RESEARCH DOSSIER
Digital Insurance Platforms and Two-Sided Markets
Abstract. Digital insurance platforms coordinate insurers, intermediaries, and prospective policyholders through search, comparison, referral, distribution, transaction processing, and increasingly embedded purchasing journeys. Their economic significance arises from the joint determination of participation, access, information, distribution fees, coverage choice, and the costs of bearing insured risk. This monograph synthesizes the economics of two-sided markets with insurance-specific asymmetric information, actuarial heterogeneity, insurance contracting, and conduct regulation. It develops transparent analytical derivations from established platform models to show when cross-group participation effects amplify demand, how the allocation of fees influences market participation, why nominal premium rankings may be poor measures of insurance value, and how insurer participation and consumer multihoming affect competition. Original applications include a restricted linear participation illustration, an explicit fee-incidence accounting identity, a quality-adjusted quote comparison, a platform-access counterfactual, and a research design that separates distribution effects from underwriting-risk changes. The evidence base combines original industrial-organization scholarship, insurance economics research, international insurance supervisory materials, European law, United Kingdom market studies, and competition remedies. Its principal conclusion is that platform efficiency depends on the interaction of search-cost reductions, contract comparability, insurer access, incentive alignment, and accountable governance. Consumer welfare cannot be inferred from the number of displayed quotes, the lowest headline premium, gross sales growth, or the existence of indirect network effects alone. The findings are analytical and evidence-bounded; they establish no new authorial doctrine or empirical causal effect unsupported by observed data.
Keywords: Digital insurance platforms; two-sided markets; multi-sided platforms; indirect network effects; price comparison websites; insurance distribution; insurer participation; commission incentives; platform fees; search costs; multihoming; gatekeeping; insurance consumer welfare; adverse selection; price parity; platform regulation; insurance economics.
Introduction
The digital organization of insurance distribution has become a distinct object of insurance economics. An insurance platform can present consumers with competing policies, transfer leads to intermediaries, embed insurance in a purchase flow, or coordinate underwriting and administration among multiple firms. Each activity can alter the cost of finding coverage, the set of risks that can be matched to insurers, and the division of economic surplus along the distribution chain. The relevant economic unit is often a system of interdependent participants rather than an isolated seller or a conventional advertisement. The foundational platform-competition analysis of Rochet and Tirole (2003) explains why the price faced by one participant group depends on the value its presence creates for another. The competition models of Armstrong (2006) sharpen this relationship by distinguishing monopoly intermediation, rival platforms, and competitive bottlenecks.
Insurance places constraints on the direct transfer of these models. A policy is a contingent claim with coverage limits, exclusions, deductibles, underwriting conditions, and obligations that may extend years beyond the sale. Insurers price the probability and severity of loss, investment and capital costs, expenses, and commercial margins; the platform supplies access and interaction, although some vertically integrated organizations perform more than one function. Insurance demand is shaped by heterogeneous risk, consumer information, and selection. Einav, Finkelstein, and Levin (2010) provides a useful bridge between platform distribution and insurance-specific demand and cost analysis.
A substantial observational record makes the question practical. The OECD intermediation study (2020) identifies both transparency advantages and conflicts involving digital aggregators, the visibility of coverage, and intermediaries' relationships with insurers. The EIOPA digitalization report (2024) examines opportunities and risks associated with digital products, distribution, and supervisory adaptation. EIOPA's third report on the Insurance Distribution Directive (2026) documents continuing variation in distribution arrangements and concerns about incentives and consumer understanding. These documents supply institutional observations rather than a single global estimate of digital-platform welfare.
This monograph asks six related questions: what makes an insurance service economically a platform; when are cross-group network effects material; how does digital comparison affect real choices among different insurance contracts; how do platform remuneration arrangements allocate costs and incentives; under what conditions can an intermediary become a gatekeeper; and what evidence is sufficient to judge welfare consequences? The purpose is a disciplined synthesis of established economic mechanisms applied to insurance institutions, with derivations expressly labeled as analytical illustrations.
The research method combines critical analysis of original journal articles, examination of official regulatory texts and published supervisory findings, explicit accounting and comparative-statics exercises, and a structured assessment of empirical identification. Regulatory examples are jurisdiction-specific and dated; mathematical examples use constructed variables and do not claim calibration to an insurer, jurisdiction, or market. Publisher abstracts are used for the claims they actually substantiate, and original full-text material or public institutional reports are consulted when available. Sources are linked at the propositions they inform and bibliographically identified in References. The research cutoff is November 14, 2026.
Insurance Platforms and Economic Roles
Economic boundaries and the architecture of intermediation
A digital insurance platform is an organizational and technological arrangement that facilitates interactions among at least two economically distinguishable participant groups in insurance-related search, access, contracting, or service provision. The term is analytically strongest when decisions by one group affect the expected value of participation by another and when an intermediary can influence access conditions, matching, pricing, visibility, or transaction rules. A website displaying one insurer's proprietary offer may be digitally sophisticated while remaining a single-firm sales channel for purposes of a two-sided-market model. Conversely, an intermediary hosting several competing insurers and a population of buyers may exert meaningful control over two-sided participation even without owning underwriting technology.
This distinction follows the organizational reasoning of Hagiu and Wright (2015): a firm can choose between a multi-sided platform and alternatives such as resale, input supply, or vertical integration, and the allocation of control affects whether dispersed professionals adapt to private information and exert effort. In insurance, the test asks who sets underwriting prices, owns the policy obligation, controls product terms, governs customer access, and retains the residual financial consequences of claims. A platform can be an intermediary without accepting the insured risk; an insurer can be a platform participant while remaining legally responsible for its insurance contracts; an affiliated technology company can supply search and infrastructure without holding the legal status of a distributor in every jurisdiction.
The principal economic interfaces can be described in a four-stage sequence. First, potential customers search or enter a commercial environment where an insurance need becomes salient. Second, a platform collects consented eligibility and preference information, requests quotes or accesses cataloged offers, and selects the products shown. Third, the customer chooses or is referred to a licensed party for the policy transaction. Fourth, renewal, claims service, cancellation, and data feedback affect the value of the relationship over time. These stages may occur within one application, or be distributed among an aggregator, a broker, an insurer, an embedded merchant, and a claims administrator. Their legal classification depends on activities performed, not the visual appearance of the interface.
The European Union's Insurance Distribution Directive (2016/97) illustrates why conduct must be mapped onto function. Its definition of insurance distribution expressly encompasses certain comparison and ranking activities where the customer can directly or indirectly conclude a contract using the website or other medium. Its exclusions also distinguish mere provision of specified information or leads from activities assisting a transaction. The European Commission's distribution explanation confirms that online comparison can fall within the sales regime. A website's claim that it is only a technology provider is therefore not dispositive; the actual interaction and legal facts matter.
