Irrational Exuberance

When Alan Greenspan used the phrase ‘irrational exuberance’ in December 1996, the Dow stood near 6,400. By January 2000 it peaked at 11,722. The spiking of prices between 1995 and 2000 looks on a long-run chart like a rocket taking off from earth — an anomaly utterly unlike anything in the prior century of U.S. market history.

I measure valuation using a ten-year moving average of real earnings — the cyclically adjusted price-earnings ratio. At the 2000 peak this ratio exceeded 45, far outside every historical precedent. The three previous highs — 1901, 1929, 1966 — all preceded decade-long stretches of poor real returns. The 2000 reading was in a class of its own.

The boom was not a U.S. phenomenon alone. Between 1995 and 2000, real stock market valuations in Brazil, France, Germany, Japan, and a dozen other countries all surged and peaked at roughly the same moment, suggesting common psychological forces at work rather than any country-specific fundamental improvement.

Even after the post-2000 correction, the price-earnings ratio at the time of the second edition remained in the mid-20s — still well above its long-run historical average. This is not a book about a bubble that has already corrected; the irrational exuberance has, if anything, migrated into the housing market.

For most of the twentieth century, real (inflation-corrected) home prices in the United States showed no sustained uptrend at all. A long series I constructed going back to 1890 reveals that real home prices were roughly flat from 1890 to the late 1990s. The idea that homes are a sure path to long-run real appreciation is a myth manufactured by the boom itself.

The term ‘housing bubble’ or ‘home price bubble’ barely appears in English-language newspapers before the 1990s — a Proquest search back to 1740 confirms this. Only in recent decades, as our public commitment to market solutions displaced rent controls and other interventions, did people begin to treat homes as speculative investments and thereby make them prone to the feedback that generates bubbles.

The post-2000 housing boom was concentrated in glamour cities — Boston, Los Angeles, London, Paris, Sydney — where prices began spiking sharply around 1997–98. These cities are part of an international market for prestige living, and the psychological salience of seeing neighbors’ home values rise every day is extremely high, making word-of-mouth contagion easy.

In our questionnaire surveys of recent homebuyers in boom cities, respondents described frenzied bidding wars, fears of being permanently priced out, and the conviction that prices in their city would rise indefinitely. These are the classic markers of a speculative bubble — and only 1% of respondents in our 2004 survey spontaneously used the word ‘bubble’ to describe what they were participating in.

No single factor caused the boom, and the list I offer is explicitly not an economic-fundamentals story. I focus instead on factors whose impact on market prices was not warranted by rational analysis: the global spread of a triumphalist capitalism after 1989, the Internet’s power as a visual symbol of revolutionary change, the Baby Boom cohort entering peak saving years, the decline of inflation that created money-illusion effects, and the massive expansion of defined contribution pension plans that channeled savings into equities.

The Internet mattered primarily as a story, not as a demonstrated profit engine. What matters for a stock market boom is the public impression a technology creates. The ease with which vivid examples of Internet disruption came to mind caused people to overestimate the probability that the revolution would translate into broad corporate profit growth — a classic availability heuristic operating at societal scale.

The decline of inflation from its 1970s highs had a subtle but powerful distorting effect. Most historical stock return data are reported in nominal terms, and investors naturally assumed that the nominal returns of the past would continue even as inflation fell below 2%. They were implicitly extrapolating a real return higher than history had ever delivered, without realizing the illusion.

The expansion of defined contribution pension plans — 401(k)s replacing defined benefit pensions — forced millions of people to make explicit investment choices and thereby become familiar with stocks and mutual funds. This learning-by-doing effect was a powerful recruitment mechanism for new equity investors, and their growing demand itself helped push prices up, which attracted still more investors in a classic feedback.

None of these precipitating factors alone, or even together, can account arithmetically for the level the market reached. They are the skin of the bubble — the reasons why excitement began. The underlying mechanism that amplified modest enthusiasm into a historic mania is a different story, one of feedback and self-fulfilling belief.

My Yale surveys of investor confidence since 1989 document precisely how price increases generated their own justification. When prices rose, measured investor confidence in future returns rose too, and the belief that any price decline would ‘surely be back up within two years’ spread even among investors who intellectually doubted the market’s fundamentals.

A Ponzi scheme is the best model I know for understanding a speculative bubble — but the key insight is that no fraudulent manager is required. When prices go up, stories spread, new investors are recruited, prices rise further, and the stories become even more plausible. The bubble is self-generating: a naturally occurring Ponzi process driven by narrative contagion rather than deliberate deception.

In feedback loop theory, initial price increases caused by precipitating factors lead to more price increases through increased investor demand. This second round feeds into a third, the third into a fourth. The original disturbance may be so remote in time that cause and effect seem impossible to connect. What looks like a market responding rationally to fundamental information is often just the echo of earlier price moves.

