RESEARCH DOSSIER

Digital Nonrival Goods and Network Economics

Abstract. Digital nonrival goods are reproducible informational objects whose simultaneous use by different agents does not, in itself, diminish the object available to any of them. Their production and economic realization nevertheless require labor, capital, computational resources, energy, institutional rights, and coordination. This Research Dossier examines the implications of that combination for price formation, investment, product diversity, network adoption, platform-mediated exchange, competition over compatible systems, data access, welfare, and economic measurement. The inquiry integrates foundational theories of knowledge and increasing returns with information-goods economics, the economics of network effects, empirical identification studies, and selected international institutional evidence. It develops transparent illustrative derivations of the fixed-cost recovery problem, the access-investment trade-off, bundling, coordination thresholds, cross-group pricing, and the welfare effects of shared use. The analysis distinguishes the replicability of an informational object from the rivalry of the services and resources used to deliver it, and distinguishes a benefit of network participation from an uninternalized external effect. An original interpretive synthesis identifies why low replication cost, realized access, concentrated economic control, and scarce complements can coexist. The evidence supports conditional rather than universal conclusions about the benefits of scale, compatibility, exclusion, and openness. Statistical production, asset values, consumer surplus, and transaction volume are kept separate. The study is an analytical monograph based on selected primary texts, available scholarly records and official publications; its numerical examples are hypothetical, not estimates, and no new named theory or independently estimated causal effect is claimed.

Keywords: Digital nonrival goods; information goods; fixed costs; intellectual property; replication; increasing returns; network effects; compatibility; interoperability; two-sided markets; data economics; product diversity; consumer surplus; digital trade; economic welfare

The economic object and research question

The object of this dossier is the allocation, production, exchange, and use of digital goods whose informational content can be shared without being physically exhausted. The economic problem arises because creating a useful informational object requires resources, whereas reproducing an additional copy often requires comparatively few resources. A useful digital file may therefore be accessible to many users even when the institutional arrangements governing access, maintenance, and continued production remain expensive.

An electronic book, a software library, a design file, a set of observations, a trained computational model, and a network protocol are different economic objects. Each contains reproducible information. Their delivery and effective appropriation differ. A software library may require interpretation, integration, and security maintenance. A trained model may be copied only by an authorized recipient and requires hardware to generate outputs. A data set may be reproducible yet legally restricted because of its contents. A protocol specification can be read by many implementers while the network employing that protocol still has finite capacity.

Economic analysis must specify whose decision is being evaluated, what object or service is being allocated, and which time horizon governs costs and benefits. This is particularly important when a provider makes software available at a zero monetary price while selling hosting or when a network intermediary charges merchants rather than consumers. The object that is copied and the object that earns revenue may belong to different components of a joint production system.

The central questions are consequently connected. What prices can sustain costly creation when efficient use favors broad access? Under what circumstances does the benefit of an additional participant depend on the adoption of compatible technology by others? How does compatibility change the effective size of a network without eliminating independent providers? Why can an abundant digital artifact coexist with dependence upon a scarce complementary service? Which observations would distinguish these mechanisms from ordinary quality differences, selection effects, and scale economies?

Goldfarb and Tucker (2019), in their review of digital economic costs, organize the field around changes in the costs of discovering, copying, moving, monitoring, and authenticating information. These costs enter different decisions. A lower copying cost affects feasible distribution; a lower search cost changes matching and consumer choice; a lower verification cost may support remote transactions. The mechanisms need not move in equal proportions.

This inquiry belongs to Economic Theory as its primary scientific domain and to Microeconomics as a shared domain. It examines the allocation mechanisms of specific kinds of digital goods and their network-mediated use. Broader theories of social networks, macroeconomic growth, platform regulation, and public expenditure enter insofar as they explain that object.

Research proceeds by conceptual distinction, conditional derivation, comparative examination of institutional mechanisms, and critical evaluation of empirical evidence. The algebraic examples use transparent assumptions and are deliberately separated from estimates obtained from observed populations. The selected literature supports the mechanisms being evaluated rather than constituting an exhaustive systematic search.

Historical genealogy of nonrival knowledge

Knowledge and distributed coordination

Economic thought addressed the role of information well before electronic distribution made copying inexpensive on a large scale. The problem of dispersed knowledge concerns how agents act on particular circumstances that other agents may know only imperfectly. Hayek's (1945) The Use of Knowledge in Society examines the coordinative function of relative prices in a world where relevant knowledge is dispersed. His argument concerns the transmission and utilization of economically relevant information through decentralized decisions, and it remains relevant when a searchable data set makes documentation more accessible.

A digitized inventory may record every unit at a warehouse and still omit information about a customer's intended use, a supplier's future constraints, or a local worker's practical judgment. Recording observations improves particular dimensions of coordination; the economic significance of an observation continues to depend on its quality, interpretation, timeliness, incentives, and context. There is therefore a difference between the availability of a recorded signal and the availability of the knowledge needed to act on it.

Invention and the allocation of research resources

The economics of invention identifies problems that arise when resources devoted to the production of knowledge yield benefits extending beyond the party that financed them. Arrow's (1962) allocation of resources for invention examines uncertainty, indivisibilities, and the difficulty of appropriating returns from information. A supplier may need to reveal information to a buyer before the buyer can assess it; revealing it can itself transfer part of the benefit for which payment was intended.

Digital technologies alter the practical possibilities for addressing this problem. Demonstrations, limited licenses, cryptographic controls, reputation, audit trails, and subscription contracts can facilitate exchange. The expense of discovery, development, evaluation, and legal enforcement remains part of the resource problem. The appropriate amount of research therefore cannot be inferred from the cost of distributing a finished file.

Increasing returns and endogenous innovation

Romer (1990), in Endogenous Technological Change, develops a growth model in which intentional investment creates reusable, partly excludable technological knowledge. The nonconvexity associated with such an input changes the equilibrium market structure in the model, and its conclusions depend on the organization and allocation of human capital and research.

This analysis makes an important distinction available to digital economics. A design is a productive input that can be applied repeatedly, while the machines, labor, and materials necessary to instantiate each application remain scarce. Observed economic growth reflects the joint contribution of reproducible knowledge and complementary inputs. The model's conclusions concern a specified economy and its mathematical assumptions; they are not an empirical prediction that every economy will grow automatically with the size of its digital archive.

Communications and interdependent demand

A parallel line of inquiry studies communications services whose value depends on the users available to communicate. Rohlfs (1974) investigated interdependent demand for a communications service. Later system-competition research analyzed expectations, incompatible systems, and firms' decisions over compatibility. Katz and Shapiro (1994) organize these issues in their review of systems competition and network effects.

These strands of economic thought converge in contemporary digital markets because a reproducible informational object may be used as a common component of a communications or transaction system. An operating system can be replicated, applications may be built for it, and the practical value of using the system may depend on the availability of compatible applications and other users. Cheap replication and interdependent demand are distinct mechanisms that can reinforce each other.

Rivalry, excludability, and the unit of analysis

The distinction between the object and its use

Rivalry is a technological or economic property of a specified unit at a specified margin: one party's use diminishes the feasible use of the same unit or resource by another party. Excludability concerns the ability to control access. The properties can vary independently.

A text file reproduced across independent devices is nonrival in the relevant informational sense. A physical storage device used to hold it occupies material resources. A license may restrict the right to use the file; a hosted reader may enforce that restriction. A service that lets users interact with the file can become congested. These facts describe different objects or levels of the same production and delivery process.

Excludability exists on a continuum in practice. Legal remedies can be available but expensive, technical access controls can be imperfect, and it can be costly to distinguish unauthorized duplication from permitted use. An economic mechanism requiring exclusion may fail if enforcing it consumes more resources than it preserves in legitimate revenues. A mechanism requiring universal access may fail if a supposedly public copy cannot be discovered or interpreted.

