RESEARCH DOSSIER
Insurance Distribution Geography and Access Deserts
Abstract. Insurance distribution is a geographically organized service that connects exposed households and enterprises to risk carriers. This monograph examines how the location, capacity, product permissions, connectivity, and continuity of distribution channels condition access to insurance in rural, remote, low-income, and disaster-exposed communities. It distinguishes the availability of a qualified channel from the willingness of carriers to insure an exposure, the affordability of equivalent contracts, and enforceable protection over time. A structured comparative review combines international supervisory guidance, administrative market evidence from the United States and European Union, digital connectivity statistics, randomized insurance-demand studies in India, China, and Ghana, and established spatial accessibility and causal inference methods. The analysis develops a reproducible measurement specification for service inventories, travel-time catchments, functional digital access, quote outcomes, and distributional disparities. It demonstrates through transparent hypothetical calculations why aggregate producer density and premium penetration can obscure localized exclusion and why observed differences in coverage cannot establish a causal effect of branch presence. Comparative evidence identifies complementary roles for assisted distribution, household liquidity, trusted information, carrier capacity, and disaster resilience. The monograph concludes that geographic access should be evaluated across the complete contract lifecycle using product-specific denominators, independently verified service capabilities, and explicit uncertainty. No new geographic microdata, worldwide prevalence estimate, or original causal effect is asserted.
Keywords: Insurance distribution; insurance intermediaries; access deserts; spatial accessibility; financial inclusion; rural insurance; digital exclusion; insurance affordability; geographic inequality; causal inference; consumer protection.
Introduction
The economic function of insurance begins before a contract exists. A prospective policyholder must encounter a trustworthy distribution channel, learn the terms of available protection, communicate information about the exposure, obtain a quotation, satisfy eligibility requirements, and complete the transaction. Existing customers must subsequently maintain coverage, communicate changes, renew contracts, and initiate claims. Each operation requires access to institutions capable of performing it. Geographic inequalities in the organization of those institutions can therefore alter the practical capacity of populations to transfer risk even where the underlying products are legally available.
An insurance distribution access desert is used here as an operational descriptive term for a locality or population in which a specified insurance transaction cannot ordinarily be completed within a stated tolerance of time, cost, service capacity, accessibility, or reliability. Classification depends on the product, customer population, transaction stage, observation period, and threshold. A remote farming community may have a reliable intermediary for crop insurance but no usable route to commercial liability advice. An urban neighborhood may have many offices but few independent sources of suitable quotations in a language understood by residents. The definition is conditional because the insurance market supplies differentiated contracts through channels with heterogeneous capabilities.
Three distinct objects govern the investigation. Distribution access concerns the ability to reach and use a competent sales or service channel. Underwriting availability concerns the existence of a carrier willing and legally able to assume the specified risk. Realized protection concerns the enforceable contract actually maintained by the customer, including its insured perils, benefit limits, exclusions, deductibles, and duration. Income, trust, and contract quality connect these objects but do not make them interchangeable. The Insurance Distribution and Intermediation Science supplies the parent disciplinary setting.
The research asks where qualified services are located relative to eligible customers; how geographic scarcity interacts with risk, resources, and infrastructure; which outcome measures reveal access barriers; and how interventions might be evaluated without mistaking endogenous market selection for a causal effect. Its scope covers voluntary and compulsory private insurance distribution, while recognizing that public or group arrangements can supply alternative enrollment and service mechanisms. Health-care provider shortages are methodological analogues, rather than the object of this inquiry. Hazard-driven withdrawal of risk carriers is addressed as a competing mechanism rather than the principal research subject.
The method is a purposive, structured review, with evidence selected for its relevance to a particular inferential task. International standards identify regulated functions; official administrative material establishes the existence and scope of registries or market observations; randomized studies identify causal mechanisms within their study populations; methodological literature guides prospective empirical designs. Published abstracts are used only for what they expressly establish. Global statistics describe the relevant background infrastructure. No systematic-review search protocol, exhaustive census of insurance distributors, direct interview study, or original empirical estimation is claimed. The evidence cutoff is November 19, 2026; legally operative local rules must be assessed for the dates of any subsequent empirical application.
Spatial Distribution Channel Inventory
The functional unit of distribution
The elementary observational unit is a service capability at a location or through a remotely usable channel, available to a defined population and product class during a defined interval. A legal entity, individual license, branch, contact point, and brand are different entities. A single broker may operate several offices; multiple producers may share one office; one telephone center may serve an entire country; a wholesale intermediary may place insurance without directly serving retail customers. Counting all these entities as interchangeable units exaggerates coverage.
