Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Cyber Insurance Protection Gaps for Small and Medium Enterprises

Abstract. Small and medium enterprises (SMEs) sustain digital losses whose financial incidence depends on security practices, access to insurance, contractual terms, claim settlement, and the capacity to finance recovery. This Research Dossier studies the gap between cyber-related economic loss and effective insurance protection, with particular attention to firms that have limited internal technical expertise, cash reserves, negotiating power, or capacity to document incidents. It integrates comparative official surveys, supervisory market reporting, original policy specimens, cybersecurity guidance, legal instruments, and empirical research concerning insurance adoption. The analysis separates a gap in enrollment from a gap in contractual adequacy, a gap in realized indemnity, and a gap in recovery finance. It shows why premium growth or counts of policies cannot establish a reduction in uninsured economic losses without compatible exposure and claim denominators. It analyzes risk perception, budget pressure, selection, minimum security controls, distribution costs, policy conditions, aggregation, and recovery liquidity. Reproducible hypothetical calculations illustrate the consequences of alternative limits, waiting periods, eligibility, dependence, and settlement timing. The central conclusion is that useful SME protection requires commercially sustainable access to verifiable risk transfer and timely incident response, alongside prevention and retained risk finance. The evidence supports institutional and economic mechanisms; it does not identify a global monetary protection gap or establish causal treatment effects of insurance. The contribution is an original analytical synthesis of established insurance, security, and economic concepts, without a new named theory.

Keywords: Cyber insurance; small and medium enterprises; insurance protection gap; underinsurance; cyber risk; insurance demand; underwriting; risk selection; cybersecurity; deductibles; cyber business interruption; incident response; liquidity; systemic risk; loss measurement; insurance regulation.

Introduction

The small enterprise is a distinctive unit of cyber risk because its digital operations, contractual responsibilities, and financial resources may have radically different scales. A firm with twenty employees can process sensitive records belonging to thousands of individuals. A workshop can depend on an external scheduling or design service for its entire productive capacity. A wholesaler can lose the ability to take orders while physical stock remains undamaged. Enterprise size is therefore an imperfect measure of the value exposed to a digital failure.

The research question is why the financing of cyber losses remains incomplete among SMEs, even where a private cyber insurance market exists. The answer has several connected parts. Insurance requires a product the enterprise can identify and obtain, a premium and control package the enterprise can afford, a contract responsive to relevant losses, and a claim process capable of delivering recovery resources when needed. These requirements depend on the organization's security state, its economic exposure, insurance supply, and institutional environment. A firm can satisfy one requirement while failing another.

This study treats the protection gap as a problem of allocation. Cyber incidents create economic damage, some of which can be reduced by technical and organizational measures. The remaining loss may be contractually transferred, absorbed by the company, recovered from another party, or ultimately passed to creditors, employees, customers, and public institutions. The allocation of financial consequences deserves analysis independently of the operational cause. Prevention changes the loss process; insurance redistributes specified consequences of that process; financing changes the timing and bearer of available funds.

The scientific object is the SME's cyber-specific protection gap, not all cyber insurance markets or all forms of SME underinsurance. The parent Science supplies the broader treatment of cyber perils, actuarial pricing, and solvency. Here those questions enter through their consequences for protection among smaller commercial firms. The inquiry asks which observations can measure underinsurance, how insurance access depends on risk and enterprise resources, why purchased cover can leave substantial residual losses, and what institutional responses might close avoidable gaps without concealing aggregate risk.

Research Design and Definitions

Population and scope

For the analytical examples, an SME is an independent or smaller commercially operating enterprise, including microenterprises where the source explicitly includes them. Empirical comparisons retain the population and size classification of each source. Employee counts are useful strata, but revenue, balance sheet, group ownership, sector, and market jurisdiction may determine official classifications. In particular, the European legal scope of regulated entities uses a specific size-and-sector test rather than an informal label; the distinction is examined below. A retail branch of a multinational group is not analytically interchangeable with a legally independent retailer employing the same number of people.

A cyber event is an occurrence involving the availability, confidentiality, or integrity of digital systems or information, including malicious and accidental events where relevant. A security event may be detected and contained without material financial loss. An incident may cause loss without an insurance claim. A claim may be accepted, denied, partially paid, or remain unsettled. These stages define different observation sets. An attack-count database measures neither insured exposure nor business failure.

Economic loss is the monetized consequence falling within a defined accounting boundary and period: restoration expenses, lost contribution margin, additional operating expense, legally attributable liability, or other specified costs. Some harm, such as lost trust, is difficult to value. Analysts should identify which costs are omitted rather than assign them unsupported amounts. Insurance indemnity is the contractual payment for a covered loss under the applicable law and terms. Incident services provided in kind are analytically separate unless valued consistently and included in a stated measure of recovery.

Method and evidence hierarchy

This is a structured narrative and comparative documentary investigation, not an exhaustive systematic review or a new survey. Source selection prioritized directly relevant original surveys, actual public contract forms, legal texts, official supervision, and original peer-reviewed empirical research. The principal UK sources are the two national survey waves and their questionnaire, supplemented by a commissioned mixed-method study with explicitly limited representativeness. The United States contributes market reporting, oversight findings, and historical regulatory filing research. Germany contributes observational evidence on demand; Singapore provides a dated insurer-sponsored survey release. European Union law and a New York supervisory circular illustrate different legal mechanisms. A comparative OECD report supplies context rather than independent confirmation of each study it quotes.

Evidence was collected from retrievable publication pages and extracted PDF text on November 18, 2026. The commissioned UK report's cover and executive summary were also viewed as page images. Some other PDF page-image requests failed despite accessible extracted text. The evidence ledger specifies access scope. Abstract-only material supports only its stated findings. No inaccessible full paper, unpublished proprietary claims dataset, or unidentified policy schedule is presented as examined. Figures are rounded as in the original source; derived numerical examples are explicitly artificial. Comparisons are analytical rather than pooled estimates.

