RESEARCH DOSSIER
Satellite Enabled Agricultural Microinsurance and Protection Evidence
Abstract. Satellite-enabled agricultural microinsurance converts remotely observed environmental conditions into evidence used in selecting, administering, or triggering insurance protection for agricultural households with limited access to conventional insurance. Its central scientific problem is the relationship between the observed environmental event, the covered economic loss, and the household's realized protection. This monograph investigates that relationship through the design of satellite-derived indices, actuarial assessment of basis risk, contractual and supervisory requirements, and the interpretation of empirical evidence concerning asset retention, consumption, investment, and resilience. It compares livestock insurance in northern Kenya with crop-index design in Ecuador, an Ethiopian index-insurance bundling experiment, and international technical and supervisory literature. Satellite observation can make geographically extensive and recurrent risks administratively insurable, yet the quality of a proxy for rainfall or vegetation is distinct from the quality of a payout rule and from its effect on welfare. Protection depends on contract fit, the joint distribution of losses and payments, affordability, settlement timing, and the institutional conditions under which a household can use insurance. An original comparative synthesis and transparent hypothetical calculations distinguish loss prediction, severe-loss nonpayment, expected indemnities, net financial protection, and household-level causal effects. The evidence supports conditional conclusions about the effectiveness of satellite-enabled insurance; it does not establish universal protection or quantify the effect of satellite sensing alone. The monograph develops an integrated assessment of existing insurance and research practices rather than proposing a new theory, doctrine, or proprietary model.
Keywords: agricultural microinsurance; satellite remote sensing; index insurance; livestock insurance; rainfall; vegetation indices; basis risk; smallholder agriculture; parametric insurance; actuarial validation; drought; household welfare; causal inference; consumer protection; financial inclusion; climate adaptation.
Introduction
An agricultural household requires protection against economically consequential losses that it cannot absorb without sacrificing future productive capacity or present consumption. Drought may reduce harvest income, destroy grazing resources, increase the cost of purchasing feed, depress livestock prices, diminish creditworthiness, and force the sale of productive assets. Flood and excessive rainfall may damage crops at specific stages while leaving the accumulated seasonal rainfall entirely ordinary. These examples demonstrate that agricultural loss is a joint consequence of the environment, the exposure of a particular household, the timing of adverse conditions, and its economic capacity to respond.
Satellite-enabled microinsurance seeks to reduce the cost and logistical difficulty of protecting dispersed agricultural populations. It uses remotely sensed information in an insurance arrangement, potentially for several different operations: identifying exposure, estimating hazard, constructing a contractual trigger, validating an insured event, or evaluating an insurance portfolio. These uses must be separated. A satellite-based exposure map supports underwriting without necessarily being the basis of payment. A contract indexed to a blended precipitation product uses satellite-derived information even though ground observations also contribute. Conversely, an index contract can operate entirely on station data or an independently sampled area yield. The use of the word satellite therefore identifies a technological component, not an independent class of legal obligations.
The inquiry addresses four connected questions. How does an observation become an enforceable insurance trigger? What does actuarial quality mean when measured losses and payouts differ for individual households? What institutional arrangements convert a technical trigger into timely protection? What empirical evidence demonstrates a change in the conditions of insured households, and how far can those results travel across populations and contracts?
Research proceeds through a targeted, interdisciplinary documentary review. Official product documentation establishes data properties; the international supervisory issues paper supplies a comparative regulatory framework; an original national enactment and policy document supply a dated jurisdictional illustration; published scientific research supplies technical and household evidence. A 2025 literature review locates the field, while original studies illustrate index construction and outcomes. Publication records and abstracts are used only for propositions explicitly established in them. Full household microdata, signed individual policies, source imagery, pricing files, and insurer claims ledgers were not obtained. The analysis is neither a systematic replication of the literature review nor a new causal estimate; its numerical examples are independently constructed hypothetical calculations.
The original contribution lies in relating three objects commonly evaluated separately: the measurement of hazard, the distribution of financial compensation relative to individual losses, and the causal evidence of protection. This synthesis is applied to the interpretation of satellite-enabled agricultural microinsurance within inclusive insurance, rather than to general remote sensing, sovereign disaster finance, or the full science of agricultural insurance.
Conceptual and Scientific Foundations
Agricultural risk and the economics of small contracts
An insurance contract redistributes resources across states of the world. The household exchanges a premium for an entitlement conditional on an insured event. The value of this entitlement depends on which states receive payment and what the payment permits the household to do. The microinsurance context emphasizes modest transactions, populations with limited access to conventional insurance, and contracts designed for their economic and administrative circumstances. Affordability and simplicity are operational requirements, but a low nominal premium is not a sufficient measure of insurance value.
Traditional indemnity assessment seeks to determine the covered loss experienced by a particular insured. Agricultural claims may require field visits, crop sampling, livestock inspection, documentation, and verification of causation. Where holdings are small and geographically dispersed, such work can consume a substantial proportion of possible premiums. Index insurance replaces some or all individual loss determination with an objectively defined observable event and a pre-specified payout calculation. The IAIS index-insurance issues paper (2018) examines the regulatory consequences of this substitution, including the possibility that the payment and the actual loss will differ.
This substitution changes the evidence required to administer the contract. The insurer ordinarily needs proof that the insured exposure and contract are valid, that the relevant observation pertains to the designated insured unit and time window, and that the trigger condition has been satisfied. It may not need to calculate every individual's harvested yield or livestock mortality. A policy can consequently be simpler to settle while presenting a different and material risk to the policyholder: economic loss may occur without a contractual payment.
The value of an index is especially dependent on covariance. A village-wide drought affecting many households simultaneously is easier to represent with a geographical environmental indicator than theft from one farmer or an isolated animal disease. An index that tracks drought exposure cannot rationally be judged for its ability to pay for excluded mortality from unrelated disease. Evaluation must specify the insured peril before measuring missed loss. At the same time, households experience their livelihoods as portfolios of crops, animals, income, debt, and consumption, so a product covering one peril may leave most economically severe states unprotected. Both contract-specific adequacy and wider vulnerability matter.
Correlated agricultural losses create an additional financial problem. A product that replaces individual inspection with automated payments can make claims administration scalable without diversifying the underlying hazard. A single regional drought can produce simultaneous obligations across thousands of contracts. Capital, liquidity, reinsurance, and risk transfer must therefore be sized for spatially correlated trigger obligations. Administrative efficiency and underwriting diversification are independent questions.