Specified object Rivalry at the relevant margin Possible access arrangement Scarce complement
Digital book content Low or absent for identical copies Copyright, subscription, open license Authoring, editing, reading time
Software source file Low or absent in copying Proprietary or open-source license Integration, security review, maintenance
Online computation Capacity-dependent Account, usage charge, contract Processing, memory, electricity
Dataset record Low in simultaneous informational use Consent, license, access governance Collection, curation, confidentiality
Communication protocol Low for the published specification Public standard or controlled specification Implementations, network capacity
Exclusive digital entitlement Rival allocation if only one party may hold an entitlement Contract or registry control Enforcement, authentication

The table is an analytical classification. A protocol can change, a data record can become obsolete, and software may require a licensed remote dependency. For each example the relevant object has to be reidentified when the transaction includes a service or entitlement in addition to information.

Distinctions relevant to exchange

A digital good can be sold as a perpetual license, a revocable subscription, a stream, or a right to perform specific operations. The consumer receives a bundle of permissions and services rather than necessarily acquiring all legal rights to the underlying informational content. Resale restrictions, update eligibility, export rights, and continuity obligations affect the value of the transaction. A payment for access to a file and a payment for a service guarantee are different economically even if the provider presents one invoice.

A digital service supplies a sequence of useful operations. Its informational code may be nonrival, but the executed operations often require scarce inputs. In an online application, copying the program across additional servers increases available capacity by employing more capital and energy. Calling an interface may therefore entail positive marginal resource expenditure. For a long-running computation the magnitude can be substantial.

There are also digital representations of scarce assets, rights, and positions. A unique seat reservation can be recorded digitally and the record copied. The booked seat remains a rival service because only one holder can use it at the same time. A digital certificate of ownership can be reproduced as information without reproducing its legally recognized entitlement. Treating every digital transaction as a transaction in nonrival goods would obscure the actual sources of scarcity.

An external effect must similarly be specified. A participant's action can change another participant's utility or production opportunities. Whether that change produces a welfare distortion depends on the extent to which contracts, prices, ownership, or rules internalize it. Increased demand for the same software because other users purchase it is an observable outcome; an external benefit or cost is a separate claim about the causal mechanism and its pricing.

Production, replication, and cost structure

The first copy and subsequent distribution

A delivered digital work is the result of a sequence of activities: discovery, preparation, testing, storage, distribution, updating, and support. Some expenditures occur before a user receives anything. Other expenditures vary with copies, accounts, operations, or time. Hal Varian's (1998) Markets for Information Goods describes scale economies and other features of information goods that influence their exchange.

A stylized cost expression for a single completed work with total delivered quantity q is:

C(q) = F + c q, for q > 0.

C(0) = 0 if development is still avoidable.

Here F is the avoidable fixed resource cost required to make the good available in the relevant period and c is the incremental cost of serving another unit. Average cost equals F/q + c, and marginal cost equals c over the uncongested range. After F is incurred and unrecoverable it is sunk; it should not be confused with an avoidable recurring service cost.

For example, a digital reference work costs 100,000 monetary units to produce and costs 0.20 monetary units per authorized distribution. At 1,000 copies, average cost is 100.20; at 100,000 copies it is 1.20. Marginal cost remains 0.20 in both illustrative states. At a uniform price of 5, the number of paid copies required to cover the fixed cost is at least 20,834, since 100,000 / 4.80 = 20,833.33. This is a break-even relationship, not a demand projection.

Fixed cost is a horizon-dependent concept. A distribution service can avoid expenditure on an additional server for a short interval but require it after adoption passes a threshold. With a fixed per-period maintenance cost M and a capacity cost K(q), the relevant total cost may be F + M + cq + K(q). If K rises steeply near capacity, marginal cost is no longer approximately constant. If monitoring or customer support grows proportionally with organizations rather than accounts, the unit of q must reflect that distinction.

Economies of scale and scope

Scale economies concern the relation between the cost of a given output and its quantity. Scope economies concern the joint cost of producing several outputs. A code library may be reused by different products, and a common distribution infrastructure may support books, software, and subscriptions. Shared components lower duplicated expenditure where their joint use is technically and organizationally feasible.

Joint costs complicate attribution. A company could allocate a common database expense across applications according to usage, revenue, or engineering effort. The resulting product margins can differ depending on that allocation. For an incremental decision, avoidable incremental cost and the relevant opportunity cost are usually more informative than a mechanically allocated share of historical expenditure.

Increasing returns do not determine the number of firms by themselves. Demand heterogeneity, product differentiation, geographic restrictions, innovation, quality, capacity, and access to complements can sustain multiple suppliers. A large market may support several costly original works even if each individual work has a declining average cost. A competition analysis must therefore identify the economic perimeter of substitutable products and the resources necessary for entry.

Risk and the cost of quality

Expenditure on reliability, auditing, documentation, accessibility, or maintaining compatibility may rise without increasing the number of bits in an artifact. The observed quantity of replicated information is a poor proxy for the cost of quality. Two executable files of equal size can have different expected usefulness because one was tested against many environments and the other was not.

Investment uncertainty matters before creation. A publisher may spend resources on projects that never yield a marketable work. The resource cost of a successful project is not the only economic cost of producing a portfolio of innovations. In a decentralized setting, expected revenues from successful projects may help finance experiments that do not produce commercial returns. Welfare evaluation must compare a feasible alternative portfolio of investments, not merely the cost of copying one known success.

Access, fixed-cost recovery, and welfare

Conditional efficient access

Suppose an already created digital good can be supplied at a constant incremental resource cost c, and potential purchasers have inverse demand P(q) = a - bq, with a > c, b > 0. Assume that buyers receive their own valuation, there are no unpriced spillovers, no distribution constraints, and money is a transferable numeraire. Conditional on provision, access maximizing total surplus satisfies P(q*) = c, hence q* = (a - c)/b.

A uniform-price monopolist chooses q to maximize (a - bq - c)q - F. The interior first-order condition yields q_m = (a - c)/(2b), and p_m = (a + c)/2. The second derivative is -2b, so this is a maximum when production occurs. The loss of gross surplus relative to conditional efficient access is (a - c)^2/(8b).

Fixed cost affects the entry decision even though it does not change the unconstrained interior quantity chosen by the monopolist. Under the stated assumptions commercial provision occurs only if (a - c)^2/(4b) >= F. Socially beneficial provision occurs if (a - c)^2/(2b) >= F. Between these thresholds a work can be worth producing at broad access while uniform monopoly pricing fails to finance creation.

Consider an illustration with a = 10, b = 0.01, c = 2, F = 1,500. Efficient quantity is 800 and efficient gross surplus net of variable resource cost is 3,200. Monopoly quantity is 400 at price 6, yielding 1,600 before fixed cost and 100 after it. Consumer surplus is 800. Total realized surplus under monopoly is 900, whereas conditional efficient provision produces 1,700 after fixed cost. The 800 difference is the access deadweight loss.

If F instead equals 2,000, the same work has an efficient net social surplus of 1,200, yet the monopoly profit from provision is -400. The difference between these thresholds illustrates financing as a central economic problem. A fee imposed upon each incremental user can generate exclusion precisely where incremental supply is inexpensive.

There is no presumption that every digital good has linear demand or constant marginal cost. The example isolates one mechanism. Market power, heterogeneous network effects, privacy, congestion, and income effects change the calculation.

Alternative financing arrangements

A lump-sum charge paid by beneficiaries before creation can support broad use without imposing a price on subsequent replication. It requires credible commitments and the ability to identify or coordinate beneficiaries. Voluntary contributions are affected by an incentive to benefit from the contributions of others. A government or consortium can finance creation using collective resources, though taxation, governance, procurement, monitoring, and opportunity costs have to be counted.

