The Price of Ideas

What does intellectual property actually cost — and what does it actually buy? A model-driven look at drugs and books, and an honest reckoning with the case for abolition

Keywords: intellectual property; copyright; patents; appropriability; non-rivalry; deadweight loss; pharmaceutical R&D; generic entry; creative industries; optimal patent length; Kinsella

---

Abstract. Intellectual property converts a non-rivalrous good — an idea, a molecule, a sequence of words — into a temporary, excludable, rivalrous one, accepting a static welfare loss in exchange for a dynamic incentive to create; whether that trade is worth making is not a matter of principle but of measurement, and the measurements differ sharply by sector, so this essay builds the trade explicitly and tests it against the evidence in the two industries where the stakes are clearest. For pharmaceuticals it shows, using the published cost estimates ($2,558M capitalised per approval in DiMasi, Grabowski and Hansen, 2016; a contested median of $985M in Wouters, McKee and Luyten, 2020), the industry cost of capital (10.5%), and measured post-expiry price collapse (US prices falling 32% within a year and 82% within eight years of patent loss, Serra-Burriel et al., 2024), that the appropriability gap is genuinely fatal: without exclusivity the asset that repays a billion-dollar fixed cost stops paying the instant generics enter, and a stylised break-even model demonstrates how steeply the required return rises as effective patent life shortens. For creative works it shows, via the Landes–Posner (1989) cost-of-expression model and the causal evidence (Giorcelli and Moser, 2020), that basic copyright did raise both the quantity and quality of output — but that protection beyond the creator’s life added nothing, that digitisation collapsed recorded-music revenue by more than half in real terms without collapsing output (Waldfogel, 2017, 2018), and that the contractual alternative favoured by the abolitionist position fails precisely where mass distribution makes appropriability hardest because it cannot bind the third party who never signed. The conclusion is neither that IP is a swindle nor that more of it is better: Kinsella’s (2008) non-rivalry premise is correct and is in fact the starting point of the orthodox welfare analysis, but his universal conclusion is falsified by the pharmaceutical and historical-copyright evidence, while the maximalist position is falsified by the same evidence pointing the other way — the welfare curve has an interior peak, and most of the live policy disputes are about how far we have wandered to the right of it.

---

1. The thing that makes ideas different

Begin with the property that every serious treatment of this subject begins with, because everything follows from it. An idea is non-rivalrous. If I teach you a method for harvesting cotton, I still have the method; your use does not consume mine. If you copy this essay, the original arrangement of words is undiminished. Kenneth Arrow put the point with characteristic economy in 1962: information, as an economic commodity, is peculiar because its use by one party does not deplete it, and once produced it can be reproduced at a cost approaching zero. Arrow framed this as the inappropriability of information — absent legal protection, any purchaser can reproduce it “at little or no cost,” so a competitive market cannot reward its creation — and drew the conclusion the rest of this essay must reckon with: a free-enterprise economy will therefore tend to underinvest in the production of knowledge. This is not a libertarian observation or a statist one. It is a physical fact about information, and it is the foundation of the entire field.

From that single fact, two opposite-looking conclusions can be drawn, and the whole debate is a quarrel between them.

The first conclusion is the orthodox welfare-economics one. Precisely because an idea can be copied at near-zero cost, a competitive market will drive the price of copies down to that near-zero marginal cost. But the first copy is not cheap. Discovering a drug, writing a novel, composing an opera, debugging a compiler — these carry a large fixed cost that does not vary with the number of copies sold. Landes and Posner (1989) gave this fixed cost its standard name: the cost of expression. If competition pushes price to marginal cost, the cost of expression is never recovered, and the rational creator, foreseeing this, does not create. The market under-provides the good. Intellectual property is the legal device that restores appropriability — that lets the creator charge above marginal cost for a limited period, so the wedge between price and cost can repay the fixed cost. We accept a static inefficiency (some buyers who value the good above its marginal cost are priced out) to buy a dynamic benefit (the good exists at all).

The second conclusion is the abolitionist one, and its sharpest modern statement is Stephan Kinsella’s Against Intellectual Property (2008). Property rights, Kinsella argues, exist for one reason only: to resolve conflict over scarce resources. Two people cannot both eat the same loaf, so we need rules about who owns the loaf. But two people can both use the same recipe. There is no scarcity, hence no possibility of conflict, hence no warrant for a property right. Intellectual property, on this view, does not protect property — it manufactures an artificial scarcity in something naturally abundant, and enforces it with state coercion that necessarily overrides real property rights in tangible things. Your copyright in a pattern of words restricts what I may do with my own paper, my own ink, my own printing press. The deadweight loss is real; the justification, Kinsella says, is not.