Institutional roles and the risk-bearing distinction
The platform ecosystem has several distinct participants. Policyholders purchase contingent protection and value accessible information, affordable coverage, reliable claims performance, and privacy. Insurers evaluate insured exposure, decide underwriting appetite, set contractual terms, pay claims, maintain capital, and may pay acquisition or renewal compensation. Brokers and agents can provide advice, act for different principals, and receive fees or commissions. Comparison businesses supply information organization, traffic acquisition, eligibility matching, ranking, and referral. Merchants embed protection into transactions for travel, mobility, electronics, finance, or other activities. Supervisors, competition authorities, and consumer-protection bodies shape the feasible set of practices.
Three accounting distinctions are essential. Gross written premium belongs to the insurer's policy economics, subject to the relevant accounting and contractual definitions. A distribution commission or platform access fee is an acquisition or servicing expense for the party obligated to pay it, with possible downstream effects on the premium. A platform's gross transaction volume is a measure of facilitated activity, not its net revenue and not the social surplus created by the transactions. Mixing these figures produces misleading assertions about platform scale, financial profitability, or welfare. Similarly, a large inventory of quotes is not identical to a large inventory of eligible, independently underwritten, comparable contracts.
Insurance introduces a multi-period relationship. The platform may influence a customer's initial purchase while the insurer retains the renewal relationship; alternatively the intermediary may organize annual switching and hold the customer interface. Insurer participation therefore depends on expected lifetime value, underwriting loss ratio, claims expenses, capital costs, persistency, and distribution attribution. Consumer utility depends on both immediate price and performance during adverse states of the world. These considerations are consistent with the insurance demand and cost heterogeneity emphasized in Einav and colleagues' empirical review.
The common description of insurance platforms as 'disintermediation' should be tested rather than presumed. A digital aggregator can replace part of a traditional broker's physical search and acquisition effort while introducing new platform fees, ranking services, data-processing intermediaries, and referral chains. The economic effect is a reconfiguration of intermediation costs, information and bargaining power. A reduction in face-to-face transactions cannot by itself establish a reduction in total distribution costs. An end-to-end cost study must count marketing, platform commission, integration, compliance, customer acquisition, customer support, payment processing, fraud controls, and policy servicing.
Platform formats, participant identities, and measurement
Four formats recur, although real businesses combine them. A comparison marketplace collects standardized risk information and displays multiple offerings; a lead-generation exchange acquires customer intent and sells or routes leads; a transaction platform concludes or facilitates contracts and may administer renewal; an embedded distribution platform places an insurance offer inside a distinct primary transaction. The same company can operate several formats, and a comparison tool may become an insurance distributor when the scope of its activities crosses a relevant statutory boundary. The EIOPA comparison-site clarification makes this distinction particularly clear for the EU.
The number of 'sides' is also a modeling decision, not a marketing label. A core two-sided model places consumers on one side and insurers on the other. A three-sided extension separately accounts for brokers or embedded merchants when they make autonomous joining or pricing decisions that influence the other groups. Search advertising platforms may introduce advertisers as another group; data or technology suppliers may remain contractual inputs if their participation does not generate a separately modeled feedback effect. More sides do not automatically make a model more realistic. Additional groups should be introduced when they explain actual strategic decisions, observably distinct costs, or relevant policy constraints.
A rigorous dataset distinguishes users, quote requests, valid quotes, insurer offers, ranked impressions, clicked results, referred customers, bound contracts, renewals, and paid claims. For every stage, the analyst should record how one customer request maps to multiple offers and how repeated requests map to unique people. Insurer participation should count independent risk carriers rather than product brands only, and ownership links should be visible. Consumer multihoming should measure the use of competing platforms or direct insurer channels. Without these denominators, conversion rates, average commissions, and apparent market shares can vary simply because of tracking definitions.
The historical experience of price-comparison channels confirms the institutional significance of these boundaries. In its 2014 thematic review of comparison websites, the UK Financial Conduct Authority treated these services as important general-insurance distribution channels and scrutinized information about cover and consumer choice. The regulator's later notice explicitly identifies that review as historical, so its measured market conditions should not be projected unchanged onto 2026. The analytical lesson survives: an intermediary's informational design is part of the economic and supervisory object.
Multi-Sided Network Effects
Two-sidedness and the price structure
A two-sided market contains cross-group interaction effects whose internalization depends on the platform's allocation of charges, services, and participation incentives. The central contribution of Rochet and Tirole (2006) is the significance of price structure: under relevant conditions, the volume of interactions changes when the distribution of a fixed aggregate charge between participant groups changes. This concept is more specific than observing that a large website is attractive. A platform may exhibit economies of scale in software or advertising while the economic two-sidedness of an interaction depends on whether additional participation on one side changes the value generated to the other.
Let the consumer group be denoted C and the insurer group I. Let n_C and n_I denote effective participating consumers and insurers, respectively. Let p_C denote a consumer-side participation charge and p_I an insurer-side access charge, measured consistently for a specified time period. A parsimonious representation is U_C = v_C - p_C + a n_I - d_C and U_I = v_I - p_I + b n_C - d_I. The positive coefficients a and b capture incremental cross-side access value; d_C and d_I are costs or frictions. The representation is illustrative. Insurance product quality, insurer eligibility, capacity, and matching costs imply that participation counts can be inadequate proxies for effective choice, and that access effects may be nonlinear or negative after congestion and low-quality additions.
A platform's accounting profit can be written as Π = p_C n_C + p_I n_I + t Q - K(n_C,n_I,Q), where Q is the number of distinct facilitated transactions, t is a per-transaction platform receipt not already included in p_I, and K records operating, acquisition, regulatory and service costs. Charges and quantities must be defined in compatible units. If p_I already includes referral commissions, the separate tQ term must be removed. This identity is a bookkeeping framework rather than a claim about any firm's audited results. Its merit is to prevent an economic interpretation of a transaction commission as an underwriting premium or of platform revenue as consumer welfare.
A platform can optimally charge one side little or nothing when that participation increases demand by the other group. Indeed the share of total costs paid by each side may be very unequal. The appropriate comparison concerns how the fee split changes the joint equilibrium, including platform profit and the incidence of costs embedded in policy premiums. The underlying theoretical sources do not imply that free consumer access guarantees a competitive or consumer-beneficial platform. A zero listed fee can coexist with an insurer-financed acquisition charge that influences the insurer's cost structure and price decisions.
Participation feedback and comparative statics
For a deliberately restricted linear illustration, assume interior participation functions n_C = A_C - b_C p_C + α n_I and n_I = A_I - b_I p_I + β n_C, with nonnegative cross-group coefficients α and β and strictly positive own-price coefficients b_C and b_I. Set Δ = 1 - αβ and assume Δ > 0, all calculated quantities remain within feasible participation bounds, and the local fixed point is behaviorally meaningful. Solving the simultaneous equations gives
Thus ∂n_C/∂p_I = -αb_I/Δ and ∂n_I/∂p_C = -βb_C/Δ. An increase in the access price on one side lowers direct participation there, which lowers attractiveness on the other side and feeds back. The denominator is a feedback multiplier, and its positive sign is part of the stability condition. If αβ approaches one, the local linear result becomes highly sensitive, making finite participation caps, insurer heterogeneity, eligibility restrictions, and dynamic entry especially important. This exercise is an algebraic consequence of the assumed equations; it is neither a new theory nor evidence that real-world insurance platforms have a particular α or β.