Adriko and Nurse's systematic review of SME cyber insurance (2024) provides a focused entry point: the authors retained 19 papers after screening an initial 451 and identified comprehension and expertise barriers in their abstract. That limited literature base motivates caution when attributing effects to insurance ownership alone. This dossier extends the analytical scope to payment adequacy and financing without claiming that its own search reproduces their systematic procedure.

The Economic Anatomy of the Protection Gap

Enrollment and actual economic loss

Let I equal one if a firm holds an identified policy responsive in principle to some cyber events during the observation period, and zero otherwise. A count-based enrollment rate is the sum of I divided by the number of enterprises in the relevant population. Its complement expresses absence of identified coverage only if status is observed for all firms. An unknown response to a survey is its own category. It cannot be silently counted as either insured or uninsured.

For an event affecting enterprise i, let L_i be the gross attributable economic loss within a consistently defined scope, P_i the ultimate indemnity from all relevant insurance contracts after coordination, and R_i the noninsurance recovery from other parties. For nonnegative losses and nonoverlapping recoveries, the insurance-only gap in money is L_i - P_i, with 0 <= P_i <= L_i. The retained net event burden is L_i - P_i - R_i, provided reimbursements do not double count the same item. Borrowing is financing, not a recovery: the principal remains an obligation. A capital injection similarly changes financial capacity without erasing economic loss.

Across a matched portfolio or population, the ratio of aggregate insurance-only gap to aggregate economic loss is 1 - sum(P_i)/sum(L_i), if the denominator is positive. This measure weights firms by realized loss. An average of firm-specific coverage percentages weights firms differently and is undefined for a firm with no realized loss. A prospective expected-loss measure uses expectations over a specified horizon and exposure distribution. It should not be equated with a realized historical ratio, particularly for a heavy-tailed peril.

Consider two businesses. One with no policy experiences a loss of 5,000 monetary units; another with insurance experiences a loss of 100,000 and receives 60,000. Enrollment is 50%, aggregate economic loss is 105,000, and indemnity is 60,000. The aggregate insurance-only gap is 45,000, or approximately 42.86%. In a different year, the same enrollment configuration may coincide with much larger uninsured losses. Policy ownership remains unchanged while loss-weighted protection changes. This illustration shows why a count cannot be converted into a monetary gap by arithmetic alone.

Distinguishing the sources of retained loss

For a firm with insurance, define C as the portion of economic loss admitted within the contract's covered categories before applying monetary retention and limits. Under a simple indemnity contract, the gap can be written as (L-C)+(C-P). The first term represents excluded or otherwise non-admitted loss; the second represents contractual retention, limit effects, and other reductions already reflected in P. When claims are disputed, separate probable ultimate indemnity from amounts paid to date. The decomposition is an accounting identity with nested definitions, not a proposed causal model.

This distinction has implications for measurement. Low enrollment suggests a possible distribution or affordability problem; a large L-C suggests mismatched risk and wording; a large C-P can arise from deliberate retention, inadequate limits, or remaining settlement uncertainty. Different interventions are appropriate for each. Reducing a deductible will not insure an excluded provider failure. Selling a policy will not necessarily prevent a firm from having an uncovered catastrophic loss.

A deliberate retained loss can be efficient. Predictable low-level restoration cost may be cheaper to finance internally than to transfer at a premium that includes underwriting and claims expenses. The policy objective is therefore not universal reimbursement of every digital expense. The more demanding question is whether firms possess an informed, credible way to finance the residual consequences of plausible severe events, including loss of income and third-party obligations.

Exposure and the Small Enterprise Balance Sheet

Operational concentration and fixed recovery costs

Digital concentration can arise from a single payment system, an outsourced email tenant, one identity provider, a booking platform, or a managed service provider. A firm using these services may gain expertise and reliability at modest cost. The same arrangement can expose the firm to a common external failure, and responsibility for maintaining backups, access rights, and incident records may be distributed across contracts.

The economic impact depends on the business process under interruption. For a store accepting alternative payment methods, one payment outage may result in inconvenience and some deferred transactions; for an online-only retailer, it can stop nearly all current receipts. For a professional service firm with recurring fees, a short outage may mainly delay work. For a manufacturer with a tightly scheduled production run, a similar outage may interrupt delivery and create additional contract costs. Sector and dependency are consequently more informative than headcount by itself.

Incident response often has a fixed component: identifying affected systems, retaining forensic expertise, assessing notification duties, communicating with counterparties, and rebuilding trustworthy access. A comparatively small technical event can therefore impose substantial costs relative to an SME's liquid resources. The fixed-cost component also appears on the insurance supply side because a smaller account may require underwriting and claims work similar in type to that required by a larger account.

Useful exposure measures include gross margin per day of critical-process interruption, sensitive record counts, dependent provider concentration, recovery time, legal responsibility for customer data, and the maximum plausible cost of urgent external expertise. These measures should be estimated separately. Employee count provides an approximate organizational scale, but it does not adequately encode them.

Financial loss, cash and continuity

Income loss is measured against a defensible counterfactual. Gross sales foregone are not automatically the economic loss because costs may be avoided, sales may be deferred rather than permanently lost, and additional expenses may enable partial operations. Insurance adjustment uses the definitions and period contained in the contract. Analysts should also distinguish the SME's own loss from lost revenue of another enterprise.

A balance-sheet vulnerability can be expressed through time. Let B(t) denote liquid cash available after ordinary committed obligations plus accessible credit at time t. Let D(t) denote the cumulative additional cash requirement caused by the incident by time t, and P(t) the cumulative insurance payments actually received. A financing shortfall occurs whenever D(t)-P(t) exceeds B(t). This is not a solvency test under accounting law. It is a simple way to represent immediate recovery funding.

An insured company may have a positive expected ultimate indemnity while lacking the cash to restart. Conversely, a company with no policy may successfully self-finance a modest loss. Access to revolving credit can reduce an immediate financing shortfall but can increase leverage and interest costs. The temporal profile of indemnity is thus important in the assessment of underinsurance, especially where the timing of forensic invoices, wages, and restoration payments is earlier than completion of claim adjustment.

