RESEARCH DOSSIER
Social Norms and Community Insurance Adoption
Abstract. Insurance adoption is embedded in relationships among households, communities, risk-sharing associations, distributors, providers, and public institutions. Community members evaluate insurance against expected losses and premiums while interpreting what significant others ordinarily do, what those others regard as appropriate, and how a purchase would affect reciprocity and membership. These influences are empirically important, but their mechanisms differ. A neighbor's insurance purchase may provide information about product quality; a peer's payout may establish contractual credibility; an association may lower transaction costs; or a community's expectations may alter the social consequences of purchasing coverage. Treating these responses as a single social-norm effect obscures both scientific identification and insurance policy. This monograph examines social norms as conditional social expectations relevant to insurance uptake, renewal, and continued participation. It distinguishes empirical expectations, normative expectations, descriptive regularities, injunctive approval, trust, social capital, network learning, informal indemnification, and social identity. Its international evidence covers agricultural and rainfall-index insurance in China, India, and Ghana; health insurance in Ghana, Kenya, Nepal, and Senegal; funeral associations in Ethiopia and Tanzania; and Australian flood insurance. The strongest insurance-specific experimental evidence establishes knowledge diffusion, trust-related barriers, effects of group-based engagement, and learning from realized payouts. Observational studies identify additional correlations with perceived norms, peer pressure, bridging social capital, and provider evaluations; these associations alone do not identify normative compliance. The original contribution is a critical synthesis of mechanism-specific interpretations and a practical identification agenda grounded in established social-norm measurement and causal-inference methods. It assesses when network interventions can complement formal contracts, when informal risk sharing competes with or enables coverage, and how social hierarchy can simultaneously produce uptake and exclusion. The resulting conclusion is that community influence should be analyzed as a set of separable processes whose effects depend on product performance, the reference group, observable behavior, institutional credibility, and distributive conditions. These are interpretive findings, not a newly proposed theory or causal estimate.
Keywords: social norms; community insurance adoption; empirical expectations; normative expectations; social learning; peer effects; social capital; trust; community-based health insurance; informal risk sharing; index insurance; insurance demand; group identity; distribution; causal identification; insurance inclusion; comparative risk science
Introduction
Insurance is a promise about uncertain future states, made at a price today, and performed through institutions that the customer may seldom have encountered. Its purchase requires assessments of probability, indemnity, liquidity, contract terms, insurer reliability, and outside risk-sharing opportunities. When a household operates inside a village, occupation, neighborhood, religiously organized association, cooperative, or family network, the same decision also communicates something to others. Joining a scheme may signal prudence, group solidarity, confidence in a provider, or willingness to participate in an institution. Refusing coverage may signify confidence in relatives, distrust of sellers, budget constraints, or adherence to local practices. The meaning of the act must be established rather than presumed.
The central research question is therefore precise: under what conditions do people's beliefs about other people's conduct and prescriptions causally alter insurance purchasing or renewal, independently of information gained through others and independently of the economic resources those relationships provide? A persuasive response must treat the counterfactual seriously. If a friend shares knowledge of an indemnity contract and enrollment rises, the event proves that networked information can change demand; it does not automatically prove conformity to a norm. If an association's members enroll more often than nonmembers, the association may reflect wealth, access, selection, mutual monitoring, political connection, or preexisting risk exposure. If a leader endorses a plan, the endorsement may act as an information signal, an institutional guarantee, an instruction, or a source of social pressure.
The theoretical starting points are the conditional-expectation account in Bicchieri's diagnosis of social norms, Cialdini, Reno, and Kallgren's distinction between norm types, and Ajzen's theory of planned behavior. They supply compatible but differently scoped analytical vocabularies: social practices and conditional compliance, salience of descriptive versus injunctive prescriptions, and intention shaped by attitudes, perceived control, and subjective norms. Their constructs must be operationalized before being carried into insurance.
Insurance-specific evidence is revealing because it often points away from pure imitation. Cai, de Janvry, and Sadoulet's randomized weather-insurance study in rural China attributes social-network effects mainly to diffusion of insurance knowledge, not to observation of purchase decisions. Giné, Townsend, and Vickery's rainfall-participation analysis in India shows that network participation and familiarity with the seller are correlated with uptake alongside wealth, credit, and basis risk. The two results are consistent: social embeddedness matters, but a link between social embeddedness and enrollment alone leaves its operative channel unidentified.
Community-based health insurance adds a different institutional problem. It offers financial protection through a membership framework whose legitimacy often rests on administrative quality, healthcare delivery, premiums, and credible claims service. The Ghana evidence of Jehu-Appiah and colleagues in their household enrollment study makes contract and provider perceptions central; it also documents adverse peer-related perceptions. The Kenyan group-presentation experiment discussed below demonstrates that meeting members through established informal groups can have lasting uptake effects under the conditions studied. Together, these cases show that social influence acquires its meaning within the actual insurance arrangement.
The research adopts a comparative, critical review of accessible original studies, journal records, primary institutional instruments, and historical accounts. It does not merge effect sizes from different populations or interpret an insurance marketing association as a causal norm estimate. Where full texts are inaccessible, the claim is limited to confirmed abstracts, publisher metadata, or author-institution records. Its original analytical synthesis rests on comparisons of research designs, mechanisms, and institutional conditions; all cited studies and limitations are recorded separately in the evidence ledger.
Insurance uptake as a temporally extended outcome
Adoption should be specified as more than a first premium. At least five distinct events matter: awareness, understanding, initial enrollment, renewal after experience, and successful use of benefits or claims. Social expectations can affect each stage differently. Membership may be publicly observable while benefit receipt is private; a paid claim may be visible while a rejection is concealed; an initial decision may follow a leader's instruction while renewal follows the household's own assessment of value. An intervention that changes one-period enrollment without changing persistence may have a different welfare interpretation from one that improves benefit comprehension and durable protection.
The distinction becomes especially important where fees are prepaid and the contingency is infrequent. A community can collectively learn that the insurer does honor payouts, yet the same community can develop skepticism when many premiums are paid without claims. No single conversion rate captures these evolving reputational dynamics. The Ghana insurance trials and rainfall-insurance research provide reasons to analyze cohort exposure and time since payout rather than aggregate community participation alone.
Scope and terminology
The expression community insurance is used comparatively for formal private or public insurance purchased within social networks, community-based health schemes, group distribution arrangements, and indigenous mutual institutions whose rules include insurance-like transfers. It does not assert legal equivalence among these arrangements. Formal insurance transfers defined contingent obligations to a provider governed by contract and regulation; informal risk-sharing relationships may rely on reciprocity, discretion, and social enforcement; statutory social protection rests on public guarantees and eligibility. The relevant similarities concern household risk management and social coordination. Legal status, capital backing, enforceability, contribution rules, and coverage differ and remain analytically explicit.