As a constructed check, take A_C = 500, A_I = 40, b_C = 10, b_I = 1, α = 2, β = 0.01 and p_C = p_I = 0, with arbitrary compatible units. Then Δ = 0.98, n_C = 580/0.98 ≈ 591.84 and n_I = 45/0.98 ≈ 45.92. An increase of one unit in p_I reduces n_C by about 2.04 and n_I by about 1.02, within this local linear specification. These are synthetic participation units and should not be read as counts of insurers in a real market. Their only purpose is to check signs, feedback, and the dependence of participation on both sides' charges.
If the platform improves quote quality or reduces onboarding friction, A_C or A_I can increase without a literal change in p_C or p_I. Accordingly, a measured rise in participation after a fee change cannot identify a causal fee effect unless technological features, product availability, advertising, risk composition, and competitor responses are held constant or credibly modeled. Likewise, comparing an insurer listing 100 unsuitable policies with one listing five well-matched contracts requires quality-weighted opportunity sets. A direct empirical analog to n_I might be an eligibility-weighted measure of effective insurer options rather than raw brand counts.
Multihoming, congestion, and market tipping
An insurer can list policies on multiple comparison services, while a consumer can seek several quotes across platforms and direct channels. Such multihoming changes bargaining positions and the extent to which any individual platform controls access. The nonexclusive-intermediation analysis by Caillaud and Jullien (2003) examines indirect network externalities alongside multihoming and differentiated charges. Armstrong (2006) distinguishes competitive bottlenecks, including a situation in which one group uses several platforms while the other makes a more exclusive choice. These models explain why nominally competing intermediaries may retain leverage over the side that must obtain access to a captive audience.
For insurance, it matters which side multihomes and whether doing so is economical. A consumer may search across several sites but incur repeated data-entry and comprehension costs. An insurer may technically integrate with several intermediaries but face channel-specific commissions, quote-API costs, compliance requirements, fraud-screening differences, and distribution conflicts. Direct underwriting capacity may be rationed to some channels. Two insurers appearing to reach the same customers may therefore experience different effective access. Where consumers search a single aggregator and insurers list everywhere, the aggregator can be a localized distribution bottleneck; where policyholders compare many sites and buy direct, insurer outside options become stronger.
Positive cross-group network effects do not prove inevitable winner-take-all concentration. Market tipping also depends on the strength of same-side differentiation, search habits, multihoming, interoperability, fixed costs, the portability of quote information, brand trust, insurer exclusivity, and how readily customers can shift their purchase. Some additional insurer participation improves coverage breadth; beyond a point, near-duplicate offers can impose comparison costs and make exclusion screening harder. A platform may also selectively suppress less profitable or less relevant products. These possibilities require a model permitting negative same-side effects and endogenous matching quality.
Consumer welfare can deteriorate if the platform's gain from another insurer listing is captured primarily through higher fees, sponsored prominence, or data exploitation while effective product choice remains unchanged. Conversely, additional insurer participation can improve access for rare or underserved risk categories where no existing participant underwrites the requested exposure. The important outcome is feasible risk transfer at acceptable total cost, not participation growth considered in isolation. The theory supports conditional mechanisms; whether the market is tipping, or who benefits from concentration, is an empirical question.
A second dynamic issue is expectation formation. Participants may join because they expect the other group to join. A sudden high-visibility campaign may make an entrant seem viable even before a large transaction base has formed, whereas an incumbent may benefit from accumulated insurer integrations and recognized consumer habits. Early advantages can persist if they create irrecoverable sunk integration costs or switching costs. Yet an entrant with superior search relevance, lower commission, or an exclusive risk specialty can overcome those advantages. The identification strategy should therefore distinguish behavioral expectations, genuine cross-group value, and cost-based scale effects instead of attributing every persistence pattern to network effects.
Search and Price Comparison
Search costs and the comparability problem
Price comparison services reduce some costs of discovering offers, requesting quotes, and contacting providers. The relevant benefit is not merely a lower observed premium but a reduction in the time, effort, uncertainty, and error associated with finding a policy that matches the purchaser's actual exposure and coverage needs. Insurance products differ on dimensions that do not fit into a single premium column: covered perils, limits, exclusions, deductibles, waiting periods, settlement basis, insurer solvency, service practices, cancellation terms, and claims handling. Search costs may decline while interpretation costs rise if a platform creates an apparently precise ranking of materially heterogeneous contracts.
The UK competition inquiry into digital comparison tools (2017) found that these services can help consumers compare options while identifying concerns involving transparency, consumer understanding, comparison effectiveness, and supplier agreements. Its scope included motor and home insurance alongside other sectors. These findings are informative about market mechanisms; a study conducted in 2017 does not establish the prevalence of every behavior in all jurisdictions today. The FCA's price-comparison review findings (2014) similarly stressed the risk of consumer attention to headline premium at the expense of cover and excess terms.
Consider two constructed insurance offers, both for the same insured exposure and policy period. Offer A costs $820, but under a specified common loss distribution the policyholder's expected retained loss is $300. Offer B costs $880 and has expected retained loss of $100. Ignoring other quality and financial dimensions, the corresponding simple expected private cost is $1,120 for A and $980 for B. The premium-only ranking selects A; the premium-plus-expected-retention ranking selects B. This example does not assert the distribution of real losses or a universal ranking method. It demonstrates that the economic object to be compared includes contractual risk retention. A risk-averse policyholder may also attach value to the dispersion and tail of retained losses, not solely their expectation.
A transparent comparison can report premium, coverage breadth, deductible, insured limits, exclusions, a clearly defined basis for any quality score, and the number of participating insurers. Where algorithmic scoring is used, the reason for placing a product ahead of a cheaper or more comprehensive alternative should be explainable at the relevant level of decision-making. A comprehensive user interface does not require identical policies; it requires explicit distinctions and a clear relation between comparison criteria and the individual customer's needs.
Insurance-specific selection and contract quality
Comparing prices across insurer offers requires separating price effects from risk selection. Insurers may refuse to quote for particular exposures, differ in underwriting appetite, or impose different excesses and exclusions. Customers with different risk probabilities, private information, and preferences self-select into products and channels. The insurance-demand estimation literature identifies the need to model demand and insurer costs together in the presence of adverse or advantageous selection. A lower average premium for policies sold through a platform can therefore reflect lower-risk consumers or narrower contracts rather than stronger price competition.
At the level of an individual quote, let x denote the risk information relevant to the underwriting decision, z the coverage design, and j the insurer. A quoted premium π_j(x,z) is an offer conditional on the underwriting information, eligibility rules, coverage definition and time. A platform's ranking of π_j(x,z_j) across changing z_j values conflates price with product choice. A more defensible comparison fixes a feasible reference coverage bundle z* and compares π_j(x,z*) where each insurer truly supplies that bundle; if exact harmonization is impossible, the analysis should use explicit adjustments or evaluate portfolios of price and cover attributes. Offer refusals are substantive observations, not missing data to be imputed as competitive quotes.
