For most of the twentieth century, economic theory ran on an assumption it rarely stated: that everyone in a market knows everything that matters. This chapter is about what happened when economists dropped it. The arc runs from the toolkit that made the move possible (Nash equilibrium), through the discovery that private information can break a market outright (Akerlof, Spence, Rothschild-Stiglitz), to the inversion that turned the discipline from describing markets to engineering them (mechanism design, then market design), with the parallel turn in how economists read data (the time-series econometrics rung) as the chapter's contained methodological section.
| Posture | The question it asks | Apparatus | Section |
|---|---|---|---|
| Assume | What is the efficient allocation when everyone knows everything? | Perfect-information general equilibrium (Arrow-Debreu) | §11.1 |
| Model | What does private information do to a market? | Asymmetric information: lemons, signaling, screening | §11.3 |
| Design | What rules produce the outcome we want, given hidden information? | Mechanism design: incentive compatibility, revelation principle, VCG | §11.4 |
| Engineer | Can we build working institutions on the design? | Market design: deferred acceptance, spectrum auctions | §11.5 |
Every theorem of postwar general-equilibrium economics rested on a sentence no one said out loud: everyone in the market already knows everything that matters. Prices, qualities, the preferences and endowments of others — all of it common knowledge, available to every agent at no cost. On that footing the Arrow-Debreu model (the completion of the general-equilibrium program, treated in the postwar-synthesis chapter) proved that competitive markets have an equilibrium and that the equilibrium is Pareto-efficient. The result was the high-water mark of the discipline's self-confidence. It was also built on a simplification that, once examined, turned out to carry more weight than any other in the model.
The perfect-information benchmark is that simplification: the working assumption, load-bearing throughout postwar general-equilibrium theory, that all agents know all relevant features of the market — prices, qualities, others' preferences. Dropping it was the most consequential micro-theoretic move of the 1970s. Not because the benchmark was stupid — it was a deliberate idealization, the frictionless plane of economic mechanics — but because once you ask what happens when one side of a transaction knows something the other does not, almost every comfortable result becomes a special case, and a new set of questions opens that the old framework had no language for.
The first person to say the assumption out loud, and to point at exactly where it broke, was the man who had done the most to perfect the benchmark. Kenneth Arrow's 1963 paper "Uncertainty and the Welfare Economics of Medical Care" is the seed of everything this chapter covers. Arrow had co-authored the existence proof for competitive general equilibrium; here he turned his attention to a market — medical care — that violates almost every assumption that proof requires. The patient does not know what treatment is needed; the doctor does and is also the seller; insurance markets sort the healthy from the sick in ways no auctioneer arranged. Arrow's recognition was the set piece: the field's most rigorous benchmark-builder named the benchmark's limit and, without proposing the apparatus that would later fill it, pointed directly at where the next two decades of work would go.
What Arrow had identified in one market was, in fact, the formal descendant of an argument the Austrians had made decades earlier in prose. Friedrich Hayek's knowledge problem — the claim that the economically relevant knowledge of a society is dispersed across countless individuals, held tacitly and locally, and never available to any central mind — is the prose ancestor of the entire information-economics program. The full Austrian argument and the calculation debate it belonged to are the subject of the Austrian chapter; what matters here is the lineage. The Austrians saw, before anyone had the mathematics, that information was the hinge on which the case for and against markets actually turned. The formalists were about to mathematize the kernel of that insight. The chapter returns to Hayek at its end, when market design has to answer the question of whether the dispersed knowledge can be elicited at all.
One framing should be fixed before the apparatus arrives, because the whole chapter is a qualification of it. The benchmark says markets coordinate on common knowledge: agents read prices that already encode everything, and trade to an efficient allocation. The chapter's verdict, stated now so the reader can watch it being earned, is that the information revolution bounded this benchmark rather than refuting it. Arrow-Debreu did not fall. It remained the welfare reference — the standard against which information-impaired markets are measured. What changed is that the action moved to the markets where information is private, costly, and strategically withheld, and the discipline built, over forty years, the tools to model and then to repair them.