Environmental observation and economic exposure
A remotely sensed observation measures a physical signal. Depending on the sensor, the signal may represent reflected radiation, thermal emission, or microwave interaction with the land surface or atmosphere. Rainfall, soil moisture, crop development, vegetation greenness, or evapotranspiration can be estimated or inferred from such measurements with algorithms and ancillary inputs. The resulting products have different spatial support, temporal cadence, error properties, and revisions.
A farmer's actual financial loss cannot generally be read directly from a satellite pixel. Two adjacent fields may contain different crops, planting dates, drainage systems, irrigation, soil depths, or management practices. Their yields can diverge under a common weather history. Grazing animals may move across administrative boundaries, making the location of the insured household an imperfect descriptor of the grazing environment. Feed purchases and changes in herd composition can alter livestock mortality and financial loss even where vegetation conditions are similar.
These distinctions explain why the scientific object is the joint chain from observation to exposure and economic outcome. Improvements in spatial resolution can reduce some sources of aggregation error but cannot remove economic heterogeneity. Conversely, a coarser index may be sufficiently protective where loss is dominated by a spatially extensive drought and local observations are sparse. The correct question is empirical: how well does the specified observable capture the contractually relevant bad states for the intended population?
The systematic quantitative literature review by Nguyen et al. (2025) identifies 89 articles covering 93 studies, of which 49 use satellite-based data and 44 do not. Its analysis of index design and basis risk documents a substantial and heterogeneous research field. These counts describe the reviewed literature, not the prevalence of insurance products, their enrollment, or the causal effect of satellite-based insurance. A technically sophisticated field of research can coexist with limited evidence concerning the consequences of individual policies.
Distinguishing index insurance from adjacent arrangements
Satellite information may support a public drought-monitoring system, an anticipatory cash transfer, a conditional credit facility, a commercial indemnity policy, or an index microinsurance contract. These arrangements can respond to the same environmental signal while generating different entitlements and incentives. A disaster transfer funded by a public budget, for example, can create a claim on a social protection program without transferring risk through an insurance contract. A premium subsidy can finance insurance participation without itself being the payout. A loan with contingent repayment can protect a borrower's cash flow through a change in debt obligations rather than a conventional insurance payment.
The legal and economic distinction matters for evaluation. Insurance performance concerns the rights established by the contract and the insurer's capacity to perform them. Social protection performance concerns statutory or programmatic eligibility, adequacy, timeliness, and funding. Credit performance includes borrowing, investment, indebtedness, and repayment. Households may benefit from a combination, but classification is required to understand each channel and avoid counting the same money or outcome twice.
Microinsurance also differs from aggregate sovereign or regional parametric coverage in the immediate unit of entitlement. A national government can receive a payout and allocate funds through a response plan. Direct household cover instead requires a defensible association between the insured household or group and the payment calculation, accurate beneficiary records, and accessible delivery. Aggregate portfolio protection or sovereign liquidity does not prove that an individual exposed household received a payment.
The research therefore distinguishes direct household-level insurance, group policies with specified beneficiary arrangements, and insurance that funds broader risk-response institutions. These distinctions govern the interpretation of the evidence throughout the monograph.
Satellite Observation and Contract Design
Data products, spatial support, and temporal integrity
Rainfall estimation illustrates the importance of data provenance. Funk et al. (2015) describe the original CHIRPS precipitation dataset, combining satellite infrared precipitation estimates with station observations. The CHIRPS version 3 documentation describes a 0.05-degree gridded dataset extending from 1981 to near the present between approximately 60 degrees south and 60 degrees north. The operational documentation distinguishes preliminary and final estimates and their different availability schedules.
These two records have different scientific functions. The original published research establishes the origin and method of a data family. Current technical documentation specifies the characteristics of a later operational version. A contract using a particular product must identify the version actually used, rather than treating all versions as interchangeable. A historical backtest performed with a revised archive could benefit from information unavailable when an operational preliminary trigger would have been calculated.
A 0.05-degree grid spacing is approximately 5.6 kilometers north to south near the equator, while east-west distance varies with latitude. This is a geometric approximation, not an assurance that rainfall is measured accurately at every point within a cell. It is particularly inappropriate to interpret a gridded estimate as an observed rain gauge at an individual farm.
An operational specification includes the data provider, product and version, observation units, grid or geographic aggregation, season, update stage, missing-data convention, source hierarchy, calculation code, rounding rules, and publication deadline. The provider's revision policy must be reconciled with the contract's finality rule. If the contract stipulates a preliminary estimate, the insurer must preserve that release. If it stipulates a final estimate, the settlement calendar must accommodate the published update schedule. Mixing the two retrospectively changes the statistical risk and potentially the contractual entitlement.
Vegetation indices provide another route. A normalized difference vegetation index, commonly expressed as the difference between near-infrared and red reflectance divided by their sum, is a dimensionless measure related to vegetation conditions. Its interpretation varies with vegetation type, soil background, atmospheric correction, saturation, cloud masking, and seasonal phenology. A low vegetation value may signal drought-related forage stress, but low greenness is not the same as livestock mortality. A seasonal anomaly relative to a location's expected vegetation often has a different meaning from its absolute value.
Other candidate inputs include satellite-derived soil moisture, snow or inundation extent, evapotranspiration, land-surface temperature, and modeled crop development. Each has a task-specific relationship to damage. Combining inputs may improve predictive accuracy yet increase the complexity of communication, the possibility of operational failure, and the difficulty of verifying a payout independently. The tradeoff must be evaluated against measured losses and contractual usability.
From observation to a legally specified trigger
An index contract requires an explicit mapping from the data to the payout. Consider a hypothetical drought contract with seasonal rainfall R, a strike of 100 millimeters, an exhaustion point of 40 millimeters, and a maximum payment of 120 monetary units. Payment P is calculated as follows:
P(R) = 120 × max(0, min(1, (100 − R) / 60)).
When rainfall is 70 millimeters, the payment is 60. At 25 millimeters it is 120. At 120 millimeters it is zero. The example is illustrative; its thresholds and nominal maximum are neither observed market terms nor a recommendation for agricultural use.
A linear payout schedule may be transparent to customers and straightforward to reproduce. Nevertheless, selecting the strike and exhaustion requires evidence on the relationship between the rainfall metric and economic damage. A crop vulnerable to a two-week dry spell at flowering may experience a total seasonal rainfall above the strike. A household with functioning irrigation may sustain little damage when the seasonal metric is far below it. A large payout in the latter case remains contractually valid; a missed payment in the former may demonstrate limited protection even if the trigger is calculated correctly.