The IAIS intermediary supervision principles identify professional and transparent intermediation as a concern of supervision and expressly recognize intermediaries without direct customer contact. This provides an institutional reason to distinguish a licensed participant in the distribution chain from a public-facing endpoint. The research inventory should separately identify the regulated firm, the natural person where applicable, the service site, and the actual transaction capabilities. It should record legal authority and product permissions as attributes with effective dates rather than permanent properties.
A complete inventory includes insurer-owned offices, independent and exclusive agents, brokers, authorized ancillary distributors, group enrollment interfaces, staffed telephone services, online platforms, and mobile or itinerant services. Mere referrals belong in a separate category. The inventory identifies whether an endpoint performs information provision, advice where required, binding, premium collection, endorsements, renewal, claim notification, and assistance with complaints. Wholesale capacity is recorded as a supply-side connection reachable through a qualified retail intermediary. Geographic availability of wholesale capacity need not imply geographic availability of retail assistance.
Administrative records and operational verification
A public register is the appropriate starting point for verifying legal identities. The UK Financial Conduct Authority's Financial Services Register records current and former authorized entities and individuals. Its documented scope makes status filtering essential. A registration address may establish a legal contact point without establishing an office that a local consumer can visit. An active license likewise says little about hours, available appointment slots, current insurer relationships, or the product classes that an individual accepts.
California provides a more specifically geographic example. Its Department of Insurance agent and broker locator allows consumers to select a line of insurance, language, locality, and distance band. The agency identifies the language information as licensee-reported. Researchers may use this type of interface to design a service discovery audit, but a listing is neither an independently verified language capability nor a guarantee of a quotation.
Construct a longitudinal registry with stable identifiers, source identifiers, physical and legal address types, valid-from and valid-to dates, geocoding confidence, operating schedules, mode of service, product permissions, accessible languages, and last confirmation time. Capture changes as events rather than overwriting prior observations. Distinguish confirmed closure from an unverified record, missing address, relocation, merger, or transition to remote service. A silent deletion from a directory is not sufficient evidence of a branch closure.
Validation can combine official records, archived directories, authorized service audits, and sampled telephone verification. Geocoding should preserve the original locality text, standardized address, match quality, and uncertainty radius. Duplicate records can be resolved by a stable enterprise identifier and service location; shared premises must not be deduplicated when genuinely independent product access remains. Measurement error is most damaging when it varies systematically with remoteness or language.
Geography and network structure
For each product, a distribution graph links customer origins to service endpoints, endpoints to authorized risk carriers, and carriers to products actually offered under the applicable conditions. Multiple offices of a single tied agency may reduce travel burdens without broadening the set of risk carriers. Conversely, a regional broker with few offices may provide broad placement capability through remote consultations.
Measures of choice should be calculated over distinct sources of quote capacity rather than brand names alone. A platform can reproduce the same underwriting offers through several storefronts. Controlled ownership, common delegation, and shared facilities create dependencies relevant to competition and operational resilience. The network inventory should therefore retain parent-company ownership, carrier appointments where observable, and the distinction between independent quotes and quote duplicates.
A useful audit reports the fraction of eligible exposures with a verified route to at least one endpoint for each transaction, the fraction with alternative routes, and the fraction with unresolved observations. This separates gaps in the service system from gaps in the researcher's information. Public reporting can aggregate observations to protect commercial confidentiality and customer locations.
Physical Access, Distance and Travel Barriers
Network travel and opportunity cost
Straight-line distance is a useful preliminary screening measure, but practical reachability depends on transport networks, ferry schedules, border rules, road conditions, accessibility of buildings, opening hours, and travelers' resources. The economically relevant burden includes the round trip, appointment waiting time, transport expenditure, lost earnings, and the chance of an unsuccessful visit. A trip that is short for a motorist may be prohibitive for a person who relies on an infrequent bus or on assistance from another household member.
For location i, eligible product k, transaction s, mode m, and period t, define T(i,j,m,t) as one-way network travel time to endpoint j. Define V(j,k,s,t) as an indicator of verified service capability. The nearest qualified service time is the minimum of T over endpoints for which V equals one and the customer's transport mode is feasible. When no qualified endpoint is observed, record the condition as unavailable or unknown according to the validation record rather than imputing a zero-minute journey.
Define the population-weighted share beyond a travel threshold H as the sum of eligible exposures in cells with a confirmed minimum time exceeding H, divided by the total eligible exposures with a known classification. Report the excluded or uncertain population alongside this proportion. The threshold must be declared beforehand or justified against the transaction, transport system, and local service expectations. Thirty or sixty minutes may be useful sensitivity values; neither is a universal standard of insurance fairness.
