RESEARCH DOSSIER
Platform Power and Digital Market Coordination
Abstract. Digital platforms coordinate exchange by linking independent groups, administering information and controlling the rules governing access, discovery, transactions and enforcement. This research dossier examines the economic conditions under which these functions generate gains from trade and under which control of their implementation creates durable market power. It integrates multisided-market economics, network effects, organizational analysis, matching, search, information economics and welfare measurement with selected institutional cases in the United States, European Union, United Kingdom and Japan. The analysis distinguishes scale from network effects, nominal affiliation from effective access, transfers from real costs, and product performance from control of distribution. It employs bounded interpretations of published studies, official records, and illustrative calculations rather than new empirical estimates. The principal conclusion is that the economic significance of platform power depends on the relationship between the value produced by coordination and participants' practical alternatives to the associated governance arrangements. Effective evaluation requires attention to cross-side participation, quality, entry, investment, institutional commitments and the welfare of affected nonparticipants.
Keywords: Digital platforms; multisided markets; network effects; market power; market design; platform governance; matching; ranking; interoperability; competition policy; welfare; microeconomics; economic theory
The digital market as an organized environment
A digital platform organizes conditions under which independent economic agents encounter, evaluate, and transact with one another. Its output consists partly of direct services and partly of arrangements that make other people's exchange possible. A payment network establishes mutual acceptance; a marketplace brings buyers and sellers into a shared catalog; a ride-hailing intermediary links geographically situated demand with available drivers. Each uses technology to administer an economic relationship.
The distinguishing issue is control over the terms of coordination. A platform may decide who may enter, what information can be displayed, which participants are visible first, how transactions are priced, what counts as acceptable performance, and how a dispute is resolved. These arrangements influence the choices that buyers and sellers make. The resulting allocation depends upon both the preferences of participants and the institutional design of the intermediary.
The organizational comparison developed by Hagiu and Wright (2015) provides a starting point: a platform is one of several possible ways to organize interdependent production and exchange. The relevant alternative may be a retailer purchasing inventory, an employer directing workers, a directory that sells information, or an interoperable network managed by multiple organizations. A business can combine these forms. Its classification should follow the transaction and the allocation of decision rights, rather than the presence of an application.
The research question of this dossier is therefore twofold. Under what conditions does centralized digital intermediation improve coordination, and under what conditions does command over the coordination process create opportunities to exercise persistent market power? The answer requires explicit alternative institutions, attention to distribution, and a distinction between creating value and controlling its allocation.
Historical foundations of networked exchange
The economic problem predates contemporary software. Payment cards, exchanges, advertising-supported publications, and communication networks all involve interdependent demand. Digital processing extends the range of transactions that can be coordinated at scale and increases the granularity of the information available to the coordinator.
The historical literature on network externalities and compatibility (Katz and Shapiro 1985) addresses the importance of a user's expected network environment. Its conceptual significance for this dossier lies in the fact that the attractiveness of a product can depend on adoption by others. Compatibility changes the set of other participants whose adoption is relevant. A supplier of compatible services can therefore preserve network usefulness without necessarily controlling the entire installed base.
Rochet and Tirole's platform competition study (2003) moved the price structure across interacting groups into the analysis of platform competition. Armstrong (2006) subsequently distinguished the pricing incentives of alternative participation regimes in competition in two-sided markets. Together these contributions provide a way to study digital exchange without assuming that the existence of several user types by itself establishes a distinct market structure.
The digital transition deepens the historical problem through programmable rules. A printed directory can allocate attention by page position; an online marketplace can allocate it separately for each visitor and revise the rule after every observed interaction. The change is in speed, observability, and the capacity to experiment. The economics of a rule that previously appeared fixed becomes an economics of endogenous design.
Market boundaries and organizational alternatives
The unit of analysis is an activity supplied to identifiable participants through a coordination arrangement. Five activities deserve separate measurement: user discovery, affiliation, matching, settlement, and after-sale enforcement. A single enterprise may supply all five; separate firms can also divide them. If competition constrains settlement but discovery is concentrated, a firm-wide market share hides that difference.
A reseller typically acquires goods or contracts for output and determines the resale terms. A service platform may leave certain decisions about price, schedule, and quality to independent providers while prescribing eligibility and monitoring rules. In a vertically integrated firm, managerial direction may substitute for bargaining among independent contractors. The organizational tradeoffs analyzed by Hagiu and Wright (2015) motivate attention to who decides what, and who bears the consequences of a decision.
Consider three hypothetical arrangements for household repairs. A local directory sells leads and leaves contracts to the parties. A managed marketplace standardizes offers, collects payment, and guarantees performance. An integrated firm employs technicians and directly assigns tasks. The same household may regard them as substitutes in obtaining a repaired appliance; technicians face markedly different contract terms and opportunities to build independent relationships. The welfare comparison must incorporate both perspectives.
Market definition is an empirical and institutional question. Substitution possibilities must be measured at the level of the service at issue. A household may choose between an app and a telephone call even though the corresponding suppliers are governed by different institutions. A provider may lack a practical alternative to the app despite consumer substitution in principle. Both observations are relevant.
Cross-side externalities and participation
A cross-side externality arises when participation on one side alters the value received by agents on another side and that effect is incompletely reflected in private decisions. More sellers can expand variety for buyers; more buyers can improve sellers' expected opportunities. The sign and magnitude depend on the activity. More advertisements can support content while also reducing the attractiveness of the user experience. More providers can shorten consumer waiting time while reducing earnings per provider.
Distinguish a direct same-side benefit, a cross-side benefit, and congestion. A messaging network may offer direct contact with additional users. A hiring marketplace may create value through applicants and employers finding one another. A delivery platform can suffer congestion in pickup queues even when more users join. Positive and negative effects can coexist.
Armstrong's participation models (2006) identify how pricing incentives change with the way agents affiliate. These comparative mechanisms suggest that the observable number of registered accounts is an inadequate measure of effective network size. Relevant participants must be active, available, compatible, and sufficiently distinct to increase the feasible set of useful interactions.
An empirical investigator should separate the effect of participation on value from common demand shocks. A favorable local economy can increase both worker registrations and customer orders without a causal externality between them. Exogenous entry, randomized exposure, capacity variation, or credible timing differences can improve identification, provided interference across groups is incorporated. The object is an economically useful increment to the feasible interaction set.