The validation unit must reflect the contract. Suppose a farmer's payment is based on mean rainfall across an administrative district. The model must be tested against losses in the insured district and against the distribution of losses within it. An excellent correlation between district-average rainfall and district-average yield may conceal systematic underpayment in peripheral locations. Changing from one district to multiple finer zones can reduce spatial mismatch, but it can also increase the sample-size burden for calibrating each zone and introduce discontinuities at borders.
The choice of season is equally important. A contract can use planting-to-harvest cumulative rainfall, independent subperiods, a weighted seasonal index, or a multiple-trigger schedule. Earlier payments may be more valuable where the objective is to finance seed replacement, forage purchases, or household consumption. Later triggers may better capture realized seasonal conditions. A schedule optimized for end-of-season correspondence can fail to protect a household against liquidity needs arising weeks earlier.
Multiple triggers can create combined or separate coverage. A deficit trigger for planting moisture and an excess-rainfall trigger for harvest damage may be calculated independently; their payments may share a maximum limit or have separate limits. Such choices change the expected payout, the probability of reaching the aggregate limit, and the degree of protection against compound events. Explicit limits avoid counting the same insured amount multiple times.
Calibration and independent validation
Index construction should start from observations of the relevant economic loss rather than from the availability of an attractive satellite product. Historical production, livestock mortality, yield, or household income records must be reconciled with insured peril definitions and exposure changes. Data exclusions, missing records, and quality flags should be documented before model selection. Records from an unusually accessible region or commercially successful farms may not represent underserved smallholders.
A development sample can support variable selection and parameter estimation, whereas temporally and spatially independent observations are necessary to test generalization. Randomly partitioning neighboring observations from the same drought across training and testing can produce misleading accuracy by allowing the same event to inform both. Leave-season-out and leave-region-out evaluations provide different insights: temporal stability and geographical portability. Neither automatically supplies evidence of consumer welfare.
The model should be compared against simpler baselines, such as a station rainfall index, an independently measured area yield, or a straightforward historical climatology. Sophistication has an implementation cost. A high-dimensional model may predict losses better in retrospective testing but provide a payout rule difficult for insurers, complainants, and supervisors to audit.
The Ecuadorian study by Valverde-Arias et al. (2020) presents a remote-sensing rice insurance design in Babahoyo, distinguishing agroecological zones and relating thresholds to crop physiology and economic consequences. Its significance for the present investigation is methodological: environmental indices become meaningful through location-specific production and financial conditions. The study is evidence of design research; it does not by itself demonstrate realized protection for customers of a marketed insurance product.
Likewise, historical vegetation and mortality records in Kenya supported the design of livestock index insurance described by Chantarat et al. (2013). This connects an environmental signal to a loss mechanism particularly relevant to pastoral systems. The calibration relationship depends on the production system, loss data, period, and geographical definition. Transplanting the same trigger to irrigated cropping or a different grazing system would require new evidence.
Operational reproducibility
Reproducibility has an operational dimension distinct from scientific replication. Two independent authorized calculations using the same frozen dataset, geographical mapping, policy version, and rounding convention should produce the same payment. A record of the data download date and checksum, the calculation software version, policy exposure units, and generated obligations provides evidence of that reproducibility.
Missing observations need an ex ante rule. Possible contractual treatments include waiting for a designated release, switching to a specified replacement source, applying a prescribed interpolation, or referring the matter to an explicitly defined dispute process. The IAIS guidance on index credibility (2018) emphasizes issues arising from the credibility and administration of the index. A discretionary replacement introduced after the result is known can alter expected payouts and compromise predictable contractual performance.
A data-provider interruption must be distinguished from a recorded zero rainfall value. Both may appear as zeros in poorly designed pipelines, with radically different meanings. Quality control should prevent missing or malformed observations from automatically generating a drought payment or suppressing a valid one. Corrections to farm coordinates or policy boundaries likewise require a documented effective date and an audit trail.
A defensible payout statement communicates the insured location or unit, coverage dates, index value, applicable strike, formula, insured amount, and resulting payment. The technical accuracy of the calculation is part of the evidence of performance. Its accessibility to a policyholder who does not have internet access or advanced technical training is a separate consumer-protection requirement.
Basis Risk and Actuarial Protection
Defining the relevant mismatch
Let L denote the monetary loss attributable to the insured peril, and let P denote the contract payout. For a specified reference loss valuation, the positive shortfall is max(L − P, 0). The signed financial mismatch L − P is different: it can be negative when payment exceeds the reference loss. An actuarial evaluation should disclose which measure is being used because averages of signed differences can conceal concentrated suffering in bad states.
The expected payment E[P] determines the pure expected claims cost per exposure for a specified payout distribution. The expected shortfall E[max(L − P, 0)] measures another object. Severe-loss nonpayment, expressed as the probability that P is zero conditional on L exceeding a declared material threshold, captures yet another aspect of quality. A policy can improve one metric without improving all others.
Several origins of mismatch are analytically relevant. Spatial mismatch occurs when the insured exposure differs from the geographical support of the index. Temporal mismatch occurs when the index window fails to reflect the timing of agricultural sensitivity. Model or design mismatch concerns the relationship between the environmental metric and financial loss, including choices of threshold, aggregation, and payout scale. The review of satellite-based index design by Nguyen et al. (2025) provides a contemporary research synthesis of such problems.
The measured mismatch further depends on the reliability of the reference loss. A recorded yield may lack information about quality degradation or price changes. Reported herd mortality may omit animals sold under distress before dying. A survey of consumption may miss informal transfers or measurement intervals that matter during a shock. Verification of basis risk should therefore document the definition and uncertainty of the benchmark, not simply report an index–loss correlation.
Hypothetical severe-state comparison
Consider 100 equally weighted household-seasons. Twenty experience a severe covered loss of 100 monetary units, and eighty have no covered loss for the purpose of the example. Two candidate contracts each pay 50 units if their respective trigger is satisfied. Contract A pays fourteen severely affected observations and eight nonsevere observations; Contract B pays eighteen severely affected observations and twenty nonsevere observations.
| Metric | Contract A | Contract B |
|---|---|---|
| Severe observations | 20 | 20 |
| Severe observations paid | 14 | 18 |
| Severe observations unpaid | 6 | 2 |
| Nonsevere observations paid | 8 | 20 |
| Total payments | 22 | 38 |
| Severe-loss nonpayment probability | 30% | 10% |
| Nonsevere payment probability | 10% | 25% |
| Mean payout over all observations | 11 | 19 |
| Mean severe-state loss shortfall before premium | 65 | 55 |
The expected payout of A is 22 × 50 / 100 = 11, while that of B is 38 × 50 / 100 = 19. The severe-state shortfall under A is (14 × 50 + 6 × 100) / 20 = 65; it is (18 × 50 + 2 × 100) / 20 = 55 under B.