Normative Expectations and Group Identity
What makes a social expectation normative
A person's empirical expectation concerns what others do. A normative expectation concerns what others think the person should do. A personal normative belief concerns what the person independently regards as right. These are three distinct answers to three different questions. A farmer may believe that few neighbors insure, that neighbors approve of prudent preparation, and that she herself ought to protect her crop. Alternatively, she may believe that most neighbors insure but that respected elders regard purchase of an outside contract as a repudiation of reciprocal obligations. The observed enrollment rate alone cannot tell which situation exists.
Bicchieri's work on diagnosing norms makes the relevant reference network and conditional preferences essential. A social norm is identified when a person's preference for a behavior depends, in the appropriate way, on expectations about what others do and what they believe ought to be done. The research requirement is stronger than observing conformity. Repeated purchase behavior could be the rational response of households who face a common price and storm probability. The analyst must determine whether changing the perceived conduct or approval of relevant others changes the focal household's preference, after informational and material influences have been addressed.
Cialdini, Reno, and Kallgren (1990) distinguish descriptive regularities from injunctive prescriptions in their focus theory of normative conduct. In insurance, descriptive communication might say that most households in a cooperative renewed their policies. Injunctive communication might report that fellow members expect contributors to make suitable provision for foreseeable losses. Either message could have an effect, but a researcher would have to distinguish norm salience from facts about insurer quality and from any implied duty to contribute financially to others.
The insurance content of an alleged norm must also be named. A community may maintain a norm of mutual funeral contribution, a norm of avoiding debt, a norm of consulting senior household members, a norm of protecting children, or a norm of fulfilling cooperative commitments. These prescriptions can influence the use of an insurance product without producing a norm that literally commands insurance purchase. A study should avoid retroactively renaming every socially mediated practice an insurance-adoption norm.
Reference groups and conditional membership
Normative expectations are relational. A household may respond to its close kin, the people with whom it exchanges labor, cooperative leaders, faith-based association members, village health workers, or peers of similar income. Treating the entire village as a single reference group can average away opposing norms. High-status members might favor formal coverage while marginalized members perceive the scheme as inaccessible. In other settings, the informal association may be the relevant normative unit while geographic neighbors provide chiefly information.
The mechanism can be represented using ordinary decision analysis. Let an individual compare expected material advantages of insurance with its price, liquidity costs, and opportunity cost. Let social approval or disapproval contribute an additional perceived consequence that depends on the individual's reference group and expectations. This is a transparent accounting of choice determinants, not an empirically established structural model. It clarifies why the same subsidy can produce different responses when contract quality is identical but public identity and obligation differ. It also clarifies why measuring perceived norms without measuring product value is inadequate.
A reference group is endogenous to social position. Membership can confer access to trusted information, collective purchasing, enforcement, or social recognition. It can also be an outcome of wealth and power. Gaining an association membership and acquiring insurance may reflect the same preexisting resources. The analyst must record who can enter the group, who speaks for it, whose expectations are salient, and which sanctions are credible. The presence of an association does not imply a unified collective voice.
Group identity, obligation, and the meaning of independence
Insurance may be understood as collective responsibility, careful household management, or private exit from mutual support. The distinction depends on the contract and on the social organization of the group. In a burial association, joining an established fund can express continued belonging. Purchasing commercial funeral coverage elsewhere could be viewed as supplementary prudence or as withdrawal from reciprocal contributions. A public health insurance scheme linked to local service providers may be evaluated as a community achievement while its fees still compete with household consumption.
Institutional identity and purchasing identity can diverge. Residents may support national health protection as a social principle while remaining unable to pay voluntary premiums. Others may enroll to access an actual medical service without sharing the association's public ideals. The research design should separate normative approval of insurance, actual purchase intention, feasible purchase, and realized coverage.
Ajzen (1991) offers a useful distinction between subjective norms and perceived control. A household that strongly perceives community approval can still lack liquidity or access to a point of sale. Consequently, a null relationship between approval and enrollment does not establish the irrelevance of normative expectations; an observed positive association does not establish their causal role either. Attitude, capacity, and norm measures must be collected with separate instruments.
Norm measurement in insurance settings
A direct diagnostic instrument asks what the respondent believes most relevant peers actually purchase, what those peers think the respondent should purchase, and how much the respondent's own decision would change if the beliefs were different. The reference group must be specified and respondents should be allowed to identify multiple networks. Questions about sanctions must distinguish tangible consequences, such as exclusion from collective aid, from diffuse reputational concerns. The response burden and privacy costs require attention.
An appropriate survey asks about actual premiums, coverage limits, basis risk, claim exclusions, and perceptions of insurer reliability in the same study. Otherwise, an observed norm coefficient can simply proxy product comprehension. The coordination-game elicitation method of Krupka and Weber (2013) illustrates how social appropriateness can be measured through incentivized judgments about what a group is likely to regard as appropriate. Its insurance application would need an insurance-specific vignette and explicit external validation; no published insurance result is imputed to their dictator-game experiments.
A particular hazard is pluralistic ignorance: participants may privately favor insurance but incorrectly believe their peers oppose it. Correcting that false belief could change purchase behavior, but the direction cannot be predicted from aggregate enrollment rates. A field experiment would need baseline assessments of both private approval and second-order expectations, carefully avoiding disclosure of identifiable household decisions.
Historical forms of collective protection
Modern contract marketing often presents peer effects as a discovery of digital distribution. The historical record shows that local insurance-like arrangements have long depended on reciprocal institutions. Dercon, de Weerdt, Bold, and Pankhurst's study of funeral associations documents formalized rules and premium-like contributions among associations in Ethiopia and Tanzania. These arrangements give social membership economic substance because claims may be supported by established obligations, contribution records, and local enforcement.
Historical coexistence of informal solidarity and formal insurance has multiple possible trajectories. Community organizations may become distribution partners and reduce acquisition costs; households may retain both forms because each insures a different loss; or formal coverage may replace a narrowly defined mutual obligation. The direction depends on risk correlation, contractual reliability, premiums, and the institution's legitimacy. The funeral evidence establishes institutional features in its specific settings, rather than a universal norm of communal insurance.
Peer Behavior and Coverage Decisions
The identification problem behind apparent imitation
A common empirical claim is that individuals purchase insurance because their friends purchase it. That sentence contains several plausible causal paths. A friend's purchase can reveal information about policy terms. It can indicate confidence in an insurer. It can alter perceived popularity and appropriateness. It can activate a conversation that improves understanding. It can coordinate enrollment, trigger a premium discount, or change shared obligations. And friends can simply face similar risks and financial conditions. These pathways require different interventions.
A credible claim about normative conformity must isolate the social value of approval or expectation from the value of information. Providing an actual claim history may change beliefs about contractual payment. Informing a respondent that a respected neighbor purchased a policy may transmit both a signal of quality and an injunctive message. Informing the respondent that most other residents have insurance can also change perceived risk prevalence. Without independent measures or a design that separates channels, "social norms" remains a descriptive category rather than an identified mechanism.