An illustrative participation calculation
The following arithmetic is a pedagogical specialization of interdependent demand, not an estimated platform model or a proposed new theory. Let n_A and n_B be masses of active users measured in arbitrary units. Let p_A and p_B be membership fees in common monetary units per active user. Suppose that within an interior linear range,
Substitution gives
The denominator follows from the cross-side product 0.4 times 0.5, which is 0.2. Under the illustrative simultaneous adjustment rule, the product below one gives convergent interdependent responses. It does not guarantee uniqueness if nonlinearities, discontinuities, expectations, or capacity limits are introduced.
With fees of 20 on each side, participation equals 130 and 125. With p_A = 10 and p_B = 30, the aggregate posted price remains 40 while participation becomes 137.5 and 118.75. Gross fee revenue changes from 5,100 to 4,937.5. If every potential pair were equally feasible, the product of group sizes would increase from 16,250 to 16,328.125; that product is merely a count-like proxy and not observed transactions or utility.
This calculation establishes that holding the sum of prices fixed need not preserve participation, the opportunities for interaction, or revenue. The distinction is central to Rochet and Tirole (2006). Actual welfare requires benefits and costs, feasible interactions, transfers, participation outside the platform, and the effect of congestion. The example deliberately leaves those elements unspecified, so a welfare ranking cannot be inferred from the arithmetic alone.
Pricing across interacting groups
A conventional posted price covers the marginal use of a service. Platform prices can also change the value of affiliation for other agents. Pricing decisions therefore include membership fees, per-transaction charges, subsidies, advertising, and indirect charges attached to payment or fulfillment.
Suppose a buyer pays 100 for a transaction, the seller receives 85, and the intermediary retains 15. The 15 is gross intermediation revenue under these stated assumptions. It is not economic profit: payment processing, fraud prevention, refunds, customer support, and technology costs may be incurred. If the intermediary separately collects a membership fee, total revenue includes it. If a tax is included in the posted amount, accounting must isolate the tax before interpreting the spread.
The side that benefits other participants may be subsidized when attracting it increases total platform revenue. Equilibrium price structure also reflects the relative elasticities of participation, marginal service costs, competitive conditions, and the extent of user bargaining. Rochet and Tirole (2003) and Armstrong (2006) give formal accounts of such interdependence. No universal rule says one side must always face a zero or negative fee.
Policy evaluation must treat the incidence of a fee change as endogenous. Lower merchant fees may be passed on as lower prices, higher quality, larger merchant margins, or some combination. A reduction in revenue might reduce buyer subsidies; competition might instead compress the platform's margin. The investigator should measure observed quantities, fees, and quality on both sides after the change, rather than assume that the party nominally charged bears the burden.
Homing, switching, and the competitive bottleneck
Single-homing occurs when a participant uses one platform for the relevant activity. Multi-homing occurs when the same participant uses several. Both are behavioral and transaction-specific. A person can install five applications yet use one for nearly all purchases. A driver may maintain several accounts but be unavailable to them simultaneously while carrying a passenger.
Armstrong (2006) studies the competitive bottleneck case, in which platforms compete to attract a single-homing group and supply access to that group to participants who use multiple platforms. In this regime, competition for one group can coexist with power over the other. The relevant question is whether a buyer or seller can bypass an intermediary to reach the specific desired counterpart.
Real switching costs include relearning software, migrating inventory and ratings, completing compliance requirements, changing workflow and settlement arrangements, losing complementary purchases, and uncertainty about future demand. Contractual terms can make these costs larger. So can useful specialization that both parties initially welcomed.
A practical test compares the cost of adding a second platform with the cost of replacing the first. A seller might duplicate a catalog cheaply but face large operational costs maintaining synchronized stock and responding to distinct dispute procedures. A buyer might install a rival cheaply but find that their preferred supplier is absent. Measures of active multi-homing, incremental transaction shares, and overlap of reachable counterparties improve the analysis.
Competition depends on what affiliation makes possible. A provider unable to transfer established customer relationships may find exit consequential even when creating a new account elsewhere takes seconds.
Expectations, entry, and market tipping
A new platform confronts expectations about the population it will attract. A consumer waits until suppliers join; a supplier waits until consumers arrive. A temporary subsidy, guarantee, or partnership can coordinate entry. The underlying reasoning is that a service which has little value in isolation can have substantial value when enough counterparties are present.
The framework of network compatibility (Katz and Shapiro 1985) and the subsequent platform-competition literature make expectations a relevant element of demand. They motivate the following conditional analysis. When benefits from additional participation are strong, switching costs are significant, and compatibility is weak, early success can reinforce itself. When capacity is constrained, preferences are heterogeneous, and networks are accessible across suppliers, several coordinated systems may coexist.
Tipping should be defined through a change in the stability or attraction of participation equilibria in a specified environment. High concentration by itself is an outcome, not evidence that a particular tipping mechanism produced it. A firm may grow because its product is better, because a rival exited for unrelated reasons, because the addressable market expanded, or because access to network participants became exclusive.
Entry policy consequently depends on diagnosis. A new entrant unable to obtain scale because of a costly standard may benefit from compatibility. One unable to replicate the incumbent's low costs may require technological innovation rather than a compatibility rule. An entrant unable to attract users because default arrangements block distribution may require a different intervention. These hypotheses imply distinct measures of user behavior, costs, and access.
Scale economies, scope, and infrastructure
Fixed expenditure on software development, fraud detection, catalog systems, and security may be spread over many transactions. Scope economies may arise when related activities share infrastructure, a merchant directory, or customer-support operations. These efficiencies differ from network effects: the former operate through the cost of supplying the service; the latter operate through the value participants obtain.
A market with substantial fixed costs can sustain a small number of efficient suppliers even without cross-side externalities. Conversely, a network with strong externalities can operate through many firms if compatible participation allows the network value to be shared. The distinction matters when considering entry remedies and industry structure.
A platform's marginal cost includes more than computing. An additional transaction may require payment services, insurance, moderation, dispute resolution, or physical delivery. Unit costs can rise when an intermediary expands into a difficult-to-serve market. Measuring marginal cost solely as server usage understates the resource requirements of verified exchange.
Economies of scope become economically significant if producing activities jointly costs less than producing them separately at equivalent quality. The comparison must include implementation, organizational complexity, and the cost of preventing conflicts of interest. Combining a marketplace with a logistics business may lower delivery costs while creating incentives to favor the affiliated service. An estimate of the first effect cannot settle the second.