Assume, solely for comparison, a flat four-unit loading on both pure premiums and no subsidy. Premiums become 15 and 23. Mean severe-state net financial burden, including the premium, is 80 under A and 78 under B. The improvement from B is ten units in gross severe-state shortfall but only two units in this premium-inclusive measure. Contract B can still provide much better protection against complete nonpayment in a severe state. The two-unit mean difference does not express the value to a household of avoiding a particularly damaging zero-payment outcome.
Nonsevere payments in the example are valid claims under the index rule. They are not evidence of fraud and are not intrinsically regulatory errors. Their significance lies in pricing, distribution of payments across states, and the possible diversion of scarce premium resources from severe losses. A household may value a nonsevere payout because the definition of covered loss used in the example is narrower than its actual needs. In real evaluation, the scope of the reference loss should be explicit.
At a subsidy rate of 50 percent applied equally to both illustrative loaded premiums, household contributions would be 7.5 and 11.5. Severe-state net burdens would become 72.5 and 66.5 respectively. The difference changes from two to six units because public support changes the household's cost, not the trigger's prediction. Subsidy cost per exposure differs between the products. A comparison of inclusive insurance designs must assess both household protection and the resources required to obtain it.
The arithmetic shows why plain accuracy can mislead. In a population with 20 severe observations, a contract that never pays is correct for eighty observations under a binary loss classification, yet fails entirely in severe states. Optimizing overall classification accuracy without appropriate weighting can privilege frequent low-loss states. Severe-state performance, conditional payment adequacy, and net costs are more informative.
Tail protection and expected utility
A household with constrained liquidity may place disproportionately high value on additional money in a bad state. In expected utility notation, terminal economic resources can be represented as W − L − π + P, where W denotes resources before the shock, π is the paid premium, and the timing of payments is assumed equivalent for this simple representation. Under risk aversion, the utility consequences of P depend on its association with L, other income, and background risks.
This expression does not account for liquidity constraints, irreversible asset sales, or the fact that the premium is paid before the loss. It is useful as an abstraction, not a fitted model of household behavior. A richer evaluation can compare distributions of consumption, productive assets, and debt, explicitly accounting for the timing of each cash flow. Estimating those consequences requires data rather than assumptions about household preferences.
Empirically, Jensen, Barrett, and Mude (2016) report in their northern Kenya basis-risk study that the insurance removed 63 percent of covariate high-loss risk across the investigated eight semiannual seasons, while 69 percent of original high-loss risk remained. These quantities concern different denominators, namely the covariate component and the total original high-loss risk. They cannot be treated as complementary percentages. The distinction supports the conclusion that successful hedging of common shocks can coexist with substantial residual household loss.
Chantarat et al. (2013) report household-level simulations in which the proposed insurance could remove approximately 25–40 percent of total livestock mortality risk in the studied setting. Their livestock insurance design study supplies an earlier quantitative indication of the difference between index-based protection and complete indemnification. The estimate is simulated performance under the design, not a causal effect of coverage on household welfare.
These two research records differ in their underlying samples, analysis, and measures. Their results should not be averaged into a general estimate of agricultural microinsurance effectiveness. They are more useful as demonstrations of the necessity of defining the loss distribution and the component of risk against which the policy is evaluated.
Portfolio aggregation, capital, and liquidity
Payment obligations may be highly concentrated in time and space. Suppose a common regional index triggers a 60-unit payment under each of 1,000 otherwise identical policies. Aggregate liabilities from that trigger are 60,000 units. The insurer may process the payments electronically, but the amount still has to be funded.
For portfolio payment random variables P₁ through Pₙ, the variance of aggregate payments equals the sum of individual variances plus twice the sum of pairwise covariances. Positive covariance increases tail exposure relative to an assumption of independence. The same environmental conditions may trigger multiple zones, insurers, or reinsurance treaties, producing correlated financial obligations.
Premium adequacy requires consideration of expected payouts, operating expenses, cost of capital, reinsurance costs, distribution costs, taxes where applicable, and uncertainty in the estimated risk. A premium subsidy can alter affordability without changing the insurer's contractual obligation. An insurer receiving the same total premium may face very different commercial risks depending on who provides the subsidy, when it is paid, and whether it is guaranteed.
Stress tests should include multi-year drought, correlated regional shocks, provider outages, data revisions, adverse changes in take-up, and payment-channel disruption. Climate nonstationarity can affect the frequency with which historical thresholds are crossed. Land use and irrigation changes can affect the relationship between a similar environmental event and economic loss. Historical expected payouts should therefore be accompanied by a clearly stated reference period and sensitivity to plausible shifts rather than represented as stable timeless quantities.
Reinsurance supporting an index book must also match its obligations. An insurer can owe payouts to customers while its own reinsurance trigger does not attach. That mismatch is portfolio-level basis risk, not a reason to alter the household contract. Evaluation of insurer resilience must accordingly compare inward policy obligations, outward risk transfer, capital, and liquid resources under common stress conditions.
Insurance Institutions and Consumer Protection
Contractual status and the regulated perimeter
Insurance is a legal arrangement whose classification and enforceability depend on jurisdiction. Some legal systems recognize index-based obligations without individual proof of damage; others require a specific relationship between insured interest, loss, and payout or establish special authorization conditions. The IAIS issues paper on index insurance (2018) discusses these differences and their implications for inclusive insurance supervision. It is an international supervisory issues paper, not a directly binding universal law.
The Kenyan illustration is particularly relevant because livestock-index insurance has an identifiable research and institutional history there. The Insurance (Microinsurance) Regulations, 2020 establish a regulatory framework in the original gazetted enactment, including microinsurance product requirements and approval provisions. The text specifically distinguishes index-based insurance in its treatment of a grace-period requirement. This analysis concerns the original enactment as gazetted on March 6, 2020; it makes no assertion about subsequent amendments, current consolidated legal obligations, licensing of particular providers, or the terms of a specific marketed policy.