Randomized network information in rural China
Cai, de Janvry, and Sadoulet's (2015) Social Networks and the Decision to Insure is especially instructive. The authors studied weather-insurance adoption in rural China and randomized intensive information sessions to subsets of farmers. Network exposure among untreated farmers thereby acquired an experimental source of variation. Their abstract reports that one additional treated friend had an effect on uptake equivalent to a 13 percent premium reduction. This comparison is a study-specific equivalence, not a universal premium elasticity or norm coefficient.
The design further varied available information about peers' decisions and randomized default options. The authors attributed the central network effect to diffusion of insurance knowledge rather than direct imitation of purchase decisions. This is strong evidence for social learning within the studied population, while cautioning against a broad claim that observable peer purchases generate normative conformity. The distinction is substantive: one policy implication is to teach well-connected farmers about real product terms, not to manufacture apparent popularity.
Even this relatively strong experiment requires careful scope. The treated network may convey both instruction and credibility; the observed outcome applies to the insurance design, information regime, and network structure under study. The randomized first-stage information exposure supports its identified effects; it does not imply that every insurance market will exhibit the same equivalence or that all interpersonal persuasion is informational.
Peer experience as evidence about contract performance
Karlan, Osei, Osei-Akoto, and Udry's (2014) Ghana agricultural insurance experiments provide a complementary dynamic interpretation. Randomized insurance and grant opportunities were used to investigate constraints on agricultural investment. Insured farmers increased agricultural investment and made riskier production decisions under the experimental conditions. Subsequent demand was associated with the household's own receipt of payouts and with payout experience among social-network members. These patterns are consistent with learning whether an unfamiliar insurer pays and how index insurance behaves under weather shocks.
A payout is both financial compensation and a public or private demonstration of reliability. If a neighbor receives money when rainfall is adverse, the experience can correct a belief that an insurer will never pay. It can also raise the perceived social acceptability of buying coverage. The study supports the importance of experience and trust, but claims about norm sanctions require an additional design. The intensity and direction of learning depend on whether a payout was expected, publicly communicated, and understood in relation to basis risk.
The demand consequences of a payout can be misleading when the household views insurance as an investment promising frequent cash returns. Transparent communication must distinguish insured contingency from a promised return on premium, explain why some severe individual losses do not trigger an index payout, and make claims and grievance channels usable. A high adoption rate achieved through incorrect expectations can turn into nonrenewal and reputational harm.
Indian evidence on social familiarity and market frictions
Giné, Townsend, and Vickery (2008), in Patterns of Rainfall Insurance Participation, observed that basis risk, wealth, and borrowing constraints predicted participation in a rural Indian product. Network participation and familiarity with the vendor were also correlated with take-up. Their evidence establishes that social relations appear alongside conventional financial variables, but the observational associations do not cleanly separate trust, social influence, and selection into networks.
Cole and colleagues (2013) extended the analysis of insurance demand using randomized interventions in India. They found meaningful price sensitivity, liquidity constraints, limited salience, and trust-related impediments. These results constrain any interpretation of insurance nonadoption as a simple failure of communal endorsement. When a family cannot pay a premium today, increasing the social importance of insurance may increase psychological pressure without creating feasible access.
The lessons from the two Indian studies are complementary. Social familiarity can matter, and so can contract-specific risk, current cash availability, and provider credibility. A program attempting to modify group norms while ignoring basis risk or payment timing is likely to misdiagnose an economic constraint as an attitudinal obstacle.
Disaster insurance and perceived norms
Lo (2013) examined 501 surveyed residents in eastern Australian cities in a flood-insurance study. Path analysis associated perceived social norms with flood-insurance purchase and found an indirect relationship involving risk perception. The results offer a plausible social-cognitive mechanism in a different insurance context from agricultural index products.
However, path coefficients drawn from a cross-sectional household survey are not experimental estimates of social norm manipulation. Perceived risk, perceived norms, and reported purchasing may be jointly determined. Respondents may infer neighborhood expectations from their own practices, or already insured households may attend more to pro-insurance messages. The study supports inquiry into norm-related mediation while leaving the causal direction and portability of the estimates open.
The flood-insurance case also emphasizes that the content of social comparison must be meaningful. Neighbors may differ in elevation, deductible, existing mortgage obligations, or local government relief expectations. A public statement about the proportion insured can be inaccurate guidance for a household whose dwelling is uninsurable or whose relevant risks fall outside the advertised product. Real coverage comparability remains a condition of responsible interpretation.
Financial peer effects beyond insurance
Bursztyn, Ederer, Ferman, and Yuchtman (2014) experimentally separated information about a peer's desired financial investment from the peer's actual possession in a brokerage field experiment. Their financial-product results indicate that social utility and social learning can be studied separately through carefully constructed treatment conditions. The experiment did not study insurance adoption and must not be cited as direct evidence that an insurance premium decision has the same social-utility mechanism.
Its methodological value is substantial. The insurance analogue would randomize knowledge of a peer's evaluation, knowledge of the peer's eventual ownership, and evidence concerning payout experience, while preserving privacy and accurate disclosure. Those conditions would help determine whether observed enrollment effects arise from information, positional consumption, or conditional expectations. They cannot be conflated with an incentive to pressure peers.
Short-term decisions and sustained demand
Initial enrollment may be strongly influenced by a trusted event, an introduced salesperson, or a subsidized first period. Renewal tests a different relationship between collective expectations and experienced contract value. A neighbor who endorses a scheme before claims are paid does not provide the same information as a neighbor who has successfully navigated a claim. Study designs that measure only first-year take-up may therefore overstate enduring social influence.
Adoption and renewal should be observed at the household-policy level with product attributes, claims events, premium changes, and membership changes recorded longitudinally. The research target is a sequence of decisions under changing information and changing group expectations. Treating the sequence as one cross-sectional binary outcome hides precisely the pathways that make insurance socially distinctive.
Community Influence and Trust
Trust is a multi-object belief
Trust in insurance may refer to a person's expectation that an agent states terms accurately, that the insurer is solvent, that a claim will be handled fairly, that healthcare providers will honor the policy, that a group treasurer will not misuse contributions, or that a regulator can provide recourse. Each expectation is institutionally specific. A household can trust its group leader while distrusting the insurer recommended by that leader. Conversely, a favorable view of a public provider may coexist with distrust of the local intermediary.
Insurance contracts are partly experience goods and partly credence goods: their value is difficult to verify before the relevant loss. Claim settlement gives the clearest signal, but only to those with a covered event. Social networks can aggregate others' experience, making reputation portable. Group identity can amplify the signal because members may treat one another as credible witnesses. These effects improve information and bargaining under some circumstances; reputational discipline can also spread misunderstandings if benefits are described inaccurately.