Search, ranking, and the allocation of attention
Search technology reduces the cost of identifying alternatives. Ranking technology establishes the order in which alternatives receive scarce attention. Search and ranking can facilitate competition by making price and quality differences visible. They can also alter the value of seller advertising, create barriers to visibility, and affect the distribution of transaction opportunities.
Dinerstein and colleagues' study of eBay search design (2018) investigates this tension using browsing records, a marketplace-design change, and an equilibrium search model. The study distinguishes directing buyers toward desired products from creating incentives for sellers to compete on price. The specific application should guide inquiry without supplying a universal ranking rule.
Imagine identical consumers who click only the first result. A change from alphabetical ordering to ordering by delivered price could strengthen price competition. If products are differentiated and consumers differ in preferences, the same change may produce worse matches. Ranking a seller by conversion probability can favor sellers with already prominent positions, because exposure itself generates observations used for prediction.
Investigators therefore need exposure, click, purchase, return, and long-run seller-entry information. Conversion conditional on exposure differs from conversion per eligible offer. A platform that improves immediate conversion by narrowing variety may reduce future choice. Experiments should measure consumer search cost and match quality as well as equilibrium price response.
The policy implication is that transparency of a ranking factor alone does not resolve its economic effect. Investigators need to know the objective pursued, the constraints imposed, the alternatives affected, and whether commercial relationships influence treatment conditional on relevant quality.
Matching and the geography of network thickness
A market is thick when economically relevant participants are available at a given place, time, and specification. Thickness is a property of feasible encounters rather than global registration. A restaurant delivery order requires a restaurant, customer, and courier within compatible distances and time windows. A remote software assignment may be matched internationally while requiring language or skill compatibility.
Cullen and Farronato's study of online task matching (2021 issue, published online 2020) links variation in matching efficiency to geographic proximity and task standardization. The study's subject is a peer-service marketplace and its implications are strongest where local and time-sensitive availability matters.
A platform can improve coverage by widening service boundaries or changing scheduling. That intervention can reduce average waiting time while increasing distance traveled. A second mechanism changes information: more accurate forecasts and commitment signals reduce uncertainty without increasing the number of workers. The two mechanisms have different resource and welfare consequences.
Take a hypothetical neighborhood with ten available providers and ten requests during an hour. Perfect geographic compatibility would permit ten matches. If five providers and five requests occur in isolated subareas in mismatched intervals, the realized number may be lower. A national increase in registrations changes nothing unless it increases relevant local availability. The arithmetic of total users is not the arithmetic of matching.
Measure the probability of fulfillment per attempted request, waiting time distribution, unmet demand, travel cost, provider idle time, and the heterogeneity of service types. The market may be highly competitive at a national level and effectively concentrated in a local time slot.
Reputation and the infrastructure of trust
A transaction can fail because participants lack knowledge of the other party's reliability. Reputation mechanisms convert recorded activity into signals that affect future exchange. An intermediary can combine ratings with identity checks, transaction guarantees and dispute resolution. The economic function is to reduce the expected cost of dealing with strangers.
Resnick and colleagues' controlled eBay reputation experiment (2006) provides an example of testing the economic significance of seller reputation rather than merely correlating ratings with sales. This dossier uses the study's experimental setting as a methodological precedent and does not extrapolate a numerical effect.
Ratings can reflect transaction quality, selection into reviewing, the buyer's expectations, exposure, or the consequences of a dispute. A seller's average score is not necessarily comparable to that of another seller if they serve different customers. A new provider can face a disadvantage even with high ability because accumulating a recorded history takes time.
Reputation can therefore become an economically meaningful asset that is partly controlled by the intermediary. If reviews are portable, an entrant may overcome some switching costs; if their verification or context is lost, portability may create new opportunities for manipulation. A credible design must preserve identity, transaction authenticity, relevant context, privacy, and contestability of errors.
Trust should be evaluated through fraudulent or failed transactions, dispute duration, restitution, entry by unfamiliar suppliers, and correction of inaccurate information. A platform's score is an intermediate measurement. The object of economic interest is the degree to which trustworthy exchange becomes feasible.
Data as an input and externality
Digital coordination generates observations about searches, prices, location, purchases, complaints and failed interactions. Data can improve matching, demand forecasts, and fraud detection. Its economic characteristics differ by purpose, quality, timeliness and substitutability. An additional observation may offer little value when the relevant prediction is already accurate, or substantial value for a rare event.
The value of data is jointly determined by complementary software, computational resources, organizational capabilities and access to the activity that generates it. Larger datasets need not produce proportionally better services. Conversely, control of highly relevant, recurring observations can provide advantages difficult to reproduce using generic information.
Acemoglu and colleagues' analysis of data-market inefficiencies (2022) studies a setting in which information about one person can inform inference about others. This provides a reason to evaluate the effects of data sharing beyond the bilateral transaction. The model is theoretical; it does not establish that a specific platform has collected an inefficient amount of data.
An institution allowing a person to export their own records can reduce a migration cost without transferring every record about other participants. Aggregation, authentication and purpose constraints determine whether the exported data are useful and whether disclosure creates risks. The economic trade-off is between the value of reuse, investment incentives, confidentiality, and external effects.
Data power is consequently a relational property. Researchers should identify the decision improved by the information, the substitutes that exist, the cost of obtaining them, and the ability of other organizations to make productive use of them. The quantity of data stored is not enough.
Algorithms, prediction, and feedback
An intermediary may use prediction systems to allocate advertising slots, choose offers, assign work, detect fraud or recommend prices. The algorithm supplies a decision rule embedded in contractual and technical arrangements. Economic analysis examines the incentives and consequences of that rule, including consequences for participants who cannot inspect it.
Feedback can connect predictions to subsequent training data. An offer ranked first receives more views and potentially more sales, producing observations that may validate the ranking. This does not establish intentional discrimination: exposure can be an independent cause of successful outcomes. Identification requires comparing comparable offers subject to different exposure or using randomized exploration where appropriate.
Calvano and colleagues (2020) examine algorithmic pricing and collusion through controlled simulations involving learning algorithms in a repeated pricing environment. Their study supports inquiry into the behavior of adaptive pricing systems under specified experimental conditions. It cannot by itself establish actual coordinated pricing, a legal agreement, or prevalence across digital markets.
