Bundling Pricing

Why everything comes in packages

Look at the things you bought this week. The McDonald's combo is a bundle (burger + fries + drink). The Spotify Duo plan is a bundle (two accounts + family billing). Microsoft Office is a bundle (Word + Excel + PowerPoint + Outlook + Teams). Your cable subscription is a bundle (ESPN + CNN + Food Network + 200 channels you don't watch). Amazon Prime is a bundle (two-day shipping + Prime Video + Prime Music + Prime Reading + photo storage). The pricing technique is so universal it feels invisible — but it is the result of a specific economic theorem, and it shapes one of the longest-running debates in antitrust law.

The theorem is simple to state. If a monopolist sells two goods to buyers whose valuations for those goods are negatively correlated — some buyers love A and merely tolerate B, other buyers reverse the preference — then a bundle priced near the average total valuation will earn more revenue than any pair of separate prices on the two goods. The intuition runs through statistics: pooling negatively correlated random variables shrinks their combined variance, and a tighter variance of willingness-to-pay means a single price reaches more of the population without leaving as much surplus on the table. The geometry was first drawn out by William James Adams and Janet Yellen in their 1976 paper Commodity Bundling and the Burden of Monopoly, and forty years of industrial-organization research has built on that diagram.

Pure bundling and mixed bundling

Modern theory distinguishes two regimes, named for what the seller offers on the menu.

Pure bundling offers only the bundle. There is no way to buy the components separately. Microsoft sells Office as a single SKU; you cannot legitimately buy Word standalone from Microsoft. A movie ticket bundles a feature film with previews; nobody offers you a fifteen-minute discount to skip the trailers. A cruise package bundles transport, lodging, meals, and entertainment with no à-la-carte option. Pure bundling is the simpler regime to analyse and to administer: one product, one price, no menu.

Mixed bundling offers the bundle and the individual components, typically pricing the bundle at a discount to the sum of standalone prices. Cable TV packages are the textbook example — you can buy ESPN as part of a tier, or buy the standalone ESPN+ streaming service, or buy individual sport packages. Adobe Creative Cloud is mixed: any single Adobe app for $22/month, or the whole suite for $60/month. Restaurants overwhelmingly do mixed bundling: the menu lists every item separately, and then lists value-priced combos. Mixed bundling weakly dominates pure bundling in theory because it adds a separating option for customers who value only one good highly. In practice it costs more to administer and is more vulnerable to antitrust challenge.

The Adams-Yellen geometry

Adams and Yellen's argument is best understood on a diagram. Imagine two goods, A and B, sold by a monopolist at zero marginal cost. Each consumer is a point in the (v_A, v_B) plane — their valuation for A and their valuation for B. A standalone price p_A buys everyone whose v_A exceeds p_A, regardless of v_B; the seller's payoff is the rectangle to the right of p_A times the price. Similarly for p_B.

Now draw a bundle price p_AB. Every consumer with v_A + v_B ≥ p_AB buys the bundle. The "buy" region is a half-plane sloped at minus one, capturing exactly the customers whose total willingness-to-pay exceeds the bundle price.

The key picture: when v_A and v_B are negatively correlated, the cloud of consumers stretches along the v_A + v_B = constant diagonal. The half-plane neatly cleaves off the high-total-valuation half of the cloud — and that half is fatter and more concentrated than any rectangle a pair of standalone prices could carve out. Bundle revenue beats separate revenue. When valuations are positively correlated, the cloud is aligned with the diagonal, and the bundle's half-plane offers no advantage over independent rectangles.

Customer 1: v_A = 90, v_B = 20
Customer 2: v_A = 20, v_B = 90

Separate pricing at p_A = $90, p_B = $90:
  customer 1 buys A only      →  $90
  customer 2 buys B only      →  $90
  total revenue                  $180

Separate pricing at p_A = $20, p_B = $20:
  both customers buy both     →  $80
  total revenue                  $80

Pure bundle at p_AB = $110:
  v_A + v_B = 110 for both    →  both buy
  total revenue                  $220

The bundle beats either standalone price configuration. The numerical case is the canonical pedagogical example, and the same logic scales to continuous distributions of valuations across millions of customers.

Variance reduction — the deeper reason

The Adams-Yellen geometry is one way to see it; a statistical lens makes the result robust to arbitrary distributions. Let X be a buyer's valuation for good A and Y their valuation for good B. The variance of X + Y is

Var(X + Y) = Var(X) + Var(Y) + 2 · Cov(X, Y)

When Cov(X, Y) is negative, the variance of the total is strictly less than the sum of the individual variances. A monopolist setting a single price on a random valuation distribution captures the most revenue when the distribution is tight — variance is the source of unmet demand on the low side and unextracted surplus on the high side. Bundling negatively correlated goods pools them into a single tighter distribution, and the monopolist's optimal single price now corresponds to a flatter, fatter mass of customers. Bakos and Brynjolfsson (1999) extended this to large bundles of information goods and showed that as the number of bundled items grows, the law of large numbers drives the bundle's per-good demand curve toward a step function — first-degree price discrimination in the limit.

