Two-Part Tariff

The pricing form

A two-part tariff is any pricing schedule of the form

T(q) = F + p · q

where T(q) is the total amount a consumer pays when she consumes q units, F is a fixed "access fee" charged once per period regardless of usage, and p is a constant per-unit price. The schedule is non-linear in q — the average price per unit, F/q + p, falls as q rises — even though the marginal price (∂T/∂q = p) is constant. That dual structure is the source of all its useful properties.

Compare this to a uniform price T(q) = p·q. A uniform monopolist who marks p above marginal cost generates a deadweight-loss triangle: consumers whose willingness to pay falls between MC and p are priced out. The two-part tariff dodges that loss by separating the extraction instrument (F) from the consumption instrument (p), letting the firm set each on its own.

The identical-consumer benchmark

Consider a monopolist with constant marginal cost MC and N identical consumers, each with downward-sloping individual demand d(p). For any per-unit price p the firm chooses, each consumer enjoys consumer surplus CS(p) = ∫_p^∞ d(p′) dp′ — the area to the left of the demand curve above the price line.

The firm's per-consumer profit under a two-part tariff is

π(F, p) = F + (p − MC) · d(p)
        subject to:  F ≤ CS(p)        (participation)

The participation constraint binds — leaving any surplus on the table is wasteful — so F* = CS(p). Substituting and differentiating with respect to p gives

dπ/dp = dCS/dp + d(p) + (p − MC) · d′(p)
      = −d(p) + d(p) + (p − MC) · d′(p)
      = (p − MC) · d′(p)
First-order condition: (p − MC) · d′(p) = 0
                       ⇒  p* = MC      (since d′ ≠ 0)

The first two terms cancel exactly: every dollar of consumer surplus the firm could squeeze out by raising p is offset by the loss in CS (and therefore in F) that the higher price imposes. The optimum drops cleanly onto p = MC, with F set to the full consumer surplus at that price. Total profit per consumer is

π* = CS(MC) + 0  =  ½ · base · height       (for linear demand)
                  =  full first-best surplus

This is the result that makes the two-part tariff so striking: it replicates the first-best outcome in an unregulated monopoly, redistributing the entire surplus from consumers to producer.

Worked numeric example

Suppose individual demand is q = 10 − p (so the inverse demand is p = 10 − q), marginal cost is MC = 2, and there is one consumer.

• Uniform monopoly. The monopolist maximises (p − 2)(10 − p), so p_M = 6, q_M = 4. Profit is (6 − 2)·4 = 16. Consumer surplus is ½·(10 − 6)·4 = 8. Deadweight loss is ½·(6 − 2)·(8 − 4) = 8.
• Two-part tariff. Set p = MC = 2. Quantity rises to q = 8. CS(2) = ½·(10 − 2)·8 = 32. Charge F = 32. Profit = 32 + (2 − 2)·8 = 32. Deadweight loss = 0.

The two-part tariff doubles the monopolist's profit (from 16 to 32) and eliminates the deadweight loss — gains come both from reaching the efficient quantity and from extracting the entire surplus.

Walter Oi and the Disneyland Dilemma

The result was formalised by Walter Y. Oi in a 1971 Quarterly Journal of Economics paper titled A Disneyland Dilemma: Two-Part Tariffs for a Mickey Mouse Monopoly. Oi used Disneyland's then-prevailing pricing — a fixed gate admission plus the famous "A-E ticket book" charging a separate fee per ride — as a real-world test of the model. The paper showed three things that the modern literature still treats as canonical:

1. With identical consumers, the firm's optimum is p = MC and F = CS.
2. With heterogeneous consumers, raising F drives marginal consumers out; the constrained second-best has p > MC, sacrificing some efficiency for higher F.
3. If demand is uncertain or consumers can be sorted into groups, multi-block tariffs (different (F, p) menus for different segments) generically dominate any single two-part schedule — the entry point to the full nonlinear-pricing literature.

Disneyland abandoned ticket books in 1982 in favour of pure flat admission (and later, a multi-tier menu of day passes). The shift was widely interpreted as evidence that the heterogeneity-driven trade-offs Oi described had become severe — high-frequency visitors had migrated to annual passes that look like nearly pure access fees, while one-day visitors prefer all-inclusive admission to per-ride accounting.