The selection problem is particularly important for embedded insurance. A policy offered alongside a device, travel booking, or financial purchase may serve a high-convenience consumer segment with a different willingness to search, a distinct risk profile, and an additional merchant relationship. The embedded channel may lower enrollment friction, but the propensity to purchase can rise even when product comparability declines. Cross-channel comparisons must therefore standardize the opportunity set and record the customer journey, the insurer and risk-bearing entity, the commission arrangement, and evidence that consumers understand the optional insurance product.
Ranking, data pipelines, and causal identification
A quote-comparison data pipeline contains at least seven analytically relevant transformations: capture of a distinct customer request; transmission and normalization of eligibility facts; receipt of insurer offers or declinations; application of product-inclusion rules; ranking and presentation; clicks or referrals; and policy binding with subsequent renewals and claims outcomes. The platform controls or influences several of these transformations. A study observing only the last click or the lowest returned premium cannot explain why another insurer was absent, why a contract was shown prominently, or whether the displayed quote survived the final underwriting step.
Ranking design is an allocation of attention. For each potential quote, a platform can compute an observed rank r_j = g(π_j,z_j,eligibility_j,quality_j,fee_j,promotion_j,customer_j). This is a generic empirical description of possible inputs, not a claim that all platforms use the same algorithm. The central identification question is whether ranking changes the probability of consideration and purchase conditional on genuine coverage quality and insurer risk selection. Ranking is endogenous when the platform uses conversion likelihood and commission arrangements to select what is seen. A simple correlation between a high rank and a high purchase rate therefore overstates the causal effect of a prominent position unless the study accounts for selection into rank.
Several empirical designs are possible. A randomized change in ranking among already eligible quotes can identify the effect of presentation on choice, provided the experiment maintains legally required suitability and fairness constraints. A randomized information display can test whether disclosure of deductibles or missing insurers improves decisions; it requires clear primary and secondary outcomes and adequate sample sizes. A staggered insurer entry into a platform can be analyzed by difference-in-differences if comparable control markets and pre-trends are credible; underwriting appetite, advertising, customer mix, and simultaneous regulation remain threats to identification. Fee changes can sometimes be studied with interrupted time series or contract-level panels, but a change in premium, coverage, and commission must be decomposed rather than interpreted as a single effect.
Selection and information are also linked to dynamic outcomes. A consumer choosing an apparently cheaper policy can face unexpected exclusions at claim time, higher renewal premiums, or more onerous service. A longitudinal design should compare first-period premiums with renewal costs and relevant claim service outcomes, while censoring and claims-selection biases are addressed. The FCA investigation of pricing practices (2020) documented renewal-related pricing problems in UK home and motor insurance, supporting the need to distinguish acquisition price from longer-run consumer value. Those historical findings cannot be treated as estimates of post-remedy behavior after subsequent rules took effect.
Data governance is part of the empirical problem. Customer-level histories can identify multihoming and switching, but privacy law, consent, commercial confidentiality, and vulnerability of sensitive risk indicators constrain collection and processing. Analysis should apply data minimization, de-identification where possible, documented data access, and specified retention. Researchers should distinguish observable demographic effects, risk-relevant attributes, and prohibited or unfair discrimination under applicable law. The permitted features and fairness tests vary by jurisdiction and insurance line, so the general research framework must not pretend to establish a universal legal safe harbor.
Platform Fee Structures
Membership, access, referral, and transaction payments
The economic allocation of platform costs is observable through the arrangement of fees and the rights supplied in return. A consumer may pay no direct fee; an insurer may pay a fixed integration fee, a periodic listing charge, a cost-per-click amount, a cost-per-qualified-lead price, a cost-per-bound-policy commission, a renewal trail, a preferred-placement payment, or a bundle of services priced together. A broker may share commission with a platform when legally and contractually permitted. An embedded merchant may be paid for distribution and customer access. Each arrangement changes the marginal return from attracting a consumer, displaying a given insurer, and completing a purchase.
A per-click price rewards attention even if the customer is ineligible for insurance. A per-lead price rewards a qualifying expression of interest under an agreed lead definition. A per-sale commission rewards conversion, and its commercial effect depends on the definition of a valid sale, cooling-off or cancellation clawbacks, chargebacks, refunds, lapse, and renewals. A fixed listing fee can be easier to budget but imposes a cost on small insurers before revenue is earned. A sponsored ranking fee can affect the order in which offers are encountered, potentially creating an incentive to promote remuneration rather than coverage fit. Such descriptions specify contract incentives, not allegations about any particular company.
A two-sided perspective asks who ultimately funds each payment and whether the fee changes both participation and final prices. Rochet and Tirole's original pricing framework and Armstrong's membership-and-usage distinction explain why a fixed access fee and a per-transaction charge can affect platform behavior differently. When a platform charges a seller for acquiring a customer, the apparent absence of a consumer price is compatible with a transfer from the insurer's margin to the platform. But pass-through to the insurance premium depends on insurer competition, demand elasticity, risk selection, capacity, and the division of bargaining power.
A simple accounting construction makes the distinction explicit. For a specified policy sold through an intermediary, define π as gross premium paid by the customer, μ as expected claims plus attributable risk-capital cost per policy, e as insurer operating expense excluding distribution, c as the commission and other policy-attributable platform payment, and m as the insurer's expected contribution margin. In a deterministic expectation-based accounting identity, π = μ + e + c + m. The equality is definitional for a consistent allocation of the quantities. It does not imply that an insurer raises π one dollar when c rises one dollar. If competition compresses m, the insurer may absorb some of the fee; if demand is insensitive or competing channels share the fee increase, a portion may reach the customer. Moreover, an increase in commission can change risk mix or conversion, altering μ and e.
For a hypothetical policy, let μ = $520, e = $110, c = $70, and m = $100, so π = $800. A negotiated reduction in c to $50 could yield a $780 premium at unchanged μ, e, and m; it could instead raise m to $120 at π = $800; or it could be partly passed through. The accounting alone cannot identify incidence. Similarly, moving an insurer's acquisition budget from an external campaign to a platform fee can be cost-neutral in an expense comparison if the same customers are reached at the same cost. Only a channel-complete calculation establishes net efficiency.
Fee incidence, insurer competition, and bargaining
The economic effects of access payments depend on the structure of insurer competition. If insurers face an elastic demand for nearly equivalent policies and have comparable underwriting costs, a high platform fee can narrow their margins and discourage participation. If platform access reaches a group of consumers who rarely shop elsewhere, an insurer may pay more to reach them, potentially passing the cost into that channel's offers where lawful. A platform enjoying a high share of consumer attention can use access terms to capture part of the insurer's expected relationship value. Conversely, where insurers can reach the same customers directly or through rival intermediaries at modest cost, the platform's bargaining position weakens.