Before the information revolution could happen, the discipline needed a way to reason about what self-interested people do when their fates depend on each other's choices. Von Neumann and Morgenstern had given economics a theory of games in 1944 — the founding treated in the marginalist chapter, where expected-utility theory is also born. It had one limitation that mattered enormously: its central solutions worked for cooperative games, where players could form coalitions and sign binding agreements. For the economy as economists actually wanted to model it — strangers transacting without enforceable side-contracts — the 1944 apparatus had no general prediction. Nash removed the contracts.
A Nash equilibrium is a strategy profile in which no player can do better by unilaterally changing strategy, given what the others are doing. Each player is best-responding to everyone else; no one has a private reason to deviate. Stated that plainly the idea sounds almost too simple to be a revolution, and Nash's 1950-51 work did not prove anything an economist would call surprising about any particular market. What it did was supply a solution concept for non-cooperative games — a precise sense of what "the players will end up here" means when no one can commit in advance and everyone is rational. The existence theorem that guarantees such a point exists under broad conditions is formal apparatus that lives in economics ch. 6; this chapter needs the concept, not the fixed-point proof.
The decisive thing about Nash equilibrium for the rest of this chapter is that it is a toolkit, not a destination. Every result that follows — the lemons market, the signaling equilibrium, the second-price auction, the entire mechanism-design program — is a game with an information structure, and the prediction for how it resolves is a Nash equilibrium (in one refinement or another). The information revolution is not a sequence of separate cleverness; it is the systematic application of one solution concept to games where the players know different things. Once you have a way to solve any game, you can start asking what games private information produces.
Two refinements made the toolkit fit for that purpose. The first, subgame perfection, ruled out equilibria that rest on threats no rational player would carry out — it disciplines what counts as a credible strategy when play unfolds over time. The second is the one this chapter leans on directly. A Bayesian Nash equilibrium is Nash equilibrium for games of incomplete information: players do not know the others' private characteristics (their "types") but hold beliefs — probability distributions — over them, and each best-responds in expectation given those beliefs. This is the technical bridge from Nash to information economics. A buyer who does not know whether a car is good, a firm that does not know whether a worker is able, an auctioneer who does not know what bidders are willing to pay: each is a player in a game of incomplete information, and each market the next sections study is a Bayesian game with an equilibrium to be found. Game-theoretic industrial organization — the Tirole-era rebuild of the theory of the firm and of imperfect competition — is the broad application surface of exactly this toolkit, and lives in economics ch. 6 and ch. 12.
You have just met Nash equilibrium and the game-theoretic toolkit. Much of it was born in a Cold War think tank.
Non-cooperative game theory, and the mechanism-design program that grew from it, were shaped by the institutions of the early Cold War — RAND, the strategic-analysis culture, the funding that flowed to formal methods. Nash's equilibrium and the auction and mechanism theory that followed were not only mathematics; they were a particular era's idea of how rational adversaries and self-interested agents could be modeled and managed. The walkthrough traces how that institutional context shaped the methods this chapter treats as pure apparatus.
In 1970, a young economist explained why you cannot reliably sell a good used car. The argument took a few pages and changed economics. George Akerlof's "The Market for Lemons" took the perfect-information benchmark and removed a single assumption — that buyers can see quality — and showed that the market could unravel to nothing. This is the information-economics revolution proper, and across three papers in six years it established a claim the benchmark had no room for: information is not a parameter you set to "perfect." It is the strategic object the market is organized around.
Start with the term the whole section turns on. Asymmetric information is a situation in which one side of a transaction knows something the other does not: the seller knows the car's quality, the worker knows their own ability, the insured knows their own risk. The benchmark assumed this away; Akerlof asked what it does. His answer runs as a spiral. Suppose buyers cannot tell good cars from bad ones ("lemons"). A rational buyer will pay only the average quality of cars on offer, since that is the expected value of a random purchase. But at a price set by the average, the owners of the best cars — whose cars are worth more than the average price — will not sell; they withdraw. With the best cars gone, the average quality of what remains falls, so the price buyers will pay falls, so the next tier of good cars withdraws, and so on. The process can run all the way down until only the worst cars remain, or the market vanishes entirely.