A meaningful empirical audit distinguishes policies that recommend price from systems that set price, and individual seller autonomy from shared pricing agents. If many independent suppliers delegate decisions to a common service, the correlations in their observed prices could result from common costs, common forecasts, or common optimization. Detecting coordinated effects requires counterfactual evidence rather than a label applied to software.
The economics of learning also creates a measurement obligation: the relevant treatment may be a changing policy with delayed effects. A short-run experiment can capture immediate click responses while missing the altered future set of suppliers and training observations.
Private governance and commitment
A digital platform coordinates through rules backed by technical enforcement, contract and dispute procedures. Eligibility requirements determine the feasible supply set; ranking determines exposure; payout rules affect working capital; return and suspension policies govern risk. These arrangements are economically substantive decisions.
Consistent and predictable governance encourages participant investment. A merchant considering a specialized listing system may need assurance that access will continue long enough to recover its cost. A worker investing in equipment may need intelligible terms for deactivation and appeal. The intermediary also needs flexibility to respond to abuse, new technology, and changes in demand.
Consider a supplier making an irreversible investment of 1,000, expected to generate an incremental return of 1,200 before any later change in the platform's fee. If the platform can subsequently capture 300 of that surplus, the supplier expects to recover only 900 and may rationally decline the investment. These are hypothetical values. Their economic purpose is to show how expected future terms can change present investment independently of the current nominal fee.
Commitment devices can include stable contractual periods, advance notice, review procedures and independent adjudication. Each has costs: rigid rules may delay legitimate changes, while discretionary processes can create uncertainty. The appropriate comparison is between achievable governance arrangements rather than between actual discretion and a frictionless ideal.
Private governance becomes a market-power issue when participants cannot reasonably escape or contest decisions that substantially alter their economic prospects. A technically sophisticated intermediary can remain disciplined by viable substitutes; dependence and the absence of practical recourse are therefore central observables.
Vertical integration and complementor dependence
A platform owner may operate a marketplace while supplying goods, advertising, fulfillment, payment services or other complements. Integration can improve quality and reduce coordination expenses. It can also introduce conflicts when the institution determining access and visibility competes with firms subject to its rules.
Zhu and Liu's Amazon complementor study (2018) investigates platform owner entry into third-party product areas, including the relationship between entry and seller efforts. The observed behavior is relevant to incentives for platform-specific investment. It does not establish that all integration is harmful, nor does the study alone identify the effect of every ranking or contractual rule.
A useful counterfactual distinguishes entry into production from preferential treatment by the intermediary. In one arrangement, the owner sells a product under the same discovery and settlement conditions as independent sellers. In another, the owner's product receives advantageous placement unrelated to the estimated value delivered to consumers. In a third, the owner supplies a common logistical service available on comparable terms. Each changes the feasible outcomes differently.
Evidence capable of distinguishing these possibilities includes comparable product characteristics, fulfillment performance, exposure by query, ranking changes after entry, cost information and seller exits. The inquiry also needs an inventory of efficiency explanations, including delivery reliability and integrated dispute handling.
A seller's short-run loss of sales can coexist with a consumer's lower price. A longer-run reduction in independent innovation can coexist with improved short-run execution. It is the magnitude, timing, and causal connection of these changes that matters for welfare.
Labor platforms and the organization of risk
Digital labor intermediaries make work opportunities available through online matching and administration. Some arrange local services, while others facilitate remote tasks. Coordination can improve access to customers and flexibility. It can also shift idle time, equipment costs, insurance obligations and demand variability onto providers.
The International Labour Organization's analysis of digital labor platforms (2021) identifies algorithmic management and the assignment of operational responsibilities as important institutional features. These observations motivate a measurement framework that distinguishes gross platform payments from net economic income.
Suppose a worker receives 120 for a day of platform activity, incurs 30 in variable operating expenses, and spends eight hours including unpaid waiting. The arithmetic produces 90 before taxes and additional fixed or insurance costs, or 11.25 per hour of total time. The figure is illustrative; it is not an estimate of actual wages. Counting only five hours spent performing paid tasks would instead yield 18 per hour on the same partial income measure.
The distinction also matters for flexibility. A worker who can freely choose productive hours has an option of economic value. A worker whose access to orders depends on accepting undesirable tasks may face a practical constraint despite formal scheduling freedom. The investigator must observe availability, acceptance, penalty, earnings variability, and opportunities outside the service.
Employment classification depends on applicable law and actual working relationships. The EU's Platform Work Directive (2024) establishes a legal framework addressing platform-work conditions and algorithmic management; national implementation and individual cases require separate examination. Economic dependence and legal employment status are related questions with different evidentiary and jurisdictional requirements.
Advertising, zero-price services, and attention
Advertising-supported services connect users, advertisers and sometimes content suppliers. A zero monetary access price is compatible with considerable economic value. Users may provide attention and information or accept restrictions; advertisers pay for the prospect of reaching relevant audiences. The monetary price observed on one side accounts for only part of the exchange arrangement.
The CMA's digital advertising market study (2020) examined user-facing platforms and advertising-related competition in the United Kingdom. Its published findings led it to recommend a new pro-competition regime. Those findings concern the investigated setting and period; they should not be treated as contemporary global market shares.
An advertising intermediary can improve allocation through more relevant impressions, lower campaign administration costs and better measurement. Problems arise when participants cannot independently evaluate measurement, compare effective prices, or switch to competing channels. Gross advertiser spending and amounts ultimately received by publishers are different economic quantities.
Attention cost is difficult to monetize. An additional advertisement can impose delay or distraction; it may also introduce useful information. Privacy costs depend on the information collected, the inferences enabled, and individual preferences. The relevant quality measure could include latency, ad load, content diversity, and user control.
For users facing no fee, a hypothetical deterioration in service quality or increase in data collection may reveal substitution responses. Unlike a conventional percentage price increase from a positive baseline, such a change requires a carefully specified quality metric. Evaluation should therefore integrate monetary and nonmonetary dimensions without declaring attention or personal data equivalent to a cash payment.
Measurement of platform power
Market power is the ability to influence economic terms relative to an appropriate competitive counterfactual. Platform power in this dossier refers to the economically consequential ability to determine conditions of access and interaction. It can manifest in prices, quality, visibility, rules, and the bargaining conditions faced by participants.
A high market share can be a starting signal but cannot establish the persistence or source of power. Shares depend upon the denominator: active users, transactions, gross merchandise value, advertising revenue, hours, or eligible searches can yield different rankings. A platform that intermediates high-value transactions may have a large value share despite a smaller transaction share.
