RESEARCH DOSSIER
Mental Accounting and Insurance Budget Decisions
Abstract. This monograph examines the economic and behavioral consequences of assigning insurance premiums, emergency reserves, precautionary savings, and possible losses to separate mental budgets. It distinguishes the cognitive allocation of money from actual legal restrictions, liquidity constraints, contractual design, risk aversion, and incomplete information. The study reconstructs the theory of mental accounting and the classical benchmark of fungible household resources, then reviews evidence on deductible choice, health insurance, warranties, agricultural risk protection, earmarked savings, and household expenditures. The most directly pertinent new empirical work is the 2026 Singapore MediSave study of how designated premium funding changes health-plan choice. A critical synthesis identifies testable distinctions among payment source, premium salience, loss framing, reimbursement form, and financially real constraints. The principal conclusion is that account boundaries can alter the subjective price and accessibility of insurance, while their welfare and causal importance depend on the contract, the household's actual options, and the institutional context. Findings are attributed at the level supported by original studies; no independent microdata reanalysis, global meta-analysis, or newly named theory is claimed.
Keywords: Mental accounting; mental budgeting; insurance demand; household finance; premium financing; earmarking; emergency savings; precautionary saving; risk transfer; deductibles; liquidity constraints; narrow framing; source of funds; health insurance; MediSave; index insurance; consumer welfare; behavioral insurance.
Introduction
Insurance is purchased within a household's financial life. The same balance sheet that contains income, liquid deposits, debt, emergency reserves, retirement savings, and expected bills also contains the premium paid to transfer a contingent loss. Households may nevertheless classify the premium as a current expense, hold emergency savings apart from routine consumption, and regard a deductible as an exceptional future payment. The resulting decisions can differ even when the underlying monetary consequences are equal. Mental accounting provides a research language for examining those classifications and their effects on insurance choice.
The central proposition of this dossier is an empirical and analytical one: the financial account from which a premium is paid, the account designated for an emergency, and the category into which a possible loss is placed can change how people value insurance. This proposition is tested against theories of risk preference, liquidity, attention, imperfect information, product design, and social institutions. The direct insurance evidence has grown substantially, most notably with Gruber and colleagues' 2026 study of premium payment from Singapore's MediSave accounts, which links the method of financing coverage to plan generosity and premium sensitivity through observed program-rule variation (MediSave plan-choice study). That finding merits close examination of identification, institutional scope, and welfare rather than universal extrapolation.
Thaler's analysis of mental accounting (1999) identifies how individuals and households organize, assess, and track financial activities. Heath and Soll's mental-budgeting experiments (1996) demonstrate that expenditure categories can constrain otherwise attractive choices. The insurance application presents a particular difficulty: premiums are paid in the present while indemnities, if any, arrive later and in contingent states. A household may therefore judge premium outflow against a monthly spending limit while judging a catastrophic loss against resources mentally earmarked for entirely different purposes. A scientifically adequate analysis follows both cash flows through time and across states.
The object of study is the behavioral organization of insurance-related household resources. The evidence covers voluntary indemnity insurance, health-plan options, deductibles, extended warranties, long-term-care coverage, index insurance, and savings mechanisms that may substitute for some insured risks. These cases differ in how benefits are determined, what funds can legally be used, whether the consumer already has coverage, and which parties bear the remaining risk. Distinct products are analyzed under their own constraints instead of being treated as interchangeable.
Methodologically, the dossier reconstructs relevant theory, identifies primary empirical designs, and evaluates alternative causal accounts. Mathematical expressions are transparent analytical illustrations rather than original estimated models. All numerical examples are hypothetical unless tied expressly to a named study. Existing data are not re-estimated; the conclusions concern what the published designs identify and what remains to be tested. The principal analytical contribution is an integrated distinction among the cost of insuring a risk, the mental category used to pay for it, and the resources available when loss actually occurs.
Mental Financial Accounts: Intellectual History and Definitions
Mental accounting entered modern behavioral economics as a way to explain decisions that depend on labels, reference points, and the organization of transactions. Thaler (1985) examined consumer choice through these decision records in Mental Accounting and Consumer Choice; his 1999 synthesis distinguishes the evaluation of outcomes, assignment of activities to accounts, and the period over which accounts are balanced. The three components matter separately for insurance. A premium can be coded as an immediate loss or a security expenditure; it can be charged to a particular category; and it can be judged monthly or over an entire coverage year. Each operation can alter perceived value without changing the insurer's contractual promise.
A mental account is a cognitive classification used to allocate and evaluate money, spending, receipts, or anticipated obligations. A mental budget is a rule, implicit or explicit, limiting flows within such a classification. Earmarking assigns particular resources to particular purposes; narrow bracketing evaluates a decision within a limited subset of its full consequences. These ideas overlap but are analytically separable. The restriction can be psychologically maintained, socially enforced, or embedded in an actual account with contractual or statutory withdrawal limits. The economic consequences of those arrangements differ. Fungibility, Labels, and Consumption (Abeler and Marklein 2017) finds that labels influence incentivized decisions, but a labelled bank account with a legal withdrawal penalty is more than a label.