Demand, Behavior, and Information

The corporate purchase decision

Under a stylized expected-utility account, risk transfer can be valuable because it replaces a contingent loss with a certain premium and a reduced remaining loss. If W is available wealth, L a random loss, P(L) the promised payment, and pi the premium, an enterprise or its owner compares the expected value of utility at W-L with utility at W-pi-L+P(L). This simplified expression omits taxes, capital-market imperfections, service benefits, transaction costs, and uncertainty about whether the contract will pay. For an SME those omissions are material.

The purchasing decision may be made by an owner who simultaneously faces payroll, lease payments, software renewals, supplier obligations, and demand for new investment. Insurance plus required controls competes with these uses of funds. The economically relevant cost is not just premium: it includes application effort, control acquisition, specialist time, recurring maintenance, and the opportunity cost of retained cash. The buyer may also consider response services, access to expertise, and commercial eligibility as benefits in addition to indemnity.

Buying decisions can be affected by the perception of financial consequences, trust, managerial attention, and product understanding. Salzberger's German SME demand study (2025) uses questionnaire data from 1,248 executives and logistic regression. Financial impact perceptions and anxiety were positively associated with purchase; perceived attack probability and prior experience did not have statistically significant relationships in the reported models. These findings concern an observational German sample. Statistical insignificance does not prove that probability or experience is irrelevant across all firms, and the associations do not establish the effect of an information intervention.

One implication is that communicating potential operational consequences can be more intelligible than giving a small business an unsupported precise attack probability. A plausible cash-flow scenario translates security into the owner's decision domain. Its assumptions should be communicated, and a strong worst-case illustration should not be presented as a typical loss.

Awareness, comprehension and trust

Product awareness, understanding, willingness to buy, ability to obtain quotations, and final contract formation are separate steps. A firm may know that cyber insurance exists but believe that its general liability policy covers all cyber incidents. A firm may obtain quotations and decline them because the responsive clauses are unclear. An owner may expect a claim dispute and assign less value to the contract than the nominal indemnity suggests.

A useful empirical study should describe which of these stages its question actually measures. In the DSIT 2025/2026 breaches survey (2026a), 39% of expressly uninsured business respondents cited lack of awareness and 34% cited budget priority among their reasons. The two percentages refer to a conditional population of 267 businesses, not all SMEs; the multiple-response design permits overlap. As a result, their sum has no interpretation as a unique share of the overall enterprise population.

The corresponding survey questionnaire (DSIT, 2026b, Q23X and Q23Y) provides an important check: coverage status is a mutually exclusive self-report, with an unknown option, while the reasons question is asked of respondents explicitly stating no cover and allows multiple choices. An awareness campaign can improve one step without resolving the cost of control compliance or ensuring adequacy of coverage.

An intermediary can reduce information costs through concrete contract comparisons. It can also produce conflicts if remuneration favors a particular insurer or high premium. The efficacy of advice should therefore be judged by the relationship between the customer's actual exposure and the proposed grants, exclusions, retentions, and services. The expected number of quotes is a process metric, not a final welfare measure.

Underwriting, Eligibility, and Affordability

Information asymmetry and selection

Insurers seek to identify differences in expected covered loss and establish premiums consistent with their expenses, uncertainty, and required capital. The enterprise understands certain features of its business better than a distant underwriter, while the insurer may understand claim wording, portfolio loss patterns, and common technology dependencies better than the buyer. Information asymmetry operates in both directions.

Adverse selection can arise if firms expecting higher loss are more likely to seek cover and those differences are not adequately reflected in pricing. A high-risk firm can also be excluded because the insurer identifies risk accurately. These are different phenomena. A market composed only of stronger risks can be profitable while leaving the most vulnerable firms unprotected. Inference from insured claims to the broader SME population is biased when eligibility and purchase depend on risk.

Historical evidence of underwriting practice is provided by Romanosky et al.'s analysis of insurance filings (2019). Their study reviewed 235 dockets and found rate-derivation explanations in 56. The regulatory filings contained various kinds of contract and application documents. The research demonstrates the value of examining actual rating material and policy forms; it does not furnish a current quoted SME premium or a universal rating equation.

Use of externally observed technical indicators may reduce questionnaire effort, but an external scan cannot reveal everything relevant to internal recovery readiness. A negative externally visible finding may correspond to an intentional safe configuration, an unrelated shared host, or an outdated domain association. Conversely, a favorable score may miss identity compromise or an untested backup. Verification requires clear mapping of the observation to the enterprise and the coverage decision, plus a way to correct errors.

The cost of meeting conditions

A basic control can change both the frequency and severity of loss. For example, effective access controls can reduce opportunities for certain forms of credential misuse; usable backups can reduce restoration time and dependence on a damaged production environment. The achievable effect depends on implementation and event type. Control purchase alone does not prove operational effectiveness.

The NIST small business guide (2024) is expressly intended to assist organizations with modest security planning and supplements the larger Cybersecurity Framework. Its relevance here is the possibility of a structured approach suitable for small organizational resources. It does not certify eligibility for any insurance contract, nor does its publication establish a quantitative loss-reduction coefficient.

The commissioned Insuring Resilience report (Grant Thornton & DSIT, 2025) illustrates underwriting friction. Its survey reached 104 respondents from 9,000 CRM-list invitations, combined with interviews involving insurers, intermediaries, and owners. Respondents identified required controls, high premiums, and complex wording as challenges. The low participation and recruitment method prevent its price or adoption figures from being treated as population estimates. Nevertheless, this evidence helps identify relevant questions for a more representative study: one must observe the initial security state, the marginal control cost, availability of assistance, quote conditions, and subsequent purchase outcome.

Suppose annual premium is 4,000 units, control installation costs 9,000 in the first year and 2,000 in subsequent annual maintenance, and the SME spends 1,000 on initial professional help. The first-year incremental outlay is 14,000; recurring outlay is 6,000 before any new advisory costs. Some control costs would have occurred without the policy, so the correct insurance incremental cost subtracts the independently justified baseline. Failure to subtract it overstates the price of insurance; treating the entire control cost as irrelevant understates the initial liquidity requirement.