Travel time should be estimated across actual road or transit networks with time-of-day and seasonal assumptions. A disaster scenario requires an alternative disrupted network. The same premises may be reachable during the dry season and inaccessible when road links are flooded. Residents who cannot use a private car require a separate mode-specific assessment. For a multi-stage transaction, a successful first visit may be followed by repeated trips; measure total expected burden and completion probability as well as initial proximity.
Catchment areas and competition for capacity
The nearest endpoint measure treats an office as an unlimited resource. In practice an office may have only a small number of staff-hours, and complex commercial or personal advice consumes more time than a routine endorsement. A capacity-adjusted catchment analysis distributes service capacity across the population that competes for it.
Luo and Wang's 2003 spatial accessibility study compares floating catchment and gravity approaches using primary care in the Chicago region and studies sensitivity to distance assumptions. Its service geography is a transferable methodological precedent, not insurance-market evidence. In the standard two-step floating catchment calculation, the first step forms a service-to-population ratio for each facility over the set of origins within its threshold. The second step sums the ratios of all facilities reachable from each origin. For insurance, capacity must be measured by relevant transactions or qualified staff-hours per period and demand by eligible customer needs, rather than by all residents indiscriminately.
For a simple threshold weight w(i,j) equal to one within the feasible travel catchment and zero outside it, let S(j) be monthly service capacity and D(i) eligible exposures in an origin. Then R(j) = S(j) / sum_i[D(i) w(i,j)] for endpoints with positive demand. The accessibility score A(i) = sum_j[R(j) w(i,j)] expresses allocated nominal service capacity per eligible exposure. It is a potential score, not predicted actual visits or a causal demand model. Where distance decay is justified, use consistent weights in both steps. Allow heterogeneous capacities by transaction and product, and test whether the same staff-hours have inadvertently been allocated several times.
Consider an illustrative example with two settlements A and B containing 600 and 400 eligible households. A has an office capable of 60 relevant consultations per month, reachable only from A; a second office can provide 40 consultations and is reachable from both settlements. The first ratio is 60/600 = 0.10 and the second 40/1,000 = 0.04. The accessibility scores are 0.14 for A and 0.04 for B. The weighted mean is (600 × 0.14 + 400 × 0.04)/1,000 = 0.10 consultations per household per month, matching total capacity of 100 divided by 1,000. This conservation check follows from using the same population and catchment weights in both steps. Aggregate density conceals a 3.5-fold difference in potential allocated capacity. These figures are constructed examples, not observed distribution data.
The calculation also demonstrates a planning tradeoff. Adding capacity to the shared office benefits both settlements, whereas adding capacity to the first improves A alone. If total monthly service demand is below the shared office's capacity, competitive congestion may be less relevant than nominal ratios suggest. If demand clusters around renewal deadlines, annual capacity can conceal severe monthly scarcity. Field validation must measure wait times, abandonment, and seasonal variation before selecting a capacity expansion.
Robust classification
Publish maps of travel times and capacity without hard classifications whenever possible. Binary classifications help target interventions but can be unstable near thresholds. Report sensitivity across plausible travel tolerances, geocoding variants, capacity scenarios, and alternative population grids. Municipal averages and unweighted area shares are especially misleading in sparsely populated regions; a population-weighted result and the unserved land area answer different questions. Neighborhood-level data in cities may reveal inaccessible services behind administrative boundaries even where a metropolitan average appears favorable.
A service desert classification should identify the affected stage, product, population, and evidence confidence. Isolated sales offices without claim support, places with severe transport constraints, and densely populated districts with appointment congestion can all be identified through this approach without pretending they require the same remedy.
Digital Connectivity and Assisted Service
From network presence to transaction completion
Digital channels change the geography of distribution by reducing some travel requirements and allowing centralized expertise to reach dispersed customers. Their relevant reach is functional rather than simply territorial. A customer needs an adequate device or accessible terminal, reliable connectivity or voice service, sufficient language and literacy support, legally accepted identity verification, a usable payment mechanism, and access to human assistance where the transaction requires it. Completion depends on the intersection of these conditions.
The World Bank's Global Findex data portal documents indicators for device ownership, internet use, digital safety, and financial services, with rural and urban and other demographic breakdowns. Its survey information is useful for identifying groups likely to experience digital barriers. It does not establish that those groups can access a specific insurance product or successfully claim benefits. The International Telecommunication Union's Global Connectivity Report reports global estimated internet use in 2025 of 85% among urban residents and 58% among rural residents. The 27-percentage-point difference describes internet use, not a measured difference in insurance enrollment or service availability.