An insurer's participation decision can be represented as expected incremental platform value V_j = Q_j(m_j - c_j) - F_j - G_j - O_j, where Q_j is incremental contracts attributable to the platform, m_j is a defined pre-platform per-policy contribution after expected claims and nonplatform costs, c_j is the incremental transaction payment, F_j is a fixed access payment, G_j is integration and compliance expense, and O_j is the opportunity cost of business displaced from other channels. This is a decision identity for incremental value, not a universally correct accounting convention. It clarifies why gross bound policies overstate a platform's value when many sales merely displace cheaper direct-channel conversions.
The causal effect of c_j is particularly difficult to infer from an insurer's platform revenues. Higher commission can coincide with better conversion because an intermediary brings qualified customers or provides better information. It can also accompany worse outcomes if ranking responds disproportionately to remuneration. A scientifically defensible fee study therefore records negotiated rates, chargeable events, insurer eligibility, ranking position, policy terms, realized persistency, and underlying customer risk. A credible experiment or policy discontinuity may identify a narrow effect; observational industry averages generally will not establish a universal pass-through rate.
Commissions also create a potential principal-agent problem when a comparison service claims to identify the most suitable product for the purchaser. Where advice or suitable-choice representations are made, differences between economic remuneration and consumer benefit become directly relevant. The OECD analysis of intermediary remuneration (2020) explores conflicts of interest and the importance of meaningful disclosure. EIOPA's 2026 IDD application findings similarly identify continued concerns about remuneration incentives in parts of the market. These sources support scrutiny of incentive structures; they do not establish that every commission compromises the consumer's choice.
International evidence and changing distribution channels
Historical UK evidence illustrates how platform expense interacts with insurer acquisition strategies. The FCA's general-insurance pricing final report, updated in December 2020, estimated substantial annual spending to acquire customers across insurers, intermediaries, and comparison websites, and documented price walking in home and motor insurance before the later pricing rules. These figures pertain to specified UK markets and study periods. They demonstrate the relevance of distribution acquisition economics; they are not contemporary global platform costs. The FCA's subsequent pricing and renewal rules took effect in January 2022 and aimed to end price walking for covered home and motor renewals through the same channel. A post-2022 evaluation must use post-reform observations rather than projecting the earlier equilibrium unchanged.
The CMA digital-comparison market study considered competition among platforms and their relationships with suppliers across several sectors. Its supporting final report described the potential for comparison to reduce search costs while stressing the importance of transparent practices. The OECD review of insurance intermediation extends the institutional picture beyond one country and reports diverse legal and supervisory approaches. This variation makes jurisdiction and product type necessary control variables in international comparisons.
Recent EU supervisory evidence also discourages premature claims of uniform digital adoption. The EIOPA distribution-market structure data, reported with the March 2026 IDD work, indicate that commissions remained prevalent among intermediaries across Member States and that online sales were still limited in many markets in 2024, though growing, especially for non-life products. These are descriptive aggregate observations with country differences. They do not imply that digital platforms lack economic significance; a relatively small share of final bound contracts can exert competitive influence by facilitating quote discovery, marketing, and switching.
Embedded distribution changes the fee question because insurance is combined with another purchasing journey. A merchant may create willingness to buy by reducing friction at an opportune time. But bundled convenience can also limit comparison, particularly for cover that is secondary to the primary purchase. Platform fee analysis should identify whether insurance is optional, which party receives distribution remuneration, whether the insurer can compete through an independent channel, and whether the customer can inspect exclusions before purchase. The current institutional shift toward integrated financial and insurance journeys is addressed in EIOPA's September 2026 discussion of digital platforms. The discussion is a supervisory contribution and prospective analysis, not measured proof of how all embedded-platform consumers behave.
Insurer Access and Gatekeeping
Participation rules, ranking rights, and market power
A platform becomes economically significant as a gatekeeper when it controls access to a valuable group of prospective policyholders under conditions that materially limit suppliers' practical alternatives. Gatekeeping is a matter of the elasticity of access and the terms of participation, rather than a conclusion drawn solely from platform size. Entry rules, insurer accreditation, technical standards, quote latency limits, ranking criteria, API integration requirements, customer-data access, and visibility policies can all shape the feasible market. Some restrictions serve legitimate reliability, fraud prevention, licensing, or quality objectives; others may protect incumbent participants or the platform's own commercial interests.
Consider two platforms with equal nominal insurer counts. One includes a broad set of independent underwriters, permits transparent comparison of material cover terms, and allows users to verify final quotes. The other carries several brands owned by a small number of underwriting groups and limits comparison information. The two have similar displayed breadth but different effective choice sets. Economists therefore need participation measures adjusted for common ownership, eligibility and underwriting independence. For competition analysis, the relevant access market might be defined around a class of risks or consumer segment rather than all insurance products in aggregate.
An insurer excluded by an aggregator may lose access to a consumer population it could otherwise serve. Yet the welfare effect of a particular exclusion depends on whether the insurer's policies are competitive and suitable, on whether the platform's selection protects consumers against unsuitable products, and on whether alternative channels are available. Fair access cannot be reduced to requiring every insurer to be listed without regard to quality and supervision. A sound governance arrangement documents eligibility standards, appeal procedures, sponsored listings, contractual conflicts, and the distinction between objective product filters and remuneration-based promotion.
The same standards apply to platform-controlled data. Historical quote-request data, behavior indicators, renewal propensity, and conversion records can improve matching and help insurers understand risks. Exclusive access to these data can also raise entry costs for rival intermediaries or insurers. Data portability, privacy safeguards, security, meaningful consent, and the commercial rights to reuse sensitive risk information can change this balance. A platform may lawfully retain some business information; whether a particular restriction is anticompetitive requires a defined relevant market, a factual assessment of alternatives, and applicable competition law.
Parity arrangements, vertical integration, and data control
Price-parity clauses can restrict an insurer's ability to offer lower prices in other channels. Their competitive effects differ according to whether a clause is narrow, concerning an insurer's own direct channel, or wide, covering competing platforms. The UK Private Motor Insurance Market Investigation Order 2015 provides a concrete remedy: it addresses agreements between comparison websites and insurers that prevent insurers from making their policies available more cheaply on other online platforms. The historical case demonstrates that contractual rules between platforms and insurers can matter independently of the consumer-facing interface.
A stylized parity mechanism is straightforward. Suppose a platform can charge a commission c while an insurer would otherwise offer a lower net acquisition-cost premium on another platform. A binding cross-platform parity obligation can restrict that response and weaken the ability of lower-cost rivals to attract customers through price. At the same time, the empirical effects depend on consumers' search behavior, product differentiation, the use of other promotional terms, and the platform's investment incentives. A contractual restriction that removes nominal price differences may alter equilibrium commissions, but it need not produce the same welfare outcome in every market. The correct assessment uses the actual contractual scope and the competitive alternatives.
In the EU, the 2022 Vertical Block Exemption Regulation supplies part of the contemporary legal context for vertical agreements and online intermediation services. Its applicability depends on the agreement, defined markets, conditions and exclusions; the mere existence of a digital insurance platform does not establish that a particular agreement benefits from the exemption or violates competition law. The CMA's digital-comparison findings likewise show why supplier agreements should be examined alongside ranking design and market transparency, rather than presumed harmful from form alone.