The asymmetry is revealing: surveys show that investors widely believed the market would ‘surely come back up’ after a crash, but did not symmetrically believe it would ‘surely come back down’ after a run-up. This is not rational Bayesian updating — it is a mood, a generalized confidence whose origins lie in experiencing many years of rising prices and finding ex-post narratives to justify each step.

The role of the news media in financial markets is not simply to relay economically significant information to investors who then rationally update their views. The media actively shape public attention and the categories of thought within which speculative events are played out. They are, in a sense, the lubricant of the bubble.

Media coverage of stock market milestones — the Dow crossing 10,000, or the S&P hitting a new high — is a form of advertising that reminds people of the option to invest, creates social proof, and lowers the psychological barrier to participation. Enhanced reporting of investing tips leads to increased demand for stocks, just as advertising for a consumer product leads to increased demand for that product.

The media thrive on superlatives, which systematically distorts investors’ sense of historical context. When record overload causes reporters to find something ‘near record’ on any given day, the cumulative effect is to make investors believe that unprecedented things are happening when often they are not — and this inflates the sense of excitement that sustains a boom.

The ends of new eras are also media events. My own first edition of this book received massive coverage in late February and early March 2000, right at the peak of the market. The media at certain moments amplify skeptical voices as readily as they had amplified the boom — the turning point in public narratives is itself a media phenomenon.

In every major bull market in U.S. history — 1901, the 1920s, the mid-1950s to mid-1960s, and the 1990s — the boom generated its own intellectual justification in the form of ‘new era’ theories. These theories did not precede the booms and cause them; rather, the boom created demand for justification, and reporters and analysts scrambled to supply it. The stock market often creates new era thinking, not the other way around.

The new era theories of the 1990s — centered on the Internet, globalization, the end of the business cycle, and a managerial revolution — had genuine partial truths embedded in them. Productivity really did accelerate in the late 1990s. But the mistake was treating partial truths as justification for price-earnings ratios four times their historical norm. The critical question is never ‘Is there something real here?’ but ‘How much is already priced in?’

The term ‘new era’ itself has appeared in virtually identical form at every major market peak going back over a century. At each peak, commentators genuinely believed that this time was different — that some structural change in the economy had permanently lifted the floor under earnings growth. Each time, the new era ended when a contradicting reality (inflation, recession, geopolitical shock) provided a focal point for the reversal.

New era thinking is part of the feedback mechanism: it makes the price increases seem permanent rather than speculative, which reduces the urgency to sell and attracts still more buyers. The theories are not just window dressing — they are a functional component of the bubble’s self-sustaining structure.

Looking across twenty countries over a century, the pattern is strikingly consistent: dramatic stock market booms occur, are accompanied by local new era stories, reach extreme valuations, and then correct. The tendency for what went up dramatically to come back down is a robust international regularity, though its timing is unpredictable.

The synchronization of the 2000 peak across so many countries simultaneously is itself strong evidence against the proposition that national fundamental factors drove the boom. The simultaneous peaks in Brazil, Germany, Japan, and the U.S. are better explained by a globally transmitted speculative psychology — carried by the same news media, the same Internet enthusiasm, and the same wave of capitalist triumphalism — than by coincident improvements in each country’s economic fundamentals.

Historical peaks in individual countries that were followed by long periods of poor returns include Japan in 1989, the U.S. in 1929 and 1966, and many others. The consistent finding is that extreme valuation ratios predict poor long-term returns — not with certainty, and not with known timing, but reliably enough to be policy-relevant.

Investors use two kinds of psychological anchors in making buy-and-sell decisions: quantitative anchors (round numbers like ‘Dow 10,000,’ or recent price levels that feel like normal baselines) and moral anchors (narratives about why it would be wrong or foolish to cash out now). Neither is derived from fundamental analysis, but both can hold prices at extreme levels for extended periods.

Overconfidence is a pervasive human trait, and it operates with particular force in financial markets. When people in experiments are certain they are right, they are in fact right only about 80% of the time — a consistent overconfidence gap. In surveys I conducted immediately after the 1987 crash, a substantial minority of individual investors said they believed they had a good idea when the rebound would occur, citing ‘gut feeling’ and ‘intuition’ rather than any concrete analysis.

Magical thinking — the intuitive belief that buying a stock will somehow cause it to go up, or that one’s own judgment is predictive — operates below the level of explicit reasoning. People do not state these beliefs aloud, but they act on them. Combined with hindsight bias (the retrospective illusion that past market moves were predictable), magical thinking makes investors systematically overestimate the reliability of their own judgment.

The fragility of psychological anchors means that when a news event breaks the current anchor — when the Dow drops below a psychologically salient round number, or when a prominent analyst turns bearish — the shift in sentiment can be rapid and discontinuous. Anchors are not stable equilibria; they are temporary resting points for a market that lacks objective grounding in fundamental value.