Supply and demand interact. Strict entry criteria can lower expected insured losses and support capacity. They can also exclude a vulnerable firm precisely when it needs incident-response resources. A staged route toward eligibility, where supported by actuarial evidence and truthful disclosure, may improve access without representing an unverified risk as acceptable. The evaluation question is whether the approach reduces losses and improves effective protection after including the costs of monitoring and service.

Pricing and product economics

A conceptual annual premium can be decomposed into expected covered loss, acquisition and administration, claims expenses not already included in indemnity, cost of risk capital, reinsurance net cost where appropriate, and profit or uncertainty adjustment. These components must not be double counted: the treatment of reinsurance and expected net loss depends on whether the model starts with gross or ceded figures.

At small policy sizes, fixed expense can represent a comparatively large percentage of premium even if expected absolute loss is low. Standardization, group purchasing, and digital distribution may lower administrative friction. However, reducing the price by replacing meaningful cover with narrow grants may worsen loss-weighted protection. Product comparison must examine effective indemnity per premium for the relevant exposure scenarios rather than equate cheaper premium with improved coverage.

GAO's 2021 cyber market report describes U.S. market changes including higher premiums and lower limits in some high-risk sectors. Its take-up measure among insurance clients is historically useful but does not estimate national SME enrollment. Supply conditions can change over time. The existence of an earlier hard market cannot prove that an SME faces identical terms in November 2026.

Contractual Adequacy and Claims Recovery

Coverage grants, exclusions and the counterfactual claim

A cyber policy is a collection of defined promises rather than a generic guarantee against digital harm. Relevant categories may include incident response, data recovery, business interruption, dependent business interruption, extortion-related expenses, privacy or network liability, and specified fraud. The bound policy, schedule, endorsements, facts, and governing law determine the result. The term cyber insurance is insufficient to classify the treatment of every consequence.

Two original documents illustrate the importance of wording. The 2022 CyberAcuView model wording distinguishes own-system business interruption from dependent business interruption through separate modules; it is explicitly adaptable by insurers. A firm holding one grant cannot infer the other solely from the product label. The sample's waiting-period and proof-of-loss provisions illustrate how temporal thresholds and documentation influence recoverability.

The public RSA Cyber Protection wording provides another example. Its general waiting-period condition treats loss incurred during the period as covered once that threshold has been exceeded, subject to other requirements. This is a franchise-like time threshold rather than necessarily an unrecoverable time deductible. The policy also separately sets terms for suppliers and schedule-specific sublimits. These examples establish textual alternatives, not the market prevalence of either structure. Neither public document is evidence of cover actually purchased by a particular SME.

Losses caused by employee payment deception may fall under crime, social engineering, or another specifically written grant. A fraudster's use of digital communication alone does not establish that a computer-security definition has been satisfied. Similarly, an accidental system interruption can fall outside a grant restricted to malicious security breaches. The practical evaluation is to trace a scenario through the grant, definitions, covered insured interest, exclusions, endorsements, waiting period, retention, limit and aggregation clauses.

Limits and adequacy

A headline limit is an upper bound under stated terms, not the payment automatically due on any event. Sublimits can cap specific coverage categories. Aggregate limits can be reduced by earlier claims; defense costs may interact with limits according to wording. Retentions can apply in multiple ways. A model of protection adequacy must therefore record whether costs erode an aggregate, whether several grants respond to the same episode, and how coordination with other insurance operates.

When loss valuation uses an indemnity period, later consequential impacts can remain with the policyholder even if the initial interruption qualifies. A relatively long maximum period cannot eliminate a waiting threshold or restore an excluded provider. Conversely, a small headline limit can fund valuable early response for a firm whose plausible covered losses are modest. Adequacy is relative to the firm's loss exposure and financial capacity.

A useful comparison of contracts starts with an identical factual scenario and computes payment under each contract without changing the loss assumptions. Changes in observed claim payments across policies can otherwise be attributed to differences in the underlying risks rather than to coverage. The model must also state whether an expense is assumed reasonable, necessary, consented to, and attributable under the specific wording.

Time to claim payment and services

A covered incident may require urgent professional assistance prior to complete loss quantification. Reporting an incident, obtaining authorization, preserving evidence, mitigating damage, and providing a proof of loss are not interchangeable steps. An SME with limited in-house staff may be disadvantaged by unfamiliar claim procedures. Prearranged expert services can have substantial value where immediate cash and specialized expertise are scarce.

Historical operations must be measured carefully. Logs may have been overwritten; accounting records may reside on the affected service; payment fraud may involve simultaneous bank recovery efforts. The ultimate economic loss can remain uncertain even after service is restored. A claim record should therefore include payment dates, outstanding reserves, denial or partial denial reasons, recoveries, expense categories, and event dates.

In a study of claim outcomes, delays should be separated into time from event to detection, detection to notification, notification to coverage decision, and decision to payment. A delay in the first stage could be attributed to security detection; the last stage may reflect claim adjustment or insurer operations. Combining them hides the mechanism responsible for cash-flow pressure.

Empirical Evidence and Comparative Measurement

British national survey evidence

DSIT's 2025 breaches survey recorded self-reported any coverage at 45% of businesses, with small and medium firms reporting 62% and 65%, respectively. Specific cyber coverage was reported by 7% overall. A fifth of business respondents did not know whether they had coverage. These are business survey responses, not a classification of actual indemnity.

The subsequent 2025/2026 official survey reported 47% overall, 55% of small businesses, and 61% of medium businesses with some insurance. Figure 3.4 reports specific cover of 15% for small and 24% for medium businesses, with bases of 241 and 154 respectively in the relevant split sample. Its rounded figures should not be forced to sum exactly; comparison of waves requires the report's statistical testing and weighting. The observed size differences may reflect affordability, product mix, or organizational purchasing needs. The survey does not establish which explanation is causal.

In the later wave, 22% of all business respondents reported uncertainty about coverage. On the published rounded figures alone, if the only unknowns were those cases and all self-reported known answers were accurate, overall true enrollment could fall between 47% and 69%. That is a simple logical bound, not a confidence interval or corrected prevalence estimate. The possible share without coverage under those same assumptions ranges from 31% to 53%. Misclassification of known responses can invalidate either bound.