A two-part tariff combines a membership fee with a per-unit price. If each customer has known identical demand, the seller can set the per-unit price near incremental cost and collect part of the resulting surplus through the fixed fee. Heterogeneity and liquidity constraints can make the membership charge exclude users, even when the per-use price is efficient conditional on joining. Providers also incur metering and contracting costs.

Advertising and sponsored distribution can transfer some of the financing burden from readers or users to parties that value obtaining their attention. The resulting price structure addresses one funding problem while creating another resource allocation problem: users' time, privacy, and the information displayed to them acquire economic significance. The full welfare calculation incorporates those consequences.

A fixed cost can sometimes be financed by selling a rival complement, such as a tailored installation, training, maintenance, physical equipment, or guaranteed computing capacity. Broad dissemination of a nonrival component can expand demand for the complement. This complement-based arrangement is sustainable when the provider can capture sufficient contribution from the activities users purchase. It does not guarantee funding for every useful component.

The welfare question distinguishes provision, access, and the burden of financing. An arrangement that reduces access losses can still be undesirable if its collection mechanism, governance cost, distributional effects, or reduction of future investment has a greater opportunity cost. A study of practical digital markets must compare available institutional alternatives rather than a theoretical marginal-cost price alone.

Differentiated products and the economics of variety

Digital products can differ in quality, language, subject matter, accessibility, compatibility, provenance, and expected continuity. Their creation may require separate fixed investments even when copying each product is inexpensive. Therefore product variety is both valuable and resource-consuming.

Dixit and Stiglitz (1977), in Monopolistic Competition and Optimum Product Diversity, examine the welfare problem of market-provided product diversity under scale economies. Their analysis specifies preference and production structures and considers the allocation of resources within and between differentiated sectors. The relation between decentralized variety and a social optimum is conditional on those structures.

A familiar illustration uses a constant elasticity of substitution aggregate of differentiated experiences:

U = (sum over i of x_i^((sigma - 1)/sigma))^(sigma/(sigma - 1)), with sigma > 1.

For this illustration x_i is use of variety i, and sigma measures substitutability. If an individual seller faces demand with elasticity sigma and takes the aggregate market conditions as given, constant marginal cost c > 0 implies a markup p = sigma c / (sigma - 1) under the usual interior profit maximization assumptions. With identical fixed cost F and free entry into symmetric varieties, break-even output solves (p - c)q = F, so q = F(sigma - 1)/c. These are familiar consequences of an isoelastic approximation, not a claim that this specification fits all digital use.

Taking c to exactly zero in the stated expression is not legitimate without reconsidering the model. Price collapses to zero and the break-even quantity formula diverges. A realistic information-good setting may require finite market demand, positive delivery cost, discrete purchases, quality differentiation, capacity limits, or a different demand specification.

Variety affects welfare through the diversity of available matches. For a small audience, producing two similar publications instead of one can improve fit but double major creation costs. In a global audience, differences of language and field may make a second specialized work valuable. Low copying costs alone determine neither excessive nor insufficient variety. The relevant comparison weighs the surplus from additional differentiation against the opportunity cost of producing it and the induced change in use of other goods.

The practical boundary of the variety problem becomes especially visible when applications depend on a common platform. A shared development environment reduces the cost of creating variants, while a compatibility restriction may increase the cost of offering a product on other platforms. Variety within one ecosystem and variety across ecosystems are related but different objects of analysis.

Pricing, versioning, and bundling

Licenses, segmentation, and dynamic access

A provider may charge for a copy, for a time interval, for the right to use many copies within an organization, or for a stream of operations. Revenue depends on willingness to pay for the contractual bundle. A consumer's benefit from a book that can be retained for years differs from the benefit of a subscription terminated after a month.

Versioning offers differentiated packages to buyers with different valuations. A basic service and an advanced service may employ common informational content but differ in precision, support, speed, available features, or permissions. Self-selection requires that the higher-value buyer prefer the advanced product at its price and that the lower-value buyer prefer the intended lower-price option to the alternatives.

Consider two buyer types. Their valuations for basic and advanced versions are respectively (4, 5) and (6, 12). At a basic price of 4 and an advanced price of 10, the low type receives a surplus of 0 from basic and -5 from advanced; the high type receives 2 from either. A price of 9.50 for advanced makes the high type strictly prefer it, while the low type still prefers basic if willing to buy at zero surplus. To obtain strict participation from the low type the basic price can be set infinitesimally below 4, provided the high type still prefers advanced. This example shows why an incentive-compatibility constraint, rather than the copying cost, determines the relative prices. Delivery and quality costs must be added to evaluate profitability.

The social effect of versioning depends on changes in access and quality. A lower-priced basic edition may reach users excluded by a uniform fee. Deliberately restricting a low-cost feature can conversely impose a loss if the restriction serves only to separate buyers and yields no offsetting improvement in provision or investment. Welfare comparison requires both the counterfactual price structure and the resources used to maintain separate versions.

Bundling and valuation dispersion

Bakos and Brynjolfsson (1999), in their study of bundling information goods, explain how aggregation of many valuations under specified statistical assumptions can improve pricing performance when unused copies are inexpensive. The result relies on the distribution and dependence of buyer valuations and is qualified by correlations, budget constraints, and complementarities.

An elementary two-good example illustrates the mechanism without estimating a market. Two equally weighted consumers have gross valuations (8, 2) and (2, 8) over two goods, and valuations add across goods. With zero incremental costs, separate uniform pricing at 8 per good yields a combined revenue of 16 from the two consumers. A bundle at 10 yields a revenue of 20 if consumers buy at zero surplus; at a slightly lower price near 10 both participate strictly. Both receive both goods, increasing use in this example. Creation costs are unchanged. The illustration exploits negatively related valuations and is not a universal bundling result.

For n independent valuations with common mean mu and finite variance s^2, the variance of the average valuation is s^2/n. Its standard deviation decreases as n grows. When positive correlations persist, dispersion need not vanish. If the pairwise correlation is a common rho, the variance of the average becomes s^2[1+(n-1)rho]/n. This expression makes explicit why systematic differences between market segments limit the concentration of bundle valuations.

A large subscription library may also introduce search costs, attention constraints, licensing payments owed to creators, and competition with specialist suppliers. Total user benefit depends on the subset actually consumed, the accessibility of relevant items, and the quality of recommendations. The cost of unused capacity in an unlimited-download library and that in a metered computing service can be very different.

Price as a signal and a contract

An observed zero price is evidence about a monetary transaction, not about all resource costs or all sources of value. It may indicate financing by another party, a promotional strategy, volunteer supply, complementary-product sales, or a publicly funded mission. A positive price may reflect quality assurance, contractual rights, or access to scarce services instead of narrowly defined copying cost.

Price comparisons should hold the relevant terms constant. A perpetual local license, a hosted subscription with updates, and a managed deployment with support are distinct bundles. Comparing the price per file across those arrangements can misidentify the market mechanism.

Investment, innovation, and intertemporal appropriation

A completed artifact may generate value over many periods. Creation requires a decision before its eventual use is known. A simple private investment test compares the initial cost F with the expected present value of future net receipts.

NPV = -F + sum from t = 1 to T of E[R_t - M_t - V_t]/(1+r)^t.

R_t is revenue, M_t is recurring maintenance expenditure, V_t is variable resource expenditure, and r is an appropriate per-period discount rate under the chosen treatment of uncertainty. The illustration uses expected cash flows and a consistent discount rate; care is required to avoid counting the same risk adjustment twice. Private profit does not include benefits accruing to other parties unless captured by payments or other modeled returns.