The OECD antitrust report (2018) examines the suitability of conventional competition tools in multisided settings. Operational analysis should incorporate cross-side feedback, indirect substitution, costs of affiliation, and the distinct functions supplied by participants.
Price-cost margins can be informative when calculated with consistent accounting boundaries. For a marketplace, the relevant price can be net intermediation revenue per transaction and the relevant costs may include the cost of providing the coordination service. Comparing a fee percentage with the marginal cost of computing is not a valid margin estimate.
An alternative empirical strategy examines a rule change and participant responses. A substantial fee increase followed by little exit may indicate limited substitution, but contract durations, delayed adaptation and common demand conditions must be considered. A change in quality or appeal processes can have similarly informative effects. Sustained profitability, returns on investment and innovation evidence contribute to interpretation but none can resolve identification alone.
Market definition and the transaction distinction
Some platforms facilitate an individual transaction simultaneously involving both sides. Others sell advertising to one group while supplying a service to another, without linking each purchase to a single joint transaction. The structure affects demand, price measurement and the analysis of substitution.
The US Supreme Court's Ohio v. American Express decision of 2018 treated credit-card transaction services as a market encompassing both sides for the case at issue. The dissent challenged the majority's reasoning. The legal holding concerns the litigation and the relevant transaction platform; applying it to every digitally mediated market would erase economically relevant differences.
Consider a card network and an online newspaper. Both have multiple interacting groups, but a card transaction requires acceptance by the merchant and use by the cardholder for that particular purchase. The sale of advertising by a newspaper can occur without a corresponding purchase from that advertiser by an identifiable reader. Their cross-side relations differ.
Market definition must begin with the disputed activity and the substitution responses of affected participants. A narrow analysis of seller access can sometimes miss a buyer-side constraint on the intermediary. Conversely, aggregating unrelated activities may obscure power over a specific gateway. The investigator should state the jurisdiction, the service, the alleged conduct, the time period, and the counterfactual.
The larger analytical conclusion is methodological: two-sidedness identifies relationships requiring analysis. It does not automatically settle the legally relevant market, establish power, or excuse a restraint.
Distributional access and the search case
Distribution can be the link between technical performance and market participation. A new search engine may be available in principle while facing difficulties obtaining prominent placement on devices, browsers or access points. Defaults can reduce search and configuration costs for users and may also channel demand toward the selected provider.
The US Department of Justice's Google remedies announcement (2025, updated 2026) reports the district court's August 2024 monopolization finding and the subsequent remedies proceedings. The announcement emphasizes distribution agreements and default access in the government's account. It is an official litigant summary: this dossier does not infer the final outcome of appeals or claim to independently evaluate the full evidentiary record.
A distinction is necessary between an efficient choice of default and a contractual arrangement that forecloses access. Default selection may serve users who prefer the selected service and reduce configuration costs. An exclusivity condition can have an additional effect by preventing rival distribution even among users who would prefer another product.
The testable economic implications include changes in search share following default variation, consumer retention after choice is offered, the revenue and cost of distribution agreements, and the effect on rivals' ability to generate observations for service improvement. Comparative data should distinguish loyalty to quality from behavioral inertia and contractual constraints.
This case illustrates why market entry involves more than product production. Access to the decision point through which demand is expressed can be a critical input to rivalry.
Welfare and total economic surplus
Platform coordination creates benefits when it reduces transaction costs, improves matches, increases useful supply, or enables exchange that otherwise could not occur. It consumes real resources in providing software, verification, logistics, moderation and other services. It may also generate external costs outside the participant set. Welfare analysis must specify whose benefits and costs enter the calculation.
For a clearly defined transaction, let V be the buyer's valuation, C the supplier's opportunity cost, K the real resource cost of coordination, and E the uncompensated external harm. The total surplus under these simplified assumptions equals V minus C minus K minus E. Transfers between these parties affect distribution and participation. The calculation omits uncertainty, future effects and distributional weights unless they are introduced explicitly.
Consider a transaction valued at 120 by a buyer. A supplier incurs 70 in opportunity cost and a platform incurs 10 in real coordination cost. With no external harm, total surplus is 40. If the buyer pays 100 and the supplier receives 85, the buyer gains 20, supplier gains 15, and platform gains 5. Changing the supplier payout to 75 transfers 10 from the supplier to the platform, holding the transaction and other quantities fixed. Total surplus stays at 40. If the supplier subsequently declines to participate, surplus may change because the transaction no longer occurs.
The static identity reveals the difference between a transfer and an efficiency effect. Distribution is economically consequential through entry, investment, labor supply, and the marginal utility of income. A credible welfare analysis tracks those dynamic mechanisms rather than dismissing them.
Cohen and colleagues' ride-hailing consumer-surplus working paper (2016) illustrates one way of studying the demand-side benefit. A consumer surplus estimate is a component of welfare. It does not include the full resource costs, provider outcomes, or effects on nonusers.
Accommodation, supply flexibility, and local costs
The accommodation market illustrates a different coordination problem from search. Dedicated suppliers maintain capacity across time, while households may offer temporary spare capacity when the expected return is sufficiently high. An intermediary can increase flexible supply by reducing the costs of finding guests and administering payment and reputation.
Farronato and Fradkin's Airbnb welfare study (2022) estimates a competition model for major US cities in which peer hosts and hotels offer differentiated accommodation. The publisher abstract reports that gains were concentrated at locations and dates with constrained hotel capacity. The result is conditional on the studied markets, data and estimated counterfactual and should not be generalized to every destination.
The economic mechanism is especially important under capacity constraints. Additional supply at a peak can reduce scarcity prices and permit trips that would otherwise be forgone. Off-peak entry may substitute more strongly for dedicated suppliers and generate smaller gains. Different patterns of land use and costs create further variation.
The scope of private accommodation welfare differs from city-wide welfare. A property owner's additional income and a tourist's reduced cost may coexist with additional noise, housing-market pressures, enforcement costs or neighborhood benefits. Such effects require independent local evidence, not an inference from the accommodation-market study.
A local evaluation should measure the nights genuinely added to market capacity, displaced long-term uses, variability across neighborhoods and dates, actual resource costs, and housing and neighborhood outcomes. The feasible alternative could be a hotel expansion, a regulated short-term rental system, or the absence of additional supply.