Heath and Soll (1996) provide a general mechanism: expenditures regarded as typical of a category can deplete its mental allotment and inhibit subsequent purchases in that category, even after controlling for income and satiation within their studies. An insurance premium may thus compete more directly with subscriptions, taxes, or other regular bills if classified as a household expense than with the amount reserved for home repair if it is classified as asset protection. Which category is selected, and whether its balance is treated as hard or adjustable, are questions for measurement.
Antonides, de Groot, and van Raaij (2011) studied a broad Dutch sample and found that mental budgeting and financial management were positively associated after adjustment for reported financial situation and other characteristics. Such associations do not imply that every restrictive budget is an error: categorization can aid monitoring and self-control, particularly under limited attention. The potential cost arises when a mental rule prevents transfers that the household could feasibly make and would prefer after considering all consequences.
The conceptual boundary is therefore precise. Mental budgeting describes how resources are represented and permitted to cross categories in decision making. Insurance budget decisions describe how those representations influence purchase, coverage level, renewal, deductibles, emergency reserves, and claims-related spending. The link must be established by observables or a credible intervention; a low rate of insurance take-up does not by itself establish the presence of mental accounting.
The Economic Benchmark: Risk Transfer, Affordability, and Fungibility
An insurance contract typically requires a certain premium in exchange for specified contingent payments. Let initial resources be W, the premium be P, and gross monetary loss in state s be L_s. If indemnity is I_s, end-of-period consumption is c_s = W − P − L_s + I_s, abstracting from timing, taxation, and other consumption. Under expected utility, a household compares the distribution of c_s across contracts, with state probabilities pi_s and a utility function u. In the simplest one-period benchmark, the chosen contract maximizes the sum of pi_s u(c_s) subject to available resources and contract constraints. Mossin's rational insurance-purchasing analysis (1968) provides the classic theoretical point of departure.
In this benchmark, a dollar held in one unrestricted cash account is financially equivalent to a dollar held in another unrestricted cash account. This fungibility property concerns the household's feasible set; it does not assert that all insurance products are equally priced or that cash and insurance always supply identical services. Insurance pools risk and can transfer purchasing power into otherwise disastrous states. A household cash reserve remains exposed to several simultaneous shocks and can be depleted before a particular loss. Conversely, insurance can exclude losses, impose deductibles, pay after proof and adjustment, or be unavailable at a required moment. Valid comparisons incorporate these contractual differences.
An apparent mental-account effect can instead reflect a genuine intertemporal resource constraint. A borrower facing high interest costs may rationally prefer a smaller deductible even if a high-deductible plan carries a lower annual expected cost. Someone with restricted savings cannot freely redirect them to premiums. A household expecting an immediate medical bill may value readily accessible cash more than the discounted expected value of a later indemnity. These cases involve shadow prices of liquidity, credit, legal restrictions, and timing, not simply cognitive compartmentalization.
A useful diagnostic is to hold wealth, liquidity, risk distribution, coverage, and administrative effort constant while varying only the designation or source of payments. If insurance choice changes under otherwise equivalent feasible resources, the response is evidence consistent with category-dependent valuation. If the funding instrument carries withdrawal limits, tax treatment, employer contributions, default features, or distinct transaction costs, the design must isolate those differences. The experimental label evidence of Abeler and Marklein (2017) is instructive because the essential comparison is between equal resources presented under different designations.
This analysis also prevents a common normative shortcut. A household that chooses no insurance may rationally self-insure, rely on public protection, prefer prevention, or find the offered policy poor value. Mental accounting is a candidate explanatory mechanism when observed classification changes the assessment of the same risk transfer. Welfare analysis asks whether the resulting choice improves or worsens feasible lifetime consumption and protection, not whether purchase rates rise.
How Mental Accounts Organize Insurance Spending
Four recurring categories make the household problem concrete. A premium account represents planned periodic payments for coverage; an emergency account represents resources available for unanticipated shocks; a savings account preserves future goals; and a day-to-day account finances current consumption. The categories may be embodied in actual deposits or operate only as planning rules. A policy premium can be perceived as reducing today's disposable income while the emergency fund is considered unavailable for ordinary transactions. The same household can simultaneously possess adequate aggregate liquid assets and experience an apparent premium shortfall within its own budget labels.
A descriptive representation assigns a planned total B across premium, emergency, long-term, and current-spending categories. If the household informally imposes a ceiling B_p on premiums, a policy priced slightly above B_p may be rejected even if B_e or B_s could be reduced at low cost. This is an interpretation of account-bound choice, not a newly introduced theory or a structural estimate. The critical distinction is the status of the constraint: a personally chosen category threshold is behaviorally significant, while a regulation blocking withdrawals creates an objective barrier.
The categories also differ in moral and social meaning. A reserve may be treated as belonging to children, retirement, education, or future medical care even when it is liquid. Family negotiation can determine whose expenses may draw on which account. Such rules can operate as commitments and protect vulnerable members from spending pressure. A quantitative analysis that counts all balances as entirely free cash may thus overstate effective discretionary resources, while an analysis that takes every self-reported category as physically inaccessible may overstate credit constraints. A careful household survey measures both formal restrictions and perceived appropriateness of transfers.