Cole and colleagues' trust-related insurance findings are relevant here because they situate distrust alongside budget constraints and attention. A community campaign that relies exclusively on exhortation cannot fix an opaque exclusion or unreliable service. Direct evidence of performance, accessible explanation, timely claims, and independent complaints mechanisms are institutional inputs into trust.
Engagement through existing informal groups in Kenya
Chemin (2018), in Informal Groups and Health Insurance Take-up, reports a series of randomized interventions with the Kenyan National Hospital Insurance Fund. Conventional information, registration assistance, and small subsidies failed to produce significant uptake increases in the studied settings, while a full subsidy induced initial purchases with no subsequent retention after the subsidy ended. A complementary intervention brought information and discussion to existing informal groups.
The reported group-presentation treatment generated approximately 12 percent take-up and approximately 7 percent after one year among participants under the described conditions, without a premium subsidy. The intervention appears able to activate discussions and strengthen product familiarity in a social setting. At the same time, it bundled multiple mechanisms: group attention, social facilitation, testimony, trust, and possible normative approval. Its positive causal treatment effect is not equivalent to a separately estimated social-norm parameter.
The welfare interpretation must also distinguish durable coverage from a temporarily induced sale. A group conversation could lower information costs while leaving provider limitations intact. A full subsidy that induces transient enrollment can solve an immediate affordability barrier but generate little post-subsidy persistence. Comparing the designs therefore requires not just initial adoption but renewal, premiums paid, claims experienced, and whether participants correctly understood the coverage.
Ghanaian evidence on community engagement
Fenenga and colleagues (2018) studied health insurance and social capital in a Ghanaian cluster-randomized trial. The study included 3,246 clients from 64 healthcare facilities with baseline and follow-up observations. The authors report that observed insured shares rose from 40.29 to 49.39 percent in the intervention group and from 36.49 to 36.75 percent in the comparison group. These are group-specific observed shares, not by themselves an unadjusted estimate transferable to another country's insurance system.
The authors also found associations between several forms of social capital and enrollment, while reporting no evidence that the intervention itself changed social capital as hypothesized. The difference is crucial. A randomized program can increase coverage through improved engagement or institutional processes while a preexisting measure of interpersonal relations predicts who enrolls. That does not demonstrate that changed social norms caused the program effect.
Because the study reports nonidentical baseline insured shares, and intervention was allocated by facility clusters, rigorous interpretation relies on the authors' design and analysis rather than a simple comparison of percentages. A monograph should not subtract the two observed changes and label the result an individual-level normative effect. Nor should the absence of a detected social-capital change be mistaken for proof that all social influences were irrelevant. It limits a particular mediation hypothesis.
Service quality and social narratives in Ghana
Jehu-Appiah, Aryeetey, Agyepong, Spaan, and Baltussen (2012) used household data to investigate perceptions of Ghana's National Health Insurance Scheme. Their analysis associated enrollment and retention with perceptions of scheme benefits, convenience, and price; it also examined healthcare providers, community beliefs, and peer pressure. Negative perceptions of peer pressure appeared alongside unfavorable views of costs and provider attitudes in the study context.
This case demonstrates that social pressure has direction and quality. If network members transmit dissatisfaction about poor service, the resulting influence may deter adoption. If a scheme is responsive and claims are honored, those same channels may support participation. Calling the former anti-insurance norms and the latter pro-insurance norms would be premature without measuring the content of the relevant expectations, but the empirical asymmetry matters for distribution practice.
It follows that a high-trust messenger cannot repair a structurally weak product. Information from a respected individual may increase attention to the policy, yet disappointed members can propagate a strong adverse experience later. Product design and claims governance are foundational to any sustainable community adoption strategy.
Trust, endorsement, and accountability
A community leader may function as an information intermediary with unusually broad network reach. Endorsement can reduce search costs and signal accountability. But it can also create a conflict of interest if a distributor rewards leaders for enrollment or if a group representative discloses members' circumstances to an insurer. Any intervention relying on a leader must document the nature of the relationship, financial compensation, claims accountability, and whether participation remains voluntary.
A responsible insurance program uses independent product documentation and plain-language illustrations, gives participants a confidential route to decline, and measures whether messages are understood. Such procedures protect both autonomy and the reliability of the resulting research. Reported group approval acquired under undisclosed compensation is a contaminated measure of normative influence.
The international supervisory context reinforces this point. The International Association of Insurance Supervisors' application paper on fair treatment addresses access, product design, distribution, advice, claims, complaints, and ongoing monitoring. Its relevance here is an institutional conduct standard, not a causal empirical finding or a universally self-executing law. Jurisdictional rules and enforcement mechanisms must be checked independently.
Institutional feedback loops
Insurance participation affects the very signals on which later participation depends. More members can increase the visible legitimacy of a scheme, expand conversational knowledge, or improve the diversity of pooled risks. At the same time, rapid growth can strain processing capacity, provoke delays, or expose previously hidden exclusions. These feedback loops have different implications for causal inference, especially where enrollment itself changes service quality.
For health schemes, clinical waiting times and provider reimbursement may shift with participation. For disaster insurance, the number of claims following a shared catastrophe can alter perceptions of insurer solvency. The proper outcome is therefore not simply the number of new policies; it also includes contract performance and continued access. Insurers and public institutions influence social expectations through observable conduct.
Informal Risk Sharing Norms
The economics of reciprocal transfers
Informal risk sharing organizes transfers across households or groups when an identifiable member experiences a shock. It may involve gifts, loans, labor exchange, funeral contributions, mutual food assistance, or a fund with prescribed benefit rules. The obligations can be supported by future reciprocity, social approval, reputation, or recognized leadership. The insurance function is partial because enforcement and payouts depend on relationships, resources, and the correlation of losses across members.
Townsend's (1994) Risk and Insurance in Village India tests a full-insurance benchmark with village consumption data. It finds substantial consumption comovement with village-wide resources and limited sensitivity of household consumption to several idiosyncratic shocks, while statistically rejecting the idealized full-insurance specification and identifying unequal protection for some landless households. This supports a nuanced understanding: village networks can smooth individual shocks without reproducing an actuarially complete insurance contract.
The counterfactual to formal insurance is therefore frequently an imperfect but meaningful social institution. A household can decline a policy because existing reciprocal arrangements cover a substantial part of the relevant loss, not because it misunderstands risk. Conversely, informal assistance may be discretionary, slow, or unable to compensate catastrophic community-wide losses. Evaluating formal adoption requires identifying what the household would otherwise receive and what social obligations it would retain.
Funeral associations as organized mutual protection
Dercon and colleagues' (2006) Group-Based Funeral Insurance documents locally organized funeral associations in Ethiopia and Tanzania. The institutions described had written or recognizable rules, contributions that resembled premiums, and benefits that could finance costly funeral obligations. They also sometimes extended assistance beyond funeral needs. Their history of resisting external capture is important to the interpretation of institutional trust.