Vertical integration adds another economic mechanism. An insurer may own an aggregator or a technology supplier may participate in underwriting through subsidiaries or commercial affiliations. Integration can reduce coordination costs, improve data standards, and make end-to-end service accountable. It can also create incentives for self-preferencing, restrictive access, or differentiated data availability. The multi-sided-platform organizational analysis helps articulate the tradeoff between centrally coordinated activities and independently controlled services. The empirical task is to observe control rights, transfer pricing, ownership, insurer ranking and competitive responses, not infer behavior solely from organizational form.
A related issue arises in platform exclusivity. If insurers are forbidden to list elsewhere, consumer choice can contract and new platforms can struggle to attract supply. Exclusivity may also provide investment incentives where a new service must recover substantial integration costs or supports a differentiated insured segment. The welfare assessment must specify the duration of exclusivity, the ability to use direct channels, the alternatives for both sides, entry and exit costs, and whether the arrangement changes quality. The conditional nature of such analysis is consistent with the cross-platform participation issues in Caillaud and Jullien's intermediation model.
Insurance intermediation and supervisory boundaries
Insurance regulatory responsibility follows the conduct and contractual chain. A technology company can support information processing while a licensed insurer bears claims liability; an aggregator can nonetheless be engaged in distribution when it selects, presents, advises on, or facilitates the conclusion of insurance contracts in a legally relevant manner. Under the EU Insurance Distribution Directive's functional definition, a comparison site is within the defined activity where it compiles or ranks insurance offerings under the specified conditions that enable contract conclusion. The EIOPA Q&A on comparison websites confirms the importance of the distinction between mere information provision and additional steps assisting the contract. Specific authorization obligations follow the facts and implementing national law.
The IAIS Insurance Core Principles, ICP 18, address professional and transparent conduct in insurance intermediation; the associated supervisory framework recognizes a broad range of intermediaries and distribution channels. These global supervisory principles supply a comparative benchmark, not a single directly enforceable worldwide distribution license. The OECD's cross-jurisdictional review similarly observes differences in the licensing, supervision, and disclosure treatment of digital intermediaries. Consequently a scientific account should compare regulatory functions before comparing formal labels such as broker, aggregator, agent or marketplace.
EIOPA's 2024 insurance digitalization survey and its 2026 analysis of AI-enabled distribution identify emerging questions about who performs distribution, how advice and product governance apply, and how customers are treated when an insurer, platform and technology provider jointly shape the journey. The economic analysis should trace the party that influences the ranking and recommendation, the party that controls the consumer relationship, and the party legally responsible for underwriting. AI-driven interfaces make these distinctions more important because conversational product filtering may be perceived as advice even when the underlying business claims to offer only general information.
An effective risk-control and accountability map records the origin of policy terms, eligibility decisions, evidence of customer consent, versioned product documents, ranking parameters, adviser or distributor identity where applicable, commission disclosures, ownership relationships, transaction timestamps, complaints and redress pathways, and the claims-bearing insurer. Such a map connects distribution conduct to insurance obligations without conflating every software vendor with a regulated insurer. It also helps researchers reproduce observed outcomes and detect discontinuities when a platform changes a recommendation system.
Because national rules differ, legal conclusions in this monograph are jurisdiction-bounded. EU legislative definitions, UK FCA conduct requirements, CMA market remedies and IAIS supervisory principles occupy different legal levels and geographic scopes. Where the question is whether a specific current service is licensed, whether a specific clause is lawful, or whether a particular AI recommender constitutes regulated advice, the decisive record is the operative law, supervisory interpretation, jurisdictional facts, and product type. The scientific comparison supplies a method for evaluating those records rather than a substitute for them.
Competition and Consumer Welfare
Welfare accounting beyond the displayed premium
Insurance-platform welfare includes the surplus from risk transfer and matching, the costs of search and participation, the administrative costs of distribution, the quality of coverage and claims service, and the possible redistribution of surplus through bargaining and fees. A consumer may obtain a lower premium by selecting narrower cover; the recorded payment falls while uninsured risk rises. A platform can raise insurer competition and still produce a poor allocation if consumers systematically misunderstand exclusions. A platform can charge a higher commission than a traditional broker yet create net value if its search, matching, and servicing costs fall by a larger amount. These are competing possibilities to be discriminated by evidence.
For a particular insured customer i and offered contract j, write a welfare-relevant private evaluation as V_ij = CE_i(L_i^0) - [π_ij + CE_i(L_ij^ret) + s_ij + h_ij], where CE_i is a consistently defined certainty-equivalent cost, L_i^0 is exposure absent the contract, L_ij^ret is the retained loss under that contract, π_ij is premium, s_ij is search and transaction cost, and h_ij represents other explicitly specified service or contractual costs. This expression is an analytical evaluation device. The certainty-equivalent convention must be defined carefully to avoid counting both expected loss and its risk premium twice, and it should not be used as a universal social welfare metric without accounting for insurer and platform resource costs. For risk-neutral customers CE equals expected cost; for risk-averse customers it also captures the cost of variability under the assumed preferences.
Aggregate welfare must also account for insurers' expected underwriting costs, capital opportunity costs, platform resource costs, externalities such as fraud prevention or risk mitigation, and any effects on parties not transacting. Premiums and commissions are transfers between the modeled participants and should not automatically be counted again as social resource costs. The analyst can summarize an allocation by W = Σ_i B_i(contract_i) - Σ_i R_i(contract_i) - C_underwriting - C_distribution - C_search - C_governance, where B_i and R_i represent specified benefits and retained losses under a nonduplicative welfare accounting convention. This is a bookkeeping structure for counterfactual comparisons, not a claim to a uniquely observable cardinal social welfare function.
The most visible metric, minimum displayed premium, measures one component of an incomplete cost vector. Better metrics include average all-in premiums among genuinely equivalent contracts, breadth of eligible independent underwriting, expected retained-loss cost, spread of policy conditions, switcher and non-switcher outcomes, claim settlement patterns, consumer comprehension, distributor remuneration disclosure, and outcomes for high-risk or otherwise underserved applicants. A platform's contribution to economic welfare is strongest when these metrics are evaluated jointly and with credible counterfactuals.
Competitive and regulatory counterfactuals
Consider four analytically distinct counterfactual environments. In the first, consumers search insurers individually and face high information costs, while insurers compete directly. In the second, an open comparison platform lowers information costs and hosts multiple independent insurers, with fees and ranking transparent. In the third, a platform with substantial consumer concentration charges access commissions and controls rankings while some customers single-home. In the fourth, public or self-regulatory measures make comparison and fee disclosure more informative, reduce certain switching barriers, or restrict a specific anticompetitive contract. Comparing these environments requires the same insured-risk distribution, coverage definitions, and sufficiently comparable market conditions. Otherwise differences in average paid premiums may be driven by composition rather than platform mechanisms.