The source also records higher detected attack or breach prevalence in medium than micro businesses. Detection capability may differ by size. Interpreting the difference as a true incidence difference without accounting for reporting error could misdirect services toward firms that are already better at detection.

United States market reporting

The NAIC cybersecurity insurance report (2026, Figure 1) records 2025 U.S. direct written premium including alien surplus lines of approximately USD 9.464 billion, compared with USD 9.144 billion in 2024. The domestic figure excluding that category was approximately USD 7.46 billion in 2025. Premium is a flow of consideration recorded under applicable reporting definitions; it is neither economic loss nor insured limit.

The number of reported policies, endorsements, insurers and insured enterprises can differ. One enterprise may purchase multiple coverages or layers, and a broader policy may be counted differently from a standalone policy. Even a correctly counted premium increase may be driven by rate, exposure mix, increased limits, or enrollment. Consequently, the NAIC totals cannot be divided by a national SME population or compared directly with the DSIT enrollment rates.

German demand and Asian survey evidence

The German evidence examined above is useful for studying association patterns within an identified survey population. It contributes an explanatory question: does perceived consequence, outside expert advice, or confidence correspond to financing decisions after adjustment for observed variables? It does not provide a monetary estimate of uninsured German cyber losses.

The original QBE Singapore survey release (2024) describes 605 surveyed SME leaders in December 2023 and January 2024 and reports cyber insurance and stated openness to purchase at 38% and 55%, respectively. These are sponsor-reported response figures. The published release is insufficient to harmonize its question, sampling, and weighting with the UK official survey; willingness and completed purchase also differ. The example supports the presence of access and perception questions outside Europe without a ranking of national market protection.

The OECD catastrophic risk assessment (2026) discusses varied cyber-insurance take-up and the use of both specific and broader policies. Its compilation shows the interpretive danger of taking results from different surveys as mutually comparable measurements. The underlying survey records have not all been independently examined here, so this chapter does not adopt their estimates as original empirical findings.

Evidence comparison and identification

Observation source Unit observed Main use Inference boundary
UK national survey waves Business response, size stratum Self-reported coverage and reasons Not audited terms or paid indemnity
UK commissioned mixed-method study Selected participating companies and interviews Mechanisms and study design Low-response nonrepresentative sample
NAIC market report Defined insurer reporting aggregates U.S. premium development No matched SME population or uninsured loss
German demand study Executives answering questionnaire Observed purchasing associations No randomized intervention
Singapore sponsor release Survey respondents Contextual reports of cover and interest Not a harmonized national estimator
Public policy forms Documented clause language Contract mechanisms Not binding terms of sampled firms

The datasets cannot be pooled because the events, countries, time periods and populations are not matched. International research is still possible: compare the mechanisms and design compatible measurements prospectively. A global estimate would require credible economic loss and ultimate indemnity data for comparable SME populations, including uncovered firms and the tail of event severity. This study has no such dataset.

Actuarial Measurement and Reproducible Illustrations

Frequency, severity and retained loss

If N is the number of relevant events affecting one firm during a year and X_j is gross loss from event j, then total annual gross loss is the sum of X_j over events. If event frequency and independent, identically distributed severity satisfy the usual compound-sum assumptions, the expectation is E[N] times E[X]. Covered severity depends on contract terms and can be materially less than gross severity. Different events may exhaust common limits, so an annual indemnity calculation must apply the actual aggregate structure before summing expected payment.

For a single eligible loss C, a simple per-occurrence retention d and maximum insurer payment M imply P = min(M, max(C-d,0)). This is a stipulated convention for the illustrations; alternative policies apply monetary limits, retention, and loss categories differently. It is not an interpretation of the actual specimens.

Assume a firm experiences exactly one loss in a year with probability 0.10 and no loss otherwise. Conditional on loss, the severity is 20,000 with probability 0.8 and 200,000 with probability 0.2. All losses are eligible in this illustration. A retention of 5,000 and payment limit of 100,000 yield payments of 15,000 and 100,000. Expected annual gross loss is 0.10 times (0.8 times 20,000 + 0.2 times 200,000), or 5,600. Expected annual indemnity is 0.10 times (0.8 times 15,000 + 0.2 times 100,000), or 3,200. The prospective expected insurance-only gap is 2,400, approximately 42.86% of expected loss. With no policy the gap would be the entire 5,600.

Increasing the payment limit to 200,000 changes the severe-event payment to 195,000. Expected indemnity becomes 5,100, leaving expected retained loss of 500. These figures abstract from premium, service benefit, contractual exclusions and financing. Reducing the retention to zero under the original 100,000 limit instead produces expected payment of 3,600 and remaining loss of 2,000. An expected-dollar comparison alone cannot describe consequences across event states; limit extension targets an upper-tail exposure while a retention reduction affects eligible smaller events.

If a risk-neutral firm pays a premium above expected indemnity, its expected monetary cost of loss plus premium exceeds the expected gross loss absent other benefits. Risk aversion, uncertain future liquidity, avoidance of financial distress, services and contractual obligations can justify this cost. The calculation should not be interpreted as a rate recommendation.

Waiting-period and dependency sensitivity

Consider an interruption lasting 72 hours and imposing a constant contribution-margin loss of 1,000 units per hour plus 12,000 restoration expense. Assume all restoration expenses are admissible, no overlap, a single retention of 5,000, a payment limit of 100,000, no sublimit, and a 12-hour interruption threshold.

Under a stipulated time deductible, eligible margin loss is (72-12) times 1,000, or 60,000. Together with restoration expense it yields eligible loss of 72,000 and payment of 67,000. Under a stipulated franchise threshold that admits the entire margin loss once exceeded, eligible loss is 84,000 and payment is 79,000. The gross event loss is 84,000 in both cases; the respective insurance-only gaps are 17,000 and 5,000.

If the same interruption concerns an external service wholly outside the covered interruption grant while the restoration costs remain covered, admitted loss is 12,000 and payment is 7,000 under the same simplified financial terms. Gap increases to 77,000. These calculations isolate three contract dimensions: timing threshold, scope of dependency, and retention. They show why generic policy counts do not reveal exposure-specific financial protection.