Notice what has happened. Both camps start from the identical premise — non-rivalry — and reach opposite destinations. That tells you the disagreement cannot be settled at the level of premises. It is an empirical question: does the absence of appropriability actually suppress the creation of valuable works below the social optimum, and if so, by how much, and in which industries? The rest of this essay is an attempt to answer that question with numbers rather than postures. I take the two industries where the answer matters most and where it turns out to differ most: pharmaceuticals, where the case for protection is at its strongest, and creative works, where it is at its most ambiguous.

2. The static cost, drawn precisely

Before the evidence, the mechanism — because you cannot weigh a trade-off you have not drawn. Figure 1 is the standard monopoly diagram, and it contains, in one picture, both the libertarian’s complaint and the economist’s reply.

Let inverse demand be P = a − bQ and let the marginal cost of reproducing a copy be a constant c. Two prices matter. The competitive price is P = c: this is what a market without intellectual property delivers, because free entry by copiers competes the price down to the cost of making one more copy. The monopoly price P_m is set where marginal revenue equals marginal cost, and it is strictly higher.

Three regions follow. The blue rectangle is producer surplus — the wedge (P_m − c) × Q_m. This is the only region that can repay the cost of expression F. Knock it out and the fixed cost is unrecoverable. The red triangle is the deadweight loss: the buyers who value the good above c but below P_m, who would have been served under competition and are not served under monopoly. This is the static cost of intellectual property, and — this is the point worth sitting with — Kinsella and Boldrin and Levine are simply correct that it exists. There is no honest version of the pro-IP argument that denies the red triangle. The green region is the consumer surplus that survives.

The orthodox claim is not that the red triangle is illusory. It is that, in the absence of the blue rectangle, the good is not produced, so the relevant comparison is not “monopoly versus competition” but “monopoly versus nothing.” If the work would exist either way, IP is pure deadweight loss and the abolitionist wins on that work. If the work exists only because of the blue rectangle, then the consumer surplus it generates — the entire area under the demand curve above price — is a benefit that would otherwise be zero, and it must be set against the triangle. Which case obtains is, again, empirical. So let us go and measure.

3. Pharmaceuticals: where the gap is fatal

3.1 The size of the fixed cost

The pharmaceutical industry is the cleanest natural experiment in the appropriability problem that exists, because it combines two features in the extreme: a colossal fixed cost of creation and a trivial marginal cost of imitation. A small-molecule drug that costs a fortune to discover and prove safe can, once its structure is public, be synthesised by a competent generic manufacturer for cents per dose.

How large is the fixed cost? The most-cited figure comes from DiMasi, Grabowski and Hansen (2016), who surveyed ten firms on a random sample of 106 compounds first tested in humans between 1995 and 2007. Their headline numbers, which I have read in the original rather than from any summary: the average out-of-pocket cost per approved compound is $1,395 million in 2013 dollars; capitalising those outlays to the point of approval at a real cost of capital of 10.5% — to account for the decade or more between spending and any revenue — yields $2,558 million; adding post-approval research raises it to $2,870 million. The capitalisation matters enormously. More than half of the headline figure is not cash spent on chemistry; it is the time-value of money tied up for years in projects that mostly fail. Indeed the engine of the estimate is failure: the overall probability that a compound entering human trials is eventually approved was estimated at just 11.83%, down from 21.5% in the authors’ earlier work. You are not paying for one drug. You are paying for the eight that died so that one could live.

That figure is contested, and intellectual honesty requires saying so plainly rather than leaning on the number that flatters the argument. The DiMasi study is funded in part by the industry whose costs it estimates, uses proprietary data that cannot be independently audited, and has drawn sustained methodological criticism. The most credible rival estimate, Wouters, McKee and Luyten (2020), used publicly available data — principally US Securities and Exchange Commission filings — for the 63 of 355 drugs and biologics approved by the FDA between 2009 and 2018 for which development costs could be reconstructed, applied the same 10.5% cost of capital, and found a median capitalised cost of $985 million (95% CI, $684–1,229 million) and a mean of $1,336 million — roughly a third to a half of DiMasi’s figure. The gap is partly explained by sample (Wouters’s public-data firms skew smaller and leaner) and partly by methodology. I do not adjudicate the dispute here. I carry both numbers through the model, because a conclusion that survives only under the larger figure is not a conclusion, it is special pleading.

Either way, the order of magnitude is the same and it is the only thing the argument needs: the fixed cost of a new drug is measured in the high hundreds of millions to low billions of dollars, and the marginal cost of a copy is measured in cents.

3.2 The cliff

Now watch what happens to the asset when its legal protection ends. Figure 2 traces the price of a branded drug over its commercial life.