Measurement beyond monetary output
Transactions at positive prices are comparatively visible in conventional economic accounts. Many digital services are offered without a direct monetary charge and financed by other activities. A user can derive substantial value from a free information service, yet that value is not identical to the market output recorded in national accounts.
Brynjolfsson and colleagues' GDP-B working paper (2019) considers welfare-oriented measurement of new and free goods. Its research agenda motivates separate treatment of output measures and valuations. The dossier does not treat willingness to accept compensation for temporary loss of access as an amount to add mechanically to national GDP.
Willingness to pay is bounded by resources and differs from willingness to accept compensation in environments with limited substitutes, endowment effects or income constraints. A valuation estimated through experimental choice may be affected by the duration of exclusion, participants' expectations, sample composition and compliance.
A digital platform's transaction volume can also be confused with its output. Gross merchandise value may primarily represent the goods sold by independent merchants; the intermediary's own revenue is a distinct flow. If a transaction is recorded as both platform output and merchant production without appropriate accounting treatment, double counting may result.
For comparative studies, publish separately the gross transaction value, platform fees and other revenue, relevant production costs, participant income, and estimated welfare measures. State the accounting perimeter, time interval, price base, and unit. This discipline allows the reader to understand how economic activity and economic well-being relate without confusing them.
Empirical identification and research designs
The major identification problem is that platform design, participant choice and market outcomes are jointly determined. A platform may change a ranking system in anticipation of a shift in demand; suppliers may adapt after learning the new rule; consumers may change their search strategy. Observed outcomes reflect the combination.
Randomized experiments can identify average effects under a specified assignment. In a networked market, a treatment assigned to one supplier may affect untreated suppliers competing for the same demand. Conventional assumptions of independent outcomes can fail. Cluster assignment across suitable geographic areas or time intervals may help, but treatment spillovers and market anticipation require explicit design.
Natural experiments may arise from changes in regulation, availability or contractual terms. Difference-in-differences designs require a defensible comparison trend; adoption may be selective and effects heterogeneous. An event study can reveal dynamics but cannot establish causality unless alternative changes are addressed. A platform's endogenous decision to modify rules is not itself an exogenous event.
Structural estimation can make a specified counterfactual explicit by estimating choice, search, costs and equilibrium responses. The approach requires model-dependent assumptions and careful validation. The eBay platform-design study (Dinerstein et al. 2018) and the accommodation welfare study (Farronato and Fradkin 2022) provide examples of data and model jointly informing market-design questions. Their estimates are not directly pooled in this dossier.
Case studies and legal records provide institutional detail and mechanisms that transaction data may omit. They also reflect selection into investigation and litigation. Combining evidence across methods is valuable when each addresses a different uncertainty, rather than when several sources merely repeat the same claim.
Evidence quality and cross-market comparison
A credible comparison has to hold the economic unit constant or explicitly account for differences. A global app store, local ride-hailing service, national advertising exchange and cross-border remote labor marketplace do not share identical capacity constraints, homing patterns, or regulatory environments.
The peer-to-peer market review (Einav, Farronato, and Levin 2016) places search, matching and reputation within market design. The task-matching and accommodation studies show why the same general coordination functions can manifest differently in geographically and temporally constrained settings.
The following table summarizes conditional mechanisms rather than numerical findings or a ranking of industries.
| Activity | Main coordination object | Salient power channel | Relevant evidence |
|---|---|---|---|
| Product marketplace | Discovery and seller-buyer transactions | Visibility, fees, access to customers | Exposure, prices, fulfillment, seller entry |
| Local task or ride service | Time- and place-specific matching | Availability, dispatch, provider dependence | Waits, idle time, net income, local substitutes |
| Accommodation | Flexible supply at particular dates | Booking visibility and access to customers | Capacity, prices, geographic substitution |
| Advertising | Allocation and measurement of attention | Access to audience and verification | Net publisher revenue, advertiser alternatives, measurement accuracy |
| Software distribution | Access to installed devices and users | Installation, payment or interface restrictions | Distribution costs, switching, security incidents |
| General search | Discovery and traffic | Defaults and distribution access | Default changes, retention, quality, entry |
The table makes differences in the counterfactual observable. A study of increased competition in advertising cannot directly identify the welfare effect of competition among couriers. Knowledge can transfer across markets at the level of mechanisms, while estimates require evidence from the corresponding activity.
The European gatekeeper approach
The European Union adopted Regulation (EU) 2022/1925 in September 2022. Its stated purpose concerns contestability and fairness in digital-sector markets involving gatekeepers. Designation is attached to identified core platform services, a distinction that matters when a single company supplies several activities.
The European Commission's DMA explanatory overview describes the regulation as complementary to competition law and identifies measures concerning certain interoperability situations, access to business-generated data, independent advertising verification, and preferential ranking. It also records the general applicability date of May 2, 2023. The dossier analyzes this institutional architecture rather than the current list of designations or individual compliance findings.
The economic rationale of advance obligations is that a case-specific intervention may arrive after distribution advantages, expectations and irreversible investments have altered the competitive environment. The advantage of an advance rule is predictability and timeliness. Its difficulty is calibrating the rule to diverse services, technical constraints and changing efficiencies.
For each obligation, an evaluation should specify the restriction being addressed and the response expected. An interoperability rule has a different intended mechanism from a prohibition on preferential treatment. A requirement for business users to access relevant data addresses a different input from a restriction on certain tracking. Success cannot be inferred solely from the fact that a formal interface or policy exists.
Indicators include the effective use of alternative services, implementation and compliance costs, the quality and security of interactions, and the trajectories of independent entry and investment. The legal design creates a testable setting; its realized impact remains an empirical matter.
The United Kingdom and Japan
The United Kingdom's Digital Markets, Competition and Consumers Act (2024), Part 1, provides for designation of strategic market status in respect of a digital activity. The statutory design identifies particular activities and confers regulatory functions on the Competition and Markets Authority. This follows the institutional concern expressed in the 2020 digital-advertising market study, while remaining a legal framework with its own requirements.
Japan's Mobile Software Competition Act overview describes Act No. 58 of 2024, which addresses specified smartphone software including operating systems, application stores, browsers and search engines. The Japan Fair Trade Commission states that the act entered fully into effect on December 18, 2025, following earlier commencement of designation provisions.
The comparison is useful because both approaches direct attention toward specific activities, but their statutory scopes and regulatory structures differ. The United Kingdom provides a regime for relevant digital activities under its designation criteria; Japan targets specified smartphone software. The European Union framework covers designated core platform services under its own legislation.