A burial association is simultaneously a mechanism of risk management and a form of recognized collective membership. Joining or contributing may be sustained by community expectations and accountability. However, the article's institutional descriptions do not identify a universal causal effect of normative expectations on buying a commercial insurance contract. The legal and financial form differs from a regulated carrier, and localized funeral costs can differ sharply from illness and harvest losses.
A formal insurer proposing partnership with such an association should examine governance, benefit liabilities, contribution collection, membership rules, and the risk of transferring control away from trusted representatives. Administrative efficiency does not automatically justify replacing existing reciprocal practices.
Network formation and selection
Fafchamps and Gubert (2007) examine the formation of risk-sharing networks in rural Philippines data. Geographic proximity and social characteristics help explain mutual-insurance links, and reported links predict subsequent gifts and loans. The findings underline that risk-sharing networks are selected and socially structured, not random collections of economically identical people.
This matters when measuring norm effects. Suppose insured households are concentrated among neighbors who have reliable financial partners. The observed correlation could reflect a preference for insurance, a common level of social trust, a shared information environment, or the selection of more economically stable households into both networks and formal coverage. Ordinary regressions with village fixed effects will not necessarily separate these explanations.
Network maps should record tie type, direction, reciprocity, and baseline relationships. A friendship connection is not interchangeable with a credit obligation, familial relation, cooperative membership, or social-authority tie. The content of a relationship determines what information and sanctions it can transmit.
Formal index insurance and informal mutual aid
Mobarak and Rosenzweig's (2013) study of informal risk sharing and index insurance analyzes randomized rainfall-insurance offers in Indian villages with preexisting information on caste-based networks. It explicitly treats informal protection and formal index coverage as interacting institutions. The article presents preliminary findings from a broader research program rather than a universal final estimate of substitution or complementarity.
A useful analytical distinction is the scale and covariance of losses. When one household faces illness and neighbors remain financially able to help, an informal network may transfer resources effectively. When an entire area experiences drought, the same neighbors may simultaneously face losses, reducing mutual capacity. An index contract backed by a wider pool or external capital can potentially pay against an area-wide event, but basis risk can leave an individual harmed without payment. Informal and formal mechanisms thus cover overlapping yet different states.
The decision to purchase can reflect several configurations. Formal coverage may crowd out reliance on neighbors when both fund the same loss. It may crowd in productive risk-taking when an externally financed payout protects shared assets. It may be complementary when formal indemnity covers a systemic event and informal aid covers unindexed household emergencies. Which occurs is empirical and contract-specific.
Basis risk and collective perceptions of fairness
Index insurance is triggered by an observable proxy, not necessarily by a household's realized loss. A disappointed member can interpret a nonpayment as a broken promise even when the contract operated exactly as written. When such experience spreads through a network, it can create shared mistrust that resembles normative opposition to the insurer. The first research task is to establish whether expectations about the index were accurate.
Giné, Townsend, and Vickery (2008) link basis risk to lower participation. The result provides a concrete economic reason for divergence between apparent group approval and actual purchase. It also shows why describing group reluctance as culture-bound resistance would conceal a contract-design limitation. Communities can make sophisticated judgments about imperfect indemnity, even when communication is informal.
Group discussion can improve contract comprehension if examples include payment and nonpayment cases. It can also become a vehicle for social pressure that discourages questions. An effective educational intervention should test understanding privately and permit informed refusal.
Poverty, risk correlation, and the boundary of community finance
Very small mutual groups face two linked constraints: limited ability to raise contributions and correlated losses. A collective promise that works for a member's funeral expense may fail during an epidemic, regional flood, or crop failure. Large-pool insurance can diversify some exposures but still requires capital and sound risk pricing. Public social insurance may rely on compulsory contributions, taxes, transfers, or government guarantees. These mechanisms operate through different legal and financial obligations.
The International Labour Organization's Social Protection Floors Recommendation identifies nationally defined basic social protection guarantees and extension to excluded populations as policy priorities. Community approval can help communicate entitlements, but its presence cannot substitute for a legal guarantee or adequate financing. A voluntary association with socially exclusionary membership rules must not be treated as equivalent to universal protection.
Culture, Gender, and Local Institutions
Culture as situated institutions and beliefs
Comparative research often invokes culture after conventional economic covariates leave variation in uptake unexplained. Such residual interpretation is scientifically weak. The practical content of culture should be specified through observable institutions, shared practices, information channels, reference groups, household decision authority, treatment of uncertainty, and expectations of reciprocity. Researchers must separate these from product availability, price, legal entitlements, and provider performance.
Two settings can exhibit similar low purchase rates for entirely different reasons. In one, community members may be willing to insure but cannot make a lump-sum payment. In another, members can pay but doubt whether benefits will be received. In a third, informal reciprocal coverage may address the relevant risk. In a fourth, normative expectations may discourage visible enrollment because contributions to an outside scheme are seen as withdrawal from existing obligations. The observed outcome is not enough to assign the mechanism.
Cross-national generalization is especially hazardous when the products differ. A voluntary index contract in China, a national health scheme in Ghana, a funeral association in Ethiopia, and residential flood coverage in Australia have different enforcement, risk bases, frequencies of observable payouts, subsidy structures, and purchase channels. A comparable social norm instrument may not have comparable meaning without local validation.
Nepal: bonding and bridging social capital
Ko, Kim, Yoon, and Kim (2018) examined social capital and willingness to join community-based health insurance through a survey of more than 4,000 Nepalese households. They report positive associations between social capital and willingness to join or pay, with differences between bonding ties and bridging connections. The willingness-to-pay outcome is a stated preference; it should not be converted into observed long-term insurance renewal.
Bonding capital can support close reciprocity, familiarity, and internal solidarity. Bridging capital can link households to more diverse associations and external institutional opportunities. These properties may produce distinct relationships with government subsidies and formal risk pooling. But their measurement is not an independent estimate of normative expectations. Networks may provide credit, knowledge, political access, or practical enrollment assistance.
The main implication is to measure the kind of relationship, not just its existence. An index labeled social capital may bundle trust, participation, reciprocity, and control; a statistically significant coefficient cannot identify which component caused uptake. Reporting the underlying items and institutional context is part of the scientific result.
Senegal: the equity costs of selective embeddedness
Mladovsky, Soors, Ndiaye, Ndiaye, and Criel (2014) investigated insurance enrollment and social capital in three Senegalese community-based health schemes through surveys and qualitative interviews. Members exhibited forms of bridging social capital and community participation that distinguished them from nonmembers. The authors caution that recruiting through existing associations can reinforce exclusion of people with fewer resources or weaker social connections.
This evidence exposes a recurring tension. An established group can lower delivery costs and disseminate credible information, making it attractive to insurers and policymakers. Yet households without membership, recognized status, stable income, or local influence may be left outside the distribution network. A successful enrollment campaign among already-connected participants can therefore widen disparities in access even as aggregate uptake rises.