Time aggregation further changes decisions. A $1,200 annual premium framed as a $100 monthly obligation may fit a mental monthly budget, while an up-front $1,200 payment can trigger rejection despite the same arithmetic total. This inference follows from Thaler's choice-bracketing framework and demands an experiment separating frequency, finance charges, and cancellation terms. Monthly payment may be superior for someone facing cash-flow volatility even without mental accounting. Empirical work must distinguish the actual value of smoothing from a mere re-description of an unchanged present value.
An important implication is that insurer statements should disclose both total annual cost and timing of payments, and comparison tools should display prospective premium, deductible, maximum out-of-pocket responsibility, and exclusions together. Such tools can let a household use category budgeting for planning while still examining the whole risk-transfer transaction. Their effectiveness remains an empirical question, since presenting more information can itself create decision burden.
Liquidity, Credit, Payment Timing, and Institutional Constraints
Liquidity is a central rival explanation of mental-account findings. It describes the ability to meet obligations as they arrive, while financial wealth aggregates assets and liabilities at different maturities. Premium obligations often precede a loss; deductibles coincide with care or repair; claim payments may arrive after documentation and adjustment. Financing an identical eventual amount at a different time can have a genuine economic value. Subsidies, tax advantages, employer deposits, or administrative rules may also make one funding channel cheaper than another.
Giné, Townsend, and Vickery (2008) found in their rainfall insurance participation study that purchase in rural India decreased with basis risk and credit constraints and increased with wealth. Familiarity with the insurance provider and social connections also mattered. That combination demonstrates why an apparent category effect inferred from survey answers would be confounded if poorer or less connected households are simultaneously subject to illiquidity, lack of trust, and contractual basis risk. The same principle applies to health-plan choice when medical savings are subject to statutory withdrawal and contribution rules.
Field experiments by Cole and colleagues (2013) examined barriers to household risk management in India and documented strong price responsiveness while revealing important nonprice barriers to weather-insurance take-up. A premium budget may be binding because the price is actually high relative to earnings; a low-cost intervention may fail because the product itself is poorly understood or weakly trusted. These are substantively distinct causal explanations. It would be inappropriate to attribute all remaining unexplained nonparticipation to mental accounting.
Payment frequency creates an additional design choice. Spreading an annual premium across months can ease constraints but may incur finance charges or alter lapse risk. A rigorous randomized experiment should report total present-value premium, fees, grace periods, defaults, and administrative ease, holding these constant where possible. An apparent preference for 'small monthly' payments could be temporal framing, valid cash-flow management, or both. Cross-country comparison additionally requires treating mandatory payroll-funded benefits and voluntary household-paid coverage separately.
The practical analytical conclusion is that every claim about a premium-account boundary must document the feasibility of transfers between accounts. Researchers should observe borrowing rates and limits, liquidity buffers, source-of-funds eligibility, and intra-household commitments before describing a payment-source response as purely cognitive.
Deductibles, Small Losses, and the Insurance Price of Certainty
Deductible choice is a particularly informative test because consumers commonly select among policies differing in both premium and retained loss. With identical underlying loss risks, a lower deductible raises the premium while decreasing payment if a claim occurs. A simple comparison weighs the premium increment against the distribution of additional retained payments, risk aversion, liquidity needs, transaction costs, and insurer restrictions. The expected amount saved on a deductible is not automatically the expected economic value of the whole contract; consumer value includes state dependence and the ability to finance the worst outcome.
Sydnor (2010) examined observed homeowner deductible selections in overinsurance of modest risks and found many low-deductible choices with prices high relative to expected payouts. A conventional expected-utility fit can imply implausibly strong curvature over small stakes. This establishes a puzzle for a restricted benchmark. It does not isolate one behavioral cause. Probability weighting, a preference to avoid an occasional large bill, loss aversion, or a separate 'repairs' account may each contribute.
Barseghyan, Molinari, O'Donoghue, and Teitelbaum (2013) used insurance deductible choice data to estimate a model allowing probability distortions and found overweighting of small probabilities important to their results. That evidence is an essential competing explanation. A consumer who pays more to eliminate a small deductible might be treating the deductible as a salient unpredictable expense rather than obeying a rigid mental category. The empirical implication differs: varying the account label at fixed probabilities should shift behavior under mental budgeting, while altering perceived event likelihood may matter more under probability weighting.
The cross-context comparison by Barseghyan, Prince, and Teitelbaum (2011) finds that home and automobile deductible preferences do not fit stable risk-preference assumptions in their analyzed data. Such heterogeneity is compatible with mental categories but also with product knowledge, differences in liquidity when claims arise, and the emotional consequences of home versus automobile loss. An identification strategy should independently vary label, timing, and loss information rather than interpret context heterogeneity as a mechanism by itself.
Fels (2020) provides an explicit theoretical insurance link through the mental-accounting access motive. In that formal account, insurance may be valued partly because it ensures access to a remedy otherwise felt unaffordable within the relevant mental category. It can explain why certain warranty and deductible choices differ from standard models, including sensitivity to the form in which benefits are paid. This is a published theoretical mechanism, not evidence that a particular surveyed buyer consciously maintained a named account. Its empirical test requires measuring remedy affordability and varying whether insurance supplies cash reimbursement, repair, or direct service.
