RESEARCH DOSSIER
Perceived Insurability and Protection Efficacy
Abstract. Perceived insurability concerns a prospective or current policyholder's beliefs about whether a specified loss can be insured and whether an existing contract would respond. Perceived protection efficacy concerns expectations that insurance will deliver economically meaningful protection when a relevant loss occurs. These judgments influence insurance purchase, the selection of coverage levels and deductibles, contract renewal, preventive behavior, and post-loss adjustment. Their empirical significance arises because the financial consequence of a loss depends on contractual scope, exclusions, claims conditions, financial strength, administration, and the timing and value of payment. This monograph reconstructs the conceptual foundations and principal mechanisms of those beliefs; examines evidence from randomized field experiments, administrative choice data, survey experiments, supervisory research, and structural modeling; and evaluates identification, measurement, and welfare inference. Its central analytical finding is that nominal insurance ownership, subjective confidence, objective legal coverage, and realized indemnification are distinct variables. Consumer comprehension can be inaccurate even after policy wording is supplied; insurer reliability beliefs can affect demand independently of price; and product design interacts with liquidity, trust, salience, and basis risk. Experimental findings from India, Ghana, and Malawi show heterogeneous demand and investment responses rather than a universal effect of more insurance. European natural-catastrophe supervision identifies coverage misunderstanding and expectations of public assistance as additional barriers. The resulting research agenda calls for contract-specific belief elicitation, independent assessment of legal eligibility and realized payment, longitudinal observation, and welfare analysis that accommodates protection shortfalls, distributional constraints, and changes in risk taking. The synthesis establishes a coherent empirical interpretation of the existing literature without claiming a new scientific theory.
Keywords: perceived insurability; protection efficacy; insurance beliefs; perceived coverage; claims expectations; insurance demand; contract comprehension; insurer trust; insurance illusion; basis risk; behavioral insurance; consumer protection; financial resilience; causal inference.
Introduction
Insurance is a contingent contractual promise whose usefulness depends on a future state of the world and the conduct of several institutions. A household may pay a premium for years before discovering whether a particular cause of loss qualifies for a benefit, whether the loss can be established with the required evidence, what deductible or sublimit applies, how the claim will be handled, and how promptly the resulting payment will arrive. Beliefs about those contingencies therefore form part of the economic substance of the decision to insure. The buyer must assess not only the probability and financial scale of damage but also the probability and value of the insurer's response.
The foundational behavioral-insurance literature documents both limited purchase of protection against catastrophic losses and purchases of policies with unattractive expected value under conventional benchmarks. Kunreuther, Pauly, and McMorrow's design principles for insurance place such choices within the interaction of buyers, insurers, institutions, and risk communication. That evidence motivates a close investigation of what a consumer believes is insurable and what a policy is expected to accomplish. A low purchase rate alone cannot establish irrationality; the price, household liquidity, exclusions, expected public aid, insurer credibility, and available alternatives may supply coherent reasons for nonpurchase.
The scientific object here has four components. First, perceived availability asks whether a specific peril and exposure can be insured at all, under conditions the consumer considers accessible. Second, perceived coverage asks whether an offered or owned policy encompasses the peril and relevant loss. Third, perceived claims performance asks whether a contractually eligible claim will actually be recognized, adjusted, and paid. Fourth, perceived efficacy evaluates the adequacy of the payment relative to economic needs, including timing, deductible, limits, and recovery costs. These are analytical distinctions for studying existing constructs, not a newly proposed psychological scale or proprietary model.
Perception is also temporally situated. A consumer can believe insurance will reimburse a flood before receiving a quote, form another belief when signing a policy, change that belief after reading exclusions, and update it again after a friend's claim. In an index-based product, the customer may understand the automatic trigger accurately yet distrust the index's correspondence to field-level damage. In a health plan, a household may understand a deductible while misjudging provider networks or total annual cost. In property insurance, the owner may accurately know a dwelling is insured while incorrectly assuming all weather-related damage is covered. The object of evaluation must always specify peril, policy version, property or person, geography, period, loss, and claim circumstance.
The evidence is drawn from original journal articles and their publisher or institutional records, accessible research manuscripts where available, and primary supervisory texts. Research designs are read at the level of their documented samples, treatments, and measured outcomes. Randomized purchase offers, observed administrative enrollment, survey responses, formal welfare simulations, and regulatory assessments answer different questions. This review treats them as complementary but noninterchangeable. It supplies no invented cross-country pooled effect, no reanalysis of restricted individual data, and no claim that an aggregate coverage figure directly measures an individual consumer's mistaken beliefs.
The economic importance extends beyond purchase. Protection confidence may influence leverage, capital investment, preventive expenditure, location decisions, contractual compliance, and willingness to rely on government recovery. Some changes increase productive risk taking because insured losses are less damaging; others shift expected cost to insurers, other insureds, or public finances. A rigorous account must evaluate the quality of protection as well as the behavioral response. The insurance-specific findings developed below therefore sit alongside the broader distinction in Rogers's protection motivation theory between threat appraisal and the perceived efficacy of a protective response. That theory originates in persuasion and prevention research; its terms provide an interpretive comparison, not direct experimental proof about insurance claims.
Perceived Scope of Coverage
Conceptual history and the object of belief
In a standard account of risk transfer, the insured trades a known premium for a contingent payment against specified future losses. The economic transaction only works as envisaged if the relevant events are covered and the insurer can perform. Technical insurability, which concerns the legal and economic feasibility of underwriting a risk, is consequently separate from subjective insurability: a customer's belief about whether coverage can be obtained or will respond. The distinction is particularly important when insurance is sold as a broad category—home, health, travel, cyber, or agricultural insurance—although the contract responds to narrowly specified causes, persons, places, and conditions.
A belief that a peril is insurable can be correct even when the currently owned policy excludes it: the consumer might find a separate endorsement or specialist contract. Conversely, a belief that an existing policy covers a peril can be false although a market for the peril exists. Perceived insurability should therefore not be coded as a single binary judgment inferred from ownership. Ownership establishes only the existence of a contractual relationship; a precise measure asks about a particular triggering circumstance.
For analytical purposes the following questions separate the relevant components: Can a policy be bought for this peril? Does this particular policy include this peril? What loss definition, excess, limit, waiting period, and geographic restrictions apply? What must the policyholder prove? What fraction of the economic loss would remain after indemnification? Each answer is potentially factual and can be assessed against policy terms or a specified hypothetical contract. The framework keeps the external benchmark explicit, avoiding the assumption that confidence implies accuracy.