This is adverse selection: the market failure that arises before a transaction, when the informed side's private information correlates with quality, so the offered terms disproportionately attract the bad types and degrade the pool. The intellectual shock of the lemons result is not its algebra (the formal model is in economics ch. 4 §4.6, on information asymmetry); it is that a market can fail completely, not partially, from information alone. No tax, no monopoly, no externality — just the fact that one side knows more. And the result generalizes far past used cars. Insurance: the people most eager to buy are the ones who know they are high-risk, so premiums rise, the low-risk drop out, premiums rise again. Credit: the borrowers most willing to pay a high interest rate are the ones least likely to repay. Labor: at a wage set for average ability, the most able workers find better options and leave. Adverse selection is everywhere one side knows its own quality and the other pays an average.
Adverse selection has a sibling worth naming for completeness, because the two are constantly confused. Moral hazard is the market failure that arises after a transaction, when one side can take hidden actions the other cannot observe: the insured drives recklessly once covered, the borrowed-against firm takes risks it would not take with its own money. Adverse selection is about hidden types before the deal; moral hazard is about hidden actions after it. Both are problems of private information, and together they are the two canonical information frictions the field organized itself around. This chapter's spine is adverse selection, but a reader who keeps the two apart will read the rest of the literature correctly.
Drive Akerlof's used-car market. Set the share of good cars and the quality gap between good and bad, then watch the spiral run: buyers pay the average quality, the best cars withdraw, the average falls, more withdraw. The point the prose claims — that asymmetric information can destroy a market entirely, not just distort it — only lands when you cross the boundary yourself. Push the quality gap up and watch survival flip to collapse.
Figure 11.2 (interactive). The lemons spiral. Buyers pay the surviving pool's average value; sellers above the price withdraw; the average re-prices each round. Drag the quality gap up to cross from a surviving thin market into total collapse.
Buyers can only pay for what they can expect — the average. When the gap between good and bad is small, the average is close enough to good that good-car owners still sell, and the market holds. When the gap is large, the average is far below what a good car is worth, so good cars leave, dragging the average toward the lemons — and once that spiral starts, nothing stops it short of collapse.
Buyers pay the expected quality of cars on offer. Good-car owners with reservation value $v_g$ stay only while the offered price $p \ge v_g$. Each round, $p$ is set to the average value of the surviving pool; sellers with $v > p$ withdraw; the pool re-prices.
Collapse occurs when, at every price, the average value of cars willing to sell at that price lies strictly below the marginal seller's reservation value: $\mathbb{E}[v \mid v \le p] < p$ for all $p$. A large quality gap $\Delta$ pushes the economy into that regime; a small $\Delta$ leaves a thin separating price where good cars still trade.
If asymmetric information can destroy a market, can the informed side do anything to save it? Michael Spence's 1973 "Job Market Signaling" gave the affirmative answer, and gave it in a form that is still uncomfortable to state plainly. Signaling is a costly action taken by the informed side to credibly reveal its private information — where the cost of the action differs across types, so that only some types find it worth taking. Spence's case was education as a signal of ability. A worker who knows they are able can acquire a credential; an employer who cannot observe ability can observe the credential. The signal works — here is the sharp part — because it is differentially costly, not because it produces anything. Education is cheaper (in effort, time, tolerance for tedium) for able workers than for less able ones. That cost gap is what lets the credential separate the types.
The result at its sharpest: a degree can be a rational, equilibrium-sustaining signal of pre-existing ability even if it adds nothing whatever to a worker's productivity. This is not the claim that education is worthless — it plainly teaches things, and the model does not deny it. It is the more disturbing claim that the labor market would reward the credential even if it taught nothing, purely as a sorting device, because the differential cost makes acquiring it informative about a trait the employer wants and cannot see. That is why the signaling-versus-human-capital distinction is real and empirically hard: when wages rise with schooling, the data alone cannot tell you how much of the return is the skill the schooling built (human capital) and how much is the signal it sent (sorting). The two stories predict the same wage premium and differ only in what the schooling did.