The existence of a category of index insurance in a regulatory framework does not establish that every satellite-based arrangement is an approved insurance product. Product classification, policy wording, eligible provider, distribution authorization, and relevant conduct-of-business rules must be checked for the particular jurisdiction and transaction. An operational dashboard created by a technology vendor cannot itself define contractual obligations unless integrated in the legally effective arrangement.
The IAIS financial inclusion resources also identify broader work on inclusive insurance supervision, including the index-specific paper and inclusive-insurance guidance subject to revision. The distinction between a dated document and a current institutional resource is important. Historical guidance can illuminate the development of an issue without being represented as the currently controlling national law.
Historical development in Kenya
Kenya's National Agricultural Insurance Policy (2023) records the livestock index insurance pilot initiated in Marsabit in 2010, subsequent public livestock insurance initiatives, and a distinct area-yield crop insurance approach. These descriptions demonstrate institutional differentiation within agricultural risk financing. Historical targeted public support for livestock coverage also illustrates how government funding can extend access to a defined vulnerable population.
The Kenyan livestock insurance program FAQ maintained by the International Livestock Research Institute explains a historical arrangement based on drought-related forage conditions, standardized livestock units, geographic index areas, and threshold-triggered payments. Its value is descriptive and educational. It is not the signed policy, proof of today's product availability, or a source of current premiums. The scientific evaluation consequently employs this material to clarify the nature of the arrangement and avoids treating its historical figures as modern contract terms.
These institutional sources must be read alongside original research. The observed protection of a household under a particular implementation can depend on the insured amount, the definition of animals or units covered, the geographical index area, the season, and the current delivery arrangements. Research from northern Kenya provides evidence for particular populations and periods, while national policy describes a wider program architecture. A conclusion about the national system's present effectiveness would require current administrative and household evidence beyond these sources.
Distribution, comprehension, and accessibility
Microinsurance commonly depends on intermediaries capable of reaching dispersed low-income customers. A sales network may include cooperatives, mobile channels, lenders, agricultural input suppliers, agents, or community organizations. Each intermediary has incentives that shape how the insurance is explained and how customer records are collected.
A seller remunerated for enrollment may emphasize affordability and benefits while neglecting the circumstances in which loss does not produce payment. A lender may be concerned with repayment protection more than household consumption. A cooperative may have incentives to support participation but may also impose group conventions that are poorly aligned with individual exposure. Contractual and institutional safeguards must make the product's risk and benefit allocation intelligible.
Effective disclosure should enable a customer to answer practical questions: What is insured? When does cover start and end? Which location determines the trigger? How much is paid at different index values? Can I lose crops or animals without payment? Who receives the money? When is payment due, and how can I dispute an incorrect calculation? The examples used at enrollment should include a severe-loss nonpayment scenario, not only cases in which the customer receives the advertised maximum.
Language, literacy, numeracy, connectivity, and cultural context influence whether these disclosures are meaningful. A short verbal explanation in a locally understood language, accompanied by accessible examples, may be more valuable than a complex digital interface. Technical simplicity does not require scientific inaccuracy. A customer need not understand the physics of atmospheric sensing to understand the conditions of entitlement.
Remote enrollment creates risks of incorrect coordinates, duplicate or disputed beneficiaries, lapsed coverage, identity mismatches, and account access problems. These operational risks can cause genuine failures of protection even when the satellite model is excellent. They should be measured independently from basis risk. Payment to a functioning mobile account does not necessarily establish that the intended household controlled or received the funds.
Complaints, data errors, and payment delivery
Three kinds of dispute require different evidence. A policyholder may allege economic loss even though the index did not trigger; a calculation dispute may allege that the published index was applied incorrectly; and a delivery dispute may allege that the correct payment was not received. The first concerns product fit and potentially disclosure or fairness. The second concerns contractual calculation. The third concerns fulfillment of payment obligations.
A complaints channel must be able to distinguish these cases, consult the frozen contract and data, correct genuine errors, and explain the result. Consumer trust can be damaged when all complaints are dismissed as basis risk. At the same time, an insurer cannot solve every uncovered loss by retroactively changing a contractual trigger, because that may undermine predictability, pricing, and equitable treatment of other customers.
There are circumstances in which public emergency transfers, separately funded discretionary assistance, or another layer of insurance can complement index benefits. Those mechanisms should remain identifiable. Labeling an exceptional public grant as an insurance payment would distort claims experience and conceal the limitations of the original contract. Honest accounting distinguishes benefits across sources.
Payment speed matters because the loss consequences evolve. Consider two identical 60-unit payments, one delivered before a household must borrow for food or livestock feed and one delivered after a costly loan or irreversible sale. Their amounts are equal while their protection may differ. A settlement evaluation should record both trigger determination and actual receipt, not merely the date a payment file was generated.
The insurer must plan liquidity and payment infrastructure for widespread triggering. Failed transfer attempts, account freezes, connectivity interruptions, beneficiary disputes, and reconciliation delays can emerge simultaneously during a disaster. Resilience of the payment channel is part of insurance performance and should be tested under the conditions in which protection is most needed.
Subsidy, equity, and sustainability
Premium subsidies can address affordability constraints or support the development of distribution networks, but their success depends on the allocation of benefits and fiscal resources. A subsidy paid to a provider may not be passed through fully to customers if the product is poorly priced or market competition is limited. A subsidy linked to a fixed amount of livestock coverage may have different distributional consequences for households holding different herds.
The distinction between enrolled people, paid premiums, valid in-force exposure, and the actual distribution of payouts is essential. It is possible to increase enrollment while leaving severe-loss nonpayment unchanged. It is also possible to improve trigger quality without increasing enrollment if premiums rise beyond the reach of the target population.
An inclusive program should consider coverage of households without formal land titles, women with limited control of mobile accounts or assets, tenants, pastoralists with mobile herds, and farmers excluded from a particular credit or cooperative channel. Exclusion can arise from the selection of the insured unit even before satellite data are applied. Evidence about the experiences of current participants will not reveal the welfare consequences for people who were unable to participate unless the evaluation separately measures access.
Empirical Evidence and Causal Interpretation
Evidence of technical and actuarial performance
Research on satellite agricultural insurance spans several levels of evidence. The potential to construct a meaningful index is one level. Out-of-sample performance relative to measured covered losses is another. Observed insurance coverage and claims delivery are further levels. Causal effects on household outcomes constitute an additional and demanding inquiry.
