Week 5: Momentum and Reversal

1. What momentum is and why it matters

This lecture covers two opposite patterns in asset prices. Momentum means that recent winners keep winning and recent losers keep losing for a while. Reversal (mean reversion) means that over longer horizons these trends turn around: yesterday's losers become tomorrow's winners. Together they describe how markets first move too slowly, then too far, and then correct.

Learning goals: what the evidence on momentum shows, how the strategy is built (J-month lookback, K-month holding), the difference between cross-sectional and time-series momentum, behavioral versus risk-based explanations, and the issues of costs, crash risk and implementation.

Momentum is in direct conflict with the Efficient Market Hypothesis (EMH). Under the EMH, prices already contain all available information, so past returns should not tell you anything about future returns. If simply buying last year's winners earns a reliable extra return, past prices do contain usable information.

"Party like it's 1999"

The lecture opens with a chart of the Nasdaq 100 index (NDX) from 1995 to early 2000, together with its 200-day moving average. The index climbed from roughly 500 points in 1995 to above 4,400 in March 2000, moving ever further above its average: the more it rose, the more investors bought. The later "Reversal Graph" shows the other half: by 2002 the index was back near 1,000.

2. History and landmark evidence

Early research detected positive autocorrelation at short and medium horizons (a positive return is usually followed by another positive one). During the 1990s a large body of work confirmed a lasting winner minus loser (WML) premium. On average, stocks with strong past performance beat those with weak past performance by roughly 1% per month. Practitioners had relied on trend-following tools for a long time: moving averages, trend lines and oscillators (indicators that signal when a market is overbought or oversold).

Jegadeesh & Titman (1993)

The landmark study. Using U.S. equities from 1965 to 1989, Jegadeesh and Titman ranked stocks on their past returns and formed winner and loser portfolios. Their best known "6/6" strategy (6-month lookback, 6-month holding period) earned about 1% per month, and the effect remained after accounting for transaction costs.

Rouwenhorst (1998) and Asness (1997)

Rouwenhorst (1998) showed that momentum also exists in European stock markets, so it is not a U.S. data-mining accident. Asness (1997) documented momentum across stocks, bonds, currencies and commodities. The conclusion on the slides: momentum is a global and persistent phenomenon.

Why it matters

• It violates the "random walk" version of the EMH: returns are predictable.
• For portfolio managers it is a potential source of alpha (return above what risk explains), but with significant risks.
• It is one of the strongest arguments in favour of behavioral finance.

Since momentum makes returns predictable, either markets are not fully efficient or our asset pricing models are incomplete. Hence two interpretations: (1) risk-based: compensation for unobserved risk; (2) behavioral: the result of psychological biases. This split is the core debate of the lecture.

3. Building a momentum strategy

Deciles and the W-L spread

The standard procedure sorts all stocks into 10 groups (deciles) based on their lookback returns. D1 holds the 10% worst performers (losers), D10 the 10% best (winners). The strategy return is WML = D10 minus D1: long the winners, short the losers. It is a "zero-investment" portfolio: its return is a pure spread.

The illustrative bar chart on the slides shows average monthly returns rising steadily across the deciles, from about 0.4% for D1 to about 1.5% for D10. Every higher decile earns a bit more (a monotonic pattern), which is stronger evidence than a spread driven only by the extremes.

Strategy parameters

• Lookback J: 3, 6, 9 or 12 months, frequently leaving out the latest month.
• Holding period K: 3 to 12 months, using equal or overlapping portfolios. Overlapping means a new portfolio starts every month and several are held at once.
• Rebalancing: monthly or quarterly, with exit rules such as a stop-loss.

Cross-sectional versus time-series momentum

Cross-sectional momentum compares assets with each other: it buys the relative winners and shorts the relative losers, even if all of them went down. Time-series momentum looks at each asset on its own: if its own recent cumulative return is positive, take a long position; if negative, take a short position. The slide shows a position line switching between Long, zero and Short over time as the sign of past returns changes.

Cross-sectional evidence and alternative markets

Asness, Moskowitz & Pedersen (2013), "Value and Momentum Everywhere", show momentum in individual stocks, across industries and countries, and in other asset classes (FX, bonds, commodities). It is persistent over time and did not disappear after publication, which is notable because many anomalies weaken once they become known. In currencies, a momentum-like strategy goes long high-yield currencies and short low-yield ones (this overlaps with the "carry" trade). In commodities, momentum is often stronger in futures markets (Moskowitz et al., 2012, the paper on time-series momentum).