For a deterministic illustration, an initial expenditure of 100 and a net receipt of 60 at the end of each of two years give an NPV at 10 percent of 60/1.10 + 60/1.21 - 100, approximately 4.13. At 20 percent the NPV is 50 + 41.67 - 100 = -8.33. The example demonstrates the importance of payment timing, even when the underlying file can technically be distributed instantly.

Intellectual property, secrecy, lead time, service reputation, and control of specialized complements can make investment privately appropriable. The associated exclusion may reduce the use of already created works. The scope, strength, duration, and enforcement cost of protection matter because future innovation may depend on access to previous knowledge.

An economic comparison therefore includes at least the expected probability and social value of new creation, the loss imposed on users who are excluded from existing artifacts, the resources devoted to enforcement, and the consequences for subsequent innovation. Broad rights over a highly reusable input may affect several downstream activities at once. Narrow protection may encourage independent adaptation but provide less incentive for expensive discovery in circumstances where alternative financing is weak.

The logic also applies to investment in maintenance. A provider may be able to sell a successful initial product yet fail to sustain its security or interoperability. A user purchases the expected durability of a service as well as its current features. When a service becomes a complement for many other organizations, the benefits of continued maintenance can extend beyond the contracting parties.

A comparison of institutions must distinguish the ex ante perspective of investors, the ex post perspective of users, and the long-run perspective of subsequent creators. A policy or contract changes the sequence of decisions. A mechanism improving use after development could discourage development if participants anticipate that mechanism before committing resources; a credible reward can sometimes preserve creation incentives while expanding access later.

Direct network effects and adoption equilibria

Individual valuation conditional on participation

Direct network effects occur when an individual's benefit from joining or using a system changes with the participation of other users of that system or of compatible systems. A communication device with no reachable correspondents may still have standalone functionality, but its communication value depends on reachable persons and organizations. A static offline document may have substantial value independent of the number of other readers.

A stylized adoption problem uses a unit mass of potential users indexed by a private standalone value v uniformly distributed on [0,1]. Adopting confers net utility v + alpha n_e - p, relative to an outside option of zero. The expected share of adopters is n_e, the per-user price is p in the same utility numeraire, and alpha >= 0 measures the strength of a homogeneous network benefit.

With an interior threshold v = p - alpha n_e, the realized adoption share is n = 1 - p + alpha n_e, truncated to the interval [0,1]. Fulfilled expectations require n_e = n. In the interior,

n = (1-p)/(1-alpha), for alpha not equal to 1.

With alpha = 0.4 and p = 0.7, n = 0.3/0.6 = 0.5. A simple iterative expectation adjustment n_(t+1) = max(0,min(1,1-p+alpha n_t)) has slope 0.4 in the interior, so local deviations decay under that particular rule.

With alpha = 1.4 and p = 1.2, n = 0.5 is an interior fixed point. Both n = 0 and n = 1 are also fixed points because the unclipped expressions are respectively -0.2 and 1.2. Under the same adjustment rule, the interior fixed point is unstable and the endpoints attract nearby initial conditions. Starting below 0.5 leads toward the low-adoption equilibrium; starting above it leads toward the high-adoption equilibrium. The result follows from the assumed update rule and utility function. Other expectation formation mechanisms and distributions can yield different equilibria.

This example explains why a supplier's ability to coordinate expectations can influence adoption without changing the replicable object. An introductory subsidy, compatibility commitment, or large anchor customer can alter the expected usefulness of the system. A network benefit does not imply multiple equilibria when it is weak relative to the dispersion of standalone preferences or the adoption response.

Benefits and external effects

Under the homogeneous network example, a change in participation affects the utility of each existing adopter. If each of n participants receives alpha n, aggregate network benefit is alpha n^2; its derivative is 2 alpha n. The extra participant's own network benefit is alpha n at the infinitesimal margin, while an additional alpha n accrues to the existing participants. Under this continuum convention the distinction is exact. It is not a universal network valuation formula.

A provider may internalize part of the external effect by earning greater revenue from the entire network. Public or cooperative ownership may internalize other parts. Negative effects such as spam, congestion, or reduced privacy must be included. A socially desirable adoption subsidy requires a remaining uninternalized benefit at the relevant margin, and financing it may impose resource and distributional costs.

The common observation that a communication network with N members can contain N(N-1)/2 unordered pairs is a combinatorial maximum, not a measurement of economic value. Many pairs have no reason to communicate, useful interactions repeat, and some additional relationships impose unwanted attention costs. Economic worth depends on weights, direction, frequency, task, and compatibility, not merely the number of possible pairs.

Network topology, switching, and expectations

The value of an additional participant can differ sharply across positions in a network. A person who is relevant to several teams may confer more benefit than a distant participant whose activity does not affect the focal user's interactions. A general way to express this in a stylized utility description is to write the network component for user i as sum over j of w_ij a_j, where a_j indicates participation and w_ij is the value to i of access to j under the relevant task and compatibility conditions.

The weights may be asymmetric and may include negative effects. The number of accounts on a platform is at best one proxy for the opportunity to contact relevant partners. A user can have multiple accounts, accounts can be inactive, and communication can occur across providers. Network measurements should distinguish registered accounts, active participants, realized connections, compatible reach, and transaction frequency.

Switching costs are additional expenditures or losses incurred when changing providers or technologies, such as reformatting files, transferring credentials, retraining personnel, replacing dependencies, or rebuilding a contact network. They can depend on decisions made long before the switching event. Investment in learning one interface may raise the subsequent cost of adopting another independent interface.

Katz and Shapiro (1986), in technology adoption with network externalities, show in a specified theoretical setting that ownership and sponsorship can change adoption advantages and that market compatibility outcomes may be inefficient in different directions. The analysis provides a reason to inspect commitment, timing, and the sponsor's incentives before interpreting the observed winner as the socially preferred standard.

Farrell and Klemperer's (2007) coordination and lock-in reviews the role of switching costs and network effects in competition. Introductory offers can be rational when acquiring users creates future revenue opportunities. The relevant welfare issue includes the subsequent availability of alternatives and the distortions generated when future terms cannot be credibly committed.

A hypothetical firm deciding between two software products may prefer the new product in features but face an immediate migration expense and the anticipated absence of compatible collaborators. A decision to remain is rational given those constraints. A comparison of licensing prices alone would omit the cost of moving a productive organization to a different environment.

Expectation effects can coexist with supply-side scale economies. Empirical growth of adoption cannot identify which caused it. A superior product might gain users steadily because it is reliable and cheap, even if participation by others has no direct influence on individual utility. Conversely, network coordination can favor an established product even if incremental supply costs are constant. The mechanisms have different implications for interoperability, entry, and consumer welfare.

Compatibility and interoperability

Compatibility describes the ability of products to participate in a shared system; interoperability describes their ability to exchange or use information and functions across organizational or technical boundaries. Two documents may have the same format yet differ in semantic interpretation, authentication requirements, or access permissions. A specification is consequently only one component of effective interoperability. Implementations, testing, governance, security, and the allocation of responsibility are also necessary.

Consider two incompatible networks with populations A and B. In a stylized model where every pair of users can communicate, the separate networks afford A(A-1)/2 and B(B-1)/2 potential undirected connections. Full interoperability adds AB potential cross-network connections. With 60 and 40 users, the two systems permit 1,770 and 780 internal pairs respectively; interoperability makes 2,400 additional pairs possible, for a total of 4,950 pairs. These are potential links, not realized transactions or monetary welfare. If cross-network contact has little value, technical compatibility can deliver little benefit; if the groups contain complementary specialists, its marginal benefit can exceed that of connections within either group.

Compatibility affects entry incentives through the expected size of the accessible network. A new provider can compete on price, reliability, privacy, or interface while preserving access to established users when interoperability is effective. A provider that must recreate the network as well as the product faces a coordination burden absent from ordinary product substitution. In both cases the engineering quality and cost of the entrant remain important.