The Ecuadorian rice design study by Valverde-Arias et al. (2020) illustrates how environmental and economic specifications can be connected at the design stage. The Kenyan livestock index design by Chantarat et al. (2013) uses observed historical loss relationships and simulates protection. Jensen, Barrett, and Mude's (2016) analysis of insurance quality examines the distribution of basis risk in northern Kenya. These studies support different propositions and should not be treated as interchangeable estimates of household welfare.
A design study can justify a larger operational pilot while leaving household effects uncertain. A backtest can reveal how the contract would have paid during observed historical events, but its relevance depends on whether the model and thresholds were chosen without looking at the same test observations. An empirical insurance-quality study can establish residual risk within a population even where it does not estimate the effect of enrollment on productive decisions.
A claims ledger can demonstrate how fast and accurately an insurer paid. It cannot, without additional evidence, identify what households would otherwise have done or whether they achieved higher welfare because of coverage. A survey showing that customers report satisfaction after payment can provide valuable information about perceived benefits but can be selected by payout receipt, program awareness, or the propensity to remain in the study.
Household protection following drought
Janzen and Carter (2019), in After the Drought, examine the effects of microinsurance using randomized evidence in northern Kenya. The published abstract reports heterogeneous protective behavior across wealth groups: insurance reduced reliance on damaging asset sales among relatively wealthier households and reduced reliance on food consumption cuts among poorer households. The study is directly relevant because it distinguishes household coping responses rather than reducing protection to the receipt of a payment.
These findings illustrate that the same financial mechanism can protect different outcomes depending on baseline resources. A household with enough productive assets may preserve them and thus sustain future earnings. A household with very few assets may have less opportunity to avoid asset sales and may instead use protection to maintain food consumption. An aggregate mean without this distinction can obscure meaningful effects at different points in the wealth distribution.
The specific household study does not establish that satellite technology, considered independently of the contract, caused the reported changes. A causal effect of insurance reflects the combined product, premiums or incentives used in the experiment, implementation, available payments, and economic behavior. Isolating the contribution of the satellite input would require a comparison that holds other features sufficiently constant.
Jensen, Barrett, and Mude (2017) present a comparison of transfers and insurance in northern Kenya. Their abstract describes panel-based analysis using exogenous variation and reports different outcome patterns for cash transfers and index insurance. Insurance is associated in the study with productive decisions and reduced distress sales, while transfer-related findings concern other dimensions such as child health and mobility. These are study-specific comparative findings, not evidence that one instrument universally dominates the other. The potential overlap in populations and programs also cautions against treating each publication as an independent replication covering unrelated settings.
Bundling with agricultural finance
Insurance may be distributed with credit or agricultural inputs. In the Ethiopian experiment, Belissa, Lensink, and Marr (2025) investigate index insurance and credit bundling with input vouchers and report effects concerning insurance demand and productive decisions. The university repository abstract establishes the comparison of bundled arrangements and its reported findings. It does not supply sufficient detail to isolate the satellite component or to verify the exact trigger, contract wording, and every treatment definition.
The evidence is therefore relevant as comparative index-insurance research rather than as a verified independent evaluation of satellite-triggered cover. The distinction prevents extending the evidence base by assigning a technological modality that has not been established from the accessed record.
Bundling may change the economic channels of impact substantially. Credit can increase the ability to purchase seed or fertilizer. An input voucher can relax a working-capital constraint. Insurance can reduce exposure to specified bad states and influence willingness to invest. Bundling these components may raise adoption and improve production, but the combined effect cannot automatically be attributed to insurance alone. The same bundle can also increase indebtedness or constrain choice if it is poorly designed.
The evaluation unit should distinguish the offer, take-up of individual components, actual insurance coverage, premium paid, credit exposure, payout, and outcome. A lender's reduced default rate is evidence about credit performance; a household's preserved food consumption or productive assets is evidence about household protection. Both matter, but their interests can diverge if insurance payments are assigned to debt service.
Identification of causal effects
Consider households randomly assigned an offer of a particular insurance product or a control condition. Let Z denote assignment, D indicate coverage, and Y denote a specified outcome measured at a predetermined time. The intention-to-treat effect is the difference in average Y between assigned groups, under appropriate randomization and follow-up. It captures the consequence of offering the intervention, including incomplete take-up.
The difference between insured and uninsured households outside the assignment framework may combine insurance effects with preexisting differences. Wealthier, better informed, or more commercially oriented households may be more likely to buy coverage. More exposed households may also be more likely to insure, producing selection in the opposite direction. Simple comparisons of purchasers and nonpurchasers are consequently difficult to interpret.
Using randomized assignment as an instrument for coverage can identify a local average treatment effect under specific assumptions, including relevance, independence, exclusion, and monotonicity. The exclusion restriction becomes particularly problematic if assignment includes credit, vouchers, training, or other components that can influence outcomes directly. In such a setting the randomized offer remains interpretable as an intervention, but attributing the entire effect to insurance would require additional design or assumptions.
Insurance may influence investment before any payout occurs. This ex ante behavior is distinct from ex post protection. A study conditioning its main analysis on actual payout receipt may undermine randomization because receipt depends on geography, realized shocks, and the contract. Better designs predefine event exposure, estimate interactions with relevant shock conditions carefully, and separate the ex ante and ex post outcomes.
Attrition is especially consequential after disasters. Households that migrate, sell assets, or become difficult to contact may have experienced the largest losses. If coverage affects mobility or the ability to remain in the study, differential attrition can bias comparisons. Follow-up rates, reasons for missing outcomes, sensitivity bounds, and appropriate handling of clustered assignment should be part of the reported evidence.
Spillovers can occur through informal sharing, livestock market prices, labor demand, migration, or credit markets. An insured household might share money with an uninsured relative, improving the latter's outcome and narrowing the measured treatment difference. At large enrollment scales, aggregate effects can differ from those observed in a small experiment. The scale and assignment unit should correspond to the expected interference.
Endpoints and denominators
An evaluation may use insurance purchase, valid coverage, claim frequency, settlement delay, investment, crop yield, livestock retention, dietary consumption, school expenditures, debt, or poverty transitions as outcomes. Each answers a different question. No single metric establishes protection for every purpose.
Coverage penetration should refer to a defined eligible or target population, with duplicates and lapsed contracts treated appropriately. Take-up among households offered a product is distinct from the percentage of all eligible households covered. The probability of a payout among insured exposures is distinct from the probability of payment among those suffering severe covered loss.