Indemnity, Reimbursement, and the Form of Insurance Benefits
Insurance benefits arrive in different forms: an insurer may reimburse documented expenses, pay a fixed benefit, repair property through a service network, settle directly with a medical provider, or transfer cash after verification. These methods can be actuarially similar yet differ materially in transaction cost, delay, service quality, and the household's ability to purchase the intended remedy. A descriptive mental-account mechanism arises when a payment labelled 'replace damaged property' is considered appropriate for that purpose while an unrestricted transfer is classified as general income or savings.
Fels's analysis of insurance access emphasizes that the value of partial coverage may depend on the method of delivering benefits. The access motive matters when a household cannot readily finance an entire remedy from the account it treats as usable. For example, a policy that arranges roof repair may be judged differently from an equivalent cash benefit because the former directly supplies the intended service and avoids bargaining or upfront payment. Testing this mechanism requires holding service quality and expected indemnity constant: direct repair could be objectively superior owing to contractor coordination, or inferior because it restricts consumer choice.
Extended warranties provide a boundary case. Their price can bundle risk transfer, service access, convenience, distribution costs, and perceived product reliability. Abito and Salant's extended-warranty market evidence (2019) attributes important demand and profitability patterns to consumers' misunderstanding of product failure probabilities and finds that probability information can change willingness to pay. Thus, a warranty's popularity cannot serve as direct proof of the mental-account access motive. Information, labels, sales context, and actual service restrictions must be separated.
Benefits denominated for a particular purpose may sometimes enhance confidence, while an unrestricted cash benefit can preserve flexibility during a crisis. Neither format is universally preferable. Comparison should ask whether the beneficiary may receive the service when it is needed, whether the policy covers the cause of loss, and whether the value of funds retained for other emergencies is counted in the household's assessment.
These distinctions also matter for index insurance, where a payout may depend on a rainfall or temperature measure rather than actual household loss. The absence of payout during an individual catastrophe is basis risk, not a mental budgeting failure. Actual contract design and basis risk must be evaluated before introducing behavioral explanations.
Measurement: Surveys, Experiments, and Administrative Data
Mental accounting is latent: the analyst does not directly observe the psychological label attached to each monetary unit. Its measurement requires more than asking whether respondents 'budget.' Instruments may elicit explicit categories, earmarking practices, account transfer rules, perceived appropriateness of withdrawals, and frequency of expense review. A survey can then contrast these measures with observed coverage choices while controlling for financial knowledge, cash flow, household characteristics, and product information. The Dutch research by Antonides and colleagues (2011) supplies a household budgeting survey as a relevant methodological example, while its observational correlations require causal restraint.
In a controlled experiment, researchers can randomly label otherwise identical funds as 'insurance,' 'emergency,' or 'general budget' and observe demand for an identical contract. The budget label must not alter real rights or taxes, or the treatment would change more than perception. A second randomization can vary whether the premium is paid monthly or annually with present value held fixed. A third can alter whether the claim is reimbursed as unrestricted cash or as a dedicated repair payment, with the equivalent service package fully described. An interaction among labels, timing, and benefit form would identify differential sensitivity, not automatically a universal behavioral law.
Laboratory studies provide control over probability descriptions and resource labels. Heath and Soll (1996) show how consumption-category budgeting can be tested against rival income and satiation accounts. Abeler and Marklein (2017) use incentivized decisions to isolate labels and fungibility. Insurance-specific experiments must additionally specify the loss probability, deductible schedule, payment rules, exclusions, and claims process. Hypothetical scenarios may capture framing responses, but real-stakes choices strengthen inference about actual payment behavior.
Administrative data improve measurement of premiums, plan selections, income contributions, and claims, while often lacking the household's stated categories. Field experiments or policy discontinuities can create exogenous variation in the permitted source of premium payments. These designs are strong only when they carefully assess concurrent changes to liquidity, administrative convenience, and eligibility. The Singapore MediSave study by Gruber and colleagues (2026) is therefore highly informative, but its particular combination of savings and insurance institutions must remain explicit.
A complete measurement record should distinguish stated willingness to pay, observed application, actual enrollment, coverage continuation, claim behavior, and post-loss welfare. These are different outcomes. A survey respondent may endorse a policy yet lack enough cash for the initial premium; an enrolled consumer may lapse when a different household obligation arises. Accordingly, experiments should pre-register the outcome of interest and report attrition, take-up denominators, balance tests, confidence intervals, multiple comparisons, and implementation details. Cross-country portability is an additional test rather than an assumed property.
Insurance-Specific Evidence across Product Lines
Evidence from actual insurance contracts is heterogeneous. The home-insurance deductible observations in Sydnor (2010) demonstrate a high willingness to avoid certain small retained losses. The auto–home comparison in Barseghyan, Prince, and Teitelbaum (2011) shows context-dependent risk choice. The structural analysis in Barseghyan and colleagues (2013) points to probability weighting. Together these studies narrow the explanatory space: mental accounting is plausible but cannot be identified merely by the prevalence of low deductibles. These alternative explanations must be treated as competing hypotheses.
For health insurance, Handel and Kolstad's (2015) administrative and survey analysis identifies information frictions and hassle costs as important predictors of plan choice and welfare estimates. Health-plan comparisons can involve uncertain use, nontransparent provider networks, and complicated cost sharing. A premium can feel expensive because the person lacks information about expected out-of-pocket spending, independently of any mental category. Ericson and Sydnor (2017) likewise explain the difficulty of choosing coverage levels under confusion and selection. Better integrated cost displays may change choices through understanding rather than source-of-funds relabelling.