The distinction between the ownership of a network and the ownership of an implementation is visible in the historical organization of Internet standards. Bradner's (1996) Internet standards process emphasizes technically competent specifications, independent implementations, and interoperability. RFC 2026 is a historical institutional document subsequently updated by other RFCs; its relevance here is the documented process rather than an assertion that every provision remains unchanged. Common protocols can facilitate entry by multiple implementers without transferring ownership of their infrastructure.

Interoperability may introduce its own vulnerabilities. A connection between two systems can transmit fraudulent messages, disclose metadata, propagate malicious content, or permit failures to spread. The costs of auditing counterparties and preventing such outcomes are real. A social evaluation compares the added gains from communication and substitution with the costs of implementation, common governance, and security, including the effects on users who did not choose the interconnection.

Portability addresses a related but distinct friction. Exporting a user's records may facilitate a change of provider; exchanging messages with users who remain at another provider addresses a continuing coordination problem. A downloadable archive may have limited usefulness if the replacement system cannot interpret its content. Conversely, a messaging bridge can leave the user's stored history difficult to transfer.

The timing of a compatibility obligation matters. An incumbent required to interconnect after making an irreversible investment confronts a different incentive problem from investors who anticipate a stable interoperability regime. Ex ante contractual expectations, reasonable cost allocation, and credible update procedures enter the investment comparison. A uniform technical requirement may also impose a disproportionate compliance burden on small providers. The appropriate comparison is between implementable regimes, each with an explicit transition path, rather than between perfect interoperability and an unchanged status quo.

Cross-group network effects and platform pricing

An intermediary serving two groups can affect the value created by interactions between them. A marketplace connects buyers and sellers, a payment network connects payers and merchants, and a scholarly repository connects authors and readers. A larger population on one side can increase the attractiveness of participation on the other. The direction and magnitude of this effect depend on transaction opportunities, search quality, and congestion.

Rochet and Tirole (2003), in Platform Competition in Two-Sided Markets, study the allocation of fees across participating groups. The price structure matters when changing the relative charges, even with an unchanged aggregate charge per interaction, changes the volume of participation or transactions. This is stronger than the observation that a company has two sets of customers: the distribution of charges has economic consequences because participants on each side cannot fully rearrange those charges through bilateral contracting.

A simplified illustrative membership model makes the interaction explicit. Let a and b denote participation shares in two unit masses, bounded between zero and one. Assume uniformly distributed reservation costs and utility before the reservation cost of 1 + alpha b - p_A on side A and 1 + beta a - p_B on side B. An interior equilibrium satisfies a = 1 + alpha b - p_A and b = 1 + beta a - p_B. For alpha = 0.4, beta = 0.5, p_A = 0.8, and p_B = 0.7, substitution gives a = (0.2 + 0.4 × 0.3)/(1 - 0.2) = 0.4 and b = 0.5. Both shares are interior. If alpha beta approaches one, the interior comparative statics become more sensitive; boundary clipping matters and the interior formula must not be extrapolated beyond its domain.

Suppose instead the platform charges 0.9 on A and 0.6 on B. The sum of charges is still 1.5, but a = (0.1 + 0.4 × 0.4)/0.8 = 0.325, and b = 0.5625. If possible matches are proportional to ab, the former arrangement yields 0.2 and the latter 0.1828125. This difference is an implication of the illustrative assumptions, not a prediction that a particular platform should adopt the first schedule. Costs, participation preferences, matching technology, and transaction fees can reverse the ranking.

Armstrong's (2006) competition in two-sided markets distinguishes market arrangements by participation and the opportunities to join multiple intermediaries. In particular, single-homing on one side and multi-homing on the other can generate a competitive bottleneck. The fee imposed on a participant must be interpreted alongside the access it conveys to participants who are otherwise difficult to reach.

The profit function of a membership platform can be written as revenues from both sides and ancillary activities minus the fixed and marginal costs of supporting participation and matching. Maximizing this expression depends on the entire response of both populations. Increasing a fee can reduce not only revenue-generating participants on the charged side but also participation and revenue on the other side. An apparently generous price may consequently be consistent with ordinary commercial optimization. A zero price by itself identifies neither efficient access nor exclusionary behavior.

A digital marketplace can facilitate transactions involving rival goods, and a nonrival digital good can be sold directly without a platform. The two economic mechanisms become jointly relevant where reproducible software supplies the intermediary and interactions among participants determine service value. Their relative magnitudes vary across payment, distribution, communications, advertising, and application ecosystems.

The welfare benchmark must specify whether sellers can pass fees through to final prices, whether users can participate elsewhere, and whether matching generates displacement of activity outside the platform. The appropriate accounting unit may be a membership, completed transaction, advertisement, or minute of attention. Changing that unit changes the interpretation of revenues, marginal costs, and participation.

Competition, openness, and organizational arrangements

Scale economies and network effects create reasons for concentrated market outcomes, but neither condition is individually sufficient to establish durable monopoly power. Differentiation, open standards, heterogeneous preferences, geographic specialization, congestion, multi-homing, and innovation can sustain several providers. Concentration is evidence requiring an economic explanation. A fixed cost spread over users and a participation benefit from reaching other users imply different entry obstacles and potentially different remedies.

The economic significance of control depends on the layer at which it is exercised. A collection of openly accessible software can depend on specialized hosting, authentication services, distribution channels, or technical expertise. Competition in the software layer can coexist with concentrated supply of complements. Conversely, a proprietary application can have many substitutes if its users can switch easily and access the same counterparties.

Open-source software illustrates an organizational form in which permission to inspect, modify, and distribute code can coexist with incentives to contribute and earn income. Lerner and Tirole's (2000) economics of open source explores voluntary contributions through labor-market incentives, including reputational and career benefits. Their working-paper argument is not a claim that such motives explain every project. Employers can also finance contributions because they use the common code or sell complementary services. A project's economic durability depends on how maintenance responsibilities, security work, and decision rights are organized.

The availability of code is different from the availability of the competence to deploy it. An organization choosing between an open implementation and a proprietary managed service must compare installation, integration, continuity, support, and opportunity costs. A zero license price combined with substantial internal engineering expenses can exceed the total cost of an alternative. The converse is possible when reproducibility, customization, and freedom from costly migration have considerable value.

The financing problem is especially acute for technical components that are essential to many unrelated businesses but individually difficult to monetize. Shared maintenance arrangements, employer sponsorship, service contracts, and procurement can provide resources. Each arrangement changes the distribution of decision rights and risks. The social significance of a component can therefore exceed the revenue directly attributable to its creators.

The relationship between innovation and exclusion must be evaluated over a specified horizon. Exclusive rights and restrictions can increase the expected return on creation while reducing diffusion, recombination, and access after creation. Open access can encourage reuse while leaving some production expenses unfunded. Neither a competitive market nor an open collective arrangement automatically solves every incentive problem. The relevant comparison considers expected investments and downstream uses under feasible alternative rules.

A credible assessment separates the ability to raise a monetary price from the ability to worsen quality, restrict access, collect more information, or impose unfavorable contractual conditions. Nonprice competition is central when the observed user price is zero. Technical alternatives, actual migration behavior, and the cost of reaching counterparties provide more information about substitutability than a simple count of available applications.

The economics of data, privacy, and learning

Data are records of observations, events, or characteristics. One party's use of an existing record need not consume its informational content. The collection, verification, labeling, storage, processing, and lawful disclosure of that record nevertheless require resources. Some records have brief economic usefulness, some remain useful for extended periods, and others become misleading when accumulated without attention to changing circumstances.