One comparison concerns total search cost. Let S_0 be aggregate search effort before a platform and S_1 after, measured with a defensible monetary or time valuation, and let C_P be the incremental real resource cost of operating and regulating the platform. Pure search efficiency yields a candidate benefit S_0 - S_1 - C_P before accounting for underwriting changes, fee transfers, product quality, and consumer choice. This can be negative when information processing and marketing absorb more resources than the search costs they save. Because commission income is a transfer, treating all commissions as an added social resource loss would double count the economic burden; the relevant resource components are actual labor, computing, advertising opportunity costs and other inputs. Distribution fees do, however, influence equilibrium behavior and may induce inefficiencies indirectly.
A second counterfactual concerns adverse or advantageous selection. If a platform brings new consumers into an insurance market, average claims costs can change. A simple difference in premiums cannot identify whether distribution became more efficient or risk composition improved. The insurance-market demand-and-cost literature emphasizes the need to infer consumer willingness to pay and insurer cost in a common framework. A proper analysis of a platform entry event should model eligibility, participation, insurance take-up, risk-relevant characteristics and pricing, and, where accessible, claims outcomes over an appropriate horizon.
A third counterfactual concerns ranking transparency. A platform can display an objectively expensive product at the top because of sponsorship, product relevance, or other contract features. If disclosure clarifies remuneration and sorting, consumers may correct their decisions, but disclosure effectiveness depends on comprehension and salience. The FCA's 2014 comparison-site findings and the OECD's 2020 intermediary review motivate evaluating communication quality rather than relying exclusively on the presence of disclosure text. For behavioral experiments, the measurable endpoint should include the chosen policy's coverage fit, not only whether a disclaimer was clicked.
A fourth counterfactual concerns renewal and switching. A low first-year premium may attract a customer even when expected renewal prices or the costs of switching are high. The FCA's original pricing market study and its post-2022 supervision of pricing rules must be treated as different policy environments. Evaluations should use observed cohorts and control for distribution channel, claims experience and underwriting changes; they should not treat an old regulatory concern as though it establishes present noncompliance. The welfare horizon should extend through the materially relevant contract and renewal period.
Empirical research program and boundary conditions
A credible research program uses linked datasets, explicit estimands, and a pre-specified sequence of tests. The first task is descriptive: estimate, for each relevant insurance line and jurisdiction, the share of quote requests, eligible insurer responses, first-year bind rates, renewals and claims flowing through each channel. The second is structural: document payments, ranking, multihoming, insurer independence, and the restrictions governing cross-channel offers. The third is behavioral: measure consumer comprehension, effective search effort, switching and coverage fit. The fourth is causal: evaluate changes in participation, all-in costs, and welfare under a convincing counterfactual.
For fee incidence, an empirical estimand is the change in a constant-quality quoted premium per unit change in platform acquisition cost, conditional on underwriting risk and contract characteristics. A regression that omits quality, insurer and time effects will generally mix fee effects with composition and competitive response. The standard difference-in-differences design requires parallel pre-treatment trends for the comparison groups; an instrumental-variable approach requires an exclusion restriction that withstands institutional scrutiny; randomized disclosure or ranking trials require that assignments do not create unlawful discrimination or unfair product presentation. A convincing paper reports exact identifying assumptions, standard errors, robustness checks and the population for which the effect is identified.
For network effects, another estimand is the causal response of effective consumer participation to an exogenous increase in eligible insurer availability, and the response of insurer participation to an exogenous change in accessible customer demand. Platform advertising, product quality and pricing can cause both sides to grow simultaneously without establishing a network effect. A stronger design uses independent shocks to insurer access or customer demand, compares eligible risk segments, and observes changes on both sides while ruling out confounding acquisition campaigns. The multiplier in the earlier linear illustration predicts a relationship conditional on its assumptions; it is not a numerical estimate for any real platform.
For ranking effects, randomized changes in the order of comparable policies can reveal how attention affects choice, provided the platform preserves relevant conduct obligations. For insurer access, changes in an onboarding rule or parity clause can provide natural experiments only if affected and comparison insurers have similar underlying trends and the policy change is sufficiently exogenous. In both cases, researchers must inspect the full customer funnel: impressions, quotes, final bound contracts, rescissions, renewals and valid claims. Click-through improvement alone cannot establish consumer welfare.
An international comparative study should normalize insurance product classifications, policy terms, consumer counts, periods and regulatory treatment. The EIOPA 2026 IDD report provides useful market-structure and conduct context within the EU, while the OECD's international intermediation review supplies a broader comparative regulatory baseline. An international dataset should not conflate legal jurisdictions, treat online sales as all online search, or assume that a platform's primary-function classification is identical across markets.
Observational studies also face survivorship, measurement and general-equilibrium problems. Platforms that close or insurers that withdraw may disappear from available logs; product terms may be recorded only for bound policies; and fees may be negotiated privately. A platform's entry can change competitors' prices and advertising, contaminating a simple untreated comparison group. Claims data emerge with delay and may require development adjustments. Longitudinal studies must identify reporting lags, missingness and ethical restrictions on sensitive data. A causal conclusion is appropriate only when these problems are addressed or its limitations are clearly bounded.
The research synthesis supports several operationally testable propositions. Where comparison is truly constant-quality, greater accessible insurer participation can increase consumers' opportunity set; the size and welfare consequences depend on fee incidence and underwriting selection. Where consumers single-home and insurers multihome, a platform may obtain greater bargaining leverage than participant counts alone imply. Where referral rewards alter ranking, transparency and ranking governance can affect effective choice. Where digital distribution reduces search effort but raises marketing or acquisition expenses, total channel efficiency is indeterminate without resource accounting. Where an insurance contract is multi-period, first-year quotes are an incomplete welfare metric. These propositions follow from established economics and the observed institutional constraints; they are bounded analytical interpretations offered for empirical testing, not claims of a newly discovered universal law.
Integrated scientific assessment
The central scientific result is a linked account of three distinct mechanisms: participation feedback among customer and insurer groups, allocation of intermediary expenses and attention, and the contractual quality of risk transfer. A platform may improve the first mechanism while worsening the second, or make quote discovery more accessible while restricting the third. The mechanisms must be evaluated together because each affects the interpretation of the others. For example, more insurers listed can raise search effectiveness when they contribute genuinely different eligible offers; it can also increase superficial quote abundance while leaving meaningful coverage unchanged. Lower advertised premiums can indicate better competition or greater uncovered risk. Higher platform revenue can indicate productive coordination or increased bargaining rents.
The evidence establishes plausible channels and documented institutional cases, while preserving important limitations. The journal literature supplies general models and their assumptions, not insurance-specific causal parameters. OECD and EIOPA sources report regulatory and market observations with particular geographic and temporal scope. The UK regulator and competition authority supply detailed case evidence and interventions, not a universal measure of platform efficiency across all lines of insurance. No proprietary policy-level panel or original randomized experiment was available for this monograph, so the constructed examples and formulas are methodological illustrations. Their transparent derivation prevents empirical numerical claims from being inferred where the necessary observations are absent.