Consumer misunderstanding is not confined to obscure words. An insurance policy may begin with an apparently wide coverage grant and later narrow its effect through definitions, exclusions, conditions, and endorsements. Schwarcz, Cude, Logue, and Marquez Alcala's homeowners insurance comprehension experiments compare respondents' general beliefs with judgments after exposure to policy excerpts. The authors report that excerpts sometimes modestly improved accuracy and sometimes reduced it. This directly challenges an intervention design that treats merely showing policy language as sufficient to establish informed protection beliefs. The work is a 2025 working paper, and its reported results should be interpreted within the survey-based scenarios and experimental materials studied.
Contract semantics and economic adequacy
Coverage is a legally and temporally qualified proposition. An occurrence provision may connect loss to the event date, a claims-made provision to a reporting date, and exclusions to particular causes of loss. Similar product labels can therefore have different consequences for an identical event. Deductibles and indemnity limits add another dimension: eligibility can be certain while the net benefit is small. In health insurance, network and service restrictions alter access and the financial incidence of care. In parametric insurance, a valid index trigger can result in payment without actual local damage, while genuine local damage can occur without the trigger. In each case the semantic object of belief differs.
Consider a household with an insured asset worth 40,000, a 30,000 and the household retains 15,000 sublimit, indemnification may fall to $15,000 depending on the contract's wording and order of operations. These are illustrative calculations, not descriptions of a universal loss-adjustment rule. A person answering yes to a question about “being insured” may hold either materially different contract and still provide the same response.
The proper empirical comparison is scenario-dependent. One can measure the subjective probability that a claim will be recognized, the subjective payment amount conditional on recognition, or the respondent's expected post-loss resources. Each conveys different information. An objective claims analysis additionally requires policy text, governing law, endorsements, facts of loss, and evidence. Legal entitlement is sometimes contested even with complete documentation, so a researcher should distinguish a clear policy benchmark from an adjudicated dispute rather than calling every contested claim an error.
The IAIS Insurance Core Principles describe fair treatment, clear and nonmisleading information, and timely claims and complaints handling under ICP 19. Those provisions connect subjective understanding with obligations in the insurance value chain, but a supervisory standard is not a measured estimate of consumer beliefs and does not make all national legal regimes identical. Product design and conduct regulation must still be interpreted within their jurisdictional implementation.
Perceived availability, affordability, and catastrophic protection
A peril can be commercially offered yet functionally inaccessible at a premium that a household cannot pay. It can also be contractually included in a basic package while the customer remains unaware of the inclusion. Empirical reports therefore should distinguish existence of an offer, price of an available offer, willingness to pay, understanding of the offer, and completed take-up. Without that separation, analyses can mistake inability to purchase for lack of perceived benefit or mistake insurance possession for effective recovery capacity.
European natural-catastrophe research illustrates these interactions. EIOPA's March 2026 discussion of climate insurance protection gaps describes perceived unaffordability, limited transparency about costs and coverage scope, expectations of state assistance, and the risk that consumers assume protection extends further than it does. The same account reports that 17 percent of respondents in EIOPA's 2025 Eurobarometer held coverage for property damage caused by natural catastrophes. The figure is a stated survey result for the studied population and coverage question, not an estimate of misperception or a global insurability rate. A gap in actual coverage can result from several mechanisms simultaneously.
The distinction between supply and belief matters for climate change. Rising hazard frequency, repair costs, reinsurance costs, and correlated accumulation can constrain feasible coverage and increase price independently of consumer psychology. Clearer information may improve selection among available products but cannot by itself make an unfinanceable risk sustainably insurable. Conversely, more capacity or a public backstop can expand formal supply while behavioral and informational frictions continue to suppress take-up. The strongest policy diagnosis tests both sides.
Identification and interpretive limits
Coverage beliefs are typically elicited through questionnaires, vignettes, or incentivized prediction tasks. Researchers should state which questions are factual, which ask confidence, and which ask intended behavior. Vignette wording can induce demand effects; familiarity with the insurance product and questionnaire translation can produce systematic group differences. If a survey asks “Is your home covered?” without specifying the peril, insurance line, and loss circumstances, the resulting percentage cannot validly be interpreted as comprehension of an exclusion.
The direct comparison between perceived and verified coverage is possible when researchers obtain the policy documents or construct an experimental contract whose terms are fixed and known. A robust design records the exact policy version, incorporates changes and renewals, and tests several substantively distinct events. Repeated observation measures learning and decay of understanding. Randomization of information can identify the effect of a disclosure format on beliefs or hypothetical choices; identifying subsequent claims and long-run welfare additionally requires administrative follow-up or a justified external model.
Coverage perception is therefore a testable contract-specific belief. Its scientific usefulness increases when the researcher can identify a factual benchmark and document the circumstances that make a policy response determinate.
Expectations of Claim Eligibility
The chain between event and payment
Insurance is often described as a claim against the insurer in a covered state. Actual claims performance contains a sequence of contingencies. The loss must occur during the relevant period; the contract must attach; the affected interest and peril must qualify; exclusions and conditions must permit recovery; documentation and notification must meet applicable requirements; and the carrier must determine and deliver payment, sometimes after appeal or adjudication. Expectations of protection should distinguish the probability of satisfying those conditions from the probability that an insurer will perform once entitlement exists.
One useful accounting expression for a specified event is the expected net payment E[Y | L] = P(E | L) × E[Y | E,L], where L denotes the specified loss event, E denotes claim eligibility under the policy, and Y is net realized payment. The identity simply conditions an expectation; it introduces no new law of insurance demand. Researchers may further decompose the second term by payment timeliness, recovery, and disputes if observed data permit. The formulation shows why “I expect my insurer to pay” can reflect high confidence in eligibility, high confidence in claim administration, or both.
Eligibility and payment expectations can move in opposite directions. A policyholder may understand that a wind peril is covered while believing claims will be delayed or reduced. Another may highly trust the insurer yet fail to notice an exclusion. In index insurance, formal eligibility is known from a measured index but the correspondence between trigger and individual loss is imperfect. Where respondents treat one of these features as the whole meaning of protection, ordinary preference and demand analyses may incorrectly attribute the effect to risk aversion.
Beliefs, contract evidence, and strategic conduct
Expected claims success can respond rationally to past experience, information about exclusions, litigation, the insurer's reputation, and claims service. It can also respond to salient but unrepresentative stories of denial or payment. Distinguishing these mechanisms requires more than a survey of stated trust. A respondent who believes a carrier will not pay may be expressing an actual concern about the policy's conditions, a perception of institutional dishonesty, uncertainty about proof, or an expectation of severe procedural cost. Each predicts different consequences from a proposed intervention.
A second analytical issue is selection into claims. Realized payment rates among filed claims are conditional on an individual deciding to submit. Small losses below a deductible, cases perceived as ineligible, and claims abandoned because of hassle are missing from that denominator. The simple fraction of claims paid cannot by itself measure the subjective probability that any loss will yield recovery. Similarly, an approval rate may mix different perils, values, and contracts. Reported claims experience must be conditioned on incident type, product, deductible, reporting behavior, and information availability.