For operational research, a digital route is accessible only when the required stages can be completed by the intended population. A publicly reachable quotation website may fail at address validation; a mobile application may be unusable with a shared handset; payment may require a bank account unavailable to the consumer; documents may be inaccessible to a screen reader. A successful first interaction does not establish the ability to maintain coverage or contest a disputed claim. Record abandonment at each stage and distinguish infrastructure failures from eligibility rejections, misunderstandings, and customer choice.
Suppose a hypothetical locality has 1,000 eligible households. Eight hundred can reliably use the mandated device or alternative terminal, and among those, 600 can complete all further digital requirements for an insurance transaction. The verified digital completion capability is then 60%, not 80%. If 250 households can complete a physical route and 150 of them are also digitally capable, the union is 600 + 250 − 150 = 700 households, or 70%. The remaining 300 illustrate why simply adding the two channel coverage percentages would double-count overlap. These are stipulated capacities; observed uptake, actual quotes, and protection require separate data.
Assisted distribution and institutional trust
Assistance can be provided through a call center, a trained local representative, a community access point with private facilities, or a periodically visiting specialist. The service may create genuine access for consumers unable to complete a transaction independently. It also creates an accountability chain. Researchers should document who is allowed to advise, who confirms consent, who holds customer data, who can bind coverage, and which entity owns complaint resolution. An assistance point that forwards a customer to a nonresponsive agent contributes little to effective reach.
The 2018 IAIS application paper, summarized by the Financial Stability Board's digital inclusive insurance guidance, situates digital technology within proportionate insurance supervision. The source supports assessing digital access together with supervisory obligations. It does not establish uniform national mandates or demonstrate that a particular mobile platform has improved inclusion.
Assistance needs careful safeguards. A local representative compensated entirely by sales may favor products that generate higher commissions. A shared kiosk may expose financial or medical data to unauthorized persons. A household member who controls a device may become an unintended gatekeeper. Test privacy, informed choice, product comprehension, and access to independent complaints, in addition to sales conversion. A multilingual operator or accessibility feature has value when it actually supports communication throughout the transaction.
Continuity and common-mode failure
The relationship between digital and physical distribution changes under stress. A remote call center may remain operational after a local disaster, but the affected population may lose power, telecommunications, identification documents, or access to payment channels. Local offices may be damaged precisely when customers need to report claims. A second digital brand using the same telecommunications or identity service does not provide full redundancy.
Assess continuity through scenarios: local branch closure, network outage, power loss, regional displacement, or widespread claim demand. Track the fraction of customers who retain a usable communication route and the maximum interruption of essential services. Prearranged offline documentation, alternative callback mechanisms, and staffed escalation can contribute to resilience where compatible with contract terms and local law. The performance criterion is the continuity of the customer's ability to exercise contractual rights.
Risk, Income and Rural-Urban Exposure Confounding
Geography as a correlated economic environment
Distribution geography coexists with differences in income, asset ownership, commercial activity, risk exposure, infrastructure, regulation, and public safety nets. Insurers and intermediaries select markets partly according to expected revenue, acquisition and service costs, product suitability, and underwriting appetite. Consumers select residences, occupations, and transport arrangements under constraints. A correlation between producer scarcity and low insurance coverage therefore admits several mechanisms.
A causal diagram can organize these relationships. Pre-existing population density, income, transport infrastructure, and risk characteristics influence both channel location and insurance outcomes. Carrier appetite affects commissions and quote availability and may influence whether intermediaries remain active. Prior insurance demand shapes anticipated profitability and therefore branch decisions. A branch opening may improve information, search costs, trust, and quote completion, which are mediating pathways to coverage. Controlling for realized quote access when estimating the total effect of an opening would block part of the effect of interest. Conversely, omitting pre-existing carrier restrictions may confound the result.
Risk is particularly important for property and agricultural insurance. The Federal Insurance Office's January 2025 homeowners market findings cover United States ZIP-Code-level policy observations for 2018–2022. Treasury reports that average premiums and nonrenewal rates were materially higher in areas of highest expected climate-related losses. The comparison demonstrates variation in the market's price and underwriting dimensions. It does not measure the number of nearby brokers, the marginal effect of broker proximity, or conditions in all countries.