A scientifically defensible future evaluation will combine effective insurer access, constant-coverage premium comparisons, observed channel payments, independent ownership, multihoming, contract renewal paths and claims performance. The minimum standard for a welfare claim is an explicit counterfactual and a resource-consistent accounting of benefits and costs. Claims about unfair conduct require the operative legal rule and the actual practice. With these conditions in place, digital insurance platforms can be studied as mature objects of industrial organization and insurance science, rather than treated solely as a technological category.
Conclusion
Digital insurance platforms coordinate insurer participation, consumer access, contractual information and distribution payments. Their economic architecture is materially affected by two-sided participation effects, but insurance-specific risk selection, differentiated coverage and continuing claims obligations shape every substantive conclusion about efficiency and welfare. The correct analytical approach begins by identifying the risk-bearing insurer and the parties controlling access, ranking, payments and the customer relationship, then traces how those functions affect actual available insurance contracts.
The theoretical literature establishes the importance of price structure, indirect network effects, multihoming, and organizational control. The insurance literature establishes the additional need to model heterogeneous demand and underwriting costs. The official record demonstrates the practical importance of transparency, quality of comparison, intermediary incentives, competition restrictions and evolving supervisory boundaries. This combination supports a conditional conclusion: search-cost reduction and broader market access create valuable opportunities for insurance customers and insurers when coverage is meaningfully comparable and fee incentives are governed transparently. A headline-price ranking, insurer count, conversion metric or platform growth rate alone is an incomplete welfare indicator.
The monograph's original contribution is an integrated, reproducible application of established theories and regulatory evidence to insurance platform economics, including clearly labeled participation comparative statics, fee-incidence accounting, a contract-quality comparison, and an empirical program for causal testing. These analyses do not constitute new authorial theories, doctrines, or general empirical estimates. The manuscript establishes a full subject-specific scientific basis for comparing platform models, designing further research, and evaluating market institutions in their appropriate jurisdictions and periods. All publication identifiers and the first-publication date remain reserved for a later authorized release.
References
Armstrong, M. (2006). Competition in two-sided markets. RAND Journal of Economics, 37(3), 668–691. https://doi.org/10.1111/j.1756-2171.2006.tb00037.x
Caillaud, B., & Jullien, B. (2003). Chicken & egg: Competition among intermediation service providers. RAND Journal of Economics, 34(2), 309–328. https://doi.org/10.2307/1593720
Competition and Markets Authority. (2015). The Private Motor Insurance Market Investigation Order 2015. UK Government. https://www.gov.uk/government/publications/private-motor-insurance-market-investigation-order-2015
Competition and Markets Authority. (2017). Digital comparison tools market study: Final report. UK Government, September 26. https://assets.publishing.service.gov.uk/media/59c93546e5274a77468120d6/digital-comparison-tools-market-study-final-report.pdf
Competition and Markets Authority. (2017). Digital comparison tools market study (case page and publication record). UK Government. https://www.gov.uk/cma-cases/digital-comparison-tools-market-study
Einav, L., Finkelstein, A., & Levin, J. (2010). Beyond testing: Empirical models of insurance markets. Annual Review of Economics, 2, 311–336. https://doi.org/10.1146/annurev.economics.050708.143254
European Commission. (2016, as subsequently amended). Directive (EU) 2016/97 on insurance distribution (consolidated text, January 9, 2024). EUR-Lex. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A02016L0097-20240109
European Commission. (n.d.). Insurance distribution. Directorate-General for Financial Stability, Financial Services and Capital Markets Union. https://finance.ec.europa.eu/banking/insurance/insurance-distribution_en
European Commission. (2022). Commission Regulation (EU) 2022/720 on the application of Article 101(3) TFEU to categories of vertical agreements and concerted practices. EUR-Lex. https://eur-lex.europa.eu/eli/reg/2022/720/oj/
European Insurance and Occupational Pensions Authority. (2024, April 30). EIOPA's report on the digitalisation of the European insurance sector. https://www.eiopa.europa.eu/publications/eiopas-report-digitalisation-european-insurance-sector_en
European Insurance and Occupational Pensions Authority. (n.d.). Insurance comparison website: Q&A 3168. https://www.eiopa.europa.eu/qa-regulation/questions-and-answers-database/3168-insurance-comparison-website_en
European Insurance and Occupational Pensions Authority. (2026, March 30). Third IDD application report. https://www.eiopa.europa.eu/publications/third-idd-application-report_en
European Insurance and Occupational Pensions Authority. (2026, March 30). EIOPA publishes third report on the application of the Insurance Distribution Directive. https://www.eiopa.europa.eu/eiopa-publishes-third-report-application-insurance-distribution-directive-2026-03-30_en
European Insurance and Occupational Pensions Authority. (2026). Structure of the EU insurance distribution market (data and 2026 IDD report materials). https://www.eiopa.europa.eu/structure-eu-insurance-distribution-market_en
European Insurance and Occupational Pensions Authority. (2026, September 16). AI, digital platforms and the future of insurance distribution. Eurofi Magazine contribution. https://www.eiopa.europa.eu/publications/ai-digital-platforms-and-future-insurance-distribution_en
Financial Conduct Authority. (2014, July 15). TR14/11: Price comparison websites in the general insurance sector (historical thematic review). https://www.fca.org.uk/publications/thematic-reviews/tr14-11-price-comparison-websites-general-insurance-sector
Financial Conduct Authority. (2014, July 16). Price comparison websites failing to meet FCA expectations. https://www.fca.org.uk/news/press-releases/price-comparison-websites-failing-meet-fca-expectations
Financial Conduct Authority. (2020, updated December). MS18/1.3: General insurance pricing practices: Final report. https://www.fca.org.uk/publication/market-studies/ms18-1-3.pdf
Financial Conduct Authority. (2020). MS18/1: General insurance pricing practices market study (case page and subsequent updates). https://www.fca.org.uk/publications/market-studies/ms18-1-general-insurance-pricing-practices-market-study
Financial Conduct Authority. (2022). General insurance pricing attestation: Multi-firm review (discussion of rules effective January 1, 2022). https://www.fca.org.uk/publications/multi-firm-reviews/general-insurance-pricing-attestation
Hagiu, A., & Wright, J. (2015). Multi-sided platforms. International Journal of Industrial Organization, 43, 162–174. https://doi.org/10.1016/j.ijindorg.2015.03.003
International Association of Insurance Supervisors. (2024, December update). Insurance Core Principles and ComFrame: ICP 18, Intermediaries. https://www.iais.org/icp-online-tool/13529-icp-18-intermediaries/
Organisation for Economic Co-operation and Development. (2020). Regulatory and supervisory framework for insurance intermediation. OECD. https://www.oecd.org/content/dam/oecd/en/publications/reports/2020/03/regulatory-and-supervisory-framework-for-insurance-intermediation_e78a9ec5/19915a09-en.pdf
Rochet, J.-C., & Tirole, J. (2003). Platform competition in two-sided markets. Journal of the European Economic Association, 1(4), 990–1029. https://doi.org/10.1162/154247603322493212
Rochet, J.-C., & Tirole, J. (2006). Two-sided markets: A progress report. RAND Journal of Economics, 37(3), 645–667. https://doi.org/10.1111/j.1756-2171.2006.tb00036.x