Portfolio selection and financial stress

Suppose 100 SMEs consist of 60 lower-risk firms each with expected annual gross loss of 1,000 and 40 higher-risk firms each with expected gross loss of 5,000. There are no limits or deductibles in the first stylized calculation; conditional on enrollment all modeled losses are insured. Population expected loss is 260,000.

In scenario A, 48 lower-risk and eight higher-risk firms obtain insurance. The enrollment rate is 56%; expected indemnity is 88,000. Expected insurance-only gap is 172,000, about 66.15% of total expected loss. In scenario B, 24 lower-risk and 32 higher-risk firms enroll. Enrollment is again 56%, but indemnity is 184,000 and gap 76,000, or approximately 29.23%. The insured portfolio's risk mix differs materially, so identical premiums per policy could not automatically be sustainable in both cases.

Neither scenario asserts an empirical SME risk distribution. The exercise establishes that a count-based measure cannot identify either the loss-weighted gap or the fair expected-loss component of premium without knowing risk composition.

For liquidity, assume a firm faces 50,000 urgent cash expense by day ten, has 15,000 available resources after ordinary obligations, and is ultimately entitled to 40,000 insurance payment on day sixty. Before the payment it faces a 35,000 shortfall at day ten. An advance of 25,000 by day ten reduces the shortfall to 10,000 while leaving ultimate indemnity unchanged. The timing dimension therefore has a mathematically independent effect on recovery finance. If the firm can finance the shortfall, associated financing costs may add to the economic burden, depending on the study's loss scope.

Regulation, Liability, and Institutional Boundaries

Regulatory obligations and insurable costs

Legal obligations after an event depend on its factual character, the organization's role, and applicable jurisdiction. The EU General Data Protection Regulation (2016, Articles 32–34) illustrates the distinction. Article 33 generally requires controller notification to the supervisory authority without undue delay and, where feasible, within 72 hours of awareness, subject to the stated risk exception. A processor must notify the controller without undue delay. Article 34 addresses communication to individuals in high-risk circumstances with specified exceptions.

These provisions can create a need for legal assessment and timely incident response. They do not establish an insurance payment or the insurability of a fine. Contract-covered notification expenses, defense costs, damages, and penalties are separate categories. The application of the regulation, national law, and contract terms must be independently assessed. The study uses the original regulation to illustrate timing and responsibility, not to provide jurisdiction-specific legal advice.

The NIS 2 Directive (EU, 2022, Article 2) generally includes relevant Annex I or II entity types qualifying as medium-sized or larger, and specifies scope extensions independent of size. This is not a universal requirement imposed on every European SME. The extent of particular obligations depends on implementing national law, sector, and entity classification. The directive illustrates how size thresholds and exceptional designations can matter for risk governance. It does not establish a general purchase mandate for cyber insurance.

Even where a customer imposes contractual security and insurance obligations on its SME supplier, those obligations are distinct from statute. Insurance sufficient to meet a contractual certificate requirement may remain inadequate for the firm's own interruption risk. Such arrangements should be studied through actual contractual obligations and policy terms rather than treating certification as evidence of indemnity.

Supervision of insurance supply

New York's Cyber Insurance Risk Framework (2021) is addressed to authorized property and casualty insurers writing cyber insurance. The circular connects risk assessment and pricing incentives with insurer financial stability. It is a supervisory framework for the insurance supply side, not a generic regulation of cybersecurity inside every SME.

This distinction matters because increasing insurance availability and maintaining a resilient underwriting market can be complementary over ordinary risks, yet conflict in scenarios of extreme accumulation. Supervisors need evidence of sustainable capital and credible contract specification. Measures forcing nominal coverage without addressing correlated exposure could transfer a loss problem from policyholders to claims-paying institutions.

Insurance market conduct can also affect comprehension. Standardized disclosure of covered events, relevant exclusions, supplier dependencies, retentions, and payment limits may help comparison, provided condensed summaries remain faithful to actual wordings. A short standardized statement cannot substitute for disclosure of material exceptions. The recommended regulatory emphasis is comparability of relevant terms, fair communication, and accurate representation of cover, rather than identical mandatory contract design across jurisdictions.

Shared Dependencies and Systemic Capacity

Common shocks and diversification

Cyber risk includes both firm-specific and shared exposures. Aggregation becomes significant when many policyholders depend on a common service or software component. An outage may cause successive individual losses for separate contracts, while a reinsurer may apply a different event aggregation definition. Correlation is relevant to portfolio capital even if no one SME has an unusually high expected loss.

Take 100 firms, each with a 1% annual probability of a 100,000 loss. Under independent Bernoulli events all firms have identical expected loss of 1,000 and portfolio expected loss is 100,000. Portfolio variance is 100 times 0.01 times 0.99 times 100,000 squared, or 9.9 billion squared monetary units. Standard deviation is about 99,499. Under a single common Bernoulli shock with the same probability, all 100 lose 100,000 together. The expected portfolio loss remains 100,000, but variance becomes 0.01 times 0.99 times 10,000,000 squared, or 990 billion; standard deviation is approximately 994,987. The latter is ten times larger.

The difference follows from covariance, not a change in the individual expected loss. These are stylized endpoints: real cyber dependence is heterogeneous and often changes as technology adoption evolves. The example establishes why an insurer may constrain limits on common dependencies even for otherwise attractive small risks. Shared technology can connect SMEs by risk topology rather than geography.

A careful exposure inventory would identify important providers, software dependencies, operational alternatives, geographic and logical service concentrations, and scenario-specific loss potential. A provider name alone is insufficient; risk depends on the precise service, architecture, recovery arrangements, and coverage trigger. Input uncertainty should be stressed rather than obscured by a highly precise aggregate estimate.

Capacity and public-private layers

Private insurers and reinsurers can diversify many risks but face capital costs and portfolio restrictions when tail events are highly correlated or difficult to model. Terms excluding particular extreme perils can leave a real gap even if enrollment is high. It is important to distinguish an exclusion for a defined event from denial of a claim based on an untested assumption about its cause.