4. The momentum factor (UMD) and factor models

The last month is skipped because very short-horizon returns tend to reverse (Jegadeesh 1990, section 10).

Momentum and the market

Momentum has a weak correlation with the market on average (low beta). In the illustrative chart it falls from about 0.4 (2000) to about −0.2 (2010) and recovers to about 0.2. It turns negative during trend reversals: after a crash the losers are high-beta stocks, so a sharp rebound hurts the short side.

Factor regressions

To test whether momentum is just a disguised known risk, researchers regress the WML return on the Fama-French three factors:

r_WML,t = α + b_M r_MKT,t + b_SMB SMB_t + b_HML HML_t + ε_t

Here r_MKT is the market excess return, SMB (small minus big) is the size factor, HML (high minus low book-to-market) is the value factor, the b's are factor loadings (sensitivities) and α is the part of the return that the factors cannot explain. Momentum shows a positive and statistically significant α even after controlling for the Fama-French factors. Market, size and value cannot explain it, so the slides conclude that momentum is an independent factor.

From CAPM to Carhart: UMD as the fourth factor

Carhart (1997) added UMD to the three-factor model, creating the four-factor model:

r_i,t − r_f,t = α_i + b_M(r_M,t − r_f,t) + b_SMB SMB_t + b_HML HML_t + b_UMD UMD_t + ε_i,t

• UMD isolates the momentum effect beyond value and size.
• b_UMD > 0 means the asset or fund benefits from trend-following exposure.
• Adding UMD reduces the unexplained α of many "growth" or trend portfolios: part of what looked like skill was simply momentum exposure.

This is why UMD improves the explanatory power of factor models for both stock returns and fund performance evaluation.

Empirical properties of UMD

• Persistence and universality: strong in the U.S., Europe, emerging markets, industries and across asset classes.
• Low correlation with Value and Size: improves diversification, especially when paired with Value.
• Market regimes: does well when markets trend; exposed to momentum crashes when trends suddenly turn.
• Implementation: keep an eye on trading costs, liquidity and limits on short selling.

5. Behavioral explanations of momentum

The EMH has trouble explaining why momentum persists. Behavioral finance offers three core explanations: underreaction (slow incorporation of new information), overreaction (excessive response to recent events) and herding (imitation and social influence).

Underreaction: slow adjustment

Investors underestimate the significance of new information and revise their expectations gradually rather than at once. As a result, prices keep moving in the same direction after news, which creates momentum. Example: news that justifies a move from 100 to 110 pushes the price only to 104 on the day; the other 6 points arrive over the following weeks.

Overreaction: excessive response

Recent events tend to get too much weight from investors. Waves of optimism or pessimism push prices further than justified. The first trend generates momentum, which in the end turns around. Key insight: put underreaction and overreaction together and you see why momentum shows up over short horizons yet fades over long ones.

Barberis, Shleifer & Vishny (1998): the BSV model

BSV build a model in which investors update beliefs with two biases:

1. Conservatism bias: people adjust too slowly to new data, so they underreact to single pieces of news.
2. Representativeness bias: after a series of similar news (for example several good earnings reports), people believe they have spotted a pattern ("this is a growth company") and extrapolate it too far, so they overreact.

Result: short-term momentum and long-term reversal. The conceptual diagram on the slides reads: New information → Belief updating (conservatism / representativeness) → Price reaction (under/overreaction). Slow updating leads to momentum; overreaction leads to reversal. Investor psychology produces self-reinforcing price patterns.

Herding

Investors mimic others, believing that "the market knows something". Capital flowing into popular stocks amplifies existing trends. Shiller describes this as "mass psychology" that drives collective overreaction. Today media, social networks and influencers act as accelerators of herding.

Overconfidence and overtrading

Overconfident investors believe they have superior information and overreact to news that confirms their prior views (confirmation bias). This leads to higher trading volume (overtrading), amplified momentum in bull markets and sharp reversals when sentiment shifts. Overconfidence helps explain why momentum crashes are sudden: many investors hold the same confident view and exit at the same time.