Observed denial rates also do not automatically determine fairness. A properly excluded loss may be legitimately denied, while an eligible loss may be wrongfully refused. This distinction requires independent contract-based or adjudicative review. The relevant empirical outcomes include objective eligibility, reason for decision, amount allowed, days to settlement, appeal, and final financial recovery. Purely perceived fairness and objective legal compliance remain separately measurable.
Behavioral feedback from claims experience
Claims are feedback events capable of reshaping beliefs. When farmers receive an index-insurance payment, they may learn both that the product can deliver money and that an observed weather trigger matters. Those are distinct sources of information. Karlan, Osei, Osei-Akoto, and Udry's Ghana field experiments report greater agricultural investment under insured-risk interventions and stronger subsequent insurance demand after personal or network payouts. The findings are consistent with learning about payment credibility, changes in perceived risk, and other treatment channels. They do not isolate a universal causal effect of trust from every contemporaneous factor.
Learning can be beneficial yet incomplete. A paid claim under one peril does not guarantee payment under another exclusion; positive experience with one company does not establish the solvency of every insurer; and failure of an index payment in a locally damaging year can be explained by basis risk rather than breach of the promised trigger. Research that collapses these cases into a single “claims satisfaction” score sacrifices crucial information about what the consumer learned.
Claims communication thus has an informational dimension independent of the cash settlement. Explicit explanations of accepted and rejected losses, clear calculation of deductibles, and intelligible appeal procedures may improve calibration. Their effect on actual trust is an empirical question; better understood coverage can rationally decrease confidence when consumers discover exclusions they had previously overlooked.
Why belief calibration matters for demand
An insurance product's perceived value depends on expected consumption in covered and uncovered states, not merely on its actuarial premium. A consumer who expects zero payment because of distrust will place little value on even an objectively generous policy. Another who anticipates complete indemnification may overpay for a contract with material deductibles or may neglect supplementary protection. Correcting a belief can therefore raise or reduce demand. An evaluation that treats greater purchase as the unique measure of successful education risks rewarding mistaken acquisition.
Handel and Kolstad's information-frictions analysis combines administrative health-plan choices and claims with survey evidence about information and hassle costs. Their findings emphasize that inferred risk preferences are sensitive to information constraints and that counterfactual insurance allocation requires welfare analysis of the underlying mechanisms. This insight generalizes methodologically: a preference estimate recovered from observed enrollment is a joint product of contract features, beliefs, choice environment, and real constraints.
Eligibility expectations warrant study as a specific variable because they connect legally enforceable contract terms to human valuation. A complete evidence chain tracks those expectations through choices, actual contingencies, claim initiation, and net realized protection.
Perceived Insurer Credibility
Credibility as a component of expected protection
An insurance contract turns present expenditure into a claim on future institutional performance. A policyholder needs confidence that policy administration will remain accessible, claim decisions will follow contractual obligations, disputes can be resolved, and valid payments can be financed. Such expectations are related to interpersonal trust and general confidence in financial institutions, but they are ultimately specific to products and the consequences of loss. Institutional confidence is therefore an input into expected protection, not a substitute for verification of the terms.
Research on life-cycle insurance choices provides unusually direct evidence of this distinction. Briggs, Rogers, and Tonetti's 2026 study of risky insurance uses newly elicited nonpayment beliefs and a structural life-cycle model covering savings, life insurance, annuities, and long-term care insurance. The authors report that survey respondents perceive material nonpayment risk and that these beliefs predict ownership. Their counterfactual simulations suggest economically important changes in participation and household allocations if perceived nonpayment risk is eliminated. Those quantitative welfare estimates depend on the study's model, calibration, and surveyed population; they are not directly observed treatment effects from eliminating nonpayment risk in an insurance market.
The interpretive importance lies in separating what households think the insurer will do from what a researcher assumes a contract will do. A model that represents all marketed policies as perfectly credible can attribute low take-up to a preference for bearing risk, even though the household anticipates nonpayment or substantial hassle. A model that allows subjective credibility can reinterpret at least part of that choice as a response to perceived quality. The empirical task is to establish whether the belief is based on verifiable experience, mistaken generalization, adverse selection across products, or other determinants.
Credibility has more than one object. Solvency confidence concerns the institution's capacity to meet obligations, conduct confidence concerns its willingness to handle claims appropriately, and process confidence concerns the ability to complete the required steps. Legal and supervisory protections change the environment in which each belief forms. A solvent insurer can administer claims poorly; a considerate claims team cannot guarantee payment for an excluded loss. Consumer confidence should therefore be elicited using separate scenarios for eligibility, administrative performance, and financial capacity. A general item such as “Do you trust insurers?” provides useful social information but cannot identify the relevant channel.
Experimental evidence on trust and demand
Cole and colleagues' randomized rainfall insurance field experiments in rural India studied the purchase of a newly offered product under price and nonprice treatments. Price sensitivity was substantial, but the published analysis identifies low trust, liquidity constraints, and limited salience as additional barriers. The study's strength is its field setting and randomized interventions; the interpretation should remain specific to rainfall index insurance and the participating rural communities. Its findings cannot be assumed to quantify credibility effects for motor insurance, commercial liability, or retirement annuities.
Trust treatments in field experiments also raise an identification issue. An intervention that increases uptake after a trusted individual explains a policy might alter perceived insurer honesty, provide technical information, make the risk more salient, reduce the social cost of purchasing, or simplify the transaction. Even when purchase is randomized or treatment assignment is random, the mechanism behind treatment effects requires separate variation, mediation assumptions, or additional measurements. A causal effect of an intervention is not automatically a causal effect of the single psychological construct named by its designers.
The Ghana results reported by Karlan and colleagues further show that actual payout experience and knowledge of a neighbor's payout predict later demand. A functioning payment can convey credible evidence of performance and illustrate the contract's mechanics. It is also possible that payouts coincide with salient weather events, income changes, or renewed attention. The paper reports patterns consistent with several interpretations, and a careful synthesis preserves that multiplicity. The correct inference is that the observed functioning of a product is economically relevant to future choices; the share of that effect mediated by pure trust must be established separately.
The Malawi randomized experiment conducted by Giné and Yang offers a particularly useful counterexample to simplistic predictions. In their insurance-and-credit experiment, offering loans bundled with mandatory weather insurance reduced loan take-up relative to a comparison loan offer. The authors discuss an explanation involving the borrower's preexisting implicit insurance from limited liability in the credit contract. An additional insurance component can therefore be judged redundant or costly even when it reduces an objectively defined risk in one part of the transaction. Institutional credibility is only one determinant of demand, and the economic baseline includes other implicit protections.