Equity evaluation must include the eligible population beyond the groups that host recruitment meetings. The denominator matters: a high conversion percentage among reachable association members says little about the share of vulnerable households served. Reporting nonparticipation by income, disability, migration status, gender, age, and geographical isolation requires appropriate privacy protection and locally relevant categories.
Ghana: peer pressure and institutional legitimacy
The Ghana NHIS findings of Jehu-Appiah and colleagues (2012) are particularly useful because peer pressure appears among unfavorable perceptions. Social information channels can propagate distrust when members have disappointing experiences with providers or administration. This challenges a policy assumption that a community's dense networks necessarily improve enrollment.
The same country contains evidence of successful engagement through institutions. Fenenga and colleagues' cluster-randomized Ghana trial supports a positive program effect on coverage, but its reported absence of induced social-capital change limits attribution to that mediator. These findings are compatible because the intervention, sample, time, institutional operation, and measured variables differ. They demonstrate why a single national cultural label has little explanatory value.
Gender, household authority, and privacy
The decision recorded as a household's insurance choice may be negotiated among several people. A policyholder may be male, while the primary beneficiary of health coverage is female; a cooperative representative may approve enrollment while another household member controls cash. Norms concerning who can sign agreements, travel to an enrollment point, discuss illness, or receive indemnity can shape both demand and realized protection.
Gender and power should be investigated as institutional variables rather than imported stereotypes. Researchers should ask who is eligible, who holds identity documentation, who supplies premium funds, who receives benefits, whether consent is required from another person, and whether social punishment follows independent financial choices. These are empirical research questions whose answers can vary within the same locality.
The 2025 IAIS fair-treatment paper emphasizes accommodating different consumer circumstances across the insurance lifecycle. For community recruitment, the relevance includes accessible communication, suitable sales practice, and the avoidance of socially coercive or discriminatory processes. The instrument does not establish a universal legal duty in every jurisdiction; it sets out supervisory guidance.
The institutional status of the scheme
Insurance adoption studies must distinguish public enrollment rights from voluntary purchase. A statutory entitlement, tax-financed protection program, cooperative membership arrangement, and commercial risk contract involve distinct options. A public scheme's eligibility can be affected by residence, employment status, or administrative registration, while a private product may price individual risks and impose exclusions. Community pressure interacts with the opportunity set that law and market structure actually provide.
The ILO Recommendation No. 202 provides a normative international policy reference for protection floors and inclusion. Its protective objective should be distinguished from empirically measured social norms about whether a particular household should purchase a voluntary commercial contract. Terminological precision prevents legal and social meanings of "norm" from being confused.
International comparison without a false pooled coefficient
The Chinese experiment identifies a causal network information channel in weather insurance. The Kenyan study identifies a group-presentation treatment effect in a particular national health product. The Ghana cluster trial identifies effects of community-engagement interventions and explicitly limits a proposed social-capital mediation explanation. Australian flood-insurance analysis associates perceived norms with purchasing through a cross-sectional design. The Senegal and Nepal studies relate social connections to enrollment or willingness, while the Ethiopian and Tanzanian evidence describes mutual institutions.
These findings do not estimate a common social-norm effect on a common outcome. The population, contract, treatment, outcome, time horizon, and identifying assumptions differ. A meta-analytic average would require harmonized measures, defensible effect transformations, and a systematic inclusion protocol. This monograph therefore interprets mechanisms across settings rather than manufacturing a general numerical estimate.
Identification of Norm Effects
Defining the causal question before collecting data
A rigorous study starts by specifying its outcome, treatment, unit, reference group, time horizon, and counterfactual. "The community affects insurance" is too broad. A suitable question is whether correcting an individual's belief about a specified reference group's approval, while preserving accurate insurance facts and keeping other treatment components constant, changes verified enrollment by a fixed deadline. A second question may ask whether revealing neighbors' actual claim payments changes renewal through provider credibility.
The two questions cannot be substituted. The first concerns perceived normative expectations; the second concerns information about contractual performance. A third study might examine whether communal distribution reduces travel time, payment friction, or transaction costs. Network contact is a delivery channel common to all three, but it is not the mechanism of interest.
The unit of analysis must also reflect exposure. Household enrollment may be an individual outcome nested within communities, while treatment spreads across friends and associations. A simple individual-level randomized experiment with controls who routinely meet treated neighbors violates the standard assumption that one household's outcome depends only on its own assignment. Identified network effects need designs and estimands that recognize spillovers.
Reflection, shared shocks, and endogenous networks
A neighbor's enrollment and a household's enrollment can be positively correlated for at least three standard reasons: one influenced the other, both responded to a common event, or households selected into a relationship because they shared latent preferences. This is the classic peer-effect identification problem in insurance-specific form. A storm, a mobile-sales visit, a premium subsidy, or a trusted doctor's message can create simultaneous take-up without any direct peer transmission.
Geographical clustering is an inadequate instrument for normative exposure when flood or rainfall risks are themselves geographically clustered. A social network may be associated with wealth, education, trade, and access to agents. Peer behavior can also respond to the focal household's decision, producing simultaneity. A regression of own uptake on the average insurance rate of peers cannot distinguish these channels by itself.
The randomized information-session design of Cai, de Janvry, and Sadoulet directly addresses parts of this difficulty because exposure to informed friends varies for experimental reasons. Yet the demonstrated channel is mainly knowledge diffusion. It is therefore a benchmark for good design and a warning against causal overlabeling.
Direct elicitation of social expectations
Measurement should distinguish: a respondent's own approval of insurance; empirical expectations concerning peers' purchases; normative expectations concerning peers' approval; expectations of sanctions or reputational consequences; perceived insurer reliability; understanding of covered losses; affordability; and actual choices. These variables should have separate, plain-language questions and be linked to specific reference groups.
Krupka and Weber's (2013) incentivized norm elicitation can inform the design of vignettes about socially appropriate decisions. For instance, respondents might predict how other community members evaluate purchasing a health policy, continuing a funeral contribution while obtaining commercial coverage, or declining an unfamiliar index contract. But a coordination-game consensus is a measure of shared appropriateness judgments; further work is required to demonstrate that a person's insurance decision is conditional on those judgments.
Sensitive questions should be asked privately where possible. Surveyors embedded in the same local association can induce social desirability bias. The identity of the interviewer, language, examples, and public or confidential survey setting may change respondents' reports. Translation requires cognitive testing so that "most people do," "people think I should," and "I think I should" remain distinct propositions.
Randomized informational and normative interventions
A clean experimental comparison might allocate communities or eligible households to independently varied, ethically accurate messages: product facts; verified descriptive prevalence; elicited or accurately reported community approval; and a neutral control with the same practical enrollment access. A factorial design would allow estimation of incremental effects and interactions, provided the normative statement was factually established, consent was obtained, and contamination was controlled. A researcher must never invent consensus or use fabricated testimonials to create normative pressure.