A risk-adjusted analysis compares equivalent insured exposures and contract terms across geographic settings. Matching should include construction or asset type, valuation, coverage limit, deductible, insured perils, prior losses where lawful and relevant, and the applicable regulatory period. For life or health products, eligibility, age, and legally permitted classification need product-specific treatment. Household income should be measured separately from the insured asset's value; high replacement costs do not imply high disposable resources.
Household resources and selection
Insurance demand depends on expected benefits, preferences, perceived contract credibility, current liquidity, and alternative methods of managing risk. Cole et al. (2013), in randomized experiments reported in Barriers to Household Risk Management, find that rainfall insurance demand in rural India responds to price and identify evidence consistent with trust, liquidity, and salience barriers. These findings justify measuring nonprice frictions in addition to geographic distance. They do not establish a universal elasticity or a causal effect of office location.
An important empirical distinction is inability to pay now versus unwillingness to purchase at an actuarially reasonable annual price. A farm household might have adequate expected annual revenue but little cash at planting time. A household may rationally prefer self-insurance for small losses while desiring catastrophic protection. Insurance literacy and prior experiences affect whether the product's expected benefits are perceived accurately. Observed nonpurchase is therefore an outcome requiring diagnosis, not direct evidence of involuntary exclusion.
Comparisons should stratify by eligibility, income band, asset exposure, rurality, and product class. Urban-rural classifications differ across national statistical systems; retain each country's original definition and use harmonized density or travel-to-center measures as sensitivity analyses. Define low income relative to meaningful local thresholds and report whether resources are measured at individual or household level. Avoid inferring individual poverty from the mean income of a census area.
Distinguishing scarcity from economic viability
A simplified intermediary location decision compares expected contribution from sales and servicing with fixed, mobility, staffing, compliance, and technology costs. Low-density markets can have smaller customer volumes and high unit service costs. Aggregation through regional service delivery may reduce fixed costs while imposing new customer burdens. Commissions may also decline when insurers restrict products in a high-risk area. The distribution network and carrier network are thus economically linked.
The relevant policy question is which cost or restriction prevents access. If qualified services already exist but carriers uniformly reject a risk, subsidizing another sales office may produce more unsuccessful applications. If carriers will write coverage but prospective customers cannot obtain a quotation, a shared regional advisor or assisted channel may solve the binding constraint. If premiums remain unaffordable after improving access, the relevant response involves the economics of coverage or household resources. These are conditional implications of conventional insurance economics, not a new theory of market organization.
Cross-National Geographic Case Studies
United States: producer discovery and underwriting availability
The United States offers geographically detailed observations and state-level institutional variation. The Treasury's 2025 homeowners release aggregated data from more than 330 insurers over the 2018–2022 period. Its published subset has minimum insurer and policy counts for ZIP Code disclosure. Consequently, absent public cells cannot automatically be classified as unserved areas. The confidentiality rule can be consequential precisely where market participation is sparse.
The California service inventory example complements the federal premium and nonrenewal data. A researcher would construct matched local snapshots of qualified agents and brokers, actual quotes, and the contemporaneous carrier appetite for comparable properties. Historical service records are necessary to align with the federal observation window: a directory observed in 2026 cannot establish producer proximity in 2018. Event-level linkage could help distinguish intermediary closures from carrier underwriting changes and test whether remaining agents become congested.
An important design feature is the geography of the eligible exposure. Postal territories are administrative units used for mail delivery. Customer residential locations, insured property locations, and producer service addresses may differ, especially for commercial insurance and rental property. The study must choose the geographic point appropriate to the mechanism being evaluated and preserve uncertainty where only an aggregate is available.
European Union and United Kingdom: registration and reach
The European Union's insurance distribution framework provides a common institutional context but national markets and supervisory systems differ. EIOPA's March 2026 third IDD report announcement describes the continued decline in registered intermediaries alongside a 12% increase in EU-passported intermediaries between 2020 and 2024. It identifies consolidation and professional workforce changes among reasons for the decline and discusses persisting shortcomings in service quality. A decrease in registrations can therefore coexist with expanded cross-border authorization. Neither measure identifies local service hours, accessibility, or quotes actually available to residents.
Study the relationship between market consolidation and local reach by distinguishing ownership changes from disappearance of consumer-facing premises. Record cross-border permissions separately from actively marketed product availability. Customer language, country-specific product law, tax arrangements, applicable consumer protections, and mechanisms for handling complaints influence functional access. A cross-border authorization is one necessary input where relevant, rather than a complete measure of service supply.