When consumers differ

The identical-consumer result is unreasonably clean. Real markets have heterogeneous demand, and a single F cannot extract everyone's surplus simultaneously. Three observations organise the heterogeneous case:

• F is bounded by the lowest-type participant's surplus. If the firm wants all N consumer types to buy, F can be no more than CS_low(p) — and the firm leaves CS_high − CS_low on the table for high-types as informational rent.
• Raising p above MC softens the trade-off. A higher p depresses every type's CS, but it depresses high-types' CS proportionally less (because their demand is less elastic at low quantities). The firm thus distorts the per-unit price upward to reduce the rent it has to give up to high-types — at the cost of some deadweight loss.
• Multiple menus dominate single schedules. Offering several (F, p) pairs and letting consumers self-select — what Mussa-Rosen (1978) and Maskin-Riley (1984) characterise as the standard screening solution — generically beats any single two-part contract.

This is the link between two-part tariffs and the broader literature on second-degree price discrimination, mechanism design, and nonlinear pricing. The two-part tariff is the simplest non-trivial menu — one block.

Where two-part tariffs show up

Industry / firm	Fixed fee F	Per-unit price p	Notes
Disney World (pre-1982)	Gate admission ($5–$15)	A–E ticket book per ride	Oi's original case study
Costco / Sam's Club	$65–$130 annual fee	Near-cost shelf prices	~70% of operating income from fees
Gyms (Equinox, Planet Fitness)	Monthly membership	Per-class / per-PT-session fees	Hi-low strategy: cheap entry, paid extras
Cell phone plans (legacy)	Monthly base	Per-GB overage, per-minute roaming	Modern unlimited plans collapse to F-only
Cloud / SaaS (AWS, Datadog)	Subscription tier	Per-API-call, per-GB-egress	Engineered nonlinear pricing
Razor-blade / printer-cartridge	Subsidised hardware	Profitable consumable	Hardware price is implicit F
Electricity / natural gas	Fixed network charge	Per-kWh / per-therm rate	F often covers regulated grid cost
Nightclubs	Cover charge	Drink prices	Cover acts as F; drinks as p
Rental car / U-Haul	Daily / hourly base	Per-mile charge	F covers asset standby, p covers fuel/wear
Country / golf clubs	Initiation + dues	Per-round green fee	Members pay reduced p in exchange for high F

The pattern is so common because it solves an everyday business problem: how to capture surplus from inframarginal users without driving marginal users out — at low cost in administrative complexity. Two prices is barely more complicated than one.

Razor-blade pricing as a disguised two-part tariff

The razor-and-blades business model — pioneered by King Camp Gillette in the 1900s, refined by HP printers in the 1980s, perfected by game consoles and (in reverse) Apple's iPhone — looks superficially different from explicit two-part pricing. Functionally it is the same trick. The firm sells a durable base good (razor handle, printer, console) at or below cost. Once owned, the base good locks the consumer into a specific consumable (blades, cartridges, games) on which the firm charges a high markup. The base good is the access fee F — paid once, granting the right to participate — and the consumable is the per-unit price p.

Two distinctive frictions arise that aren't present in classic two-part pricing:

• Aftermarket competition. If third-party blades, off-brand cartridges, or unlicensed games are technically and legally permitted, the consumable price gets competed down and the model collapses. Firms therefore invest in interoperability locks (chip authentication on cartridges; region encoding on games; proprietary connectors on power tools).
• Consumer myopia. The base-good purchase decision involves comparing a known F to an uncertain stream of future p × q payments. Behavioural research shows that consumers systematically under-weight future consumable costs, which lets firms set p further above marginal cost than a fully rational model would predict. This is the empirical "printer-ink price puzzle": ink margins of 50–80% persist despite the model's prediction of competitive arbitrage.