Mass psychology and narratives

Prices respond to stories as well as to numbers. With "narrative momentum", upbeat stories (AI, crypto, green tech) drive buying. Once the story changes, sentiment flips and prices correct. The slides give the dot-com bubble, the 2021 crypto boom and the meme stocks GameStop and AMC as examples.

Momentum and investor sentiment: Baker & Wurgler (2006)

Baker & Wurgler (2006) constructed an index of sentiment and found that it forecasts how strong momentum will be. In high-sentiment periods investors overreact to good news, so prices stray further from fundamentals. Once sentiment collapses, momentum crashes. The example chart is hump-shaped: WML returns climb from roughly 2.5% at sentiment level 1 to roughly 3.5% at level 3, drop to about 2% at level 4 and end near 0% at level 5. So momentum peaks when sentiment is fairly high and fails when sentiment is extreme.

Experimental evidence

Experiments in the tradition of Kahneman-Tversky and Barberis-Thaler show that people overestimate the continuation of trends and react more strongly to salient, vivid signals (salience bias). Even in laboratory markets, where the true value is known, participants generate price momentum. The tendency to "chase performance" appears to be a natural psychological trait.

6. Measuring momentum and its risks

Quantitative measurement

To test whether winners' outperformance is significant: (1) compute each stock's 12-month return excluding the most recent month; (2) rank stocks into deciles D1 to D10; (3) form WML = Return(D10) − Return(D1); (4) repeat every month and compute the average spread (and its t-statistic). A positive and persistent WML is the hallmark of momentum.

Rolling five-year momentum return

The momentum premium is not constant: it moves in cycles. In the illustrative chart the five-year average WML return is about 1% per month in the 1980s and 1990s, falls to slightly below zero around 2010 (after the 2009 crash) and recovers to about 0.6 to 0.8% afterwards. Momentum is strong in bull markets and weak after sharp reversals.

Momentum across asset classes

The same kind of pattern shows up in currencies (buy high-yield currencies, sell low-yield ones), in commodities (buy recent winners, sell recent losers) and in equity indices (buy strong markets, sell weak ones). In the example bar chart, average yearly returns are roughly 10% for stocks, 8% for FX, 9% for commodities and 6% for bonds: "Momentum is everywhere" (Asness, Moskowitz & Pedersen, 2013).

Transaction costs

Momentum requires high turnover: the winner and loser lists change every month, so trading costs rise. Despite these costs, momentum remains profitable for large- and mid-cap stocks, and low-turnover variants cut costs with little loss of return. Implementation needs careful rebalancing and liquidity filters.

Momentum crashes

Momentum breaks down when the market turns sharply (2009 and 2020 are examples). Holding the winners turns into a crowded position that is unwound in a rush. From a behavioral angle, an abrupt change in sentiment sets off panic among investors on the long side. In the example chart, WML drops from roughly −5% in month 1 to roughly −20% in month 3 and climbs back to 0 by month 6. Slogan: a momentum crash is a collective behavioral reversal.

A more crowded trade is a more fragile trade. When many funds cut their leveraged positions at once (mass deleveraging), sudden reversals follow, and investors, worried that "the story is over", drop the strategy. The slides cite 2009: an abrupt swing in risk appetite caused an 80% drawdown in momentum portfolios in the space of three months.

Risk management

• Volatility targeting: scale position size to volatility. Example: with a 12% annual volatility target and current momentum volatility of 24%, invest 12 / 24 = 0.5 times the normal position.
• Dynamic hedging: reduce exposure when the strategy's market beta rises (after a crash the short side is full of high-beta losers).
• Diversification: pair it with value or carry to get steadier returns.

Market behavior changes, so momentum strategies must adapt.

7. Momentum in practice and combining factors

Implementation: sort securities on their 6 to 12 month return while leaving out the last month; buy the winners and short the losers using equal or value weights; screen for liquidity and rebalance each month or quarter. Applications: managed futures and CTA (Commodity Trading Advisor) strategies, which mostly follow trends, long-short equity funds, and approaches that mix momentum signals with fundamental or macro screens.