An informational arm should explain the insurer, premium, coverage limits, exclusions, basis risk, and claims procedures in equal detail across relevant comparison groups. A descriptive-norm arm could communicate a verified local purchase statistic, but only if privacy and representativeness permit. An injunctive-norm arm poses additional ethical and epistemic difficulties: it needs a valid measure of approval from an appropriate reference group, and it must avoid coercion by authority figures. Where these conditions cannot be met, vignette or observational work may be preferable.
Claims-performance treatments offer another mechanism probe. Providing audited examples of actual payouts and covered nonpayments can alter reliability beliefs without implying that everyone else approves of the policy. Separately observing responses to such messages helps identify when people are learning about contract performance rather than conforming to presumed norms.
Experimental estimands and clustering
Let Y be an accurately measured outcome such as enrollment by a prespecified date. The average treatment effect is the difference in expected Y under assigned exposure to a precisely defined treatment and under its comparator. The design determines whether that contrast represents a direct effect, an intention-to-treat effect, or a combined direct-and-spillover effect. No estimate is supplied in this monograph because no new data or experiment has been conducted.
Where communities are randomized, analysis must account for the number and sizes of randomization clusters. Standard errors calculated as though thousands of members were independently assigned would overstate precision. Prespecifying primary outcomes, dealing with attrition, measuring baseline coverage, and correcting multiple comparisons where warranted are ordinary requirements. A null finding may result from low statistical power or weak manipulation of expectations; it should not be transformed into proof that normative effects are absent.
Random saturation designs can identify how exposure to treated peers changes untreated individuals, provided links and interference assumptions are explicit. Encouragement designs may estimate the effect of assignment to outreach even when not all recipients attend. An instrumental-variable interpretation requires credible exclusion restrictions: for example, a leader's invitation may influence trust directly, invalidating a claim that it operates exclusively through improved factual knowledge.
Differentiating information from normative compliance
Four observable patterns deserve separate interpretation. An increase following verified clarification of benefits suggests an information mechanism. An increase following a truthful payout demonstration suggests belief updating about institutional performance. An increase following accurately measured peer approval, after factual content and material incentives are held constant, is stronger evidence for a normative channel. An increase following group sales access may reflect transaction-cost reduction or coordinated logistics.
Each pattern remains subject to experimental and measurement assumptions. A respondent's publicly announced intention can change without a private policy purchase; an enrolled individual can withdraw after the first contribution; a survey response can reflect courtesy to a fieldworker. Observed paid premiums, contract issuance, policy status, and claims realization should be linked wherever consent and privacy protection allow.
The study should test whether the intervention changed its intended mediator. A norm intervention that moves enrollment but fails to change measured normative expectations cannot be confidently described as proof of norm-induced adoption. Conversely, measured expectations can change without observable take-up if premiums remain unaffordable. Mediator analysis introduces additional assumptions even under randomized treatment assignment and should be identified as such.
Heterogeneity and distributional effects
Average effects can conceal opposite responses. A well-connected household may benefit from group endorsement, while a socially isolated one remains unreachable. An influential individual may publicly endorse coverage while quietly benefiting from commission income. A person facing stigma regarding illness may avoid an enrollment event even if coverage would be valuable. A household with strong reciprocal support may substitute away from formal insurance, whereas another may purchase complementary protection.
Heterogeneity should therefore be examined using prespecified, substantively grounded characteristics such as initial access to information, contract comprehension, credit constraints, type of social tie, prior claim experience, gendered decision authority, and membership in existing mutual organizations. Some dimensions will have small cell sizes, and exploratory patterns should remain labeled exploratory. Broad claims about a national culture or a demographic group should not be extrapolated from one site.
An ethical evaluation compares both who gained and who was omitted. The Senegal community-insurance findings are a warning that associational recruitment can concentrate benefits among already-advantaged network members. Inclusive outreach may need channels outside existing groups, along with premium subsidies, accessible administration, and enforceable consumer protection.
Claims integrity, consent, and supervisory standards
Researchers and insurers should make the factual status of every normative claim auditable. If a campaign asserts that most comparable households hold a given policy, the denominator, place, reference date, renewal status, and product class should be documented. If a campaign quotes a member's claims experience, explicit permission and accurate context are necessary. An apparently pro-insurance social norm based on misleading information should not be accepted as a policy success.
The IAIS fair-treatment application guidance underscores that access, communication, claims handling, and complaints deserve scrutiny throughout the product lifecycle. A survey or group-enrollment intervention must meet applicable local law, research ethics review where required, and privacy requirements. The ILO social protection standard further directs attention to whether vulnerable populations have practical access to protection rather than merely being exposed to persuasion.
Evidence hierarchy and comparative assessment
The research evidence supports different strengths of conclusion. Randomized studies identify the effects of specified intervention packages subject to compliance and spillover assumptions. Network-information experiments can separate particular informational channels from purchase imitation. Cross-sectional norm measures and associational studies identify patterns that invite causal testing but do not establish causal pathways. Institutional ethnography and descriptive case studies explain how local rules and membership operate without implying a pooled treatment effect.
The international pattern is coherent at the level of mechanisms. The Chinese work identifies informational social learning; Ghana and Kenya show the relevance of trust and organized engagement in health coverage; the India and Ghana agricultural studies identify affordability, basis risk, and payout experience; Nepal and Senegal illuminate the differing social capital and equity dimensions of community schemes; the funeral and village-insurance studies show that reciprocal institutions may provide real protection before formal policies arrive. The evidence does not justify a single causal parameter called the social-norm effect on global insurance adoption.
Practice implications derived from the evidence
Insurance distributors should diagnose an observed protection gap before selecting a campaign. If consumers do not understand the contract, provide accurate, independent education. If the product is priced beyond feasible budgets, address affordability or payment timing through legitimate means. If claims are unreliable, fix contractual and operational performance. If norm-related stigma is demonstrated, consider confidential access and locally validated communication. If existing mutual institutions are valuable, assess partnership without appropriating or undermining them.
For researchers, a credible study begins by declaring which mechanism is hypothesized, who belongs to the reference group, what information is held constant, and what a successful insurance outcome means. For policymakers, successful community outreach is judged against durability, fairness, quality, and household protection. For insurers, a trusted group relationship is an institutional responsibility as well as a distribution opportunity.
These are evidence-based analytical recommendations, not an original theoretical system. No new named model, doctrine, or distinct social-norm law is asserted. Their value lies in disciplined differentiation of existing mechanisms and transparent identification of the evidence each claim can sustain.
Conclusion
Social relationships are constitutive of many real-world insurance decisions. Households learn from peers, attend to the approval of relevant others, depend on reciprocal institutions, evaluate leaders' endorsements, and draw inferences from the claims experience of friends and relatives. The scientific task is to identify which influence operates in a given contract, population, and institutional setting.