The UK is treated separately from the EU reporting population. Its FCA register supplies a source for verification of authorization, including former participants. It provides an institutional basis for de-duplicating regulated firms and checking permissions, in conjunction with actual service addresses and operational observations. Do not substitute EU passport counts for permissions in the UK. Comparative work must identify the relevant legal framework, national register definition, and observation period instead of imposing a single transnational license concept.
India, China and Ghana: trusted information and rural demand
Rural index insurance experiments in India, China, and Ghana demonstrate that physical proximity operates within institutional and social settings. Cole et al. (2013) identify price and other demand barriers. Cai, de Janvry, and Sadoulet (2015), in Social Networks and the Decision to Insure, use randomized information interventions in rural China and report that the insurance adoption influence of treated friends operates through diffusion of knowledge. Their evidence supports investigating interpersonal information channels and exposure to explanations of products. It should not be interpreted as geographic mapping of brokers.
The Ghana experiments discussed by Karlan et al. demonstrate the connection between risk-transfer opportunity, trust, and productive activity within the study's settings. The experiments also reinforce the importance of index insurance quality: the relationship between the index payout and loss experienced by a farm affects the protection obtained. A service access point that sells a poorly matched product may increase transactions while leaving meaningful exposure unaddressed.
Across these experiments, researchers should distinguish the effect of a randomly assigned information, price, or insurance treatment from the independent effect of intermediary location. The relevant unit of assignment and the mechanism differ across studies. They support hypotheses about liquidity, trust, information diffusion, and investment incentives. They do not establish a pooled geographic elasticity or the relative prevalence of distribution deserts across these countries.
Comparative implications and intervention choice
The institutional examples illuminate different measurement layers: California demonstrates discovery of licensed distributors; EIOPA's reporting describes market organization and cross-border permission; rural field experiments establish demand-side and informational mechanisms; global connectivity statistics characterize a background constraint. Each provides a different kind of evidence. Combining them into a numerical cross-country ranking of access deserts would be methodologically unsound without harmonized product-specific inventories and exposure denominators.
An intervention can be matched to the observed failure stage. Where travel burdens dominate, scheduled outreach, regional service locations, or accessible transport may help. Where capacity is insufficient, staffing or pooled qualified services may reduce waiting. Where trusted information is scarce, appropriately supervised assistance and understandable disclosures can improve informed engagement. Where underlying carrier appetite is absent, market capacity and risk financing require separate assessment. Where connection instability is the relevant barrier, offline and low-bandwidth options can be tested for reliable completion. These measures should be compared on their effect on qualified quote access, suitability, retention, and net service cost rather than assumed beneficial solely because they expand the number of outlets.
Spatial Causal Designs and Measurement Uncertainty
Data architecture and estimands
A reproducible empirical project begins with a versioned population and exposure frame. For each product and transaction period, preserve customer origin, exposure type, eligibility, geography definitions, historical distribution sites, actual channel capability, carrier availability, price and coverage attributes, contact and abandonment events, and relevant pre-treatment covariates. Retain definitions, coverage of participating institutions, collection errors, and privacy constraints in a data dictionary.
The estimand must correspond to a specific intervention. One example is the average effect on eligible households of losing a local branch that previously provided a specified service, compared with the counterfactual in which that branch remained open. A separate estimand is the effect of offering assisted digital access to households with a specified observable deficiency. The outcome might be probability of obtaining an equivalent quotation within fourteen days, not premium among customers who successfully purchase. Outcomes conditional on purchase have a treatment-sensitive selection mechanism and require separate interpretation.
Branch closures can arise from low projected demand, insurer withdrawals, changing demographics, or disasters. Their timing is therefore potentially endogenous. Researchers should document why the particular branch ceased operations and whether service migrated to another channel. An event cannot be interpreted as an exit from geographic service simply because a business registry changed ownership.
Difference-in-differences and interference
Where credible comparison locations exist, a difference-in-differences design can compare changes in outcomes among exposed and unexposed locations. The simple two-group, two-period estimator subtracts the comparison group's change from the treated group's change. If quote success in a treated locality falls from 60% to 45% after closure while it falls from 55% to 50% in a comparison area, the illustrative estimate is (45 − 60) − (50 − 55) = −10 percentage points. Under an appropriate parallel-trends assumption, no differential simultaneous shock, and stable measurement, the contrast can estimate a causal effect. These invented figures are not an empirical conclusion.
Staggered timing complicates estimation. Callaway and Sant'Anna's multiple-period difference-in-differences, subsequently published in 2021, provides an approach organized around treatment timing and relevant comparison groups. A prospective insurance study can define cohort-specific effects, use suitable never-treated or not-yet-treated controls as justified by the chosen method, and aggregate the effects transparently. The archived working-paper version and journal bibliographic record differ in publication format; this monograph draws only on their shared methodological scope, not on a replication of the estimator.