For long-term care, Gottlieb and Mitchell's (2020) narrow-framing insurance analysis links survey-based framing measures to a lower probability of purchase in the Health and Retirement Study setting. The mechanism is related to isolated evaluation of premiums and benefits rather than an experimentally manipulated bank account. Narrow framing can make a premium appear as certain expenditure judged separately from future care protection. The finding underscores the need to measure the particular cognitive operation instead of treating all nonparticipation as a generic bias.
For extended warranties, the product misperception study by Abito and Salant (2019) demonstrates a direct role for mistaken failure beliefs, while Fels (2020) supplies a theoretical account of access through insurance. One product category can therefore support multiple mechanisms. Disentangling them may require randomizing probability information, earmarking labels, and the form of the benefit separately. The relative explanatory power can vary across consumers and products.
For weather insurance, the 2008 Indian participation research and 2013 randomized India interventions emphasize basis risk, affordability, and trust. These studies are highly relevant to household budgeting but do not isolate a specifically mental premium account. Cross-line synthesis consequently supports a structured conclusion: labels, price, liquidity, complexity, trust, and probability beliefs interact, and evidence for one does not eliminate the others.
Rural Risk Protection and the Emergency-Savings Interface
The rural India and Kenya studies show why the same research question has different institutional meanings. Kenyan experiments on health savings test savings technologies and earmarking, including trade-offs between commitment and liquidity. Indian rainfall insurance research tests purchase of a contingent contractual payout in agricultural communities. The comparison is useful precisely because the interventions are not equivalent. A household savings balance finances its own expenditures; an index contract transfers specified weather risk according to a trigger that can diverge from actual damage.
Dupas and Robinson (2013) report that a simple safe savings place can facilitate health-related investment and improve vulnerability outcomes within their field setting. Earmarking for emergencies proved more useful than tying funds to certain preventive expenditures. That result suggests account designation may guide attention and self-control while rigidity may be costly when urgent needs are uncertain. It supplies evidence on household financial design, not a universal prediction that earmarked insurance-premium funds increase insurance adoption.
Giné, Townsend, and Vickery (2008) report that wealth, basis risk, and constraints help explain agricultural insurance participation. Cole and colleagues (2013) then use randomized price and nonprice interventions in rural rainfall-insurance markets. The apparent gap between exposure to risk and willingness to insure arises in a setting where contractual basis risk and unfamiliar products can make refusal economically comprehensible. Treating these constraints as mere psychological budgeting would mischaracterize the empirical evidence.
Cross-context inference can nevertheless be productive. A study could first elicit whether households classify crop insurance as an input cost, a household emergency expense, or a financial investment. It could then test whether identical offers with random but truthful budget framing affect purchase after fully measuring trust, knowledge, price, and liquidity. Such an experiment would evaluate transferability of the mental-account mechanism rather than assume it from the existing Indian findings. It should also measure actual renewals and claims because a one-time voucher response does not establish durable risk protection.
The broad lesson from these studies is methodological discipline: savings interventions and insurance interventions reveal related household allocation choices, but their causal parameters are tied to the exact device, product, and population. They become part of a comparative account of insurance budgeting only when the differences are stated.
Evidence from General Household Finance and Categorical Transfers
General household finance supplies evidence that categorization changes financial behavior outside insurance. Abeler and Marklein (2017) show in their incentivized labels experiment that otherwise fungible money may be spent differently when the budget source carries a label. Hastings and Shapiro (2018) exploit detailed retail transactions and institutional variation to analyze SNAP benefit spending, reporting that food purchases respond differently to SNAP resources than to cash even when nominal fungibility should render the source irrelevant. These studies show that money designation can matter, while their treatment of food consumption cannot be imported as a measured insurance premium effect.
The Dutch household survey of Antonides and colleagues (2011) finds mental budgeting widespread and related to expense monitoring. This finding also challenges a uniformly negative interpretation of mental accounting. Mental categories can reduce complexity and improve household control; the relevant welfare question is whether boundaries remain adaptable when risk and needs change. Insurance demands long-term persistence, making a clearly labelled premium reserve potentially useful for keeping coverage active even if other rigid boundaries cause underinsurance.
Karlan, McConnell, Mullainathan, and Zinman (2016) document in field experiments on savings reminders that messages tied to goals and incentives can improve attainment of savings commitments. This is evidence for attention and reminders in a savings context. It does not show that insurance renewal reminders, category labels, and bank earmarking have identical effects; their application to premium persistence remains a distinct research hypothesis. Reminder salience may improve planning yet could become manipulative if the message obscures the cost of coverage or its exclusions.
The synthesis across general financial behavior and insurance creates an important measurement opportunity. Researchers can combine administrative expenditure records with category labels reported in surveys and vary the account from which insurance premiums are drawn. Where source-of-funds rules generate exogenous variation, they should test the consequences for actual contract choice. Where labels are randomized, they should measure whether participants perceive different rights to spend the same cash. The strongest evidence comes from changes in behavior under controlled availability of resources, not from retrospective self-characterizations alone.