Agricultural yields can increase after investment but still fail to translate into higher income if output prices decline or input costs rise. Asset retention can indicate protection of productive capacity but might be undesirable in a setting where selling some animals or reallocating resources is optimal under changing conditions. Food consumption measures may miss nutritional composition, household allocation, or later repayment burdens. Outcome definitions need to follow the economic mechanism rather than an assumption that a favorable movement in any indicator is conclusive.
A protective effect in one drought does not guarantee protection in subsequent seasons. A study population may face another climatic regime or a different combination of market conditions. Multi-season follow-up is particularly valuable where the objective is to prevent persistent asset depletion and poverty traps. Evidence of resilience requires attention to the trajectory of resources and adaptation, not simply one moment after payment.
Limits of generalization
The literature reviewed here includes direct satellite-linked design research, analysis of realized index quality, randomized household evidence, comparative social protection research, and index-insurance bundling evidence. The sources differ in observation units, insured hazards, research methods, time periods, and access scope. They cannot responsibly be pooled into an overall numerical benefit of satellite-enabled agricultural microinsurance.
The geographical focus of the primary household evidence in this monograph is northern Kenya, with comparative evidence and design material from Ethiopia and Ecuador. These locations represent meaningful but nonexhaustive contrasts. Outcomes in other production systems, climates, supervisory regimes, and markets require locally validated evidence. The international conclusion is about the conditions of defensible assessment, not a universal estimate of realized impact.
Integrated Evaluation and Research Requirements
Connecting observable evidence to protection
The findings of the preceding chapters can be combined through a sequence of existing insurance and evaluation tasks. First, an insured exposure and peril must be defined. Second, a relevant observation and index calculation must be specified. Third, the relationship between the payout and the distribution of losses must be validated. Fourth, the contract must reach eligible households at a manageable cost and generate enforceable rights. Fifth, payments must be calculated and delivered accurately and on time. Finally, protection outcomes must be measured using an appropriate empirical comparison.
None of these tasks can substitute for the others. A high-resolution satellite product is evidence of a measurement capability, not of financial protection. An enforceable contract creates rights but does not show that its payment addresses the economically severe states. A paid claim proves fulfillment of a specified obligation when correctly calculated and received; it does not establish the causal welfare effect of the product. Conversely, a well-designed randomized experiment can demonstrate an implementation's effect without furnishing evidence that the technical trigger would work in another market.
The contribution of the present synthesis is to identify what conclusions can be drawn at each stage and what further evidence is necessary to make the next claim. This approach uses ordinary actuarial, economic, and regulatory concepts. It establishes no new named theory or proprietary evaluation methodology.
| Research question | Minimum relevant evidence | Conclusion supported |
|---|---|---|
| What environmental conditions were observed? | Provider, version, geographic support, dates, data-quality records | Specification of the environmental measurement |
| Was the contractual payment correct? | Policy version, exposure mapping, index value, formula, settlement ledger | Reproducible contractual performance |
| Did payment correspond to loss? | Matched reference loss and payout data, severe-state definitions, independent validation | Distribution of basis risk and adequacy |
| Could the customer afford and receive protection? | Eligible population, pricing, subsidies, enrollment, coverage and receipt records | Access and delivery in the observed program |
| Did coverage change household outcomes? | Appropriate comparison design, baseline and follow-up data, defined estimands | Scoped empirical effects |
| Is the program financially resilient? | Portfolio obligations, correlated stress tests, capital, liquidity and reinsurance | Financial capacity under declared scenarios |
These distinctions also clarify the hierarchy of evidence. Technical product documentation supports data properties, but not behavioral effects. Supervisory literature establishes concerns and regulatory approaches, but not the binding meaning of a country's law. A scientific abstract may establish its reported finding while leaving the design details needed for replication unavailable. A full paper with accessible methods is stronger evidence for methodological interpretation but still does not replace operational claims data.
A reproducible research specification
Future empirical studies should link four classes of records under appropriate legal and ethical safeguards. The first contains coverage records: policy identifier, exposure location, insured unit, peril, period, insured amount, premium, subsidy and beneficiary. The second contains environmental data and index computation: source, version, dates, geographic aggregation, quality indicators, calculation and frozen output. The third contains loss and payment information, including loss benchmarks, complete and failed settlement records, actual receipt, delays, corrections and complaints. The fourth contains economic outcomes and comparison variables.
These records need not be centrally accessible to all parties. Privacy-preserving linkage and access control may permit researchers to evaluate protection without publishing exact household locations, identities, account numbers, or individually sensitive loss values. Public transparency should focus on the contractual specification and aggregate validation, while protecting the household data required to investigate unfairness and heterogeneous outcomes.
A research plan should declare primary endpoints, the covered-loss benchmark, the severe-state threshold, the validation strategy, the population, the causal estimand if applicable, and how missing data and revisions will be handled. Data retention should permit independent examination of a disputed trigger while complying with applicable retention and privacy requirements. Decisions concerning threshold changes must be versioned so that historical observations are evaluated against the contract in force at the time.
A portfolio research program can complement household analysis with climatic sensitivity and correlated-loss scenarios. Multiple assumptions should be reported, including changes in hazard frequency, exposure distribution, customer retention, reinsurance recoveries and settlement delays. It is valuable to stress the model precisely where records are sparse or calibration uncertainty is greatest rather than relying only on average historical performance.
Interpreting selection among alternative instruments
A satellite-index product competes for household resources and public funding with savings, insurance using other indices, indemnity insurance, credit arrangements, public transfers, investments in irrigation or water access, and other risk management interventions. The relevant counterfactual depends on the policy question. Comparing an affordable index product with no protection addresses a different decision from comparing it with an area-yield product or a targeted transfer scheme costing the same amount.
In locations with sparse historical loss data, an easily observed satellite signal may be technically attractive. Yet the lack of loss data also makes validation difficult. Pilot programs can address that gap through purposeful collection of household losses and outcomes, provided they maintain defensible disclosure and do not represent unvalidated performance as established.
In locations where conventional indemnity assessment is feasible and inexpensive, the administrative advantage of an index product may be smaller. A hybrid arrangement can use remote sensing to prioritize claim investigation while reserving individual assessment for obligations. Such a product has its own costs, incentives, and contractual rules; it should be evaluated rather than assumed superior.
Risk prevention and insurance are likewise complementary in some circumstances. A farmer may improve drought resilience through water storage, crop choice, or management while using insurance to transfer residual risk. Whether an intervention crowds out or encourages adaptation is empirical. A payment tied to a fixed geographical hazard may have limited sensitivity to individual mitigation, while a premium or underwriting structure can create other incentives. The allocation of adaptation benefits and insurance costs should form part of program assessment if it is an intended objective.