Jones and Tonetti (2020), in Nonrivalry and the Economics of Data, develop a theoretical account of repeated data use and the implications of different allocations of control rights. Their model links access, innovation, and privacy and demonstrates conditions under which decentralized data hoarding restricts socially valuable reuse. Its welfare conclusions depend on its assumptions about preferences, property rights, competition, and technology. They do not establish a universal requirement to share personal data.

A record can generate benefits for multiple firms while also creating a privacy cost for the person described. A firm considering sharing may take into account the value of losing exclusive information and contractual revenues; the affected individual may experience an additional cost from unwanted disclosure. The potential gains from shared analysis and the harms arising from disclosure belong in the same social comparison. Ownership of a copy does not by itself establish a right to use it for every purpose.

Data production is often joint with other activities: a retail transaction creates a receipt, a transport journey creates location records, and a digital service produces usage logs. Some collection expenditures would have occurred even without subsequent analysis. An economic estimate should distinguish incremental data production expenses from joint costs and separate the expected value of a reusable collection from the revenue of the business that happens to produce it.

Accumulation of data can improve quality if the additional observations carry relevant information and the system can use them effectively. Diminishing marginal gains arise when new observations duplicate previous ones; biased coverage, stale observations, and changing environments may make accumulation insufficient to improve quality. A feedback through which usage generates data and data improve a service is therefore a hypothesis about production and incentives, not an automatic consequence of an expanding user base.

A hypothetical estimator provides a useful distinction. If independent observations have common variance 100, the variance of their sample mean is 100/n. Increasing sample size from 100 to 400 reduces the variance from 1 to 0.25 and the standard error from 1 to 0.5. If observations are dependent, selectively observed, or systematically biased, the formula does not apply without adjustment. More records cannot eliminate a persistent bias merely by raising n.

Access arrangements can include licensing, controlled computation, anonymization, aggregation, purpose limitations, and audited research environments. Each arrangement has implementation costs and risks. A user who encounters an apparently useful statistic may also be affected by its subsequent use in classification decisions. A realistic welfare assessment considers the accuracy of inferences, the distribution of affected persons, security risks, and the opportunity cost of limiting access. This chapter provides an economic analysis of these trade-offs rather than legal advice or a technical claim that any single privacy measure is sufficient.

Empirical identification and research designs

Positive correlation between adoption and the number of existing users does not establish a causal network effect. Common tastes, advertising, a superior underlying product, geographic proximity, and simultaneous shocks can generate the same pattern. A relevant empirical strategy requires variation that changes the value or availability of connections without simply selecting people who would already have adopted.

Manski's (1993) reflection problem establishes a central difficulty in identifying social effects: group behavior and individual behavior can determine each other, while group characteristics and common circumstances can generate additional correlation. A regression of individual participation on average participation among peers may combine endogenous interaction, contextual influences, and correlated unobservables. Excluding the individual from the group average does not in general resolve these problems.

One prospective research design assigns an invitation or access subsidy to randomly selected groups and measures the response of untreated participants. Treatment may change the probability that a prospective user's counterparts are available. A credible interpretation requires a definition of exposure, documentation of the assignment process, and allowance for contamination across groups. When effects depend on the assignments of other individuals, conventional individual-level treatment and control comparisons need a design that explicitly incorporates interference.

Tucker (2008) examines influence in technology adoption using an organization's communications network and variation in the usefulness of video communications. The study distinguishes formal organizational influence from other connections. Its design illustrates why researchers should observe the identity and role of potential contacts rather than rely exclusively on aggregate installed-base statistics. The findings arise from a particular enterprise and technology; their magnitude cannot be directly transferred to consumer communications platforms.

Gowrisankaran and Stavins (2002) study electronic payment adoption using bank-level observations on automated clearinghouse participation. Their investigation uses several empirical approaches to distinguish shared adoption patterns from other influences. The institutional setting matters because financial intermediaries have different incentives and operational constraints from individual software users.

The required outcome variable depends on the mechanism. A new connection may increase transaction frequency without materially changing registration. Interoperability may expand the fraction of feasible contacts while reducing the need to maintain duplicate accounts. A change in software licensing may alter installation rates yet have little effect on actual use. Studies should therefore distinguish adoption, retention, active participation, completed interaction, transaction costs, and welfare outcomes.

A difference-in-differences comparison around the introduction of interoperability can be informative when treated and comparison units would have followed sufficiently similar trends without the reform. Pre-reform trends, staggered adoption, simultaneous changes in pricing, heterogeneous effects, and spillovers require explicit examination. Effects estimated from participating firms alone may omit substitution from other channels. An instrumental-variable design requires an instrument that moves the relevant exposure and an argument that it affects the outcome only through that exposure, subject to the estimand and assumptions used.

Empirical work must also confront measurement of relationships. A registered network may contain inactive accounts and automated traffic; one person may maintain multiple identifiers. A contact graph can represent historical communications rather than the connections relevant to a present decision. Directed, weighted networks may encode economic influence more accurately than undirected binary links. The appropriate model and data resolution depend on the research question and the privacy constraints on observation.

The dossier uses these studies as evidence that careful identification is possible in particular settings and that common adoption patterns alone are insufficient. It does not independently reproduce their regressions or reestimate their parameters from underlying microdata.

International institutional comparisons

Digital network economies operate within distinct regulatory, technological, and organizational environments. International comparisons should identify the institutional mechanism rather than infer the effect of an entire country's digital development from one intervention.

Payment interoperability offers a well-defined case because the completion of a payment requires compatible identification, routing, authorization, and settlement arrangements. Copestake, Kirti, Martinez Peria, and Zeng (2025), in an IMF study of interoperability in money and payments, investigate fragmentation and the value of connecting payment networks in India, using data from the Unified Payments Interface and an established fintech provider. Their reported evidence is consistent with increased digital payment use following integration, with stronger effects where fragmentation was greater. As working-paper evidence, the results should be read in relation to the institutional settings and identifying assumptions examined by the authors.

A payment network combines nonrival software and protocols with rival operational resources and settlement obligations. The value of making an additional counterparty reachable is economically distinct from the expense of processing an individual transfer. Wider reach can reduce the need to maintain separate payment relationships while the underlying institutions retain separate balance sheets. Security, fraud allocation, dispute procedures, and compliance costs remain relevant.

The European Union's Digital Markets Act is an example of a regulatory response to persistent dependencies on digital intermediaries. The official EUR-Lex Digital Markets Act summary describes the framework established by Regulation (EU) 2022/1925, including obligations concerning designated gatekeepers and certain interoperability and access arrangements. The analytical interest lies in the separation between the benefits of participating in an established network and the difficulties of contesting particular services supplied by its operator. This is a description of the framework as documented in the cited official summary, not an assessment of individual firms' compliance or a current legal determination.

Historical Internet standardization provides a contrasting institutional arrangement. The standards process described by Bradner (1996) relies on published protocols and testing of implementations rather than designation of particular commercial gatekeepers. These mechanisms can coexist: a common protocol can operate under voluntary technical governance while certain commercial intermediaries face separate statutory requirements.

Comparing these arrangements demonstrates that interoperability is an economic and institutional variable as well as a technical property. A common message format alone may be insufficient if authorization, commercial terms, or dispute settlement prevent effective exchange. A regulatory mandate may fail if users cannot discover compatible services. A successful experiment in one payment environment cannot establish the same effect in a market with different user preferences, payment alternatives, and risks.

International comparison therefore requires comparable definitions and an explicit counterfactual. The same increase in digital transaction volume may indicate displacement of cash, substitution from another digital channel, or additional economic activity. Only the last necessarily implies an increase in total transactions, and none by itself measures the change in social surplus.

Digital production, trade, and welfare measurement

Digital activity is measured through several accounting systems with distinct purposes. National accounts measure production and associated income within a defined boundary. Trade statistics record cross-border transactions according to applicable residence and transaction conventions. Business data may record gross transaction value, revenue, registrations, or advertising impressions. Welfare research estimates benefits and costs experienced by people, including effects outside observable market payments.