Drive Spence's signaling model. Move the cost-of-signal gap between able and less-able workers, and the productivity gap between them — including all the way to zero. Watch a separating equilibrium form (the signal sorts the types) or collapse into pooling (the signal stops working). The load-bearing experiment: set productivity to zero while keeping the cost gap, and watch the equilibrium still separate — the market rewarding a credential that produces nothing.
Figure 11.3 (interactive). Net payoff by type as the signal level varies. A separating equilibrium exists when the able type wants the credential and the less-able type does not. Set the productivity gap to zero, keep the cost gap, and the equilibrium still separates.
A signal sorts types only when it is differentially costly — cheaper for the able than the less-able. If the cost gap shrinks too far, the less-able worker can afford to mimic the able one, the signal stops distinguishing them, and the market pools. Crucially, the productivity gap does no work in this: set it to zero, keep the cost gap, and the credential still separates the types — it sorts purely as a signal, rewarding a piece of paper that builds no skill at all.
Signaling has a mirror image. Where signaling is the informed side acting to reveal itself, screening is the uninformed side designing a menu of options so that the informed side self-sorts by choosing among them — an insurer offering a low-deductible, high-premium contract alongside a high-deductible, low-premium one, knowing that high-risk customers will pick the first and low-risk the second. Michael Rothschild and Joseph Stiglitz's 1976 analysis of exactly this competitive-insurance setting produced the deepest of the three results. They asked whether a stable equilibrium of such contracts exists, and found that under asymmetric information a separating equilibrium — one in which the different types choose different contracts, so the insurer can tell them apart — may fail to exist at all. The alternative, a pooling equilibrium in which everyone takes the same contract, can always be undercut by a competitor cherry-picking the low-risk customers; but the separating contracts may be undercut by a pooling offer in turn. The market can have no equilibrium.
That last result is worth dwelling on, because it reaches back to the benchmark of §11.1. The postwar achievement was an existence proof: competitive markets have an equilibrium. Rothschild-Stiglitz showed that the existence guarantee depends on the assumption the chapter is about. Drop common knowledge of types, and a perfectly competitive market — many firms, free entry, price-taking — can fail to have any stable configuration at all. Asymmetric information does not merely make markets inefficient; it can make the equilibrium concept itself come apart. The three papers together — lemons, signal, screen — established the unifying move of the field: information is the strategic object, and the question is no longer "what is the efficient allocation" but "what can the players learn about each other, and what will they do given what they can learn."
One boundary has to be drawn before leaving this section, because it is the most common confusion in the modern literature. Information economics is not behavioral economics. The two are the two great 1970s departures from the postwar benchmark, and they depart in opposite directions. Information economics relaxes the assumption of common knowledge while keeping the agents fully rational: Akerlof's buyers, Spence's workers, Rothschild-Stiglitz's insurers all optimize flawlessly — they simply do not know each other's private types. Behavioral economics, the subject of a later chapter, does the reverse: it keeps full information and relaxes rationality, studying agents whose judgments and choices depart systematically from the optimizing ideal. Lemons (1970) and prospect theory (1979) are time-siblings and method-opposites. Conflating them — treating "the market got it wrong" as one undifferentiated idea — loses the precise thing each revolution actually changed.
You have just seen that the welfare theorems assume away the thing this chapter is about — common knowledge of quality. This walkthrough's efficiency case has to answer it.
The strongest case for market efficiency rests on the welfare theorems — and those theorems assume every trader knows every quality. Akerlof, Spence, and Stiglitz showed that when one side knows more, the efficient outcome can fail to exist at all: the lemons market unravels, the insurance market may have no equilibrium. The efficiency defense does not collapse, but it now has to specify where information is good enough for it to hold — which is the whole question the walkthrough pursues.
You now have the asymmetric-information trio. Health insurance is the market it was really about.