Credibility as a dynamic belief
Insurance credibility can change through renewal cycles, claims experience, media coverage, financial stress, litigation, and intervention by supervisors. Researchers should distinguish a temporary public reaction to a highly salient claim denial from a lasting assessment of insurer conduct. A panel survey can examine within-person belief changes before and after new information. An event study based on public claims events must address concurrent changes in prices, availability, hazards, macroeconomic conditions, and local media attention. Treating all changes after an event as learning about claims quality would be a causal overreach.
Measurement should capture both probabilities and reasons. Respondents can estimate whether a specified valid claim will be paid within a stated period, what fraction of the eligible amount they expect to recover, and how likely they believe it is that they would need a dispute procedure. Calibration can then be compared with observed data if a suitable matched claim population exists. For long-duration life and retirement products, direct validation may require decades, and current ratings or past payment experience are incomplete proxies. These limits make careful source attribution especially important.
Supervisory texts offer institutional benchmarks. The IAIS conduct-of-business principle requires insurers and intermediaries to treat customers fairly through fulfillment of contract obligations, and its guidance addresses clarity, suitability, claims handling, complaints, and supervisory assessment of consumer outcomes. These requirements illuminate why reliable administration and intelligible communication belong in a protection efficacy study. They supply normative obligations rather than direct evidence that consumers believe insurers are reliable.
The value of credible protection
Credible insurance can transform productive choices by making financial outcomes more predictable. It can also change borrowing, precautionary savings, and dependence on public assistance. These adjustments may be beneficial if they reduce wasteful self-insurance and enable economically productive investments. They can create costs where the additional confidence arises from exaggerated promises, concealed exclusions, or misunderstanding of indemnity limits. A consumer-welfare test therefore needs an assessment of actual indemnification and downstream consumption alongside subjective confidence.
One implication follows from the joint consideration of contractual clarity and credibility. Improving comprehension can increase trust when the contract is genuinely reliable, but it can reduce trust when it reveals unexpected gaps. Likewise, publicity about prompt payouts can raise perceived credibility without changing contractual scope. Both developments should be recorded rather than evaluated solely through the direction of purchase. Effective disclosure seeks calibrated confidence: a belief appropriately responsive to the protection actually available.
Overconfidence and Misunderstanding
The insurance illusion
A common insurance-specific form of error is overestimating the coverage provided by an existing contract. It can arise from a broad product label, an unqualified sales description, failure to read exclusions, the difficulty of integrating multiple documents, or confusion between policy ownership and full indemnification. A related error is underestimating the amount retained through a deductible or sublimit. Such beliefs may create a perceived safety net that affects financial planning even if no claim is ever made.
EIOPA explicitly discusses this problem in its 2026 demand-side protection-gap assessment. It identifies the “insurance illusion” in which consumers assume they are protected against natural catastrophe losses despite consequential exclusions or gaps. This is a supervisory assessment of documented information and protection issues in Europe. The term is useful descriptively, but its prevalence should be measured through contract-specific survey evidence rather than inferred from the size of total uninsured catastrophe losses.
The 2025 homeowners insurance study by Schwarcz and colleagues makes the measurement problem concrete. Their policy-language experiments present policy excerpts and compare judgments about coverage. The finding that some excerpts worsen accuracy suggests that complexity can arise from how clauses interact, not solely from a lack of exposure to text. Broad grants and later restrictions are especially difficult to integrate. The intervention that works for a standalone disclosure statement may fail when the recipient must resolve a hierarchy of conditions and exclusions.
This observation has direct implications for scientific testing. If participants are asked only whether they have read the terms, the response cannot establish understanding. If they correctly answer a generic question about deductibles, that does not show they can apply an exclusion to a factual scenario. And if they correctly answer a scenario after a researcher highlights the critical exclusion, it does not show that they would locate it independently during an ordinary purchase. A comprehensive evaluation tests recognition, application, unaided retrieval, confidence, and subsequent behavior separately.
Ambiguity, complexity, and confidence
Uncertainty about insurance is multidimensional. Consumers may be unsure about event probabilities; uncertain about contractual meaning; uncertain about claim treatment; or unclear about the insurer's future condition. Those sources of uncertainty require different remedies. Hazard mapping helps the first problem. Plain-language, contract-specific illustrations address some of the second. Transparent claim standards and effective dispute channels address parts of the third. Prudential and conduct supervision address institutional performance. A general information leaflet cannot reasonably be expected to solve all four.
The distinction also matters for modeling. A person who chooses a higher premium may be avoiding coverage uncertainty rather than demonstrating high curvature of utility over final wealth. Handel and Kolstad's health insurance evidence demonstrates that information frictions and hassle costs influence revealed choices and the inference of risk preferences. If an analyst treats every purchase as a choice made with correct information, the resulting preference estimates can absorb differences in knowledge, search costs, and familiarity with plan rules.
Bhargava, Loewenstein, and Sydnor's dominated-plan study examines 23,894 employees' health plan choices at one US firm. The published results report that a majority selected financially dominated plans, with excess spending equivalent to 24 percent of chosen plan premiums. The authors' experimental follow-up points to deficits in health insurance understanding. The finding is strong evidence of misvaluation in that particular menu and setting. It should not be extrapolated into a statement that most consumers everywhere make dominated insurance choices, and it does not separately identify every component of perceived claims efficacy.
Choice errors are particularly consequential when consumers possess high confidence in a mistaken interpretation. A useful experiment therefore elicits not only an answer but the probability respondents assign to their answer being correct. A respondent who is uncertain can be targeted with clarifying information. A respondent who is highly confident may not search for any additional explanation. Yet confidence is an empirical variable, and differences by age, education, language, or experience need to be tested rather than presumed.
Overprotection, underprotection, and asymmetric errors
Perception errors do not always point in one direction. Some individuals incorrectly believe a peril is covered and consequently fail to buy supplemental protection. Others incorrectly assume a covered event is excluded and pay for overlapping products or remain unnecessarily anxious about the loss. A third group accurately understands that cover exists but mistakes a deductible for a prohibition on claiming. Analyses should distinguish false-positive beliefs, false-negative beliefs, and errors about payment magnitude.
The direction of a belief error changes the welfare implications of correction. Revealing an exclusion may reduce confidence yet improve financial planning and increase demand for suitable coverage. Revealing existing protection may reduce duplicate spending and improve overall welfare even if nominal purchases fall. Consumers facing liquidity constraints may remain uninsured after accurate information because the affordable set of choices has not changed. An information program should therefore be evaluated through calibrated beliefs, feasible decisions, and financial outcomes—not an unconditional preference for higher take-up.