The OECD (2026) analysis of catastrophe finance places cyber risk in a broader examination of government-supported financial protection. Public-private arrangements can be considered for severe consequences that exceed feasible private risk capacity, but a useful arrangement requires a bounded trigger, allocation of layers, pricing or fiscal funding, transparent eligibility, and rules for claims and catastrophic exhaustion. Otherwise a nominal guarantee may create financial exposure without reliable access to funds.

An SME-focused arrangement should distinguish routine risk transfer from catastrophe support. Subsidizing all routine indemnity can distort prevention incentives and impose poorly targeted fiscal costs. Extreme-event support with vague scope can create disputes or encourage private withdrawal. The design challenge is not solved by declaring cyber risk uninsurable in general or by promising unlimited coverage. The evidence in this dossier supports analysis of the constraint, not a quantified proposal for any particular national backstop.

Distribution, Prevention, and Policy Responses

Improvement of access

The basic commercial pathway consists of identifying exposure, obtaining information and quotes, meeting eligibility requirements, selecting a suitable policy, and maintaining the conditions needed for coverage. Measuring the pathway as a process can distinguish firms that never inquire from firms denied coverage or offered terms they cannot afford. Each transition has a denominator. For example, declination rates among applicants exclude firms that never apply; a purchase rate among quoted firms excludes denied applicants.

Interventions should be evaluated at the stage they target. General awareness work may increase appropriate inquiries. Clear coverage comparisons may decrease misunderstanding or buyer disengagement. Shared specialist services can reduce information and implementation costs. Proportionate verification can improve underwriting without requiring unnecessary repetitive questionnaires. Measures that merely increase the number of policies issued should also be tested against excluded losses, limits, claim payment, and overall recovery outcomes.

Insurance services can help make risks legible and provide access to response experts. However, claims service use may be constrained by contractual terms, geographic availability, or capacity during a widespread event. A prevention arrangement should specify what the insurer or its vendor actually supplies, who pays, whether participation is required, and what evidence demonstrates that the control functions. A promotional statement is not equivalent to an operational guarantee.

Incentives and consumer value

Prevention and insurance can be complementary when an insurer recognizes risk-reducing controls or subsidizes access to effective advice. They can compete for the same constrained budget. An owner facing an uninsurable residual scenario may rationally prioritize a recovery system that protects the continuity of several otherwise uncovered processes. In another firm, a small premium may transfer a risk that a much more expensive control cannot eliminate.

Any subsidy or discount should be evaluated against counterfactual behavior. A discount given to firms that already had the relevant control need not produce any additional risk reduction. A program that helps previously excluded firms implement a control may produce additionality, but only if the control works and new cover responds to their material losses. Conversely, an observed decline in claims after introducing a requirement can arise from removing high-risk firms from the insured pool rather than improving their security.

An evaluation should examine the total change in both insured and uninsured losses. Selection effects can make an insurer's loss ratio improve while population protection deteriorates. A mechanism that improves aggregate welfare must account for costs borne by the firm, insurer, service provider, and any public funder.

Practical institutional assessment

The relevant measures differ for participants. SMEs need exposure-specific evaluation and recovery funding plans; brokers can document comparisons and explain contract fit; insurers can monitor application-to-bind transitions, corrected technical indicators and claims outcomes; public authorities can support evidence collection and consider whether barriers are due to information, liquidity, lack of expertise, or correlated risk capacity.

Publication of anonymized aggregate data should protect commercially sensitive incident information. Detailed forensic data can reveal exploitable vulnerabilities and personal information. Research access therefore needs secure governance, minimization of unnecessary details, and clear restrictions on reuse. Privacy protections need not prevent reporting of harmonized anonymous denominators.

A proposed data table for access research would include unique enterprise units, sector, size definition, coverage status, application date, quote result, broad declination reason, chosen limits and retentions, control requirements, and status at renewal. A matched loss table would separately include event date, detected event type, identified technology dependence, gross loss categories, insurance notices, covered or denied loss, ultimate paid or reserved amounts, and dates of actual transfers. With appropriate consent and linkage, these records would support analysis of transition barriers and realized protection.

Critical Discussion and Research Agenda

The strongest challenge to studying protection gaps is that absence of insurance is sometimes rational. A firm with limited digital dependency and sufficient liquid reserves may prefer to retain risk rather than purchase coverage that adds expense and administrative complexity. This possibility means the enrollment rate is not a normative insurance target by itself. A serious assessment must identify exposure, capacity to absorb loss, available contract terms and the net value of formal transfer.

A second challenge concerns prevention. Insurers are often assumed to improve cyber hygiene by conditioning coverage on controls. This is a plausible mechanism, yet observational evidence cannot separate insurer influence from preexisting differences between insured and uninsured firms. An insured organization may be safer before it ever applies because it has more resources and qualified staff. Underwriting may also remove weaker risks. A study comparing insured and uninsured loss records without controlling for these selection mechanisms does not measure the causal effect of insurance.

A third challenge is that insured loss may be easier to observe than uninsured loss. Insurers require documentation for settlement; an uninsured firm may restore operations informally and never record its total costs. Conversely, insured firms can report events with little financial consequence when a service grant facilitates response. This difference changes the observation process. A valid comparison would need a common event identification and loss-valuation protocol.

Fourth, contract adequacy has an unobservable counterfactual: what would a rejected or uninsured firm have recovered under a product it did not buy? Estimating this requires applying a documented policy and schedule to a consistent hypothetical event, clearly distinguishing that exercise from actual claim adjudication.

A future longitudinal study could recruit firms at the quote-request stage, collect pre-application controls and financial resources, observe coverage offers, and follow subsequent controls, incidents, losses and recoveries. The design must address self-selection into applying. Parallel sampling of nonapplicants would help identify demand barriers. A randomized offer of independent educational or technical assistance, where ethically and practically permissible, might help isolate a specific intervention's effect; random assignment of insurance itself may be impractical. Claims-tail estimates would need long observation windows and treatment of censoring, firm exit, repeated incidents, and exposure changes.