During the protected period the brand sustains a price far above marginal cost; this is the blue window, the only interval in which the fixed cost F can be repaid. Then the patent expires, generics enter, and the price does not drift downward — it falls off a cliff. The two marked points are measured, not assumed. Serra-Burriel et al. (2024), in a difference-in-differences study of 505 drugs losing protection across eight high-income countries between 2011 and 2020, found that in the United States prices fell by 32% within the first year of patent loss and by 82% within eight years (95% confidence intervals 24–39% and 71–89%). The decline varied across countries — from 64% over eight years in Australia down to 18% in Switzerland — which is itself informative: the speed of the collapse is a function of how competitive the generic market is allowed to become. (One caveat the authors flag and I repeat: these are list prices, not net of the confidential rebates that pervade US drug pricing, so the true net collapse may differ in level though not in direction.)

For specific blockbusters the collapse is starker still. When atorvastatin (Lipitor) lost US protection in late 2011, the price of a month’s supply fell by more than 80% within roughly a quarter once multiple generic manufacturers were cleared to compete. The general pattern is well established in the health-economics literature: a single generic entrant produces a discount of perhaps 20–40%; the steep declines arrive as the second, third and later entrants pile in, because the post-entry market for a chemically identical molecule is close to the textbook ideal of perfect competition, and perfect competition does exactly what the textbook says — it drives price to marginal cost and economic profit to zero.

This is the appropriability problem of Figure 1, made concrete and given a clock. The generic firm did not pay the $985 million, or the $2,558 million, or whatever the true figure is. It free-rides on a fixed cost it did not bear, and in doing so it competes away the very wedge that was supposed to repay that cost. If generics could enter immediately, on the day of approval, the blue window would have zero width, the wedge would never open, and — by the logic of Figure 1 — the drug would not be developed in the first place. The patent does not enrich the innovator at the consumer’s expense in some gratuitous way. It is the device that makes the blue window wide enough to clear the fixed cost. Take it away and you do not get cheap drugs; you get no drugs.

3.3 The break-even, and why effective life is everything

We can be quantitative about how wide the window must be. Suppose, for the sake of a transparent illustration — and I flag every assumption rather than smuggling it, because an unstated assumption is a defect — that a drug earns a constant annual gross margin M (price minus marginal manufacturing cost, times volume) for L years of effective on-market exclusivity, after which the cliff takes the margin to zero. The fixed cost F is already risk-adjusted in the DiMasi and Wouters methodologies, since both link the cost of failures to the successes, so I treat the cash flow as deterministic. To break even at the industry real cost of capital r = 10.5%, the present value of the margin stream must equal the fixed cost:

F  =  M · [ 1 − (1 + r)^(−L) ] / r

which rearranges to the required annual margin

M  =  F · r / [ 1 − (1 + r)^(−L) ].

Figure 3 plots this hurdle against effective patent life, for both cost estimates.

The lesson is in the convexity. With a ten-year effective life, a $2,558 million drug must clear about $425 million in gross margin every year, simply to return its cost of capital; a $985 million drug must clear about $164 million. Stretch the effective life to fourteen years and the hurdles fall to roughly $357 million and $137 million. Compress it toward the shaded region on the left — a long clinical programme that eats years off the patent term before launch, or an early cliff from a successful patent challenge — and the required margin climbs steeply, because there are fewer years over which to amortise the same fixed cost.

This is not an abstraction. The patent term is twenty years from filing, but filing happens early, before the decade-plus of trials; what remains at launch is the effective life, and it is the only thing that pays. Budish, Roin and Williams (2015) showed in the American Economic Review that this design has a perverse consequence the cliff model predicts. Because the patent clock starts before clinical development and runs at a fixed length regardless of how long that development takes, drugs that take longer to bring to market — preventives and treatments for early-stage disease, whose trials must run for years to show a mortality benefit — enjoy a shorter effective life and a correspondingly higher break-even hurdle. The result is a documented distortion in the direction of research: over a recent five-year window, eight new drugs were approved to treat advanced lung cancer, every one of them for the most late-stage patients whose trials read out fastest, while not a single drug has ever been approved to prevent lung cancer, and only six have ever been approved to prevent any cancer at all. The market is not failing to invest because prevention is scientifically impossible. It is failing to invest because the fixed-term patent makes the long-horizon project uneconomic. Note what this establishes: patents matter so much in this industry that the precise shape of the patent rule bends the trajectory of medical research. That is the strongest possible evidence that the appropriability constraint binds.

So for pharmaceuticals the verdict is unambiguous, and it cuts against the abolitionist. The fixed cost is real, enormous, and audited (if disputed); the marginal cost of imitation is trivial; the post-expiry price collapse to near marginal cost is measured; and the historical record contains no example of a research-intensive pharmaceutical industry sustaining itself on free-ridden molecules. Here the blue rectangle of Figure 1 is load-bearing. Knock it out and the building falls.