The interpretation of overconfidence also depends on the loss benchmark. A household may reasonably expect small everyday damage to be funded through savings rather than insurance. An interviewer who equates declining an unattractive policy with a mistaken sense of security commits an identification error. Likewise, an affluent consumer's willingness to bear a deductible may reflect an optimal allocation of risk. The relevant question is whether the individual understands the risk retained and chooses knowingly under the available budget and alternatives.
A practical measurement sequence
A well-specified study can begin with a short neutral scenario that specifies the asset, loss cause, timing, policy premium, deductible, limit, and relevant endorsements. Before presenting the text, the researcher elicits the respondent's spontaneous coverage judgment, expected net payout, confidence, and intended purchase or prevention behavior. Participants then receive one of several randomly assigned explanations: the complete ordinary policy, a standardized summary, a scenario-based coverage table, or a carefully designed counterfactual. The study repeats the judgment, tests recall after a delay, and records choice where possible.
Contract specialists should construct and independently check the answer key. The grading rubric must account for conditions such as sublimits, additional deductibles, and plausible disputes. If the scenario leaves a legal ambiguity, forcing a binary correct answer would generate false precision. For international comparisons, instruments require validated translation, local contract adaptation, and testing of measurement invariance. Results should be reported by policy type and scenario, not as an undifferentiated global “insurance literacy” score.
A further design improvement separates comprehension from persuasiveness. A clear disclosure might accurately lower the stated willingness to buy an inadequate policy; that response is evidence of improved selection rather than failed communication. Retention of understanding, correct valuation of alternatives, and realized coverage at renewal are the more meaningful endpoints.
Protection Efficacy and Behavior
From expected payment to economic action
The perceived efficacy of insurance is the expected contribution of a policy to preserving resources and sustaining recovery under a specified loss. The concept requires attention to probability of payment, its magnitude, timing, relation to total harm, and the household's alternative means of coping. A fully credible payment of a small amount can have limited recovery value; a large nominal entitlement delivered after the crucial rebuilding period may preserve wealth but fail to prevent liquidity distress. The relevant unit of analysis is therefore the anticipated and realized financial consequence of the insured event.
Protection motivation research offers a historically important distinction between perceived threat and expected benefit of an action. Rogers's 1975 theory of protective responses and the subsequent self-efficacy extension by Maddux and Rogers (1983) examined the effectiveness of preventive responses and perceived ability to enact them. In insurance markets, a similar separation is analytically useful: perceived probability or severity of harm differs from perceived ability to obtain coverage, and both differ from perceived efficacy of a purchased policy. These are analogical applications of established behavioral constructs. The original experiments concerned protective behavior and persuasion rather than policy eligibility or insurance claims.
Where insurance is available, greater expected efficacy can raise willingness to pay. It can also reduce precautionary saving and allow a household or enterprise to undertake activities with higher expected returns and higher gross risk. The direction of these responses depends on what the insurance replaces. If the buyer already has a loan with limited liability, a substitute public guarantee, an employer benefit, or accessible emergency savings, the incremental value of a new policy can be small. The Malawi experiment reported by Giné and Yang demonstrates why a contract's value should be judged relative to the preexisting risk-sharing arrangement rather than an imaginary state of entirely uninsured exposure.
An insured household's behavioral change cannot be identified from two cross-sectional variables alone. A positive correlation between high protection confidence and risky investment may reflect the effect of anticipated insurance, but it may also arise because wealthier and better-informed households both understand insurance more deeply and can afford more investment. Causal inference requires variation in insurance access, contract reliability information, payouts, or premiums that is plausibly independent of the underlying risk-taking propensity. Random assignment is particularly informative when the outcome is observed after assignment and attrition is transparently measured.
Agricultural investment and the value of protection
Karlan and colleagues' randomized Ghana insurance interventions furnish an instructive empirical case. Farmers offered weather-related risk protection increased agricultural investment and selected riskier productive activities. The analysis suggests that uninsured risk constrains productive choices, with subsequent insurance demand affected by individual and socially observed payments. These results establish that a functioning protection mechanism can change production behavior. However, because the treatment includes contract access and financial consequences, it cannot automatically isolate each subjective belief or establish the same marginal response for all farms, countries, or perils.
An index policy also introduces a distinctive mismatch: payment depends on an observed indicator rather than a loss adjuster's determination of individual financial damage. The index can trigger without significant damage, or fail to trigger when an individual suffers serious harm. An observed reduction in perceived efficacy may therefore reflect rational expectations about basis risk rather than misunderstanding or generalized distrust. This possibility is central to interpretation of the Indian rainfall product as well. Product education should explain both the trigger and its imperfect correspondence to the beneficiary's exposure.
Three outcome levels should be examined in such settings. At the contract level, test whether the trigger and payment function as promised. At the belief level, test whether the farmer expects the trigger to respond to the relevant local hazard and understands exclusions or payout thresholds. At the economic level, observe investment, consumption smoothing, indebtedness, crop choices, and recovery. These levels are connected but not identical. A contract can perform exactly as written while falling short of a producer's financial needs.
Credit, liquidity, and retained risk
The economic relationship between insurance and credit is not one-directional. Insurance can protect borrowers against events that would otherwise undermine repayment, thereby increasing demand for financing. At the same time, a loan already offers explicit or implicit risk-sharing through contractual recourse limitations, lender forbearance, or collateral. Giné and Yang's Malawi field evidence found lower loan take-up under a bundled-insurance offer, illustrating that a product may be perceived as adding costly protection to an already partly protected arrangement. The experiment is specific to the bundled terms and farm credit setting studied.
A useful contract comparison must therefore specify the borrower’s payoff in the same adverse state under every available arrangement. Insurance premiums reduce cash at the beginning of the period; payouts may protect repayment or recovery later; and loan covenants may shift loss between parties. The timing of these cash flows matters when liquidity is scarce. A household can recognize substantial future protection while declining coverage because the present premium competes with food, housing, or production inputs. In that case improved risk communication without affordability changes has a limited opportunity to alter take-up.
Risk transfer may also reshape preventive effort. Where an indemnity policy reduces the policyholder's marginal exposure to losses, some forms of prevention may become less privately attractive. Deductibles, exclusions, inspection requirements, experience rating, and contractual obligations can counteract such incentives. Yet the magnitude of prevention responses must be measured; the presence of insurance does not prove reduced safety behavior. Insurers sometimes reward verified mitigation or provide information that makes prevention more effective. Research must examine the actual institutional design and observed effort rather than treating moral hazard as an automatic result.
Catastrophe experience, salience, and memory
Perceived loss likelihood interacts with perceived protection efficacy. A person who knows a policy pays for flooding may still regard the premium as wasteful if severe flooding seems impossible. Another may be highly alarmed by a recent event but incorrectly believe an ordinary home policy will pay for it. A third may buy insurance immediately after a disaster and allow it to lapse as the event fades from memory. These patterns require separate measures of hazard beliefs and contractual beliefs.
