A single-product monopolist faces a well-understood problem. It sets price where marginal revenue equals marginal cost, and the resulting markup obeys the Lerner condition: the price-cost margin equals the inverse of the own-price elasticity of demand. The rule is elegant and testable. It is also almost entirely inadequate for describing how most real firms set prices, because most real firms do not sell one thing.

Firms sell inkjet printers near or below cost and recover the money on cartridges. Razors are subsidised by blades, and game consoles are sold at a loss to sell games. Airline fares are unbundled from baggage and seat selection. None of these is an anomaly. Each is what optimal pricing looks like once a firm internalises the effect that each price has on the demand and cost of its other products. A good priced below marginal cost can be perfectly rational as part of a portfolio. A good priced well above its standalone monopoly level can be rational for the opposite reason.

This creates a real puzzle for regulation. Regulators usually apply predation tests, cross-subsidy rules and cost-based price controls one product at a time, yet multi-product pricing works at the level of the portfolio. Three forces drive it: demand interdependencies, cost complementarities and strategic interaction with rivals and potential entrants. Each modifies the Lerner benchmark in a distinct way, and each changes how we should judge pricing conduct.

Demand Relationship Exploitation

Consider a firm selling two goods. When it raises the price of the first, it loses some sales of that good, and it also shifts demand for the second. If the goods are substitutes, some of the lost customers move to the firm's other product, which softens the revenue loss. If they are complements, the firm loses sales on both. The optimal markup on each good therefore depends on more than its own elasticity. It also depends on the margins earned on related products and the cross-price elasticities linking them.

Formally, the multi-product Lerner condition adds a correction term. The markup on good i equals its inverse own-elasticity, adjusted by the margins on related goods, weighted by cross-elasticities and relative revenues. For substitutes, the correction pushes the optimal markup above the independent-pricing benchmark. The same mechanism underlies merger analysis based on diversion ratios and upward pricing pressure, as formalised by Farrell and Shapiro. A merger turns a rival's product into a sibling, so cannibalisation that was once someone else's problem becomes an internal one.

Complements push in the opposite direction. Cournot observed in 1838 that independent monopolists supplying complementary inputs, such as copper and zinc for brass, would together charge more than a single integrated monopolist. Each ignores the harm its markup does to the other's demand. Integration internalises that externality and lowers prices. Inside a single firm, the same logic can push the margin on one good below zero. The console, the printer or the base airfare becomes a vehicle for selling higher-margin complements.

Behavioural considerations sharpen this result. Gabaix and Laibson's analysis of shrouded attributes shows what happens when some consumers fail to anticipate add-on prices. Firms then price the visible base good low and the hidden add-on high. Competition does not necessarily fix this. A firm that reveals its add-on pricing educates myopic consumers, who may then exploit rivals' loss leaders rather than switch to the honest firm. The equilibrium can survive with many competitors, which sets it apart from classical complementarity pricing.

The distinction matters for policy. Loss-leader pricing driven by genuine complementarity is efficient. It moves prices toward the integrated optimum and typically benefits consumers. Loss-leader pricing driven by shrouding transfers surplus from naive to sophisticated buyers and distorts purchase decisions. The observable price structure may be identical in both cases. Telling them apart requires evidence on whether buyers forecast lifetime costs at the point of purchase. Inspecting margins alone cannot answer that.

Takeaway

A negative margin on a single product tells you almost nothing on its own. What matters is whether the cross-effects that justify it reflect real complementarity or consumers' failure to see the whole price.

Cost Scope Economics

On the supply side, multi-product firms exist partly because producing goods together can be cheaper than producing them separately. Baumol, Panzar and Willig formalised this as economies of scope: a cost function has them when producing a bundle of outputs jointly costs less than the sum of standalone costs. Sources include shared inputs with spare capacity, common distribution networks, joint marketing and transferable know-how. Railways carrying both freight and passengers are a canonical example, as are telecom networks carrying voice and data.

Scope economies create a problem with no clean theoretical solution. Common costs cannot be allocated to individual products in any economically meaningful way. When one network is built to serve two services, there is no fact of the matter about how much of its cost belongs to each. Regulators long relied on fully distributed cost methods, which impose an accounting allocation that can produce prices with little relation to efficiency. A product may look loss-making under one allocation and profitable under another, while the underlying economics stay the same.