These literatures provide a basis for interpreting budget decisions in a wider context of financial management. They do not establish a single globally invariant 'mental accounting effect' on insurance demand. Heterogeneity in account rigidity, financial autonomy, institutions, and policy quality remains an empirical object.
Identification, Rival Explanations, and Causal Limits
Identification begins by specifying the treatment: is it the label assigned to a budget, the source permitted to pay a premium, a shift in the real cash balance, or a message about insurance value? A label manipulation can be randomized independently of account restrictions in a laboratory. Institutional eligibility changes in the field often affect several dimensions simultaneously, so credible causal interpretation requires showing why changes unrelated to mental classification do not explain the outcome. The Singapore study confronts this issue by embedding and testing modeled hassle and liquidity alternatives (Gruber and colleagues (2026)). That strengthens the case for mental accounting within their design while leaving open questions about other mechanisms and external validity.
At least seven rival explanations are material. Conventional risk aversion values insurance because it reduces consumption volatility. Probability weighting alters perceived likelihoods, as examined in deductible-based risk-preference research. Loss aversion and reference dependence can change the evaluation of a premium versus a claim, as distinguished in prospect theory. Liquidity and credit limits change what can be financed when bills fall due. Information frictions and hassle costs change how well a person understands or completes a purchase (Handel and Kolstad (2015)). Product trust, basis risk, exclusions, and claims quality change the objective or perceived value of the offer. Social and intra-household rules can make even physically liquid funds unavailable for a particular purpose.
These mechanisms may coexist. A source-of-funds intervention may work partly through attention and partly through a true change in borrowing opportunity. A household may compartmentalize expenses precisely because its cash flow is volatile. The task is not to force a mutually exclusive classification but to estimate which additional explanatory content comes from mental-category rules after the feasible-set factors are accounted for.
Causal designs should test pretrends around policy changes, eligibility manipulation, discontinuities in age limits, sample composition, and differential attrition. Experiments should separate the right to withdraw funds from their descriptive label, report actual payouts rather than only intentions, and avoid inferring welfare from enrolment alone. Administrative records should include premiums, plan attributes, claims, and household liquidity where lawful and practicable. The measurement of mental budgets should be conducted before outcomes to limit ex post rationalization.
Selection further complicates comparison. People who choose a dedicated savings account may already be more organized or more risk averse; individuals receiving employer-paid coverage may differ in employment and income; survey respondents may remember recent catastrophes. Observed correlations between earmarking and purchase cannot resolve these differences without a defensible strategy. A well-specified null effect is also informative: it might indicate flexible budgeting, binding real constraints, or an offer whose objective value is low.
Insurance Welfare, Consumer Protection, and Distributional Effects
Welfare evaluation requires accounting for all relevant states of the world and all household costs. Choosing a more generous plan can reduce out-of-pocket volatility but can also raise premiums, alter incentives, or create less liquid resources. Choosing a high deductible can save premium money yet expose the household to bill shocks. Mental-account interventions may help households see these consequences together; their effects must be evaluated against informed preferences and realistic risks rather than against a predetermined ideal coverage level.
The Singapore evidence links payment-source design to modeled welfare and selection implications under its institutional assumptions (MediSave welfare analysis). That does not license automatic transfer of the estimated effect to other health systems. Separately, Ericson and Sydnor (2017) describe how health insurance choice complexity can limit the benefits of large product menus. A consumer-protection design can show premiums, deductible ranges, maximum liability, and expected payments under representative claims scenarios, while keeping the difference between guaranteed and estimated values explicit.
Distribution matters. Account labels may be protective for people facing pressure to spend earmarked money and restrictive for households with scarce cash. Mandatory earmarking can protect future health expenditure yet deprive a household of liquidity for food or housing. A premium-financing intervention can favor people already holding balances in eligible accounts and fail to reach those without them. Policies designed from average treatment effects should therefore report heterogeneous consequences by income, debt, age, household size, asset access, and health status without presuming a single optimal account architecture.
Transparency is essential. Insurance providers should not exploit mental category boundaries by describing a high-loaded warranty as 'only a few dollars' while hiding exclusions or deductibles. Evidence from extended-warranty misperceptions shows how distorted failure beliefs may affect demand; the implication is to disclose likelihoods and product coverage as well as headline premiums. Similarly, sales messages invoking an 'emergency fund' should preserve consumer understanding that reserves and insurance have different payout structures.
One scientifically defensible design principle for research and practice is to evaluate insurance options as integrated household cash-flow profiles across time and risk states, while making it straightforward to maintain familiar budgeting labels. This is a comparative editorial recommendation derived from existing evidence, not a named new doctrine. It favors factual disclosures, realistic payment schedules, and independent validation of behavioral effects.
Comparative Interpretation Across Insurance Systems
Different insurance systems assign premium payment and residual loss to different institutions. In voluntary property insurance, the household usually pays a premium from its own budget and receives indemnity after a covered event. Employer-sponsored health plans may obscure the full cost because contributions are deducted from pay and plan menus contain numerous contingent benefits. Public insurance may rely on taxes or payroll contributions and provide entitlements independent of individual purchase choice. Agricultural index policies pay according to specified measures, creating basis risk. These structural distinctions matter to mental-account interpretation.