People who communicate regularly with one another think similarly. This is not a failure of individuality — it is a basic fact about human social cognition, and it means that the precondition for epidemic-style contagion of investment ideas is always present. If the millions of investors were truly independent of each other, as some models assume, irrational exuberance would cancel out. But they are not independent.

Epidemic models from epidemiology — with infection rates and removal rates — provide a better mathematical framework for understanding speculative bubbles than equilibrium models. The rate at which investment ideas spread through word of mouth follows logistic curves. A major national news story unrelated to markets can lower the ‘infection rate’ of speculative ideas by deflecting social attention, which helps explain why markets are not notoriously volatile during genuine crises.

Word-of-mouth transmission is especially important because it carries emotional contagion that media reports cannot. In studying how institutional investors choose stocks, John Pound and I found that even professional investors learned about their holdings primarily through direct conversation with other professionals — not from formal analysis. The ideas that spread are the vivid, easy-to-repeat ones, not necessarily the analytically sound ones.

People simultaneously hold conflicting ideas about the market — ‘prices might be too high’ alongside ‘prices will keep rising’ — and a shift in public attention can bring either belief to the fore without any new fundamental information arriving. This is why markets can turn rapidly on events that seem disproportionate to any rational reassessment of value.

The efficient markets theory — that prices accurately reflect all public information at all times — is the leading intellectual basis for dismissing concerns about bubbles. I take it seriously: the theory has genuine empirical support for short-horizon forecastability (day-to-day prices really are close to random walks), and it correctly identifies that naive exploitation of patterns is harder than it looks.

But the theory fails at the level that matters most — long-horizon aggregate market valuation. In my 1981 paper in the American Economic Review, I showed that actual stock price movements are far more volatile than the present discounted value of subsequent dividends would justify. If prices were rationally set, they should be smoother than dividends, not more volatile. This ‘excess volatility’ finding is one of the most robust in financial economics.

There are too many examples of what look like obvious mispricings — the 1999 Palm/3Com spinoff, the extraordinary valuations assigned to profitless Internet companies — for the efficient markets defense to be fully convincing. Merton Miller’s response, that such anomalies are ‘too interesting’ and should be abstracted away, is an aesthetic judgment, not an empirical refutation.

Price-earnings ratios predict long-run returns reliably across international markets. When the ratio is extremely high, as it was in 2000, subsequent ten-year real returns have historically been very poor or negative. This is not a random walk property — it is mean reversion at long horizons, and it is inconsistent with the efficient markets claim that prices are always correctly set.

A popular explanation for the high market of the late 1990s was that investors had finally ‘learned’ that diversified stock holdings are safe for the long run — that they had absorbed the lesson of Jeremy Siegel’s Stocks for the Long Run and were rationally bidding up prices to reflect lower perceived risk. Glassman and Hassett’s Dow 36,000 was the most prominent statement of this view.

I find this learning story unpersuasive. People did come to believe that stocks were safer than they used to think — but this belief was itself a product of the bull market, not independent evidence about fundamental risk. They ‘learned’ from an extended stretch of rising prices that rising prices are normal. This is not learning a truth; it is extrapolating a trend and calling it enlightenment.

The sense that investors have just discovered important truths and arrived at a new enlightenment has appeared at virtually every major market peak in history. It is itself a predictable component of irrational exuberance. The lesson of history is not that stocks are risk-free for the long run; it is that at certain price levels, the expected real return on stocks is low or negative for very long holding periods.

Glassman and Hassett were simply wrong — we now know this. Their error was conflating the historical fact that stocks have outperformed bonds on average with the inference that any current price, however high, is therefore justified. The equity premium is a compensation for real risk; eliminating it in theory does not make the risk disappear.

We cannot completely protect society from waves of irrational exuberance or irrational pessimism — these are part of the human condition. But we are not without tools. The question is not whether to tolerate speculative markets but how to design institutions that reduce the harm their excesses cause.

Retirement plans urgently need to be put on a sounder footing. The shift from defined benefit to defined contribution plans transferred market risk from employers to individual workers who are poorly equipped to bear it. Current 401(k) plan design — offering multiple stock categories without strongly worded diversification advice — invites serious errors by people who are acting on popular-culture myths about surefire profit opportunities rather than personal financial judgment.

Privatizing Social Security by investing balances in the stock market would be a serious mistake. It would expose the retirement security of the entire population to the same speculative volatility I have been describing, at precisely the moment when they can least afford it. Social Security’s value lies partly in its role as a risk-sharing institution across generations; dismantling that role in exchange for exposure to market returns that are far from guaranteed would be irresponsible.

Opinion leaders — Federal Reserve chairs, Treasury secretaries, prominent academics — have historically played a role in calling attention to over- and underpricing. The Fed chair issued warnings at the peaks of 1929, 1966, and 2000. These warnings had imperfect effects, but they were not without impact. Greater willingness among institutional voices to speak plainly about valuation — to exercise moral authority on behalf of long-run stability — is a legitimate and underused policy instrument.