Theory offers two more defensible approaches. Faulhaber's cross-subsidy framework bounds acceptable prices. Revenue from any subset of products should exceed that subset's incremental cost and fall below its standalone cost. Within those bounds, no group of customers subsidises another, and no group would prefer to leave for a separate supplier. Ramsey-Boiteux pricing then chooses among these subsidy-free price vectors. It sets markups inversely proportional to demand elasticities, which recovers common costs while minimising the deadweight loss of pricing above marginal cost.

Ramsey pricing has an uncomfortable distributional implication. The customers with the fewest alternatives bear the largest share of common costs, so efficiency and equity can diverge sharply. Regulators have often softened Ramsey structures for that reason. Ramsey pricing also requires the regulator to know demand elasticities that the firm understands better, which turns the problem into one of mechanism design. Littlechild's price caps on a basket of services partly sidestep this. Laffont and Tirole showed that caps with suitable weights let the firm move relative prices toward Ramsey levels using its own information.

Scope economies also shape where product lines end. Teece observed that scope economies alone do not require a multi-product firm, since a shared input could in principle be sold or licensed to other firms. Integration becomes necessary when that input is tacit knowledge, highly specific capital or otherwise costly to contract over. A firm's product line therefore reflects both technology and transaction costs. A product is worth keeping when its incremental revenue covers its incremental cost, whatever share of common cost an accountant assigns to it.

Takeaway

Common costs belong to the portfolio, not to any single product. Incremental and standalone cost set the defensible bounds, and every price in between is a choice about who bears the burden.

Strategic Entry Deterrence

Multi-product pricing is not only about extracting surplus from existing demand. It is also a strategic tool for shaping the competitive environment. An incumbent with a broad product line has three ways to make entry less attractive. It can occupy positions in product space. It can link the pricing of its products so entrants must compete on several fronts at once. And it can signal how aggressively it will respond. The portfolio itself becomes part of the barrier to entry.

Schmalensee's study of the ready-to-eat cereal industry gave the classic account of brand proliferation. Suppose products are differentiated along a spatial dimension and entry requires sunk costs. An incumbent that fills product space densely leaves no niche large enough for an entrant to recover its investment. The incumbent accepts some cannibalisation among its own brands as the price of keeping entrants out. The Federal Trade Commission pursued this theory against major cereal producers in the 1970s, ultimately without success.

Judd later identified a crucial limitation. Once an entrant appears next to one of the incumbent's products, prices fall sharply in that segment. Through substitution, they also fall on the incumbent's neighbouring brands. A multi-product incumbent may therefore withdraw the contested product to relieve pressure on the rest of its line. If exit is cheap, proliferation is not a credible deterrent. Its force depends on commitment: sunk costs, irreversible capacity or reputational stakes that make withdrawal expensive.

Bundling and tying provide another channel. Whinston showed that tying a monopolised good to a competitive one commits the incumbent to aggressive pricing in the tied market, because every lost sale there now costs it the margin on the bundle. That commitment can deter a single-product entrant that would otherwise have been viable. Nalebuff extended the argument, showing that mixed bundling can shrink the range of prices at which an entrant can profitably operate, even without outright foreclosure. Multi-product loyalty rebates raise the same issue, as litigated in LePage's v. 3M and addressed by the discount attribution test in PeaceHealth.

Yet deterrence is not always optimal. Fudenberg and Tirole's taxonomy shows that when prices are strategic complements, an incumbent may prefer a 'puppy dog' stance. It prices softly, which encourages the entrant to price softly too, rather than committing to aggression. A multi-product incumbent facing a niche entrant may rationally give up a segment and keep high prices elsewhere. For antitrust, this means the same pricing structure can deter or accommodate depending on the strategic setting. Only analysis at the portfolio level can tell which.

Takeaway

A product line is also a commitment device. Its strategic force depends less on how many products the incumbent sells than on how costly it would be to stop selling them.

The common thread is that the product line, not the individual product, is the natural unit of economic analysis. Demand interdependencies, scope economies and strategic interaction each break the link between a product's price and its own cost and elasticity. Margins on individual goods result from portfolio-level optimisation and cannot be read in isolation.

This is a challenge for regulation built around product-level tests. Below-cost pricing may reflect efficient complementarity, cost recovery on Ramsey principles, exploitation of shrouded add-ons or entry deterrence. Telling these apart requires information about cross-elasticities, consumer foresight, cost structures and commitment. Much of that information is private to the firm.

That information gap is the deeper lesson. Hurwicz's framing applies directly. Good policy toward multi-product firms is less about specifying the correct prices than about designing rules whose outcomes stay acceptable when the regulator cannot see everything the firm sees. Incremental cost tests, basket price caps and bundled discount attribution are all examples of such rules.