Two pieces of causal evidence close off the obvious escape — the hope that the break-even arithmetic is a paper constraint firms would simply absorb. They do not absorb it; entry tracks the expected reward on both sides of the equation. On the demand side, Acemoglu and Linn (2004) used exogenous variation in market size driven by US demographic shifts and found that a one percent increase in the potential market for a drug category raised the number of new drugs entering that category by about six percent — roughly four percent for new non-generic drugs and four to six percent for new molecular entities, the most research-intensive class. Bigger expected markets pull more medicines into existence. On the supply side, Gaessler and Wagner (2022) exploited a natural experiment — drugs whose underlying patents were invalidated in European Patent Office opposition proceedings, an exogenous shock to expected exclusivity — and found that the loss of a single year of market exclusivity lowers the likelihood that a drug development project reaches approval by about 4.9 percentage points, against an unconditional approval rate of 30.8 percent; firms overwhelmingly abandon the project the moment the patent falls. The first result says reward draws entry; the second says shortened exclusivity drives exit. Both are exactly the sign the break-even model predicts, measured rather than assumed.

4. Creative works: where the gap is real but smaller, and shrinking

4.1 The cost of expression, and its discontents

Now the harder case, and the one Kinsella’s argument fits more comfortably. Landes and Posner (1989) — whose model I have read in the original and which remains the most careful economic statement of the pro-copyright position — apply exactly the Figure 1 logic to books. The author and publisher incur a cost of expression (the author’s time, editing, typesetting) that is independent of the print run. The marginal cost of a copy is the cost of printing, binding and distribution. Without copyright, a reprinter buys one copy and undersells the originator down to marginal cost; foreseeing this, the originator does not write. The structure is identical to the drug case.

But here is the crucial difference, and Landes and Posner are scrupulous about it in a way that maximalist copyright advocates who invoke them are not. They devote pages to the qualifications that limit the need for copyright, and the qualifications are substantial:

• Inferior copies. When a copy is a poor substitute — a blurry photostat, a cheaply bound reprint — the originator can price above marginal cost without legal protection.

• The head start. Copying takes time. There is an interval during which the originator faces no competition, and for faddish works whose demand is front-loaded, that interval alone may suffice.

• Indirect appropriability. The originator can sometimes capture the value of copies by charging more for the original that enables them — the mechanism by which a publisher of academic journals prices in the photocopying it cannot prevent.

• Non-pecuniary and indirect returns. Many authors are paid for writing in coin other than royalties: prestige, a professorship, consulting income, the platform a book confers. Where these dominate, the royalty stream — and hence the copyright that protects it — is not what motivates creation.

This sceptical line is not new, and it is worth registering its pedigree because it shows the appropriability-without-copyright argument is older than the platforms now accused of self-interest in making it. Arnold Plant (1934), in the first sustained economic analysis of book copyright, pressed it with historical evidence: through the nineteenth century anyone in the United States was free to reprint a foreign work, yet American publishers paid English authors handsomely for “advance sheets,” and authors sometimes earned more from their uncopyrighted American sales than from their protected British royalties — priority in the market, not legal monopoly, secured the return. Plant’s verdict was that copyright raises the variety of titles published while reducing access to them — more books are written, but fewer copies of the ones readers want reach them — and that, beyond a point, it bends authors toward writing whatever maximises monopoly profit rather than what they would otherwise write. That last point anticipates, in the creative domain, the direction-distortion this essay has already met in pharmaceuticals.

And then they make the observation that detonates the maximalist position from within. Beyond some level, more copyright protection is counterproductive, because creating a new work requires borrowing from old ones — characters, situations, chord progressions, prior results — and the stronger and longer the protection on existing works, the higher the cost of expression for every future creator who must license, search around, or substitute. Every author is simultaneously a borrower from the past and a creator for the future; the borrower wants weak protection on others’ work, the creator wants strong protection on her own, and the social optimum balances the two. This is not a critique imported by opponents of copyright. It is built into the foundational economic defence of copyright. It means the policy question was never “copyright: yes or no.” It was always “copyright: how much,” and the answer is an interior optimum, not a maximum.

4.2 The causal evidence: basic copyright works; extensions do not

Theory can establish that an optimum exists without telling you where it sits. For that you need an experiment, and creative-industry experiments are rare because copyright is usually changed by lobbying, not by chance. Giorcelli and Moser (2020), in the Journal of Political Economy, found a clean one. Napoleon’s military campaigns brought French copyright law — France’s copyright statute of 1793, granting composers exclusive rights tied to their own lifetimes — to two Italian states, Lombardy and Venetia, in 1801, while the remaining Italian states, sharing the same language and culture and exposed to the same French influence, did not adopt copyright. Comparing the production of new operas across eight Italian states using newly assembled data on 2,598 operas created between 1770 and 1900 in a difference-in-differences design, they found three results, and all three matter.

First, the adoption of basic copyright caused a significant increase in the number of new operas — their estimates imply a roughly five-fold increase in the number of historically popular operas in the states that gained protection. Second, it raised their quality, measured by how popular and how durable they proved — copyright induced not just more work but better work, because it let composers profit from a piece’s continued performance rather than a single commissioning fee. Third — and this is the result the maximalists never quote — extensions of copyright beyond the composer’s life produced no measurable benefit at all. The incentive to create responds to basic protection; it does not respond to the prospect of one’s heirs collecting royalties decades after one’s death.