The literature establishes several robust findings at appropriate levels of inference. In rural China, randomized network exposure facilitated knowledge diffusion concerning weather insurance, and the study separated that channel from simple purchase imitation. In India, contract basis risk, liquidity, price, trust, and social familiarity all appeared relevant to low rainfall-insurance uptake. In Ghanaian agriculture, insurance provision affected investment and subsequent demand responded to payout experience. In Kenyan health insurance, engagement through existing informal groups improved take-up and persistence under the tested conditions, while more conventional inducements often performed poorly. In Ghanaian health coverage, cluster-randomized engagement increased enrollment without demonstrating the hypothesized change in social capital. Comparative work in Nepal and Senegal identifies associations between social connections and coverage while exposing serious selection and equity concerns. Funeral associations and informal village insurance document practical risk-sharing functions that must be recognized in any comparison with formal contracts.
The evidence also clarifies what remains unidentified. A measured correlation between peer insurance and own insurance does not establish normative compliance. A trial that increases enrollment through group meetings does not establish whether information, social approval, trust, or access generated the effect. A path analysis linking risk perception and perceived norms is informative about association, not proof of mediated causality. These boundaries are not editorial reservations; they are the substantive foundations of a credible science of social influence.
An appropriate empirical program measures private attitudes, empirical expectations, normative expectations, reference groups, material constraints, trust, contract comprehension, and actual behavior separately. It uses designs that account for network selection, shared shocks, interference, and dynamic renewal. It discloses how group-based outreach affects socially isolated or financially vulnerable populations, and it evaluates claims performance as part of the outcome rather than treating enrollment as sufficient.
The central interpretation is that community influence affects insurance through identifiable and potentially competing channels: information transmission, expectation-sensitive conduct, institutional trust, transaction-cost reduction, and existing informal protection. Their strength and direction depend on observable contract attributes and social arrangements. Accurate diagnosis makes possible insurance practices that improve meaningful protection and preserve informed, voluntary, equitable participation. The work closes as a critical international synthesis of established evidence, with no claim to a new original theory, new dataset, pooled estimate, or external peer review.
References
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
Bicchieri, C. (2017). Norms in the Wild: How to Diagnose, Measure, and Change Social Norms. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780190622046.001.0001
Bursztyn, L., Ederer, F., Ferman, B., & Yuchtman, N. (2014). Understanding mechanisms underlying peer effects: Evidence from a field experiment on financial decisions. Econometrica, 82(4), 1273–1301. https://doi.org/10.3982/ECTA11991
Cai, J., de Janvry, A., & Sadoulet, E. (2015). Social networks and the decision to insure. American Economic Journal: Applied Economics, 7(2), 81–108. https://doi.org/10.1257/app.20130442
Chemin, M. (2018). Informal groups and health insurance take-up: Evidence from a field experiment. World Development, 101, 54–72. https://doi.org/10.1016/j.worlddev.2017.08.001
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social Psychology, 58(6), 1015–1026. https://doi.org/10.1037/0022-3514.58.6.1015
Cole, S., Giné, X., Tobacman, J., Topalova, P., Townsend, R., & Vickery, J. (2013). Barriers to household risk management: Evidence from India. American Economic Journal: Applied Economics, 5(1), 104–135. https://doi.org/10.1257/app.5.1.104
Dercon, S., de Weerdt, J., Bold, T., & Pankhurst, A. (2006). Group-based funeral insurance in Ethiopia and Tanzania. World Development, 34(4), 685–703. https://doi.org/10.1016/j.worlddev.2005.09.009
Fafchamps, M., & Gubert, F. (2007). The formation of risk sharing networks. Journal of Development Economics, 83(2), 326–350. https://doi.org/10.1016/j.jdeveco.2006.05.005
Fenenga, C. J., Buzasi, K., Arhinful, D. K., Duku, S. K. O., Ogink, A., & Poortinga, W. (2018). Health insurance and social capital in Ghana: A cluster randomised controlled trial. Global Health Research and Policy, 3, Article 35. https://doi.org/10.1186/s41256-018-0090-y
Giné, X., Townsend, R., & Vickery, J. (2008). Patterns of rainfall insurance participation in rural India. The World Bank Economic Review, 22(3), 539–566. https://doi.org/10.1093/wber/lhn015
International Association of Insurance Supervisors. (2025). Application Paper on Fair Treatment of a Wide Range of Consumers. Published July 7, 2025. https://www.iais.org/2025/07/iais-publishes-application-paper-on-fair-treatment-of-a-wide-range-of-consumers/
International Labour Organization. (2012). Social Protection Floors Recommendation, 2012 (No. 202). International Labour Conference, 101st Session, adopted June 14, 2012. https://normlex.ilo.org/dyn/nrmlx_en/f?p=NORMLEXPUB%3A12100%3A0%3A%3ANO%3A%3AP12100_ILO_CODE%3AR202
Jehu-Appiah, C., Aryeetey, G., Agyepong, I., Spaan, E., & Baltussen, R. (2012). Household perceptions and their implications for enrollment in the National Health Insurance Scheme in Ghana. Health Policy and Planning, 27(3), 222–233. https://doi.org/10.1093/heapol/czr032
Karlan, D., Osei, R., Osei-Akoto, I., & Udry, C. (2014). Agricultural decisions after relaxing credit and risk constraints. The Quarterly Journal of Economics, 129(2), 597–652. https://doi.org/10.1093/qje/qju002
Ko, H., Kim, H., Yoon, C.-G., & Kim, C.-Y. (2018). Social capital as a key determinant of willingness to join community-based health insurance: A household survey in Nepal. Public Health, 160, 52–61. https://doi.org/10.1016/j.puhe.2018.03.033
Krupka, E. L., & Weber, R. A. (2013). Identifying social norms using coordination games: Why does dictator game sharing vary? Journal of the European Economic Association, 11(3), 495–524. https://doi.org/10.1111/jeea.12006
Lo, A. Y. (2013). The role of social norms in climate adaptation: Mediating risk perception and flood insurance purchase. Global Environmental Change, 23(5), 1249–1257. https://doi.org/10.1016/j.gloenvcha.2013.07.019
Mladovsky, P., Soors, W., Ndiaye, P., Ndiaye, A., & Criel, B. (2014). Can social capital help explain enrolment (or lack thereof) in community-based health insurance? Results of an exploratory mixed methods study from Senegal. Social Science & Medicine, 101, 18–27. https://doi.org/10.1016/j.socscimed.2013.11.016
Mobarak, A. M., & Rosenzweig, M. R. (2013). Informal risk sharing, index insurance, and risk taking in developing countries. American Economic Review, 103(3), 375–380. https://doi.org/10.1257/aer.103.3.375
Townsend, R. M. (1994). Risk and insurance in village India. Econometrica, 62(3), 539–591. https://ideas.repec.org/a/ecm/emetrp/v62y1994i3p539-91.html