Singapore's MediSave rules supply a concrete case in which the right to use designated medical savings for health premiums is observable and variable; the observed purchase response is specific to that environment (Singapore insurance-financing evidence). US deductible research supplies a separate setting in which a premium-versus-retained-loss trade-off can be directly examined (home-insurance deductible evidence). Indian agricultural studies show price sensitivity alongside basis risk, familiarity, and credit barriers (India risk-management experiments). Kenyan health saving experiments show the role of secure storage and differentiated earmarking under resource constraints (Kenya savings evidence).
A cross-national comparison cannot validly rank populations by the size of a single mental-account bias without harmonized instruments. Survey wordings, legal restrictions on savings, medical payment obligations, risk profiles, social coverage, and access to credit must be aligned or explicitly adjusted. The meaning of an 'emergency fund' varies where public health coverage absorbs major expenses or where catastrophe support arrives through informal networks. A typical annual premium also has a different opportunity cost at different income and inflation levels.
Temporal effects deserve separate study. Insurance bought annually may be reviewed in a single narrow transaction; monthly payments may be evaluated in a recurrent account. Losses remembered after a disaster may temporarily change perceived probabilities. Long-term-care insurance demands the evaluation of distant care needs in the face of current premium expenditure, as illustrated by narrow framing in later-life coverage. Cross-sectional snapshots can therefore conceal changing labels, resources, and perceptions over the life course.
The generalizable research claim is limited but useful: account designation is a plausible determinant of the subjective price and perceived accessibility of insurance, and its effect is conditional on institutions and opportunity sets. Direct causal evidence is most compelling where the designation changes while competing determinants are measured or controlled. International scientific coverage requires reproducing that logic in each institutional setting.
Analytical Synthesis and a Testable Research Agenda
The evidence supports an ordered synthesis. First, mental financial categories are a documented feature of consumer choice and financial management, supported by experimental and observational findings (budget-category experiments; household finance survey). Second, insurance presents a particularly consequential application because premiums, retained losses, reimbursements, and alternative reserves occur in different states and periods. Third, direct research on health insurance financing finds that eligible payment from a designated health savings account changes plan choice in Singapore (2026 insurance funding study). Fourth, a large adjacent insurance literature shows that probability weighting, liquidity, information, and basis risk are substantial competing influences. The evidence therefore favors a conditional account, not a blanket diagnosis of underinsurance or overinsurance.
Three empirical distinctions organize future work. The first is accounting label versus legal restriction: researchers should vary descriptions and withdrawal permissions independently. The second is premium salience versus total-contract valuation: participants should evaluate annual premiums, retained losses, and coverage across matched representations. The third is access versus expected-value transfer: cash benefits, service benefits, and post-loss reimbursement should be compared with the same underlying remedy and realistic delivery frictions. These are research-design distinctions, not a proposed authorial theoretical system.
A pre-registered multicountry experiment could randomize the premium's stated funding account, the label of an equally liquid emergency reserve, and the framing of the benefit. Participants would face genuine, transparently priced small-stakes insurance choices or incentives tied to real decisions, with financially identical conditions where ethically and legally feasible. Primary endpoints would be actual enrollment, selected deductible, subsequent renewal, and ex post realized spending. A preregistered heterogeneity analysis would test whether effects differ with liquidity, budgeting rigidity, financial literacy, prior claims, and compulsory coverage. The design would report null findings and effect sizes rather than prioritize statistical significance alone.
For field identification, administrative variation in funding eligibility can be combined with survey measures collected before plan selection, while disentangling actual funds from labels. The Gruber et al. setting offers a published example of institutional variation; future replications should scrutinize eligibility rules, simultaneous reforms, and selection into accounts. Investigators should publish code and sufficiently documented data where law and privacy permit, verify key tables independently, and preserve a source-to-claim audit trail.
The welfare agenda is equally important. A premium-funding change may increase enrollment without improving household outcomes if it directs resources from urgent needs into poor-value insurance. Conversely, a reduction in a rigid premium-account barrier may allow valuable catastrophic protection. The right outcome measure depends on price, contract quality, risk exposure, resource accessibility, and informed preferences. The research program should therefore combine behavioral measures with actuarial and household-finance outcomes rather than treat any single choice as the benchmark of success.
Conclusion
Mental accounting is a documented mechanism by which people organize, evaluate, and restrict financial decisions through categories. In insurance, those categories influence the apparent affordability of premiums, the perceived availability of emergency reserves, the acceptability of deductibles, and the value attached to cash reimbursement or direct access to a remedy. The mechanism is most clearly identified when choices differ across financially equivalent funding descriptions or eligibility changes whose alternative effects are carefully addressed.
The evidence provides a coherent but bounded picture. Classical insurance theory establishes the importance of loss distributions, coverage, and risk aversion. Behavioral experiments establish that categories and labels can affect otherwise fungible spending. Insurance-specific studies show context-dependent deductible choice, competing probability distortions, information frictions, and narrow framing. Evidence from Singapore now connects the funding source of health insurance premiums directly to plan choice, while studies from India and Kenya demonstrate the material roles of affordability, basis risk, trust, and savings architecture.