Read those three findings against Figure 1 and Landes–Posner and they lock together perfectly. Basic copyright moved output up the rising part of the welfare curve: the blue rectangle was load-bearing for opera as it is for drugs. But term extension is the flat-to-falling part: it adds nothing to the dynamic incentive (a composer’s behaviour in 1805 is not altered by his estate’s revenues in 1860) while it does add to the static cost and the cost of expression for later borrowers. Figure 4 is the unifying picture.

The dynamic gain from inducing new works rises with protection but saturates — past a point, more protection induces no further creation, exactly as Giorcelli and Moser found for extensions. The static and access costs rise without saturating. Net welfare therefore peaks at an interior T* and declines thereafter. The question “is intellectual property good?” is, on this picture, malformed. Protection below T* is too little; protection above T* is too much; and the entire substance of intellectual-property policy is locating T* and asking which side of it we are on. The causal evidence says basic copyright and basic patents sit to the left of T* (they bind, they induce real creation), while life-plus-seventy copyright terms, evergreened drug patents, and dense patent thickets sit to the right (they impose cost without buying incentive). Both the naive abolitionist and the reflexive maximalist are answering the wrong question.

4.3 The Amazon problem: when the marginal cost of a copy goes to zero

The reader asked, specifically, what happens when a distributor like Amazon — or anyone with a server — can copy and distribute a work at will. The model answers precisely. Everything in Figure 1 turns on the marginal cost of a copy, c. For a physical book, c is positive but modest. For a digital file, c is approximately zero. As c → 0, the competitive price goes to zero, the blue appropriability rectangle collapses, and — on the pure theory — the work cannot recover its cost of expression at all. Digital reproduction does not weaken the appropriability problem; it perfects it. This is why Landes and Posner, writing in 1989, already saw the trajectory: improving copying technology, they noted, steadily raises the need for copyright over time, because it erodes every one of the practical frictions (inferior copies, the head start) that once substituted for legal protection. By their logic, the digital era should have been an extinction event for creative production.

It was not. And the reason it was not is the single most important empirical finding in this whole literature, because it complicates the tidy theory in a way the theory did not anticipate. Joel Waldfogel (2017, in the Journal of Economic Perspectives, and at book length in 2018) documented what actually happened when digitisation — Napster and its successors — devastated the appropriability of recorded music. United States recorded-music revenue began a precipitous slide in 1999 and, on the series Waldfogel uses — the RIAA’s reported value of US music shipments, inflation-adjusted to constant 2016 dollars — fell to roughly 25 percent below its 1999 peak by 2005 and, in real terms, by more than half by 2012, with international sales off by a similar fraction (Waldfogel, 2018). By the orthodox model, that revenue collapse should have produced a collapse in the quantity and quality of new music. It did not. On balance, Waldfogel finds, digitisation increased the number of new products created and made available to consumers; and because product quality is unpredictable ex ante, more releases mean more high-quality draws, so the quality of the best new music — measured by critics’ best-of lists and by what listeners actually consume — held up or improved rather than declining. The shift shows up structurally too: the share of top-selling albums released by independent labels grew from roughly 12% to 35% between 2000 and 2010. The same pattern, he found, held for film, books and television.

Why? Because the very technology that destroyed appropriability also destroyed the cost of expression. The fixed cost F that copyright exists to protect was itself a creature of the predigital production and distribution system — the studio, the pressing plant, the distributor, the shelf space, the gatekeeper. Digitisation let a musician record at home, a novelist self-publish, a filmmaker distribute globally without any of that apparatus. F fell at the same time as c fell. And if F falls far enough, the appropriability problem shrinks even as appropriability itself evaporates, because there is simply less fixed cost that needs recovering. Waldfogel’s “gatekeepers” — the studios and labels whose high fixed costs the old copyright bargain subsidised — turned out to have been, in significant part, a cost the system imposed rather than a value it created.

This is the right place to guard against over-reading that result, because the aggregate survival of creative output is not the same as copying being costless to the individual creator, and the honest evidence cuts both ways. Reimers (2016), in the Journal of Law and Economics, isolated the substitution effect directly: using a difference-in-differences design on the e-book and physical sales of RosettaBooks titles between 2010 and 2013, she found that private anti-piracy enforcement raised legal e-book sales — the closest substitute for an unauthorised download — by more than fourteen percent, with most of the gain coming from deterring casual rather than professional infringement. So copying does displace paid sales where the copy is a genuine substitute; the displacement is measurable and not small. What the music evidence shows is not that θ, the substitution rate, is zero, but that the fall in F outran it in the aggregate. Both facts are true at once, and a serious account holds them together rather than choosing the convenient one: protection still does real work for the marginal creator whose costs are not trivial, even as digitisation lowered the costs that made protection necessary in the first place.