Gallagher's US flood insurance take-up study uses a national panel of flood events and policies to document a sharp rise in purchase following floods followed by a return toward prior levels. The paper also identifies take-up effects in geographically unaffected communities sharing a television media market. Its analysis supports learning and forgetting as explanations for time-varying uptake. The study does not by itself establish that consumers have learned to interpret every flood exclusion correctly. It concerns changes in insurance purchasing in a particular institutional environment.
The distinction matters for communication programs. Hazard maps and disaster narratives can increase awareness of loss probabilities without clarifying insurance limitations. Contract-specific examples can improve understanding without increasing perceived hazard. The strongest interventions identify the principal missing link and evaluate the predicted intermediate outcome. Repeated, dated measures of perceived probability, expected claim eligibility, intended recovery spending, and actual insurance holdings are needed to distinguish persistent learning from transient salience.
Insurance and the distribution of post-loss resources
Realized indemnification can have uses that are not obvious from the policy's initial marketing. Gallagher and Hartley's Katrina household-finance analysis documents substantial mortgage paydown using flood insurance proceeds among affected owners in the circumstances they studied. A payment that relieves debt may improve the household's balance sheet while leaving substantial reconstruction needs. Accordingly, judging protection efficacy solely by whether money arrived can ignore whether the payment restored housing services, secured business continuity, or met medical needs.
Recovery is also affected by the claims timetable. Households with little liquid wealth may need emergency credit before indemnification is available. High deductibles and temporary exclusion of some costs can create a gap between the monetary value of an insurance policy and its immediate ability to prevent hardship. Conversely, well-timed partial payments can have high practical value where an early cash constraint is binding. These distinctions suggest measuring consumption, borrowing, housing or production continuity, and debt as separate outcomes instead of constructing an unexamined single “recovery success” indicator.
Public disaster assistance adds further complexity. If households expect compensation after severe shocks, their expected private protection includes a state-contingent transfer even where no private policy exists. EIOPA's 2024 demand-side staff paper identifies barriers and tested measures related to natural-catastrophe uptake. Its 2026 assessment of protection gaps includes expectations of government assistance among the reported demand-side factors. A belief that the state will help may be grounded in actual policy or historic relief practice. It becomes potentially misleading when the conditions, size, speed, or reliability of relief are misjudged.
Interpreting purchase and nonpurchase
Observed purchase is an action under a budget and a particular opportunity set. It is not a direct psychometric measure of belief, and nonpurchase is not proof of inadequate risk awareness. Buyers face premium loadings, coverage limits, liquidity needs, transaction costs, and the presence of other risk-sharing mechanisms. In some states private coverage cannot be obtained at a feasible price. In others the policy offers only partial risk transfer. A rigorous model of demand permits the consumer to value the expected policy payoff while declining it for coherent reasons.
A particularly important inference is that accurate knowledge can produce either more or less insurance demand. Information about broad coverage can increase willingness to purchase a useful contract; information about a restrictive exclusion can lower willingness to purchase an unsuitable one. The welfare objective is informed allocation of risk under real opportunities. Both effects are consistent with improved protection literacy.
Empirical Assessment and Welfare
Research strategy and evidence hierarchy
The central empirical problem is to link a statement about expected protection to a verifiable event, choice, or contract outcome. This requires the researcher to keep the source of variation and object of inference explicit. A randomized educational intervention identifies the effect of that intervention on the outcomes measured under the experimental conditions. An administrative comparison can document persistent choices and real financial outcomes but may require stronger assumptions for causal claims. Survey belief elicitation records expectations and stated confidence, not necessarily actual behavior. A structural life-cycle model can integrate many decisions and simulate counterfactuals, but its outputs remain conditional on the stated behavioral equations, identification strategy, and calibration.
Existing studies provide complementary evidence. Handel and Kolstad document the consequences of information and hassle costs for health plan choice and welfare; Bhargava, Loewenstein, and Sydnor document dominated plan choice within a specific menu; Cole and colleagues identify price and nonprice frictions in Indian rainfall insurance; Karlan and colleagues examine investment and learning from payouts in Ghana; Giné and Yang find a negative take-up response to bundled insurance in Malawi; Briggs and coauthors analyze beliefs about nonpayment in long-horizon products. Their populations, contract terms, estimands, and outcomes differ. The scientific conclusion is a set of mechanism-specific findings, not a single average treatment effect of “perceived efficacy.”
One can organize evidence by its relation to the actual causal process. Product terms and legal texts establish what the insurer promises. Documented offers show whether the product was available at a price. Contract comprehension tests reveal what respondents understand about the promise. Elicited probabilities reveal expected performance. Recorded take-up and renewal reveal decisions. Administrative claims establish realized performance for selected incidents and claimants. Financial outcomes show the extent of risk smoothing and recovery. Each step requires its own denominator and data-quality assessment.
Measurement of beliefs
A reproducible belief instrument identifies the reference event, respondent, coverage instrument, information set, and response scale. Examples include the subjective likelihood that a $20,000 loss caused by a named peril is contractually eligible, the expected dollar payout after a stated deductible, and the anticipated number of days to payment. A more general question about confidence in “insurance protection” is useful for comparisons of attitudes but weaker as an indicator of contract understanding.
Belief elicitation can be made concrete through proper scoring rules or other incentives, although incentives may introduce complexity and their effects on reported probabilities require separate evaluation. Enumerators should avoid supplying an answer through the form of the question. A respondent's numerical estimate and explanation can be collected independently so that researchers can tell whether a low payment expectation arises from an exclusion, an expected dispute, or concerns about solvency. Confidence in accuracy can then be elicited on a separate scale.
Where objective scoring is intended, expert adjudication of the scenario is essential. A “correct” response must be determined using the specific contract and jurisdiction, including definitions, conditions, coverage periods, and relevant endorsements. Ambiguous contract language should be documented and treated explicitly. Policy updates between baseline and follow-up make the benchmark time dependent. The analyst should predefine rules for classifying uncertainty, partial payments, and unresolved legal interpretations to avoid correcting the scoring standard after observing results.
Calibration, discrimination, and learning
Calibration asks whether predicted frequencies correspond to observed frequencies across comparable cases; discrimination asks whether beliefs distinguish higher- from lower-protection states. A population can have the correct average expected payment yet severe individual-level misunderstanding. Conversely, respondents can distinguish which events are more likely to be paid while systematically overestimating every payout. Average confidence alone identifies neither property.
For a fixed experimental policy, a straightforward test compares the distribution of stated payouts with the contractual payout schedule for a balanced set of incidents. For actual insurance books, researchers must address selective purchase and claim reporting before comparing beliefs with claim outcomes. A paid-claim rate among submitted claims is affected by the propensity to file, and file completion may depend on literacy and perceived success. Without observing losses that did not generate filed claims, naive calibration estimates can be misleading.