Worked example: software suite vs separate sales

Suppose a software firm has four buyers and two products, Word and Excel. Marginal cost is zero. Valuations are:

Buyer	v(Word)	v(Excel)	Total
Writer	$100	$10	$110
Analyst	$10	$100	$110
Manager	$60	$60	$120
Student	$30	$30	$60

Compare three pricing strategies.

Separate pricing, p_W = $60 and p_E = $60. Writer buys Word ($60). Analyst buys Excel ($60). Manager buys both ($120). Student buys nothing. Revenue = $240.

Separate pricing, p_W = $30 and p_E = $30. All four buyers take both products. Revenue = 4 × 60 = $240.

Pure bundling, p_AB = $110. Writer, Analyst, and Manager all have v_W + v_E ≥ $110 and buy. Student does not. Revenue = 3 × 110 = $330.

Mixed bundling, p_W = $100, p_E = $100, p_AB = $110. Writer buys Word standalone at $100 (paying more than $110 − 10 = $100 would have given him a strict consumer-surplus loss; the deal works only because we are at indifference). Analyst takes Excel at $100. Manager buys the bundle for $110. Student is priced out. Revenue = $100 + $100 + $110 = $310. With a marginal cost or with a slight shift in the menu, mixed bundling can be tuned to beat $330 — but the point is that pure bundling already extracts substantially more surplus than separate pricing in the negative-correlation case.

This is the exact logic Microsoft used when stapling Excel, PowerPoint, and Word into Office in the early 1990s, killing WordPerfect and Lotus 1-2-3 in the process.

Where bundling shows up in the real economy

Bundle	Type	Mechanism
McDonald's value meal	Mixed	Pools low-burger-only and low-fries-only customers; production complementarity in the kitchen
Microsoft Office	Pure	Negatively correlated Word vs Excel valuations across writers, analysts, designers
Cable TV tier	Mixed	Pools sports fans, news watchers, lifestyle viewers; supply-side enforced by network wholesale contracts
Spotify Duo / Family	Mixed	Discounts for multi-account households; pools heterogeneous listening preferences
Amazon Prime	Mixed	Shipping + video + music + reading + photos under one annual fee; two-part tariff layered on bundling
Adobe Creative Cloud	Mixed	Photoshop + Illustrator + Premiere etc.; pools designers, illustrators, video editors
Hospital all-payer rates	Pure (DRG)	Diagnosis-related groups bundle all care for an episode at one price
Hotel resort fees	Pure	Wifi + gym + pool included whether you use them or not
Cruise vacation package	Pure	Transport + lodging + meals + entertainment unseparated
iPhone (iOS + Safari + App Store)	Pure	Operating system tied to browser and store; current antitrust focus in EU and US

Tying and the antitrust risk

Bundling becomes tying when a seller with market power in one product conditions its sale on the purchase of a second. The economic theory of harm is the leverage hypothesis: the seller uses the tying-good monopoly to capture or foreclose the tied-good market. The classical view (the Chicago school) was sceptical — if the seller has one monopoly already, they earn one monopoly profit, and tying merely re-prices the bundle, not the rents. The modern post-Chicago view (Whinston 1990, Carlton-Waldman 2002) restored the leverage theory by showing that tying can credibly commit to predatory pricing in the tied market and so deter entry there.

The most famous case is United States v. Microsoft (2001). Microsoft was alleged to have used its Windows operating-system monopoly to tie Internet Explorer into Windows, foreclosing Netscape in the browser market. The DC Circuit's en-banc decision agreed that the conduct violated Section 2 of the Sherman Act on monopolisation grounds, while moving technological tying to a rule-of-reason analysis rather than per se illegality. The remedy was a behavioural decree, not a structural breakup — but the case set the template for every subsequent platform-tying case. Apple's bundling of Safari and the App Store with iOS is the current European Digital Markets Act target; Google's tying of Search to Android is being litigated in US v. Google (Android).

The legal test now turns on (1) market power in the tying product, (2) substantial foreclosure of competition in the tied product, (3) the tied product being a separable product (not just a feature), and (4) absence of a procompetitive justification. The economics is rarely about whether the bundle exists — it almost always does — and almost entirely about whether the bundle is the firm's only way to monetise complementarities or a tool to deny rivals scale.