Parallel trends concern unobserved counterfactual outcomes, so favorable pre-event patterns do not prove the assumption. Anticipation matters if clients move before a closure. Changing area population can produce apparent differences unrelated to the experience of a fixed cohort. Include a pre-event event study, attrition analysis, checks for concurrent changes in underwriting rules or major employers, and sensitivity to alternative comparison definitions. Uncertainty intervals should reflect the level at which treatment varies and relevant spatial dependence.
Spillovers present a substantive challenge: the closure may redirect customers into neighboring service areas, increasing wait times among nominal controls. The research design should estimate exposure to the full channel network, define direct and indirect treatment groups, and test buffer distances or network-based spillover definitions. The absence of observed spillovers is not a default assumption when facilities have overlapping catchments. Spatial panel models can describe residual geographic structure, but an autoregressive term by itself does not establish exogeneity.
Alternative designs and uncertainty
Random assignment of outreach availability or assisted access, where ethically and operationally feasible, provides a direct way to estimate intervention effects. Cluster randomization can reduce contamination while capturing community-level information diffusion; sufficient independent clusters are necessary for valid precision. Transparent allocation and equal baseline entitlements should be maintained. Report intention-to-treat effects and distinguish them from estimates for actual users, whose participation may be selective.
A regulatory change or eligibility boundary may support a quasi-experimental design if it changes distribution access independently of other determinants of outcomes. A geographic boundary that simultaneously changes compulsory coverage rules, rates, or hazard classifications cannot be treated as an instrument for distribution alone without defending the relevant exclusion assumptions. An instrumental-variable estimate, where justified, pertains to the population whose exposure changes under the instrument and should not be generalized automatically.
Measurement uncertainty includes geocoding error, outdated registries, address mismatches, unknown hours, unobserved capacity, imprecise transport times, and disclosure suppression. Publish results across alternative inventory verification rules and thresholds; compare household-weighted estimates to exposure-weighted estimates, and preserve uncertainty bounds for missing classifications. For example, if 80 of 900 classified households are confirmed beyond a service threshold and 100 of 1,000 eligible households are unknown, the classified estimate is 8.9%. The all-population worst-case bounds are 8% to 18%, depending on the unknown group. Missingness may be selective and these bounds make no missing-at-random assumption.
Spatial dependence should inform variance estimation, sample design, and interpretation. Small localities may have unstable rates; suppression and controlled aggregation protect privacy but reduce resolution. Time alignment across datasets is crucial. Geographic boundaries should be frozen to a version or crosswalked with documented weights, and the results should be tested under alternative plausible aggregation schemes. All code, source versions, and exclusions used for a subsequent quantitative investigation should be archived with appropriate confidentiality restrictions.
Scientific and operational evaluation
A distribution-access intervention should be judged against a declared baseline and an independent counterfactual when possible. Suitable endpoints include the time and cost of contacting a qualified intermediary, probability of obtaining an appropriate quote, breadth of independent carrier options, rate of binding and renewal, and reliability of essential post-sale services. Consumer comprehension, choice, complaint resolution, and protection adequacy should accompany measures of acquisition efficiency.
The full evaluation records incremental operating cost, consumer travel or assistance cost, capacity utilization, effects on neighboring service points, and resilience in disruption scenarios. A centralized digital solution may be efficient for certain transactions while increasing the burden on customers who need assisted completion. A local office may improve continuity and trust but have low utilization. A comparative assessment should identify who benefits, which customers remain unserved, and whether a less costly intervention achieves a similar quality-adjusted result.
This monograph's original contribution consists of linking the distinct measurement layers across the insurance contract lifecycle and specifying conditions under which geographic research can distinguish distribution scarcity from underwriting and income constraints. Its numerical illustrations have explicit assumptions and are independently reproducible from the text. The actual magnitude of distribution deserts, causal impacts of branch closures, and cost-effectiveness of particular interventions require appropriately linked local data. The absence of these estimates is a boundary of the review design, not a claim that the phenomena are absent.
Conclusion
Insurance distribution has a measurable geography because the services through which risk transfer is arranged and maintained have finite capacities, distinct functions, legal permissions, and requirements for customer participation. The location of an intermediary is analytically significant when that intermediary can actually deliver the transaction required by an eligible customer. Digital access changes travel geography but remains conditioned by devices, connectivity, identification, payments, accessibility, and assistance.