Learning studies should distinguish immediate comprehension, retention, and subsequent application. A policyholder can understand a rule during a survey yet forget it by the next renewal. Repeated observation also helps separate transient hazard salience from durable knowledge of contractual provisions. Gallagher's flood study shows the importance of the time dimension for insurance take-up; it should motivate belief-panel designs rather than be treated as direct proof about claims-coverage comprehension.
Interpreting consumer heterogeneity
Income, education, age, claims experience, occupation, prior financial shocks, language, location, and contract type may influence observed beliefs. Their relationships with insurance decisions can operate through affordability, opportunity, market access, information, and risk exposure as well as cognition. Reporting a group difference is therefore a descriptive starting point. Causal interpretation requires separating those channels where the data support it.
Kousky and Netusil's Portland flood literacy study reports that general insurance terms are comparatively familiar while detailed flood policy provisions are poorly understood by many surveyed residents. The authors also report that living in a mapped high-risk floodplain is associated with better knowledge that ordinary homeowners coverage excludes flooding, while detailed policy literacy remains limited. The findings apply to the communities and survey examined; they do not establish the same literacy distribution in other cities or national insurance systems.
A cross-country comparison must also account for contract differences. Public disaster compensation may be stronger in one jurisdiction; standard flood coverage may be included in one policy type and excluded in another; consumer redress and prudential guarantees can differ. An identical answer to “Does home insurance cover flood?” could be correct in one legal and product setting and incorrect in another. Cross-national measurement should therefore combine functional equivalence of scenarios with accurate local documentation instead of pretending that policy labels have identical meaning.
Welfare beyond simple take-up
The welfare value of better calibrated beliefs depends on actual attainable contracts, changes in premiums and deductibles, the allocation of retained risk, transaction costs, insurer behavior, and consequences for saving and investment. A higher purchase rate may reflect correction of mistaken pessimism, but it may also represent marketing-induced overconfidence. A lower purchase rate may reflect recognition of a poor contract rather than reduced welfare. Welfare analysis needs a benchmark of feasible decisions and outcomes rather than a normative assumption that more private insurance is always beneficial.
For an illustrative fixed contract, let gross loss be L, premium be π, net payment be Y, and post-loss resources before financing and transfers be W - L + Y - π. The relationship is an accounting identity, not a proposed model of individual psychology. To evaluate protection, one must examine how Y is determined and how the consumer's belief about Y differs from the contractually available and realized payment. Timing, discounting, taxes, credit access, and external support may need to be added to the relevant welfare calculation. Expected utility, distributional welfare weights, or consumption-smoothing measures can then be applied under their explicit assumptions.
The standard expected-utility comparison can be informative if beliefs and objective risks are separated carefully. But insurance-market welfare can also depend on insurers' administrative costs, risk selection, moral hazard, solvency capital, premiums required for sustainability, and the effects of mandatory participation. The social optimum need not coincide with the coverage level that an individual would choose when relying on inaccurate beliefs or bearing information costs. Neither the analyst's preferred coverage share nor a consumer's stated confidence is a sufficient welfare criterion by itself.
Structural simulations deserve special treatment. The 2026 Briggs–Rogers–Tonetti life-cycle insurance paper reports counterfactual effects of eliminating nonpayment risk on ownership, savings, and welfare under its calibrated assumptions. These are scientifically relevant demonstrations of potential magnitude in a specified model. They should be presented as simulations and compared with alternate parameterizations and credible empirical targets; they cannot be treated as measured consequences of an actual policy eliminating insurer nonpayment.
Disclosure quality and regulatory evaluation
Meaningful conduct regulation aims to make contractual offers and claims service comprehensible and fair. The IAIS ICP 19 standards and guidance provides an international benchmark for pre-sale information, appropriate treatment, and fulfillment of obligations. EIOPA's 2025 study of NatCat product information documents examines a sample of information disclosures and finds both useful practices and scope for clearer explanations. A supervisory finding that some disclosures are confusing is evidence about their presentation and potential for consumer misunderstanding, not a precise estimate of the causal effect of one new information format on purchasing or loss.
The central policy design principle is to test comprehension of salient contingencies instead of relying only on the presence of a standardized document. A scenario-based product explanation can display what is covered, what is excluded, how much remains after a deductible, and what evidence must be provided. Independent testing should establish whether readers can use the document to answer those questions. The outcome should be assessed across literacy levels and access conditions, including assistive technology and translations where relevant.
Information can also be made timely. Coverage illustrations at the point of sale, reminders before renewal, clarification after substantive endorsements, and accessible explanations during claim submission may each address a different belief. A single mandatory disclosure delivered at purchase may be forgotten by the time a rare loss materializes. Regulation of claim handling, complaints, and public communication can influence actual credibility as well as expectations; the two outcomes need separate monitoring.
There is a limit to disclosure as policy. When household incomes are inadequate, premiums are unaffordable, or carriers cannot sustainably underwrite escalating correlated hazards, clearer explanation alone cannot create feasible protection. The 2026 EIOPA protection-gap discussion explicitly links transparency with affordability, climate adaptation, and potential risk-sharing interventions. A credible policy response combines understandable contracts with genuine capacity, appropriate incentives, and transparent treatment of public liabilities.
Research synthesis and directions for verification
The strongest established conclusion is that subjective assessment of insurance protection has several empirically distinct sources and consequences. Comprehension experiments reveal errors about terms, market-choice studies document the effects of information frictions, field experiments reveal the importance of trust and liquidity, payout experience informs subsequent demand, and supervisory studies identify consumer-facing disclosure weaknesses. No single one of these sources measures every element of perceived insurability. The synthesis gains explanatory force by joining results at their proper levels of inference.
Several empirical priorities follow. One is to obtain linked microdata combining survey beliefs, exact policy terms, premium and underwriting offers, claim eligibility determinations, payment amounts, and household recovery. Another is to compare experimental disclosure designs on factual comprehension, choice quality, retention, and realized financial outcome rather than immediate stated intent alone. A third is to measure insurer credibility as contract-specific expectations, distinguishing claims administration, solvency, and uncertain eligibility. A fourth is to investigate how public assistance expectations interact with private coverage, risk mitigation, and distributive capacity under changing hazard conditions.
Future studies should preregister primary outcomes when feasible, document survey instruments and answer keys, test attrition and nonresponse, report uncertainty intervals and heterogeneity without selective emphasis, and share replication materials subject to privacy and contractual restrictions. International generalization requires locally accurate product and legal comparisons. Where only supervisory summaries or published abstracts are accessible, a review must describe that evidence level and avoid unsupported fine-grained estimates. Negative and mixed results, including the Malawi bundled-insurance response, are essential components of a balanced synthesis.