The OECD (2023) Digital Supply and Use Tables handbook distinguishes the transaction process, the products involved, and the kinds of producers. This distinction is valuable because an ordinary physical good can be ordered through a digital interface, while an informational product can be supplied by a firm whose primary activity lies outside a narrowly defined digital industry. A single digital economy share consequently depends on its classification boundary.

The U.S. Bureau of Economic Analysis (2023), in its digital economy estimates for 2017–2022, distinguishes infrastructure, electronic commerce, priced digital services, and federal nondefense digital services. Its definitions separate total output from value added. The publication is a dated statistical release containing estimates and revisions for a specified period. Its categories extend beyond nonrival goods; applying its totals as an estimate of the size of nonrival digital production would be a classification error.

Suppose a marketplace reports 100 monetary units of sales between independent buyers and sellers and charges sellers a 10-unit commission. In a simplified accounting example where the marketplace acts as intermediary, its service output is 10, not 100. The 100 records the value of the mediated sales under the assumptions stated. If intermediate inputs to the platform's own service cost 4, its value added is 6. The sellers' contributions belong to their own production accounts. Double-counting arises when gross mediated sales are added to platform output as if the intermediary had produced the entire underlying product.

The IMF, OECD, United Nations, and WTO (2023) Handbook on Measuring Digital Trade provides a statistical framework for digitally ordered and digitally delivered international trade. The two classifications can overlap. A physical book bought from an overseas seller through an online ordering system is potentially digitally ordered but physically delivered. A cross-border software subscription delivered over a network may satisfy both classifications. Overlap must be accounted for before aggregating totals, and the actual transaction must satisfy the underlying international trade definitions.

The welfare associated with an unpriced digital service is not necessarily zero. Brynjolfsson, Collis, and Eggers (2019) employ online choice experiments to estimate the compensation users would accept for giving up access to selected services. Their study illustrates a direct route to measuring benefits beyond observed expenditures. Such estimates depend on the elicitation design, the sampled population, the time period, available substitutes, and the extent to which stated decisions are consequential. They are not automatically additions to GDP, whose production boundary and valuation framework serve different statistical purposes.

The distinction between willingness to pay and willingness to accept is also consequential. A participant asked what compensation would justify losing an existing service may give a different answer from one asked how much they would pay to obtain it. Budget constraints, attachment, substitutes, and the wording and duration of the choice matter. Multiplying one survey estimate by a global account total disregards sampling, duplicate accounts, heterogeneity, and the equilibrium effects of many users simultaneously leaving a network.

A complete interpretation reports the unit, denominator, price basis, reference year, territorial coverage, and source of a statistic. Monetary valuations of data assets, annual output, cumulative investment, operating profit, consumer surplus, and gross mediated sales cannot be summed without an explicit accounting framework that addresses overlap and temporal consistency.

Distribution, infrastructure, and environmental constraints

Reproducibility removes a particular form of scarcity from the use of existing information. It does not ensure that people can acquire, interpret, or act on that information. Devices, network connections, skills, language, accessible interfaces, and institutional authorization determine realized opportunities.

The International Telecommunication Union's (2025) Facts and Figures release estimated six billion people online and 2.2 billion offline in 2025. The same release reported Internet use of 94 percent in high-income countries and 23 percent in low-income countries, and 85 percent in urban areas compared with 58 percent in rural areas. These estimates describe the specified 2025 release, not the latest possible 2026 figures. The income-group difference is 71 percentage points and the urban-rural difference is 27 percentage points. They demonstrate considerable disparities in a fundamental complement required to access online information; they do not identify the causal contribution of any individual barrier.

The relevant cost for a person is the complete cost of effective participation. An unpriced online textbook may require a device, connectivity, time, and language competence. An enterprise using a freely available computational component may need engineering staff and reliable electricity. Removing the price of the informational object reduces one barrier while leaving others in place. Equal permission to obtain a file and equal ability to benefit from it are different economic outcomes.

Environmental resource use arises in the production and operation of computing and communications infrastructure. A file's reproducibility does not determine the energy required for processing it. The workload, hardware, utilization, cooling, network configuration, electricity generation, and time horizon all matter. An inference from the low cost of duplicating stored information to the low resource use of every computational service would be invalid. This dossier makes no numerical estimate of emissions or energy intensity.

Distributional analysis must identify the bearers of benefits and costs. A service can save users time while shifting revenue among workers, suppliers, and infrastructure owners. Wider access to a reproducible educational resource may yield greater gains for those with complementary skills. The evaluation of a policy requires an explicit welfare criterion if benefits to different individuals are to receive different weights. Aggregate net surplus and its distribution answer related but different questions.

Integrated analytical synthesis and counterexamples

The preceding analysis supports an integrated economic interpretation. Digital nonrivality concerns the use of an informational object; fixed creation costs concern its supply; network effects concern the dependence of participation value on others; institutional control determines feasible access and appropriation. These properties can occur together, but one does not logically imply the others.

Consider an academic article downloadable as an ordinary file. It can be read independently of how many other readers possess it, although wider readership may produce indirect scientific benefits through citation and discussion. Its central allocation problem concerns creation, preservation, accessibility, and attribution. A proprietary messaging service also uses reproducible code, but the usefulness of its service depends strongly on whom users can reach. A compute-intensive analytical service may permit reuse of underlying code while charging for rival processing capacity. Applying a single marginal-cost or competition prescription across these three cases conceals their different economic mechanisms.

The appropriate starting point is the actual transaction and the object supplied. If the object is a license to existing information, the analysis should isolate irrecoverable creation costs and the avoidable cost of dissemination. If it is participation in a network, the analysis should identify the relevant counterparties and the effect of additional participation. If it is an ongoing service, marginal resource requirements and service quality must enter the calculation. Many contracts combine all three.

Suppose a hypothetical platform invests 100 in development, spends 1 per participating user, and can serve 80 users whose willingness to pay is 3 each at a particular network size. At a uniform charge of 3, receipts are 240, variable costs are 80, and net receipts after the development cost are 60. If the development has already occurred, a charge of 1 would recover variable costs but not the original expenditure. If a separate financing arrangement covers creation and expands the accessible network, the surplus comparison depends on the added users' benefits, costs, and any benefits conferred on existing participants. The numerical example deliberately holds the initial willingness to pay and quantity fixed for the accounting calculation; it does not imply that quantity remains unchanged when the price changes.

An additional question concerns the locus of control. Suppose the code becomes freely reproducible while participation remains restricted by control of a common identity directory. The effective entry barrier may remain substantial unless users can authenticate and reach counterparties through independent implementations. Releasing the code solves a permission problem without necessarily solving a coordination problem. Conversely, independent implementations of a shared protocol may compete even when the production of each implementation has substantial fixed costs.

The original synthesis of this dossier is the ordering of these considerations into a conditional interpretation of concrete transactions. It combines established economic arguments rather than asserting a novel theorem or a new named theory. Its practical analytical significance is to prevent the characteristics of one economic object from being attributed indiscriminately to another, and to make the resource, incentive, participation, and welfare assumptions visible before drawing conclusions.

Counterexamples clarify the boundaries. A nonrival informational product with negligible fixed costs, no network effect, and readily available substitutes may be supplied competitively at a very low price. A rival transportation service can possess strong indirect network effects when riders attract drivers and drivers attract riders. A network can be fragmented by users' preferences rather than technical incompatibility. A provider with high fixed costs may recover them through a competitive market if demand and differentiation are sufficient. A large installed base may reflect superior service rather than persistent exclusion. A freely shared database may impose harms through inaccurate or unwanted inferences. Each example breaks a commonly assumed implication without contradicting the underlying economics of reproducibility.