Relation to bundling, versioning, and nonlinear pricing

Two-part tariffs sit inside a broader family of monopoly-pricing instruments, each picking a different point on the efficiency-versus-extraction frontier:

• Uniform pricing. The simplest. One price, one schedule. Leaves a DWL triangle; captures only the rectangle below price and above MC.
• Perfect (first-degree) price discrimination. Charge each consumer her exact reservation price. Theoretical extreme; requires impossible information.
• Two-part tariff. One menu, two prices. Achieves first-best with identical consumers; pretty good with mild heterogeneity.
• Multi-block tariff / declining-block pricing. Sequence of price thresholds, each cheaper than the last. Better separation across types; common in electricity, telecoms.
• Versioning. Multiple products at different (F, p, quality) bundles; consumers self-sort. Microsoft Office tiers, airline cabin classes.
• Bundling. Sell goods in combinations to exploit demand correlations. Cable TV channel packages.
• Mechanism design / optimal nonlinear pricing. Solve for the menu that maximises profit subject to incentive-compatibility and individual-rationality constraints. The Mussa-Rosen-Maskin-Riley framework; the gold standard.

Each instrument trades menu complexity (and administrative cost) against revenue. A two-part tariff is the cheapest non-trivial mechanism that captures first-order gains over uniform pricing — which is why it shows up so much in practice.

Welfare and distribution

The welfare verdict on two-part tariffs depends on whose welfare you care about:

• Efficiency. With identical consumers, two-part pricing is fully efficient — the entire potential surplus is realised, just redistributed from consumers to producer. With heterogeneous consumers it is still typically more efficient than uniform monopoly pricing, because p is closer to MC even when not equal to it.
• Consumer welfare. Strictly worse than competition. A monopolist using two-part pricing extracts more surplus than one using uniform pricing — the very property that makes the instrument attractive to firms.
• Distribution. Regressive when F is large relative to typical consumption. Light users pay an effectively high average price per unit; heavy users a low one. Costco illustrates: a heavy weekly shopper amortises the $65 fee over thousands of dollars of groceries; an occasional shopper finds the fee a prohibitive fraction of her spend.
• Discrimination class. Pigou (1920) classifies a two-part tariff as second-degree price discrimination — the menu is the same for all consumers, but consumers self-select different average prices by choosing how much to consume. Unlike third-degree (group-based) discrimination, it requires no demographic sorting and is therefore harder to regulate against.

Common pitfalls

• Forgetting the participation constraint. The textbook optimum F = CS(p) holds only when CS is non-negative — i.e., the firm cannot set F higher than the consumer's willingness to participate. Overshooting drives buyers out entirely and the firm collects nothing.
• Treating heterogeneous demand like homogeneous. Applying the identical-consumer formula F* = CS(MC) when consumers differ leads to dramatic under-pricing of low-types or over-extraction from high-types, depending on direction. The constrained second-best is materially different from the unconstrained ideal.
• Confusing fixed cost with fixed fee. F is a payment from consumer to firm. It is not the firm's fixed production cost (rent, capital). The two move independently — F can be far above or far below the firm's fixed cost.
• Ignoring resale arbitrage. If consumers can resell access (one Costco member buying for ten friends), the F-based extraction model breaks down because the firm only sees one F per resale group. Mitigations include membership cards, photo IDs, and per-account purchase limits.
• Setting F to extract the average surplus. A common mistake: average-surplus pricing under-extracts from high-types and excludes low-types. The correct heterogeneous-case strategy is to set F at the lowest-participating-type's surplus and recover the rent gap through a markup on p.
• Ignoring outside options. The model implicitly compares "consume q via the two-part tariff" to "consume zero". Where good substitutes exist (other gyms, other clubs, other cloud providers), F is bounded by net-of-substitute surplus, which can be much smaller than the unconstrained CS.

Mechanism-design viewpoint

From the mechanism-design perspective, a two-part tariff is a particularly simple direct revelation mechanism: consumers reveal their type by choosing q, the menu maps (F, p, q) to a payment, and we ask whether the menu is incentive-compatible (each type prefers its assigned (F, p, q) point) and individually rational (each type's net surplus is non-negative).

With identical consumers, every type has the same preference, so incentive compatibility is trivial and individual rationality binds — giving Oi's result. With heterogeneous types, the firm's optimisation problem becomes a constrained allocation: maximise expected profit subject to all types preferring their own menu item and earning non-negative net surplus. Solving this problem in general yields the Mussa-Rosen and Maskin-Riley results — and shows that the optimal menu often involves more than one (F, p) pair. The two-part tariff is then a benchmark, not the answer, in any realistic setting.