Arrow's 1963 paper that opens this chapter was about medical care; Akerlof's lemons, Spence's signaling, and Rothschild-Stiglitz's screening are the apparatus that grew from it. Health insurance is the asymmetric-information market par excellence: the insured always knows more about their own risk than the insurer, so adverse selection is not a friction at the edge but the central design problem. Every feature of health-insurance policy — mandates, community rating, risk adjustment — is an answer to the unraveling this chapter's first section described.
Every economist before the 1960s asked the same question of a market: what will it do? Mechanism design asks the opposite one. Given that participants are self-interested and hold private information they will reveal only when it serves them, what rules would produce the outcome we want? The field is sometimes called reverse game theory, and the name is exact: instead of taking a game and solving for its equilibrium, you fix the equilibrium you want and search for the game that delivers it. This is the inversion that turned economics from a science that describes markets into one that engineers them — the chapter's central claim, and the hinge between the problem the last section posed and the working institutions the next one celebrates.
Mechanism design is the theory of designing the rules of a game so that its equilibrium produces a chosen social outcome, subject to participants acting in self-interest and holding private information. The binding constraint of the whole enterprise is one Leonid Hurwicz named: incentive compatibility. A mechanism is incentive-compatible when participating as intended — in the cleanest case, telling the truth about your private information — is in each participant's own interest. You cannot simply ask people for their private values and expect honest answers; a well-designed mechanism must make honesty the rational choice. Incentive compatibility is the line every workable mechanism has to clear, and Hurwicz's framing of it — the formal sense in which a decentralized system can or cannot elicit the information it needs — is what made the design problem precise enough to attack.
Attacking it directly looked hopeless, because the space of possible mechanisms is unimaginably large — any rule mapping any messages to any outcomes is a candidate. Roger Myerson's revelation principle is the result that made the field tractable. It says that any outcome achievable by any mechanism is also achievable by a direct mechanism — one in which participants simply report their private information — in which truthful reporting is optimal. The search over all conceivable mechanisms therefore collapses to the search over incentive-compatible direct mechanisms, a far smaller and well-defined set. The revelation principle is to mechanism design what Nash equilibrium was to game theory: not a result about any one market, but the move that made everything after it possible to compute. The third member of the 2007 trio, Eric Maskin, supplied the existence side of the question — implementation theory asks when a desired outcome can be made an equilibrium of some mechanism at all, and Maskin characterized the conditions; the formal treatment is in economics ch. 12 §12.1.
The abstract apparatus lands best on a concrete, genuinely surprising case, and the field has a canonical one. In a second-price (Vickrey) auction, the highest bidder wins but pays the second-highest bid. William Vickrey's 1961 result is that in such an auction, bidding your true value is a dominant strategy — the best thing to do regardless of what anyone else does. The intuition is exact and worth stating because it carries the whole idea: your bid determines only whether you win, never what you pay (the price is set by someone else's bid). So shading your bid below your value can only cost you a win you would have wanted at the price you would have paid; bidding above your value can only win you an auction at a price exceeding what the item is worth to you. Either way, honesty weakly dominates. This is a theorem, not a trick, and auctions descended from it — eBay's proxy bidding, the generalized second-price auctions that price search advertising — run a large fraction of the world's online commerce.
Try to beat truth-telling in a second-price auction — and fail. Set your true value, then shade your bid up or down against simulated opponents drawn from a value distribution. Sweep the bid and watch your expected payoff peak exactly at honesty and never rise above it. That is the mechanism-design surprise: truth-telling is a theorem, not a trick. Toggle the VCG extension to add a participant and see each one charged the externality they impose.
Figure 11.4 (interactive). Expected payoff as you shade your bid away from your true value in a second-price auction. The curve peaks at b = v and never exceeds that peak. Sweep the bid; honesty wins.
In a second-price auction your bid sets only whether you win, never what you pay — the price is the runner-up's bid. So shading down only forfeits wins you would have wanted at a price you would have happily paid; shading up only buys wins at prices above what the item is worth to you. Every deviation from your true value can only hurt. That is why bidding honestly is dominant: it is the best move no matter what anyone else does.
With value $v$, bid $b$, and highest rival bid $m$: you win iff $b > m$ and then pay $m$, for payoff $v - m > 0$ exactly when $m < v$. Bidding $b = v$ wins precisely the auctions worth winning ($m < v$) and skips the rest.