This finding must be handled carefully, and not oversold, because it has been weaponised. The result is genuinely inconvenient for the maximalist case: it shows that creative output is far more robust to the loss of appropriability than the simple model predicts, at least where production costs have fallen in step. But Waldfogel’s argument has also been seized upon by parties — large technology platforms among them — with a direct commercial interest in weak copyright, and it has fair critics. It measures aggregate quantity and average quality over a window of years; it does not measure the long-run effect on the professional creative class of removing the income floor, nor what happens to expensive, slow, fixed-cost-heavy forms (the big-budget film, the multi-year investigative book) for which the production cost did not fall. The honest reading is narrow and it is this: the digital era falsifies the claim that strong copyright is necessary for a flourishing creative output in the forms where production costs collapsed alongside reproduction costs. It does not establish that copyright is unnecessary everywhere, and it says nothing about pharmaceuticals, where the fixed cost has done the opposite of fall.

4.4 The international and out-of-contract problem, and why it sinks the contractual alternative

The reader also asked about international distribution, online works, and the case where a copier is “outside of contract.” This is not a footnote; it is the precise point on which the most sophisticated version of the abolitionist position breaks.

Intellectual-property rights are territorial. A copyright or patent is granted by a state and enforceable within its jurisdiction; a work freely copied in a country that does not recognise or enforce the right escapes the creator’s protection entirely. Online distribution makes every work instantly transnational, so the weakest enforcement regime anywhere on earth tends to set the effective ceiling on appropriability everywhere. This is a real and serious erosion, and it is one reason the empirical appropriability of digital works is so much lower than the law on the books would suggest.

Now, the standard libertarian rejoinder — Kinsella’s, and the broader Rothbardian tradition’s — is that creators do not need state-granted monopolies at all, because they can protect their works by contract. Sell the book or the film under a licence that forbids copying; enforce the licence against violators; no artificial scarcity, no state IP, just ordinary freedom of contract. The proposal is coherent in a world of few buyers and durable relationships. It collapses in a world of mass digital distribution, for a reason that is purely economic and that Landes and Posner identified in 1989: contracts bind only the parties to them. This is the doctrine of privity, and it is not a technicality. If I sell you a file under a no-copying licence, that licence binds you. It does not bind the stranger who later downloads the file from a server, because the stranger never agreed to anything. There is no privity between the creator and the marginal copier. Contract reaches the second party; it cannot reach the third. And in mass distribution the third parties are the entire problem — they are millions, anonymous, and contractually untouched.

This is why only a property-style right — a right good against the world, which is precisely what intellectual property is and what a contract is not — can reach the unconsenting copier. The contractual alternative does not avoid the need for IP; it quietly presupposes a world small enough that everyone who touches the work is in privity with the creator, which is exactly the predigital world whose frictions made copyright less necessary in the first place. Remove those frictions — go online, go global, go to zero marginal cost — and the contractual solution fails in the same breath and for the same reason that the appropriability problem becomes acute. The abolitionist cannot have it both ways: the conditions under which contracts suffice are the conditions under which copyright was least needed, and the conditions under which copyright is most needed are the conditions under which contracts fail.

5. Answering Kinsella on the merits

We can now state, precisely and without either deference or dismissal, what is right and what is wrong in the case for abolition — because the reader asked for the economics of it, not the rhetoric.