The relevant research questions are whether designation identifies durable bargaining and access problems; whether measures are targeted to the mechanism causing them; and whether compliance changes actual participant choices rather than formal policies alone. Cross-country comparisons require consistent definitions of the regulated service, baseline switching costs, technological differences, existing enforcement and the date of exposure.
No claim is made that one jurisdiction has already achieved superior welfare outcomes. The cases identify different institutional experiments and the measurements needed to compare their results.
Interoperability, portability, and access remedies
Interoperability enables participants using different systems to interact under a compatible interface or protocol. Portability enables a participant to move some records or assets between systems. Multi-homing enables simultaneous affiliation. These mechanisms can reduce different costs and should be evaluated separately.
A seller who exports a catalog may save setup time but still lose access to the original platform's buyers. A messaging user who can exchange messages across systems may preserve counterpart access but lose certain product-specific functionality. A worker who can export verified records of completed tasks might reduce the cost of building a reputation elsewhere without changing which platform supplies orders.
The European Commission's gatekeeper overview illustrates several distinct access obligations. A general economic assessment requires the exact scope of a proposed interface, its authentication and privacy safeguards, implementation cost, quality and security implications, and how participants actually use it.
Compatibility can preserve benefits that depend on network reach while permitting competition in services built over the network. However, a common technical requirement may impose fixed costs on small entrants; poor design may create vulnerabilities or reduce the opportunity for differentiation. These costs are particularly relevant when protocols must change rapidly.
A remedy should address the demonstrated bottleneck. If users cannot move information, portability may help. If the obstacle is reaching users who remain elsewhere, interoperability or alternative distribution may be more relevant. If the obstacle is exclusionary contract terms, a narrower contractual remedy may suffice. A well-specified evaluation should measure the actual reduction in switching or access costs and the accompanying effects on resource use and service quality.
Counterexamples and unintended effects
Economic mechanisms should be evaluated against cases where their predicted effect is absent or reversed. Strong network benefits need not produce a single supplier if systems interoperate. Free access need not imply either vigorous competition or market power. Vertical integration can improve quality and reduce cost when conflicts are effectively controlled. A high commission may finance valued services.
The observed price response to a fee cap is ambiguous without further assumptions. If the intermediary has durable market power and fixed costs are already covered, reducing a fee may transfer surplus to business users. If the fee finances services valued by consumers, the intermediary may reduce them or charge the other side. The magnitude depends on competition, cost and contractual arrangements.
Mandatory transparency can improve participants' understanding of decisions. It can also make ranking easier to manipulate or reveal information that facilitates coordination among competitors. A design requiring meaningful reasons and audit records can sometimes address accountability without revealing every operational parameter.
Likewise, a consumer's ability to install an alternative application should not be treated as proof that a market is contestable. Installation is only one component of effective substitution. The alternative must supply sufficient quality, counterpart access and economic opportunity. Conversely, a dominant service may face a credible threat from a differentiated product even with low observed switching in a stable period.
These counterexamples do not invalidate the underlying economic mechanisms. They specify the boundary conditions under which an inference about welfare or power is justified.
Original analytical synthesis
The preceding research supports an integrated interpretation of platform economics. Digital coordination combines transactions, information and institutional rules. Its achievements and its risks emerge from the same practical capacity: the ability to shape a sustained environment in which independent activities become mutually compatible.
The first analytical distinction is between the value of the organized interaction and the source of the intermediary's bargaining position. Better search, verification or availability can increase the gains from trade. Access restrictions, contractual dependence or exclusive distribution can alter the division of those gains and influence future competition. A single observed price or profit measure cannot identify the contribution of each mechanism.
The second distinction concerns time. A price that maximizes participation today may encourage specialized investment and subsequently increase dependence. A remedy that increases immediate switching can alter investment in common infrastructure and participant trust. Evaluation of digital coordination requires a sequence of decisions, not merely a static comparison of market shares.
The third distinction concerns the object of substitution. Participants can replace an application, a particular transaction, a reputation environment, or access to counterparties. These are different substitutes. Competitive discipline is strongest where the consequential economic functions can be replaced at reasonable cost. It may be weak where only the interface can be replaced.
The fourth distinction is between knowledge and authority. An intermediary may have superior information because it observes transactions, and authority because it can enforce rules. Better information can justify particular improvements; it does not establish that every allocation of authority is efficient. Evaluating the two together requires measuring prediction accuracy, the distribution of opportunities, and the incentives created by the rule.
These are authorial analytical conclusions synthesized from established mechanisms and the illustrative calculations in this dossier. They are not offered as an independently validated new theory, named doctrine, or claim to priority. Their practical contribution is a disciplined way to ask what exactly a platform supplies, how that contribution changes feasible exchange, and which control rights determine the availability of alternatives.
Research agenda and standards of inference
Future research should connect participant-level information to evolving rules, rather than relying solely on cross-sectional correlations between firm size and economic outcomes. The most useful evidence would observe the same activity before and after a change in access rules while tracking its affected groups, substitutes, costs and quality.
One research question is whether an alternative distribution channel provides meaningful access to the same users. A suitable study would measure adoption conditional on default variation, repeated use, search or matching quality, and provider investment. The alternative explanations include product quality, learning costs and selection into the new channel.
A second question concerns the portability of verified economic histories. A field design could compare the entry trajectories of providers with different opportunities to transfer authenticated records. Participation and earnings should be measured alongside disputes, manipulation attempts and privacy harms. Transfer must be distinguished from the subsequent ability to obtain transactions.
A third question concerns the welfare effects of changes in ranking. Observational data should record eligible offers as well as exposed offers. The experiment or natural variation should distinguish quality and commercial affiliation. Long-run measurement should include provider investment and exit, since short-run conversion is only one component.
This review uses publicly retrieved bibliographic and abstract-level evidence wherever full texts were unavailable, official summaries for institutional comparisons, and transparent hypothetical calculations. It does not reproduce econometric estimates, independently verify confidential platform data, or adjudicate pending disputes. Those boundaries limit empirical and legal claims. The scope is economic reasoning and comparative institutional research, with no investment, financial or legal service offered.
Conclusion
A digital platform supplies an organized setting for economic interaction. It alters both the costs of finding and completing exchanges and the terms under which independent participants can obtain access to them. Its economic power is consequently best investigated at the level of specific activities and consequential rules.