Shading down to $b < v$ forfeits the cases $b < m < v$ (a positive-payoff win given up); shading up to $b > v$ wins the cases $v < m < b$ (a negative-payoff win taken on). Either deviation weakly lowers expected payoff, so $b = v$ is a dominant strategy. VCG generalizes this: charging each agent the externality $\sum_{j \ne i} \big(W_{-i} - W_{-i}^{*}\big)$ makes truthful reporting dominant in any allocation problem.
The Vickrey auction is the simplest case of a general construction. A VCG mechanism (Vickrey-Clarke-Groves) extends the second-price logic to any allocation problem: each participant pays the externality their participation imposes on everyone else — the loss in value to the other participants caused by this one's presence. Pricing each agent's externality, rather than what they receive, is exactly what makes truthful reporting a dominant strategy, because it severs the link between what you say and what you pay for what you get. VCG is the canonical efficient, truthful mechanism, the formal heart of the design program, and the proof and its revenue properties are in economics ch. 12 §12.3 and §12.4. The 2007 Nobel to Hurwicz, Maskin, and Myerson marked the inversion's maturity: the discipline now had a worked-out, awarded science of rule-design, not a collection of clever auctions.
You have seen that mechanism design can repair some information-broken markets by engineering the rules. That reframes the efficiency question from 'do markets clear' to 'can we design markets that clear'.
Roth's matching and the Milgrom-Wilson auctions are the constructive answer to the lemons problem: where a market fails on its own, a designed mechanism can make it work — the move from describing markets to designing them. The bound is Hayekian: design works where the private knowledge can be elicited, and degrades where it cannot. Efficiency, on this view, is partly something you build, not only something you find.
Then the math left the blackboard. The most striking thing about the mechanism-design program is that it did not stay a theory; over the 1990s and 2000s economists used it to build real institutions that allocate real things — and the institutions worked better than what they replaced. Market design is the applied subfield that turns mechanism-design theory into working allocation systems: auctions, matching markets, school-choice systems. It is the discipline's clearest case of cumulative, agreed-upon success, and it is what makes "economics as engineering" a description rather than a boast.
The engine of the matching side is an algorithm from 1962. The deferred-acceptance algorithm of David Gale and Lloyd Shapley solves two-sided matching markets that have no prices — markets where, say, doctors and hospitals, or students and schools, must be paired and money cannot do the allocating. Everyone proposes to their first choice; each receiver tentatively holds its best applicant and rejects the rest; the rejected propose to their next choice; and so on until the music stops. What it produces is a stable matching: a pairing in which no unmatched pair would both prefer each other to their assigned partners — no one can defect into a better arrangement that the other side would also accept. The algorithm is also strategy-proof for the proposing side, which is incentive compatibility made operational: participants do best by submitting their true preferences. The formal account is in economics ch. 12 §12.5.
Alvin Roth turned the algorithm into institutions. The National Resident Matching Program — which assigns graduating American medical students to residency positions — was redesigned on deferred-acceptance principles in the late 1990s after the old system broke down under strategic gaming. School-choice systems in Boston and New York were rebuilt the same way, replacing mechanisms that punished families for ranking honestly with ones where honest ranking is the dominant strategy. And the case that makes the abstraction unforgettable: kidney exchange. A patient with a willing but incompatible donor is, on their own, stuck. Roth's matching theory pairs such couples into chains — my donor's kidney goes to your compatible patient, your donor's to a third patient, and so on — so that a cycle of otherwise-impossible transplants happens at once. A real kidney finds a real recipient who would never have received one under the old institution. This is a Nobel-winning piece of abstract mechanism design saving lives that the previous system could not.