The evidence establishes the importance of disaggregating registered market participants, functional endpoints, carrier willingness to insure, household affordability, and protection in force. California's discovery systems illustrate the difference between licensure and actual underwriting availability. European supervisory reporting demonstrates that consolidation and expansion of cross-border authorization can occur together. United States homeowners insurance statistics illustrate the independent role of geographic risk in affordability and nonrenewal. Rural insurance experiments demonstrate that trust, information, risk transfer, and liquidity influence decisions where insurance products are offered. Global connectivity evidence makes the assumption of universally usable digital channels untenable.
The appropriate unit for measurement is a defined service capability relative to eligible exposures and a particular transaction in a declared place and time. Network travel measures, capacity-adjusted catchments, and digital completion audits reveal different dimensions of potential access. Quote and policy records illuminate realized access, while coverage terms and claims outcomes describe substantive protection. Population-weighted distributions, unknown-case bounds, and sensitivity analyses should accompany threshold-based classifications.
Causal research must address endogenous channel location, spatial interference, concurrent risk and economic conditions, and changes in the composition of the observed population. Randomized interventions and appropriately identified longitudinal designs allow stronger inference than comparisons of producer density with insurance participation. Their conclusions remain bounded by the studied contracts, institutions, and populations.
The institutional implication is that improving geographic insurance access consists of finding and resolving the specific impediment to a suitable, enforceable, and serviceable contract. Distribution expansion is valuable when it increases substantive access without compromising conduct, consumer rights, or contract quality. The geographical study of insurance must ultimately connect market organization with the lived capacity of exposed populations to acquire and maintain risk protection.
References
Cai, Jing, Alain de Janvry, and Elisabeth Sadoulet. 2015. Social Networks and the Decision to Insure. American Economic Journal: Applied Economics 7 (2): 81–108. DOI: 10.1257/app.20130442.
California Department of Insurance. n.d. Find an Agent or Broker. Official consumer service guidance. Accessed November 19, 2026.
California Department of Insurance. n.d. Home Insurance Finder. Official consumer search interface and disclaimer. Accessed November 19, 2026.
Callaway, Brantly, and Pedro H. C. Sant'Anna. 2021. Difference-in-Differences with Multiple Time Periods. Journal of Econometrics 225 (2): 200–230. DOI: 10.1016/j.jeconom.2020.12.001. Author working-paper record.
Cole, Shawn, Xavier Giné, Jeremy Tobacman, Petia Topalova, Robert Townsend, and James Vickery. 2013. Barriers to Household Risk Management: Evidence from India. American Economic Journal: Applied Economics 5 (1): 104–135. DOI: 10.1257/app.5.1.104.
European Insurance and Occupational Pensions Authority (EIOPA). 2026. EIOPA Publishes Third Report on the Application of the Insurance Distribution Directive. Official report announcement, March 30.
Financial Conduct Authority (FCA). 2026. Financial Services Register. Official register guidance, updated February 13.
International Association of Insurance Supervisors (IAIS). 2018. Application Paper on the Use of Digital Technology in Inclusive Insurance. November 12. Financial Stability Board official compendium summary.
International Association of Insurance Supervisors (IAIS). n.d. ICP and ComFrame Online Tool. ICP 18, Intermediaries. Accessed November 19, 2026.
International Telecommunication Union (ITU). 2025. Global Connectivity Report 2025. Chapter: Connectivity divides. Geneva: ITU.
Karlan, Dean, Robert Osei, Isaac Osei-Akoto, and Christopher Udry. 2014. Agricultural Decisions after Relaxing Credit and Risk Constraints. Quarterly Journal of Economics 129 (2): 597–652. DOI: 10.1093/qje/qju002. University of Ghana repository record.
Luo, Wei, and Fahui Wang. 2003. Measures of Spatial Accessibility to Health Care in a GIS Environment: Synthesis and a Case Study in the Chicago Region. Environment and Planning B: Planning and Design 30 (6): 865–884. DOI: 10.1068/b29120. Bibliographic and abstract record.
OECD. 2026. Global Insurance Market Trends 2026. Paris: OECD Publishing. DOI: 10.1787/c8be6585-en.
U.S. Department of the Treasury. 2025. U.S. Department of the Treasury Report: Homeowners Insurance Costs Rising, Availability Declining as Climate-Related Events Take Their Toll. Official Federal Insurance Office report announcement, January 16.
World Bank. 2025. The Global Findex Database 2025. Download Data and Indicator Coverage. Washington, DC: World Bank.