The principal interpretive contribution of this monograph is the disciplined separation of belief, entitlement, administration, and realized welfare across the insurance decision cycle. These categories are drawn from established contractual, behavioral, and empirical research. Their systematic application makes conflicting findings intelligible without multiplying theoretical labels or asserting an untested universal mechanism.
Conclusion
Perceived insurability and protection efficacy are central empirical questions in insurance economics because a policy's financial value is contingent on what is insured, the circumstances of eligibility, the conduct of the insurer, and the recovery ultimately achieved. A person's confidence in a policy, actual legal coverage, the handling of a filed claim, and realized indemnification can diverge. An account of consumer demand that measures only premium and purchase can consequently miss a substantial part of the transaction: what protection the buyer thinks is being purchased.
The research supports five connected conclusions. First, beliefs must be defined for specified contracts, events, and amounts. A generic question about whether one is insured cannot distinguish a coverage exclusion from a large deductible or a payment delay. Second, contract wording and the manner in which it is presented materially affect comprehension. The homeowners-policy experiments reviewed here show why simply supplying the relevant text may leave mistaken interpretations intact. Third, insurer credibility and payment expectations influence the value consumers assign to products, although experiments and observational data do not always isolate pure trust from information, salience, and transaction costs. Fourth, effective risk transfer can change productive investment, borrowing, and savings; the direction of demand responses depends on preexisting financial arrangements, affordability, and basis risk. Finally, the welfare consequences of more accurate beliefs can include either more or less insurance purchasing. The relevant outcome is the allocation of risk and the adequacy of financial protection under feasible contracts.
The cited evidence spans different empirical designs and populations. Health-plan choice research shows how dominated selections and information frictions can distort observed demand. Agricultural insurance experiments in India, Ghana, and Malawi reveal different responses to contract terms, credibility, and credit arrangements. The US flood-insurance literature documents the time dependence of demand following disasters. European supervisory work identifies gaps between consumer assumptions and the scope of catastrophe protection. Life-cycle simulations show how beliefs about nonpayment may influence long-horizon insurance holdings. These results cannot be merged into one universal elasticity or interpreted as evidence of identical behavioral causes across insurance lines.
An adequate research design records subjective expectations before and after information or experience, establishes an independently reviewed contractual benchmark, observes the real purchasing opportunity, and follows the financial consequences when the relevant event occurs. Contract version, claim eligibility, claimant selection, payment timing, retained loss, public support, and distributional constraints all belong in the analysis. Regulatory evaluation should assess accurate understanding, fair claims administration, sustainable capacity, and recovery outcomes together.
The original contribution of this monograph is an evidence-based synthesis of the connections among subjective coverage beliefs, claims expectations, perceived institutional credibility, behavioral decisions, and actual protection. Its conclusions are analytical interpretations of established research and supervisory evidence. They do not constitute a new named theory, original empirical data set, or experimentally verified universal causal law. A scientifically meaningful measure of insurance protection ultimately joins what people expect to receive with what the contract, the insurer, and the surrounding institutions can actually deliver.
References
Bhargava, S., Loewenstein, G., & Sydnor, J. (2017). Choose to lose: Health plan choices from a menu with dominated option. The Quarterly Journal of Economics, 132(3), 1319–1372. https://doi.org/10.1093/qje/qjx011.
Briggs, J., Rogers, C., & Tonetti, C. (2026). Risky Insurance: Life-Cycle Insurance Portfolio Choice with Incomplete Markets (NBER Working Paper No. 35122). National Bureau of Economic Research. https://www.nber.org/papers/w35122.
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://www.aeaweb.org/articles?id=10.1257/app.5.1.104.
European Insurance and Occupational Pensions Authority. (2024). Revised staff paper on measures to address demand-side aspects of the NatCat protection gap. EIOPA. https://www.eiopa.europa.eu/publications/revised-staff-paper-measures-address-demand-side-aspects-natcat-protection-gap_en.
European Insurance and Occupational Pensions Authority. (2025). Natural catastrophes, is your home covered? Findings from a sample-based analysis on the clarity of NatCat coverage in Household Insurance Product Information Documents. EIOPA. https://www.eiopa.europa.eu/publications/natural-catastrophes-your-home-covered_en.
Gallagher, J. (2014). Learning about an infrequent event: Evidence from flood insurance take-up in the United States. American Economic Journal: Applied Economics, 6(3), 206–233. https://www.aeaweb.org/articles?id=10.1257/app.6.3.206.
Gallagher, J., & Hartley, D. (2017). Household finance after a natural disaster: The case of Hurricane Katrina. American Economic Journal: Economic Policy, 9(3), 199–228. https://www.aeaweb.org/articles?id=10.1257/pol.20140273.
Giné, X., & Yang, D. (2009). Insurance, credit, and technology adoption: Field experimental evidence from Malawi. Journal of Development Economics, 89(1), 1–11. https://doi.org/10.1016/j.jdeveco.2008.09.007.
Handel, B. R., & Kolstad, J. T. (2015). Health insurance for “humans”: Information frictions, plan choice, and consumer welfare. American Economic Review, 105(8), 2449–2500. https://www.aeaweb.org/articles?id=10.1257/aer.20131126.
Hielkema, P. (2026, March 25). Climate insurance protection gaps: A demand-side challenge. Eurofi Magazine; published on the EIOPA website. https://www.eiopa.europa.eu/publications/climate-insurance-protection-gaps-demand-side-challenge_en.
International Association of Insurance Supervisors. (2024, December). Insurance Core Principles, including ICP 19: Conduct of business. IAIS online standards and guidance. https://www.iais.org/icp-online-tool/.
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.
Kousky, C., & Netusil, N. R. (2023). Flood insurance literacy and flood risk knowledge: Evidence from Portland, Oregon. Risk Management and Insurance Review, 26(2), 175–201. https://doi.org/10.1111/rmir.12242.
Kunreuther, H., Pauly, M., & McMorrow, S. (2013). Design principles for insurance. In Insurance and Behavioral Economics: Improving Decisions in the Most Misunderstood Industry (pp. 185–207). Cambridge University Press. https://doi.org/10.1017/CBO9781139050319.014.
Maddux, J. E., & Rogers, R. W. (1983). Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change. Journal of Experimental Social Psychology, 19(5), 469–479. https://www.sciencedirect.com/science/article/abs/pii/0022103183900239.
Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91(1), 93–114. https://doi.org/10.1080/00223980.1975.9915803.
Schwarcz, D., Cude, B., Logue, K., & Marquez Alcala, G. (2025). Read But Not Understood? An Empirical Analysis of Consumer Comprehension in Homeowners Insurance (University of Michigan Law & Economics Working Paper, paper 284). University of Michigan Law School. https://repository.law.umich.edu/law_econ_current/284/.