Multisided demand means that the structure of prices and participation can matter independently of the aggregate price. Compatibility, homing, expectations and localized capacity determine whether scale creates persistent advantages. Search, reputation and data reduce information costs while changing the distribution of opportunities and investment incentives. Governance determines who can enter, remain, and contest decisions. Welfare analysis must include resource costs, participation responses, benefits, risks, distribution and realistic alternatives.
The evidence discussed in this monograph identifies mechanisms and selected applications, with differing levels of empirical and institutional verification. It supports neither an automatic inference from market concentration to harm nor an automatic inference from coordination benefits to competitive discipline. Economic assessment is persuasive when its counterfactual explicitly preserves or evaluates the useful functions of the intermediary while examining restrictions and their consequences.
Digital market coordination is thus a fundamental subject of economic theory and microeconomics. Its investigation connects the organization of exchange to competition, information, institutional commitments, and the distribution and creation of economic surplus.
References
Acemoglu, Daron, Ali Makhdoumi, Azarakhsh Malekian, and Asu Ozdaglar. 2022. “Too Much Data: Prices and Inefficiencies in Data Markets.” American Economic Journal: Microeconomics 14 (4). DOI: 10.1257/mic.20200200. Source.
Armstrong, Mark. 2006. “Competition in Two-Sided Markets.” RAND Journal of Economics 37 (3): 668–691. DOI: 10.1111/j.1756-2171.2006.tb00037.x. Source.
Brynjolfsson, Erik, Avinash Collis, W. Erwin Diewert, Felix Eggers, and Kevin J. Fox. 2019. “GDP-B: Accounting for the Value of New and Free Goods in the Digital Economy.” NBER Working Paper 25695. DOI: 10.3386/w25695. Source.
Calvano, Emilio, Giacomo Calzolari, Vincenzo Denicolò, and Sergio Pastorello. 2020. “Artificial Intelligence, Algorithmic Pricing, and Collusion.” American Economic Review 110 (10): 3267–3297. DOI: 10.1257/aer.20190623. Source.
Cohen, Peter, Robert Hahn, Jonathan Hall, Steven D. Levitt, and Robert Metcalfe. 2016. “Using Big Data to Estimate Consumer Surplus: The Case of Uber.” NBER Working Paper 22627. DOI: 10.3386/w22627. Source.
Competition and Markets Authority. 2020. “Online Platforms and Digital Advertising Market Study.” Final report and accompanying materials, July 1, 2020. London: CMA. Source.
Cullen, Zoë, and Chiara Farronato. 2021. “Outsourcing Tasks Online: Matching Supply and Demand on Peer-to-Peer Internet Platforms.” Management Science 67 (7): 3985–4003. DOI: 10.1287/mnsc.2020.3730. Online October 2020. Source.
Dinerstein, Michael, Liran Einav, Jonathan Levin, and Neel Sundaresan. 2018. “Consumer Price Search and Platform Design in Internet Commerce.” American Economic Review 108 (7): 1820–1859. DOI: 10.1257/aer.20171218. Source.
Einav, Liran, Chiara Farronato, and Jonathan Levin. 2016. “Peer-to-Peer Markets.” Annual Review of Economics 8: 615–635. DOI: 10.1146/annurev-economics-080315-015334. Source.
European Commission. n.d. “About the Digital Markets Act.” Official explanatory page on Regulation (EU) 2022/1925 of September 14, 2022. Accessed November 17, 2026. Source.
European Parliament and Council of the European Union. 2022. Regulation (EU) 2022/1925 of September 14, 2022, on Contestable and Fair Markets in the Digital Sector. Official Journal L 265, October 12, 2022, 1–66. Source.
European Parliament and Council of the European Union. 2024. Directive (EU) 2024/2831 of October 23, 2024, on Improving Working Conditions in Platform Work. Official Journal L, 2024/2831, November 11, 2024. Source.
Farronato, Chiara, and Andrey Fradkin. 2022. “The Welfare Effects of Peer Entry: The Case of Airbnb and the Accommodation Industry.” American Economic Review 112 (6): 1782–1817. DOI: 10.1257/aer.20180260. Source.
Hagiu, Andrei, and Julian Wright. 2015. “Multi-Sided Platforms.” International Journal of Industrial Organization 43: 162–174. DOI: 10.1016/j.ijindorg.2015.03.003. Source.
International Labour Organization. 2021. “The Role of Digital Labour Platforms in Transforming the World of Work.” Research resource, June 1, 2021. Geneva: ILO. Source.
Japan Fair Trade Commission. n.d. “Approaches in the Digital Market.” Official summary of the Mobile Software Competition Act, Act No. 58 of 2024. Accessed November 17, 2026. Source.
Katz, Michael L., and Carl Shapiro. 1985. “Network Externalities, Competition, and Compatibility.” American Economic Review 75 (3): 424–440. Source.
OECD. 2018. Rethinking Antitrust Tools for Multi-Sided Platforms. Paris: OECD Publishing. DOI: 10.1787/a013f740-en. Source.
Resnick, Paul, Richard Zeckhauser, John Swanson, and Kate Lockwood. 2006. “The Value of Reputation on eBay: A Controlled Experiment.” Experimental Economics 9 (2): 79–101. DOI: 10.1007/s10683-006-4309-2. Source.
Rochet, Jean-Charles, and Jean Tirole. 2003. “Platform Competition in Two-Sided Markets.” Journal of the European Economic Association 1 (4): 990–1029. DOI: 10.1162/154247603322493212. Source.
Rochet, Jean-Charles, and Jean Tirole. 2006. “Two-Sided Markets: A Progress Report.” RAND Journal of Economics 37 (3): 645–667. DOI: 10.1111/j.1756-2171.2006.tb00036.x. Source.
Supreme Court of the United States. 2018. Ohio v. American Express Co., 585 U.S. 529, No. 16-1454, decided June 25, 2018. Majority opinion and dissent, Cornell Legal Information Institute reproduction. Source.
U.S. Department of Justice. 2025. “Department of Justice Wins Significant Remedies Against Google.” Press Release 25-902, September 2, 2025, updated April 15, 2026. Source.
United Kingdom. 2024. Digital Markets, Competition and Consumers Act 2024, c. 13, Part 1. As enacted. Source.
Zhu, Feng, and Qihong Liu. 2018. “Competing with Complementors: An Empirical Look at Amazon.com.” Strategic Management Journal 39 (10): 2618–2642. DOI: 10.1002/smj.2932. Source.