The main limitations of this study are the absence of matched global microdata and the lack of bound policy schedules and SME-level paid and denied claim files. The comparative legal discussion is illustrative, with original EU acts examined through retrieved text and without a comprehensive audit of national transposition. The UK national survey gives self-reported coverage, while the smaller commissioned study, German analysis, and sponsor-reported Asian survey each have distinct selection and measurement constraints. Some primary PDF text was available when page-image retrieval failed; the evidence ledger records these access boundaries. The work is a source-based narrative synthesis, not a systematic meta-analysis. Its conclusions are confined to mechanisms, measurement requirements, and conditional illustrations.

The original contribution consists of applying established financial identities and insurance principles jointly to SME enrollment, risk selection, contract response, and payment timing. The numerical examples demonstrate logical differences under specified assumptions, rather than estimating the effect of insurance in an observed population. No independently established new theory, novel empirical law, or worldwide percentage is asserted.

Conclusion

The cyber insurance protection gap for SMEs reflects a chain of interconnected economic and contractual conditions. Potential losses arise through a firm's own digital activities and through dependencies on other organizations. Prevention changes the frequency and severity of loss; insurance allocates specified residual consequences; the firm's liquid resources and payment timing determine its capacity to finance recovery. A scientifically meaningful assessment must distinguish these processes.

The documentary evidence demonstrates that firms report diverse coverage arrangements and limited certainty about their insurance status, that awareness and budget constraints can impede demand, and that eligibility and policy complexity merit attention. Primary policy specimens establish that the scope and financial consequences of interruption grants can differ. Official insurer reports show the scale of a market while not measuring underlying SME exposure or uninsured damage. Comparative behavioral evidence identifies relationships worthy of further causal research.

The monetary gap depends on the joint distribution of economic loss, insurance ownership, coverage terms, and ultimate payments. The same enrollment rate can coincide with radically different loss-weighted outcomes; the same ultimate coverage can coincide with different short-term financing outcomes. This is the central analytical finding. It explains why an expansion of policy counts is an incomplete criterion of success.

Effective access requires comprehensible contracts responsive to the insured's material exposure, affordable and proportionate risk controls, viable insurance pricing and capital, dependable response and claims arrangements, and realistic plans for retained losses. International comparison requires harmonized measurements and transparent treatment of unknown status and missing losses. Addressing the protection gap is therefore a question of improving the reliability of financial protection alongside cybersecurity and institutional capacity, while explicitly recognizing the risks that remain outside the adopted contracts.

References

  • Adriko, R., & Nurse, J. R. C. (2024). Cybersecurity, cyber insurance and small-to-medium-sized enterprises: a systematic review. Information and Computer Security, 32(5), 691–710. DOI: 10.1108/ICS-01-2024-0025. Original record

  • CyberAcuView, LLC. (2022). CyberAcuView Cyber Insurance Wording. Model wording, first-party business interruption and dependent business interruption modules, pp. 21–28. Lloyd's Market Association hosted copy. Original record

  • Department for Science, Innovation and Technology. (2025). Cyber Security Breaches Survey 2025. UK official statistics, chapter 3, Figure 3.6. Original record

  • Department for Science, Innovation and Technology. (2026a). Cyber Security Breaches Survey 2025/2026. UK official statistics, chapters 3 and 4, Figures 3.4–3.5 and 4.1. Original record

  • Department for Science, Innovation and Technology. (2026b). Cyber Security Breaches Survey 2025/2026: technical report. Questionnaire items Q23X_INSUREX and Q23Y_WHYNOINSURE. Original record

  • European Parliament & Council of the European Union. (2016). Regulation (EU) 2016/679 (General Data Protection Regulation). Official Journal L 119, 1–88, Articles 32–34. Original Official Journal PDF. Original record

  • European Parliament & Council of the European Union. (2022). Directive (EU) 2022/2555 (NIS 2 Directive). Official Journal L 333, 80–152. Consolidated text dated December 27, 2022, Article 2. Original record

  • Grant Thornton & Department for Science, Innovation and Technology. (2025). Insuring Resilience: Adoption of Cyber Insurance by UK Small and Medium Sized Enterprises. April 2025; publicly released August 11, 2025. Executive summary, sections 1.3–1.4 and 3.3. Original record

  • National Association of Insurance Commissioners. (2026). Report on the Cybersecurity Insurance Market. 2025 reporting year, section 1, Figures 1–2. Original record

  • National Institute of Standards and Technology. (2024). NIST Cybersecurity Framework 2.0: Small Business Quick-Start Guide. NIST SP 1300, final February 26, 2024. DOI: 10.6028/NIST.SP.1300. Original record

  • New York State Department of Financial Services. (2021). Insurance Circular Letter No. 2 (2021): Cyber Insurance Risk Framework. February 4, 2021. Original record

  • Organisation for Economic Co-operation and Development. (2026). Financial Protection Against Catastrophic Risks: Examining the Need for Government-Supported Financial Protection for Natural Hazard, Cyber and Infectious Disease Risks. OECD Publishing, chapter of report DOI 10.1787/32a3acd7-en. Original record

  • QBE Insurance (Singapore) Pte Ltd. (2024). Less than half of SMEs fully aware of cyber risks, while the number without protection grows, finds QBE Singapore annual SME survey. News release, April 23, 2024, p. 1. Original record

  • Romanosky, S., Ablon, L., Kuehn, A., & Jones, T. (2019). Content analysis of cyber insurance policies: how do carriers price cyber risk? Journal of Cybersecurity, 5(1), tyz002. DOI: 10.1093/cybsec/tyz002. Original record

  • RSA. (n.d.). Cyber Protection Policy Wording. Public form, insuring clauses 5–6 and general condition on waiting period, pp. 3, 8. Retrieved November 18, 2026. Original record

  • Salzberger, A. (2025). An empirical analysis of the behavioral influences and information sources affecting the cyber insurance decisions of German SMEs. The Journal of Risk Finance, 26(2), 213–240. DOI: 10.1108/JRF-05-2024-0151. Original record

  • U.S. Government Accountability Office. (2021). Cyber Insurance: Insurers and Policyholders Face Challenges in an Evolving Market. GAO-21-477, May 2021. Official report summary. Original record