The principal scholarly conclusion is that household insurance budgeting is jointly determined by objective financial constraints, institutional design, and the categories through which people interpret resources. A mental-account explanation becomes empirically persuasive only when it survives comparison with those competing mechanisms. The analytical contribution of this monograph is the integrated source-based interpretation of that interaction, its identification criteria, and its implications for evidence-based consumer analysis. No new theory, proprietary model, or universal causal coefficient is asserted.
The next research priority is to measure account boundaries before insurance decisions, experimentally vary labels separately from cash access, and follow actual renewal and loss outcomes. Coverage, reserves, and premium financing should be evaluated together through the household's feasible consumption across states and time. This produces a rigorous basis for comparing behavioral evidence internationally while preserving the distinct economic and institutional conditions of each insurance market.
References
Thaler, Richard H. 1985. “Mental Accounting and Consumer Choice.” Marketing Science 4 (3): 199–214. Primary publication.
Thaler, Richard H. 1999. “Mental Accounting Matters.” Journal of Behavioral Decision Making 12 (3): 183–206. Primary publication.
Heath, Chip, and Jack B. Soll. 1996. “Mental Budgeting and Consumer Decisions.” Journal of Consumer Research 23 (1): 40–52. Primary publication.
Antonides, Gerrit, I. Manon de Groot, and W. Fred van Raaij. 2011. “Mental Budgeting and the Management of Household Finance.” Journal of Economic Psychology 32 (4): 546–555. Primary publication.
Abeler, Johannes, and Felix Marklein. 2017. “Fungibility, Labels, and Consumption.” Journal of the European Economic Association 15 (1): 99–127. Primary publication.
Hastings, Justine, and Jesse M. Shapiro. 2018. “How Are SNAP Benefits Spent? Evidence from a Retail Panel.” American Economic Review 108 (12): 3493–3540. Primary publication.
Mossin, Jan. 1968. “Aspects of Rational Insurance Purchasing.” Journal of Political Economy 76 (4, Part 1): 553–568. Primary publication.
Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica 47 (2): 263–291. Primary publication.
Tversky, Amos, and Daniel Kahneman. 1981. “The Framing of Decisions and the Psychology of Choice.” Science 211 (4481): 453–458. Primary publication.
Johnson, Eric J., John Hershey, Jacqueline Meszaros, and Howard Kunreuther. 1993. “Framing, Probability Distortions, and Insurance Decisions.” Journal of Risk and Uncertainty 7: 35–51. Primary publication.
Sydnor, Justin. 2010. “(Over)insuring Modest Risks.” American Economic Journal: Applied Economics 2 (4): 177–199. Primary publication.
Barseghyan, Levon, Jeffrey Prince, and Joshua C. Teitelbaum. 2011. “Are Risk Preferences Stable across Contexts? Evidence from Insurance Data.” American Economic Review 101 (2): 591–631. Primary publication.
Barseghyan, Levon, Francesca Molinari, Ted O’Donoghue, and Joshua C. Teitelbaum. 2013. “The Nature of Risk Preferences: Evidence from Insurance Choices.” American Economic Review 103 (6): 2499–2529. Primary publication.
Fels, Markus. 2020. “Mental Accounting, Access Motives, and Overinsurance.” Scandinavian Journal of Economics 122 (2): 675–701. Primary publication.
Dupas, Pascaline, and Jonathan Robinson. 2013. “Why Don’t the Poor Save More? Evidence from Health Savings Experiments.” American Economic Review 103 (4): 1138–1171. Primary publication.
Cole, Shawn, Xavier Giné, Jeremy Tobacman, Robert Townsend, Petia Topalova, and James Vickery. 2013. “Barriers to Household Risk Management: Evidence from India.” American Economic Journal: Applied Economics 5 (1): 104–135. Primary publication.
Giné, Xavier, Robert Townsend, and James Vickery. 2008. “Patterns of Rainfall Insurance Participation in Rural India.” World Bank Economic Review 22 (3): 539–566. Primary publication.
Handel, Benjamin R., and Jonathan T. Kolstad. 2015. “Health Insurance for ‘Humans’: Information Frictions, Plan Choice, and Consumer Welfare.” American Economic Review 105 (8): 2449–2500. Primary publication.
Ericson, Keith Marzilli, and Justin Sydnor. 2017. “The Questionable Value of Having a Choice of Levels of Health Insurance Coverage.” Journal of Economic Perspectives 31 (4): 51–72. Primary publication.
Gruber, Jonathan, Mengyun Lin, Haoming Liu, and Junjian Yi. 2026. “How You Pay Drives What You Choose: Mental Accounting in Health Insurance Plan Choice.” Review of Economic Studies, advance online publication, August 10. Primary publication.
Gottlieb, Daniel, and Olivia S. Mitchell. 2020. “Narrow Framing and Long-Term Care Insurance.” Journal of Risk and Insurance 87 (4): 861–893. Primary publication.
Karlan, Dean, Margaret McConnell, Sendhil Mullainathan, and Jonathan Zinman. 2016. “Getting to the Top of Mind: How Reminders Increase Saving.” Management Science 62 (12): 3393–3411. Primary publication.
Abito, Jose Miguel, and Yuval Salant. 2019. “The Effect of Product Misperception on Economic Outcomes: Evidence from the Extended Warranty Market.” Review of Economic Studies 86 (6): 2285–2318. Primary publication.