujin, which can be distinguished from intensity and attitude.We refer to this factor as "financing." The results of proportional hazards regression showed that the hazard ratio (HR) per 1-SD increase in the mujin component for intensity and attitude was 0.82 (95% CI: 0.68-0.99;Table 6).In contrast, the HR for the "financing" aspect of the mujin score was higher than unity (1.21, 95% CI: 1.07-1.38).These HRs were slightly altered by adjustment for demographics, physical health, educational attainment, and income (Model 1 in Table 6).In addition, when adjusting for   social activity score, the adjusted HR for intensity and attitude was largely attenuated, to 1.01 (0.81-1.25), whereas that for the financing aspect remained statistically significant, at 1.20 (1.07-1.35;Model 2).As indicated by the results of Model 3, further adjustments for health behavior did not alter the HRs for either mujin factor.The results of bivariate proportional hazards models for mortality were almost identical to those for the onset of functional disability (Table 7).Although HRs were not statistically significant in multivariate models, the financing aspect was similarly associated with mortality, with adjusted HRs ranging between 1.09 and 1.11.In contrast, the adjusted HRs for intensity and attitude were very close to unity.

DISCUSSION
In a representative sample of elderly residents of Yamanashi, we found that individuals who intensively and enthusiastically  participated in mujin were more likely to maintain functional capacity.The protective effect of mujin activities can be partially explained by socioeconomic status and fully explained by social activity in general.However, the contribution of health behavior to explaining the effect of mujin was not as strong as we expected.A striking finding of our study was that participation in mujin that were actively involved in group financing was positively associated with incident disability, suggesting that such mujin might be hazardous to health.
In recent years, microfinance-one of the more famous examples being the Grameen Bank in Bangladesh 17 -has emerged as a financial intervention in impoverished rural communities in developing countries.9][20] Although, strictly speaking, microfinance managed by commercial banks or nongovernmental organizations is not a term formally applied to structures such as ROSCAs, the two nevertheless share many characteristics in terms of the roles of individual members within the group.Thus, the findings of our study are consistent with those of studies on microfinance.
Individual social capital could be the vector of various forms of social support. 12Our prospective evidence suggests that mujin activities serve the community by fostering strong individual (structural) social capital 2 and protecting elderly adults against progression of what would otherwise be the natural course of ADL deterioration.In the present study, adjustment for the level of social activity support attenuated the beneficial effect of mujin, which suggests that mujin have a role similar to social activity in general, which is a known component of structural individual social capital.Health behavior did not fully explain the association between enthusiastic intensive participation in mujin and lower incidence of disability, which suggests that psychosocial support (such as the provision of emotional support, prevention of social isolation, and maintenance of morale) is the central role of mujin rather than promoting healthier behavior (such as smoking cessation and regular exercise), at least for older Japanese adults. 21t is especially interesting that we also found a higher health risk associated with the financial aspect of mujin.This is probably a reflection of the "dark side" of social capital. 10ccording to Kawachi et al, this potential negative aspect of social capital includes (a) excessive demands placed upon members of cohesive groups to provide support to others, (b) expectations of conformity that may result in restrictions on individual freedom as well as intolerance of diversity, (c) the exercise of in-group solidarity to exclude members of out-groups, and (d) down-leveling of norms within a tightly knit group, which may obstruct potential upward social mobility" (Kawachi et al, 2008; page 5). 10 Given the important responsibility associated with financial transactions among mujin members, it is logical to presume that members are subject to high demands for and expectations of conformity to the wishes of other group members. 5In fact, ethnological researchers have described the strongly closed nature of "serious" mujin and have suggested potential negative effects on group members due to the daunting scale of mutual responsibility. 22,23he validity and reliability of this study are supported by the very high response rate, the use of highly standardized outcome measures, and the factor analysis-based scoring approach used to evaluate social capital and mujin.Nevertheless, some limitations of this study should be noted.First, our definition of disability onset may differ from actual levels of disability.However, we addressed this in part by specifying the need for level 3 LTC as the cutoff for having full ADL support.Similarly, although we used multiple approaches to ensure the accuracy of information on participant deaths, our strategy to identify deceased cases did not rely on official death records, potentially causing information biases.For example, some decedents might have been treated as censored, which would result in underestimation of the number of deaths and lower statistical power in our analyses.Second, the evaluation of 2 mujin factors might not be sufficiently valid because the variables in our factor analysis included both continuous and ordered specifications.However, their validity and reliability were partially supported by high internal consistency and the fact that our findings did not contradict those of previous theoretical and empirical studies. 2,9Finally, although some studies found sex differences in the association between social capital and health, our analyses were not stratified by sex due to the limited sample size.
In conclusion, because social capital is developed in specifically local cultural and historical contexts, determining the health effects of community-specific social capital-such as that afforded by mujin-would increase understanding for public health research on social capital.Our study of the Japanese equivalent of ROSCA activities among elderly adults shows that such financial self-help groups, which are common in many parts of the world, could have beneficial effects on member health.These effects may be due not only to financial empowerment, but also to strengthening of interpersonal ties.Nevertheless, our striking findings regarding the potential negative effect of mujin that have an aggressive financing role warrant further study.Understanding the potential positive and negative effects of social capital is critical, especially when promoting community-based financial interventions such as microfinance programs, which have been steadily gaining popularity worldwide. 18,19

Table 1 .
Definitions of levels used in the Japanese public long-term care insurance system ADL: activities of daily living, LTC: long-term care.

Table 2 .
Incidence of functional disability and mortality by characteristics of adults aged 65 years or older in Yamanashi, Japan(2003) a Geometric mean.

Table 3 .
Incidence of functional disability and mortality by baseline mujin characteristics among 562 adults aged 65 years or older in Yamanashi, Japan(2003) a For the mujin meeting that is most important for the respondent (if he/she participates in >1 mujin group).

Table 4 .
Correlation coefficients for mujin variables

Table 5 .
Standardized scoring coefficients yielded by factor analysis of mujin characteristics: principal component method with promax rotation For the mujin meeting that is most important for the respondent (if he/she participates in >1 mujin group). a

Table 6 .
Hazard ratios (95% CI) for onset of functional disability a HR is per 1-SD increase.

Table 7 .
Hazard ratios (95% CI) for all-cause mortality a HR is per 1-SD increase.