Policy comparisons should therefore test at least four counterfactual consequences: whether creation would be undertaken, whether additional use would occur, how participation and compatibility would change, and which costs or harms would be transferred among economic actors. These are established questions of welfare and industrial economics applied jointly to digital settings. An intervention can improve one outcome and worsen another, and the net effect must be established for the relevant population and horizon.

Research limitations and open scientific questions

The research is a selective analytical synthesis of foundational publications, empirical study records, official statistical materials, and historical institutional documents. Directly accessible texts and official summaries were examined where available; some journal contributions were checked through abstracts or bibliographic records rather than through complete restricted articles. The research evidence records access scope for each reference. No independent replication of unavailable microdata, comprehensive systematic review, or econometric estimation is claimed.

The mathematical illustrations are deliberately small and transparent. Their parameter values are hypothetical. Linear demand, uniform reservation costs, homogeneous network benefits, independent statistical observations, and complete potential graphs facilitate derivation but are not universal empirical descriptions. Changing these assumptions can alter existence, stability, pricing, and welfare results. The examples establish conditional relationships and permit arithmetic verification; they do not calibrate actual markets.

Important open questions concern the persistence of digital lock-in after a compatibility reform, the allocation of ongoing maintenance expenses across users of shared code, the social value of data under heterogeneous privacy preferences, and reliable welfare measurement when network participation is both a cause and a consequence of individual benefits. New evidence could distinguish gains from lower duplication costs, greater reach, improved matching, and changes in competition within a common empirical setting.

Future research would benefit from observation of actual utilization, costs, and switching behavior before and after changes in access arrangements, with suitable comparison groups and appropriate privacy safeguards. Comparative institutional studies should specify the technical scope of interoperability and the governance mechanisms in effect. Production and welfare estimates should retain separate accounting boundaries. Building these distinctions into research designs improves the prospects for interpreting results beyond one product or institution.

Conclusion

The economic significance of digital nonrival goods lies in the possibility of repeated use of existing information without its consumption. Creating and sustaining useful informational objects requires scarce resources, and their availability depends on arrangements for financing, control, maintenance, and complementary production. The economics of networks adds an analytically distinct source of interdependence: the benefit of participation can depend on the presence and accessibility of other economic actors.

Together these features explain why digital economies exhibit diverse pricing systems, concentrated and competitive structures, open and proprietary production arrangements, and persistent disparities of access. Fixed costs require a feasible financing mechanism; network effects require attention to expectations and compatibility; data reuse requires assessment of rights, incentives, and harms; economic measurement requires a defined production or welfare boundary.

The scientific result is an integrated conditional account of these mechanisms, supported by established theory, selected empirical studies, official statistical frameworks, and worked economic illustrations. The account identifies how abundance at the level of reproducible information can coexist with scarcity of complementary resources and restricted participation, while the benefits of shared use depend on the actual institutional and technical arrangements through which economic actors interact.

References

Armstrong, M. (2006). Competition in two-sided markets. RAND Journal of Economics, 37(3), 668–691. DOI: 10.1111/j.1756-2171.2006.tb00037.x.

Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. In The Rate and Direction of Inventive Activity: Economic and Social Factors (pp. 609–626). Princeton University Press for the National Bureau of Economic Research.

Bakos, Y., and Brynjolfsson, E. (1999). Bundling information goods: Pricing, profits, and efficiency. Management Science, 45(12), 1613–1630. DOI: 10.1287/mnsc.45.12.1613.

Bradner, S. (1996). The Internet standards process — Revision 3. RFC 2026, BCP 9. Internet Engineering Task Force. DOI: 10.17487/RFC2026.

Brynjolfsson, E., Collis, A., and Eggers, F. (2019). Using massive online choice experiments to measure changes in well-being. Proceedings of the National Academy of Sciences, 116(15), 7250–7255. DOI: 10.1073/pnas.1815663116.

Copestake, A., Kirti, D., Martinez Peria, M. S., and Zeng, Y. (2025). Integrating fragmented networks: The value of interoperability in money and payments. IMF Working Paper No. 2025/126. International Monetary Fund. DOI: 10.5089/9798229013857.001.

Dixit, A. K., and Stiglitz, J. E. (1977). Monopolistic competition and optimum product diversity. American Economic Review, 67(3), 297–308. JSTOR stable identifier 1831401.

European Union, EUR-Lex (n.d.). Digital Markets Act. Official summary of Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022, Official Journal L 265, 12 October 2022, pp. 1–66. Accessed November 17, 2026.

Farrell, J., and Klemperer, P. (2007). Coordination and lock-in: Competition with switching costs and network effects. In M. Armstrong and R. Porter (Eds.), Handbook of Industrial Organization (Vol. 3, pp. 1967–2072). Elsevier. DOI: 10.1016/S1573-448X(06)03031-7.

Goldfarb, A., and Tucker, C. (2019). Digital economics. Journal of Economic Literature, 57(1), 3–43. DOI: 10.1257/jel.20171452.

Gowrisankaran, G., and Stavins, J. (2002). Network externalities and technology adoption: Lessons from electronic payments. Working Paper No. 2002-16. Federal Reserve Bank of San Francisco, May 1, 2002.

Hayek, F. A. (1945). The use of knowledge in society. American Economic Review, 35(4), 519–530. Online reprint, Library of Economics and Liberty.

International Monetary Fund, Organisation for Economic Co-operation and Development, United Nations, and World Trade Organization (2023). Handbook on measuring digital trade. Second edition. IMF, OECD, UN, and WTO. DOI: 10.5089/9789287073600.071.

International Telecommunication Union (2025). Global number of Internet users increases, but disparities deepen key digital divides. Facts and Figures 2025 press release. Geneva, November 17, 2025.

Jones, C. I., and Tonetti, C. (2020). Nonrivalry and the economics of data. American Economic Review, 110(9), 2819–2858. DOI: 10.1257/aer.20191330.

Katz, M. L., and Shapiro, C. (1986). Technology adoption in the presence of network externalities. Journal of Political Economy, 94(4), 822–841. DOI: 10.1086/261409.

Katz, M. L., and Shapiro, C. (1994). Systems competition and network effects. Journal of Economic Perspectives, 8(2), 93–115. DOI: 10.1257/jep.8.2.93.

Lerner, J., and Tirole, J. (2000). The simple economics of open source. NBER Working Paper No. 7600, March 2000. National Bureau of Economic Research. DOI: 10.3386/w7600.

Manski, C. F. (1993). Identification of endogenous social effects: The reflection problem. Review of Economic Studies, 60(3), 531–542. DOI: 10.2307/2298123.

Organisation for Economic Co-operation and Development (2023). OECD handbook on compiling digital supply and use tables. OECD Publishing, Paris. DOI: 10.1787/11a0db02-en.

Rochet, J.-C., and Tirole, J. (2003). Platform competition in two-sided markets. Journal of the European Economic Association, 1(4), 990–1029. DOI: 10.1162/154247603322493212.

Rohlfs, J. (1974). A theory of interdependent demand for a communications service. Bell Journal of Economics and Management Science, 5(1), 16–37.

Romer, P. M. (1990). Endogenous technological change. Journal of Political Economy, 98(5, Part 2), S71–S102. DOI: 10.1086/261725.

Tucker, C. (2008). Identifying formal and informal influence in technology adoption with network externalities. Management Science, 54(12), 2024–2038. DOI: 10.1287/mnsc.1080.0897.

U.S. Bureau of Economic Analysis (2023). U.S. digital economy: New and revised estimates, 2017–2022. Survey of Current Business, December 6, 2023.

Varian, H. R. (1998). Markets for information goods. University of California, Berkeley. Author manuscript, April 1998, revised October 16, 1998.