The auction side has an episode just as concrete. In 1994 the United States Federal Communications Commission needed to allocate spectrum licenses, and rather than the old method — comparative hearings or lotteries, which gave the public nothing — it ran an auction designed by economists including Paul Milgrom and Robert Wilson. The simultaneous multiple-round design they built handled the hard part — that licenses for adjacent regions are worth more together than apart, so bidders need to assemble packages — and the auctions raised tens of billions of dollars while putting spectrum in the hands of those who valued it most. Economics had, quite literally, earned the public billions by getting the rules right. Milgrom and Wilson's auction-design work was recognized with the 2020 Nobel, and Roth's matching work (shared with Shapley) with the 2012 Nobel; the field had become a working applied science with a string of prizes to mark it.
Here the chapter owes its one honest calibration, and it returns to Hayek to make it. The success of market design is real and consensual — but it is bounded, and the boundary is precisely the one the Austrians drew. Market design works where the relevant private knowledge can be elicited through the mechanism: bids reveal values, rank-order lists reveal preferences, and the mechanism's incentive-compatibility makes the revelation honest. It degrades where the knowledge is genuinely tacit, unstable, or simply not the kind of thing a bid or a ranked list can carry. Read carefully, incentive-compatible design is a partial solution to Hayek's knowledge problem rather than a refutation of it: instead of presuming the designer possesses the dispersed knowledge, a good mechanism makes the participants reveal it. That is a genuine answer to Hayek — but only over the domain where the knowledge is elicitable, and the kernel of his objection is conceded for the rest. This is the chapter's one calibration, and it is a real one, not a manufactured both-sides: the engineering is a triumph in its showcase domains and honestly limited outside them.
Which closes the loop the chapter opened. Information economics is the formal, mathematical descendant of Hayek's knowledge problem. The Austrians identified the dispersed-private-information problem in prose and lost the institutional fight over central planning; the formalists mathematized the kernel of what they had seen and built a research program — and then a working technology — on it. The lineage is not a coincidence of theme; it is the same problem, first stated in words and lost, then stated in mathematics and, over the elicitable domain, partly solved. The full Austrian argument remains the Austrian chapter's to narrate; this chapter's contribution is to show where the kernel went.
The 1980 generation of macroeconometricians had stopped trusting the structural models they had inherited. The Cowles Commission program (treated in the postwar-synthesis chapter) built large systems of equations whose identification rested on theoretical restrictions — assumptions about which variables could be excluded from which equations. By the 1970s those restrictions looked, to a rising cohort, simply incredible: assumed for convenience, not believed. This section is the chapter's contained methodological rung — the shift in how economists handle time-series data, which sits between the Cowles structural program and the credibility revolution that follows it. It is a parallel story to the chapter's spine: not what private information does to markets, but what counts as a credible empirical claim, and how that standard kept rising.
Christopher Sims's 1980 "Macroeconomics and Reality" was the manifesto. Its target was exactly the incredible identifying restrictions of the structural program, and its alternative was the vector autoregression, or VAR: a model in which each variable is regressed on lagged values of itself and of every other variable in the system, with minimal theoretical restriction imposed. The posture is deliberately atheoretical — let the data describe the joint dynamics rather than forcing them through a theory's exclusion assumptions. The VAR became the workhorse of empirical macroeconomics, and the move it embodied — distrust the theory's identifying claims, model the data's own correlations first — is the methodological turn the section is named for.
The other pieces of the turn each gave the discipline a sharper tool. Clive Granger's 1969 work gave a precise, testable sense of a word economists had been using loosely: Granger causality. X Granger-causes Y if past values of X improve the forecast of Y beyond what Y's own past already provides. It is explicitly a statistical, predictive notion — "helps predict," not "metaphysically produces" — and naming it precisely let economists test for directional influence in time series rather than assert it. Robert Engle's 1982 ARCH model addressed a different feature of real data: that volatility itself moves over time, clustering in turbulent periods and subsiding in calm ones. Modeling time-varying volatility became the foundation of financial econometrics, where the size of tomorrow's swing is exactly the quantity at issue. Granger and Engle shared the 2003 Nobel for this body of work.