The à-la-carte cable debate

For two decades, US consumer advocates and FCC commissioners have pushed for à-la-carte unbundling of cable TV: let viewers buy ESPN, HBO, and CNN individually instead of paying for 200 channels. The economics is more ambiguous than the policy rhetoric. Industry-funded studies (and several independent ones) consistently find that à-la-carte raises projected consumer prices: when ESPN is sold standalone instead of bundled, its price has to rise to recover the cross-subsidy from low-intensity viewers, and households that watched ESPN occasionally end up paying more for less. The bundled tier worked as a Pareto-efficient cross-subsidy from light to heavy viewers.

Streaming has unwound much of this from the supply side. ESPN+, Disney+, HBO Max, Apple TV+, and Netflix are à-la-carte. Surprise — they have not turned out cheaper than the cable bundle they replaced. By 2024 the average US household subscribed to five or more streaming services and paid more in aggregate than a basic cable subscription would have cost. The pricing has not gone away; the bundle has fragmented but the same surplus is being extracted by the new gatekeepers. Disney+ now bundles Hulu and ESPN+ back together; Amazon Prime bundles video, music, and shipping; the wheel turns and bundling re-emerges in every generation of platform competition.

Variants and extensions

• Mixed bundling with menus. Adobe-style "individual app at $22, suite at $60" menus implement second-degree price discrimination via self-selection. The premium-tier discount versus the sum of standalone prices is the metric to watch.
• Tying without bundling. Requirement contracts (you must use only IBM punchcards in your IBM machine) achieve the same foreclosure without a single price. Banned by IBM v. United States (1936) and a long line of cases since.
• Two-part tariffs with bundles. Amazon Prime — fixed annual fee for unlimited access to the bundle's contents. Combines bundling with subscription pricing and gets the entry-deterrence benefits of both.
• Bakos-Brynjolfsson large bundles. When N is large and goods are independent or negatively correlated, the law of large numbers turns a bundled subscription into near-first-degree price discrimination. The basis for Spotify's flat-fee music catalogue and Netflix's flat-fee video catalogue.
• Bundling under competition. Nalebuff (2004) showed that asymmetric bundling — one firm bundles, the rival does not — can be a credible entry-deterrence weapon even without per se monopoly power.
• Bundled discounts (the loyalty rebate). A seller offers a percentage discount on good A only if the buyer also takes good B. The de Beers diamond and pharmaceutical-rebate cases (the AstraZeneca and Intel proceedings) all turn on whether such schemes amount to exclusive dealing.

Welfare consequences — winners and losers

Bundling redistributes surplus along three dimensions.

Producer vs consumer. The monopolist captures more total surplus under bundling than under separate pricing whenever the Adams-Yellen condition holds. Consumer surplus generally falls. Total welfare can rise (because more units are sold) or fall (because some buyers are forced to consume goods they don't value); the sign is empirical.

High-valuation vs low-valuation consumers. Buyers in the diagonal "high-A, high-B" corner gain little from bundling because they would have bought both at any reasonable separate price. Buyers in the "high-A, low-B" and "low-A, high-B" corners — the Adams-Yellen sweet spot — bear most of the surplus extraction; they pay for goods they don't really want. The à-la-carte cable complaint is exactly this distributional point.

Rivals foreclosed. Tying can drive out efficient single-product rivals (Netscape, WordPerfect, Lotus 1-2-3, RealNetworks). The lost-rival cost is the long-run competitive cost of bundling and is the locus of antitrust concern. Where the bundle is the only viable distribution channel, market structure tips toward a single integrated provider — and once concentrated, prices typically rise.

Common pitfalls

• Confusing complementarity with negative correlation. Bundling toothbrushes with toothpaste exploits complementarity (both used together). Bundling Word with Excel exploits negative correlation of valuations (different users want different ones). The mechanisms are distinct: complementarity bundles save on transaction costs, negative-correlation bundles extract surplus.
• Assuming bundling always works. When valuations are positively correlated — luxury watches and luxury cars — buyers who want one also want the other, and the bundle offers no variance-reduction gain. Separate pricing dominates.
• Treating pure bundling as the benchmark. Mixed bundling weakly dominates pure bundling and separate pricing in almost every theoretical setting (McAfee-McMillan-Whinston 1989). In practice, fixed costs of administering the menu sometimes tip the balance toward pure.
• Forgetting marginal cost. Adams-Yellen assumes zero marginal cost. With positive marginal cost, bundling is less attractive because the seller loses money on buyers whose valuation for one good is below cost. Software, media, and information goods (near-zero MC) are the natural home of bundling for this reason.
• Treating tying as automatically illegal. Tying requires market power in the tying good. Hot dog vendors who sell only buns with their dogs are not antitrust violators. The Microsoft case turned not on the bundling per se but on Microsoft's Windows monopoly being used to foreclose Netscape.
• Ignoring entry deterrence. A bundle priced just above the cost of producing both goods can deter a single-product entrant who would otherwise have served the high-A, low-B niche. The dynamic competitive effect is often larger than the static surplus extraction.