Research integrity and the boundaries of the present evidence
The most informative primary evidence identified here supports bounded statements. The Kenya livestock research establishes that an index can hedge a meaningful part of covariate risk while leaving substantial residual total risk. Household research reports heterogeneous coping effects. The Ecuadorian design research demonstrates the relevance of crop physiology and economic thresholds. The Ethiopian trial illustrates the importance of interventions bundled with insurance. Technical documentation establishes properties of a precipitation data family. Dated supervisory and national sources clarify institutional and regulatory questions.
The evidence does not establish a global protection rate, a universally optimal spatial resolution, a uniformly superior index, or an isolated causal effect of satellite sensing. Direct household studies are geographically concentrated. Several journal papers were inspected via their published abstracts or institutional records; their unexamined estimation details and datasets cannot be treated as independently validated. National legal analysis is limited to the original Kenyan enactment inspected, rather than a current multi-jurisdictional legal survey. No present-day insurer claims ledger, customer policy, or independent satellite processing archive was available for empirical reconstruction.
These limits guide the interpretation of the monograph. Findings are supported at the level of the observed evidence and clearly identified analytical derivations. Further empirical work should address these gaps through transparent operational verification and stronger household-level comparison, with local adaptation of the contract and research design.
Conclusion
Satellite-enabled agricultural microinsurance is a means of converting environmental observations into insurance entitlements for agricultural populations whose risks are expensive to assess individually. The technology broadens the possible geography and regularity of hazard observation. Its economic significance depends on the contract that converts observations into payments and on the institutions that deliver those payments to households.
Three distinctions determine the scientific interpretation. Measurement accuracy concerns the observed environmental signal. Insurance quality concerns compensation relative to the distribution of covered losses and the costs of obtaining that compensation. Household protection concerns what the household is able to preserve or accomplish because of the coverage. These distinctions explain why a better proxy for weather does not necessarily produce more valuable insurance and why evidence of timely contractual payment does not independently establish a welfare effect.
The available Kenyan research demonstrates that index design can reduce common drought-related risk while leaving substantial household-level risk, and that insurance may change coping strategies in ways dependent on baseline resources. The Ecuadorian and Ethiopian evidence illuminates distinct aspects of agricultural index design and the integration of insurance with financial access. Together these materials support a context-dependent, empirically disciplined account of protection. They do not support a single quantitative conclusion for all products or populations.
An effective assessment requires transparent observation provenance, independently validated payout schedules, severe-state basis-risk measures, affordable and enforceable coverage, reliable settlement, sound financial capacity, and credible evidence of household outcomes. These are connected requirements of inclusive insurance. The scientific purpose of satellite-enabled microinsurance is realized when remotely observed conditions support protection that is demonstrably meaningful in the actual economic lives of the insured.
References
Belissa, T. K., R. Lensink, and A. Marr. 2025. “The Impact of Bundling Index Insurance with Credit and Input Vouchers: Experimental Evidence from Ethiopia.” Journal of Economic Behavior & Organization 234: 107013. doi:10.1016/j.jebo.2025.107013. University repository record.
Chantarat, S., A. G. Mude, C. B. Barrett, and M. R. Carter. 2013. “Designing Index-Based Livestock Insurance for Managing Asset Risk in Northern Kenya.” Journal of Risk and Insurance 80 (1): 205–237. doi:10.1111/j.1539-6975.2012.01463.x. Institutional research record.
Climate Hazards Center, University of California, Santa Barbara. n.d. “CHIRPS: Rainfall Estimates from Rain Gauge and Satellite Observations.” Version 3 product documentation. Accessed November 19, 2026. Official dataset documentation.
Funk, C., P. Peterson, M. Landsfeld, D. Pedreros, J. Verdin, S. Shukla, G. Husak, J. Rowland, L. Harrison, A. Hoell, and J. Michaelsen. 2015. “The Climate Hazards Infrared Precipitation with Stations—A New Environmental Record for Monitoring Extremes.” Scientific Data 2: 150066. doi:10.1038/sdata.2015.66. USGS publication record.
International Association of Insurance Supervisors. 2018. Issues Paper on Index-Based Insurances, Particularly in Inclusive Insurance Markets. Approved May 25, 2018; released June 18, 2018. Official issues paper.
International Association of Insurance Supervisors. n.d. “Financial Inclusion.” Accessed November 19, 2026. Official supervisory resources.
International Livestock Research Institute. n.d. “Frequently Asked Questions: Index-Based Livestock Insurance.” Accessed November 19, 2026. Program explanation.
Janzen, S. A., and M. R. Carter. 2019. “After the Drought: The Impact of Microinsurance on Consumption Smoothing and Asset Protection.” American Journal of Agricultural Economics 101 (3): 651–671. doi:10.1093/ajae/aay061. Journal article record.
Jensen, N. D., C. B. Barrett, and A. G. Mude. 2016. “Index Insurance Quality and Basis Risk: Evidence from Northern Kenya.” American Journal of Agricultural Economics 98 (5): 1450–1469. doi:10.1093/ajae/aaw046. Institutional study record.
Jensen, N. D., C. B. Barrett, and A. G. Mude. 2017. “Cash Transfers and Index Insurance: A Comparative Impact Analysis from Northern Kenya.” Journal of Development Economics 129: 14–28. Institutional research record.
Kenya, Ministry of Agriculture and Livestock Development. 2023. National Agricultural Insurance Policy. December 2023 (official file posted in 2024). Official national policy.
Kenya, Republic of. 2020. Insurance (Microinsurance) Regulations, 2020. Legal Notice No. 26, Kenya Gazette, March 6, 2020. Original gazetted regulations.
Nguyen, T. T., S. Mushtaq, J. Kath, T. Nguyen-Huy, and L. Reymondin. 2025. “Satellite-Based Data for Agricultural Index Insurance: A Systematic Quantitative Literature Review.” Natural Hazards and Earth System Sciences 25: 913–927. doi:10.5194/nhess-25-913-2025. Open research article.
Valverde-Arias, O. R., P. Esteve, A. M. Tarquis, and A. Garrido. 2020. “Remote Sensing in an Index-Based Insurance Design for Hedging Economic Impacts on Rice Cultivation.” Natural Hazards and Earth System Sciences 20: 345–362. doi:10.5194/nhess-20-345-2020. Open research article.