The piece that points forward is James Heckman's. His 1979 correction addressed selection bias: the error that arises when the sample whose outcomes you observe is not random. You observe wages only for people who chose to work; if the choice to work is related to the wage on offer, a naive regression of wages on characteristics is biased, because the sample selected itself. Heckman's method estimates the selection process jointly with the outcome equation, correcting the bias. Heckman shared the 2000 Nobel (with Daniel McFadden). His selection correction is the bridge to what comes next: the worry about whether the people you observe are a representative sample is the same worry that the credibility revolution — the identification-first methodology of natural experiments and randomized trials — would make the discipline's reigning evidence standard. That story, and the work of Card-Krueger, Angrist-Imbens, and Banerjee-Duflo, belongs to the New Keynesian chapter; this section carries only the rung before it. The thread that runs through both is a single one: methodology is an evolving standard for what counts as a credible empirical claim, and the time-series turn raised the bar that the credibility revolution would raise again.
You have just seen Heckman's selection correction. It is the bridge to the credibility revolution this walkthrough is about.
Before the credibility revolution decided what counts as evidence, the time-series turn changed how economists read data — and Heckman's selection correction is the rung between them. The worry that the people you observe are not a representative sample is the same worry that natural experiments and randomized trials would later make the discipline's reigning standard. Heckman's structural-econometrics defense of modeling the selection process, rather than only designing it away, is the live counter-position this walkthrough engages.
The perfect-information assumption did not die in 1970. It was promoted — from a description of how markets work to a benchmark for how well they could. That is the chapter's verdict on what the revolution did. Arrow-Debreu remained standing as the welfare reference; what the information economists added was the apparatus to model the markets that fall short of it, and what the mechanism designers added was the apparatus to repair some of them. "Approaching the competitive ideal" stays a coherent design target precisely because the ideal was bounded, not refuted.
The relational view of this lineage lives on the intellectual-history timeline, and it is worth saying plainly what the timeline does and does not yet carry. The chapter's named thinkers — Nash, Akerlof, Spence, Stiglitz, the 2007 trio of Hurwicz, Maskin, and Myerson, Vickrey, Roth, Milgrom, and the econometricians Granger, Sims, Engle, and Heckman — are largely not yet nodes on the graph. What the timeline does carry, and what the reader can explore now, is the surrounding structure: the von Neumann-Morgenstern founding that Nash built on, the Hayek node whose knowledge problem this chapter traces forward, and the mechanism-design school node that the inversion produced. The embed below shows that available skeleton; the fuller figure-level lineage is a graph-enrichment the platform has flagged and not yet built. Naming the gap honestly is part of the reconciliation: the prose carries the full lineage; the graph carries, for now, its anchors.
Arrow, "Uncertainty and the Welfare Economics of Medical Care" (American Economic Review, 1963). Nash, "Equilibrium Points in N-Person Games" (1950); "Non-Cooperative Games" (Annals of Mathematics, 1951). Akerlof, "The Market for Lemons" (Quarterly Journal of Economics, 1970). Spence, "Job Market Signaling" (QJE, 1973). Rothschild & Stiglitz, "Equilibrium in Competitive Insurance Markets" (QJE, 1976). Vickrey, "Counterspeculation, Auctions, and Competitive Sealed Tenders" (Journal of Finance, 1961). Hurwicz, "On Informationally Decentralized Systems" (1972). Gale & Shapley, "College Admissions and the Stability of Marriage" (1962). Myerson, "Optimal Auction Design" (1981). Maskin, "Nash Equilibrium and Welfare Optimality" (1977/1999). Grossman & Stiglitz, "On the Impossibility of Informationally Efficient Markets" (1980). Granger, "Investigating Causal Relations by Econometric Models and Cross-Spectral Methods" (Econometrica, 1969). Heckman, "Sample Selection Bias as a Specification Error" (Econometrica, 1979). Sims, "Macroeconomics and Reality" (Econometrica, 1980). Engle, "Autoregressive Conditional Heteroscedasticity" (Econometrica, 1982). Roth & Sotomayor, Two-Sided Matching (1990). Roth, Who Gets What — and Why (2015). Milgrom, Putting Auction Theory to Work (2004). Royal Swedish Academy of Sciences, Prize in Economic Sciences citations 2000, 2003, 2007, 2012, 2020.