What is right. The non-rivalry premise is correct, and it is not a minor concession. Kinsella is right that ideas are not scarce in the physical sense, right that intellectual property manufactures an excludability that does not occur in nature, and right that this manufacture imposes a real cost — the red triangle of Figure 1, which no honest economist denies. He is also right, directionally, that much of the intellectual-property edifice we actually have is unjustified: the causal evidence (Giorcelli and Moser on copyright extensions; Budish, Roin and Williams on the distortions of fixed patent terms) places a great deal of existing protection to the right of T*, where it imposes cost without buying incentive. Boldrin and Levine (2008) marshalled the broader case that patent systems frequently retard rather than promote innovation in industries characterised by cumulative, follow-on invention — they go further than this essay will, calling intellectual monopoly “an unnecessary evil,” a consequence of innovation rather than its cause — and on those industries the case has force. Moser (2005) bears on the same question from the opposite end of history. Using roughly fifteen thousand innovations exhibited at the world’s fairs of 1851 and 1876, she shows that patent laws shaped the direction of innovation rather than merely its level: inventors in countries without patents clustered in industries where secrecy did the work patents would have done — scientific instruments, food processing, dye stuffs — while patent-granting countries spread across the spectrum, and when the Netherlands abolished its patent system in 1869 the share of Dutch innovation in food processing rose from 11 to 37 percent. That is the same lesson as Budish, Roin and Williams, read backwards: patents do not simply turn invention up or down; they bend what gets invented. There is a second strand the abolitionist gets right, and it is the strongest empirical case against over-broad rights: when the protected object is itself an input into later work, protection can measurably suppress the follow-on. Williams (2013), in the Journal of Political Economy, used the natural experiment of the human genome — genes sequenced first by the private firm Celera and held under its contractual IP, against genes sequenced by the public effort and placed immediately in the public domain — and estimated that Celera’s gene-level IP reduced subsequent scientific research and product development on those genes on the order of 20 to 30 percent, an effect that persisted even after the IP lapsed. (Williams is scrupulous that she does not measure net welfare: Celera’s entry was itself spurred by the prospect of IP and may have sped sequencing, so the follow-on loss is one side of a ledger, not the whole of it — a caveat the honest reader keeps.) The copyright analogue is just as clean. Biasi and Moser (2021), in American Economic Journal: Microeconomics, studied the wartime Book Republication Program, under which the United States licensed cheap reprints of enemy-owned German science books: the average title’s price fell about 25 percent, and they find that a ten percent decline in price raised follow-on citations by English-language authors by roughly 45 percent, with treated books drawing on the order of 80 percent more citing works than comparable controls. Weaker protection of an input — patent or copyright — raised the cumulative science built on it. An abolitionist who claimed only that current intellectual property is far too strong would be defending a position the mainstream evidence substantially supports.

What is wrong. The universal claim — that intellectual property never generates net social benefit, that creation would proceed undiminished without it — is falsified, and not by theory but by measurement. The premise that non-rivalry implies no warrant for exclusion does not survive contact with the fixed-cost problem, because the orthodox argument was never that ideas are scarce; it was that the incentive to produce them is, and that incentive is scarce precisely because ideas are non-rivalrous and therefore unappropriable in a competitive market. Kinsella’s argument treats the non-rivalry of the finished idea as decisive, when the economically relevant scarcity is in the costly, uncertain, failure-ridden process that produces it. The pharmaceutical evidence shows that where that process is expensive enough and imitation cheap enough, the absence of exclusion does not merely reduce output at the margin — it can eliminate the category, as the eleven-percent success rate and the billion-dollar fixed cost and the eighty-percent price cliff together demonstrate. And it is no accident that pharmaceuticals are the sector where the constraint bites hardest: Cohen, Nelson and Walsh (2000), surveying 1,478 US manufacturing R&D labs, found that firms in most industries rank patents the least important of their appropriation mechanisms, relying instead on secrecy and lead time — but that patents are decisive in a small set of industries, “most notably pharmaceuticals,” precisely because a regulator-disclosed, chemically reverse-engineerable molecule cannot be protected by secrecy or a head start at all. This is the empirical hinge of the whole essay: appropriability mechanisms are sector-specific, so the case for patents is strong exactly where the alternatives fail and weak where they do not. The historical-copyright evidence shows that basic protection caused measurably more and better creative work. You may argue about how strong protection should be; you may not argue, against this evidence, that it is never productive.

The deepest error in the abolitionist argument is methodological, and it is worth naming because it is the error this essay has tried throughout to avoid. It derives a universal empirical conclusion — IP never helps — from an a-priori premise about the nature of ideas. But whether a particular institution, in a particular industry, produces net benefit is not the kind of question an axiom can answer. It is the kind of question only data can answer, and the data answer it differently in different industries. That is the finding. There is no single correct verdict on intellectual property, because intellectual property is not one thing doing one job. It is a deadweight loss imposed on opera that paid for itself, an extension imposed on opera that did not, a twenty-year monopoly on a molecule without which the molecule would not exist, and an evergreened thicket on a molecule that distorts research away from the patients who need it most — all at once.

6. What the models actually license

Strip away the rhetoric on both sides and the four pictures in this essay license a small number of conclusions, stated flatly.

The static cost of intellectual property is real and is correctly identified by its critics (Figure 1). It is justified only where the dynamic benefit exceeds it, which is an empirical condition, not a presumption.

In pharmaceuticals the condition is met decisively. A fixed cost in the high hundreds of millions to low billions (DiMasi et al., 2016; Wouters et al., 2020), a marginal imitation cost near zero, and a measured post-expiry price collapse of 32% in a year and 82% in eight (Serra-Burriel et al., 2024) together make exclusivity the precondition of the asset’s existence (Figures 2 and 3). The live problem here is not too much protection in the aggregate but its design: the fixed patent term bends research toward fast-reading late-stage treatments and away from slow-reading prevention (Budish, Roin and Williams, 2015), a distortion that better-designed protection — endpoint reform, term structured around effective rather than nominal life, targeted subsidy for long-horizon work — would relieve. Stronger protection is not the answer; smarter protection is.

In creative works the condition is met for basic protection and fails for extended protection. Copyright at its founding level raised both the quantity and quality of output (Giorcelli and Moser, 2020); copyright beyond the creator’s life did not (the same study); and the welfare curve’s interior peak (Figure 4) tells you that the long terms and broad scope of contemporary copyright sit on its falling side. Digitisation, by collapsing the cost of expression at the same time as the cost of copying, has shrunk the appropriability problem in the forms where production costs fell, without abolishing it and without touching the forms where they did not (Waldfogel, 2017, 2018). The contractual alternative that abolitionists offer in copyright’s place fails on privity exactly where mass distribution makes it most needed.

And the unifying conclusion, the one the reader asked for and the one the evidence actually supports rather than the one any prior would prefer: the right quantity of intellectual property is neither zero nor unbounded. It is an interior optimum that differs by industry, and the principal pathology of the actual system is not that it exists but that, in most of the domains where we can measure, it has wandered to the right of its own peak — long after the marginal work it was meant to call into being has already been written, and long past the point where the molecule it was meant to fund has already been found.

---

References

Acemoglu, D., & Linn, J. (2004). Market size in innovation: Theory and evidence from the pharmaceutical industry. Quarterly Journal of Economics, 119(3), 1049–1090. https://doi.org/10.1162/0033553041502144

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

Biasi, B., & Moser, P. (2021). Effects of copyrights on science: Evidence from the WWII Book Republication Program. American Economic Journal: Microeconomics, 13(4), 218–260. https://doi.org/10.1257/mic.20190113

Boldrin, M., & Levine, D. K. (2008). Against Intellectual Monopoly. Cambridge University Press.

Budish, E., Roin, B. N., & Williams, H. (2015). Do firms underinvest in long-term research? Evidence from cancer clinical trials. American Economic Review, 105(7), 2044–2085. https://doi.org/10.1257/aer.20131176

Cohen, W. M., Nelson, R. R., & Walsh, J. P. (2000). Protecting their intellectual assets: Appropriability conditions and why U.S. manufacturing firms patent (or not) (Working Paper No. 7552). National Bureau of Economic Research. https://doi.org/10.3386/w7552

DiMasi, J. A., Grabowski, H. G., & Hansen, R. W. (2016). Innovation in the pharmaceutical industry: New estimates of R&D costs. Journal of Health Economics, 47, 20–33. https://doi.org/10.1016/j.jhealeco.2016.01.012

Gaessler, F., & Wagner, S. (2022). Patents, data exclusivity, and the development of new drugs. Review of Economics and Statistics, 104(3), 571–586. https://doi.org/10.1162/rest_a_00987

Giorcelli, M., & Moser, P. (2020). Copyrights and creativity: Evidence from Italian opera in the Napoleonic age. Journal of Political Economy, 128(11), 4163–4210. https://doi.org/10.1086/710534

Kinsella, N. S. (2008). Against Intellectual Property. Ludwig von Mises Institute. (Originally published as Kinsella, N. S. (2001), Against intellectual property, Journal of Libertarian Studies, 15(2), 1–53.)

Landes, W. M., & Posner, R. A. (1989). An economic analysis of copyright law. The Journal of Legal Studies, 18(2), 325–363. https://doi.org/10.1086/468150

Moser, P. (2005). How do patent laws influence innovation? Evidence from nineteenth-century world’s fairs. American Economic Review, 95(4), 1214–1236. https://doi.org/10.1257/0002828054825501

Plant, A. (1934). The economic aspects of copyright in books. Economica, 1(2), 167–195. https://doi.org/10.2307/2548748

Recording Industry Association of America. (2024). U.S. recorded music revenue database (1973–2024). RIAA. https://www.riaa.com/u-s-sales-database/

Reimers, I. (2016). Can private copyright protection be effective? Evidence from book publishing. The Journal of Law and Economics, 59(2), 411–440. https://doi.org/10.1086/687521

Serra-Burriel, M., Martin-Bassols, N., Perényi, G., & Vokinger, K. N. (2024). Drug prices after patent expirations in high-income countries and implications for cost-effectiveness analyses. JAMA Health Forum, 5(8), e242530. https://doi.org/10.1001/jamahealthforum.2024.2530

Waldfogel, J. (2017). How digitization has created a golden age of music, movies, books, and television. Journal of Economic Perspectives, 31(3), 195–214. https://doi.org/10.1257/jep.31.3.195

Waldfogel, J. (2018). Digital Renaissance: What Data and Economics Tell Us about the Future of Popular Culture. Princeton University Press.

Williams, H. L. (2013). Intellectual property rights and innovation: Evidence from the human genome. Journal of Political Economy, 121(1), 1–27. https://doi.org/10.1086/669706

Wouters, O. J., McKee, M., & Luyten, J. (2020). Estimated research and development investment needed to bring a new medicine to market, 2009–2018. JAMA, 323(9), 844–853. https://doi.org/10.1001/jama.2020.1166