Momentum + Value: "better together"

According to Asness, Moskowitz & Pedersen (2013), momentum and value have almost no correlation (they actually tend to move against each other), so holding both lowers total portfolio risk. In behavioral terms, value captures earlier overreaction (prices driven too far down) and momentum captures ongoing underreaction (trends that persist). Both anomalies come from one and the same psychological mechanism, seen from opposite sides.

Value, momentum and quality form the basis of "smart beta" investing (rule-based factor portfolios). Each factor has its own performance cycle: momentum does best in trending environments, value in mean-reverting ones. In the illustrative chart momentum rises to about 13% and value more unevenly to about 7%.

8. Technical analysis as visible momentum

Trendlines

Technical analysts use charts to identify trends. Trendlines are simple geometric representations of market dynamics; the goal is to spot support (a level where falling prices tend to stop) and resistance (a level where rising prices tend to stall) that signal continuation or reversal. An uptrend line connects a sequence of higher lows; a downtrend line connects a sequence of lower highs. Breaking a trendline signals a possible change in direction.

The example chart shows prices moving within a range, with support around 24,400 and resistance around 25,370: prices bounced several times off both lines. Prices often revert toward this "equilibrium zone".

Psychological interpretation: trendlines serve as emotional reference points. Resistance levels count because they mark earlier highs that people remember (investors who bought at that level want to "get out at break-even", which adds selling pressure; think of anchoring and the disposition effect). Once the price breaks through resistance, optimism grows and fresh momentum appears. Trendlines are not laws of nature; they are shared expectations drawn onto a chart.

Moving averages

A moving average (MA) smooths prices to reveal the underlying trend and filters out noise.

• Simple Moving Average (SMA): the arithmetic mean of the last n prices. SMA_t = (P_t + P_t−1 + ... + P_t−n+1) / n.
• Exponential Moving Average (EMA): gives more weight to recent prices. Standard form: EMA_t = k × P_t + (1 − k) × EMA_t−1, with k = 2 / (n + 1).

Golden Cross: a short MA (say the 50-day) moves above a long MA (say the 200-day), read as a bullish momentum signal. Death Cross: the short MA drops below the long MA, read as bearish. Traders often use both to decide when to enter or exit.

Behavioral perspective: an MA converts uncertainty into a sense of "visual certainty" about the moment to act. Crossovers often trigger overreaction because they create an illusion of control. At its core the MA is a heuristic: "follow the line". Moving averages are behavioral momentum made visible. When lots of traders respond to the same crossover, the signal partly fulfils itself.

The Relative Strength Index (RSI)

Oscillators measure the strength and intensity of momentum. The best known is the RSI by Welles Wilder (1978), on a scale of 0 to 100: above 70 means overbought, below 30 means oversold.

1. Daily price change: ΔP_t = P_t − P_t−1
2. Split into gains and losses: G_t = max(ΔP_t, 0), L_t = max(−ΔP_t, 0)
3. 14-day averages (standard window): Ḡ = avg(G_t−13, ..., G_t), L̄ = avg(L_t−13, ..., L_t)
4. Relative Strength: RS = Ḡ / L̄
5. RSI = 100 − 100 / (1 + RS)

The slides give RSI = 80 as a potential downward reversal and RSI = 20 as a potential upward reversal. The RSI can be combined with other signals (for example moving averages) for confirmation. In the example chart the Nasdaq 100's RSI mostly stays between 30 and 70: prices often revert to this "equilibrium zone". So the RSI measures momentum, but extreme values are read as reversal signals.

Other indicators: MACD and the Stochastic Oscillator

The MACD (Moving Average Convergence Divergence), developed by Gerald Appel in the 1970s, measures the difference between two EMAs:

MACD_t = EMA₁₂(P_t) − EMA₂₆(P_t)
Signal_t = EMA₉(MACD_t)
Histogram_t = MACD_t − Signal_t

It captures convergence and divergence between short- and long-term trends. Buy (bullish) when the MACD crosses the signal line from below; sell (bearish) when it crosses from above. A rising histogram means strengthening momentum; a falling histogram means weakening momentum. A divergence (price and MACD moving in opposite directions, for example new price highs while the MACD makes lower highs) warns that a reversal may be near.

The Stochastic Oscillator compares the current price with the recent high-low range (a price near the top of the range means strong upward momentum) and is a good measure of the "speed" of a trend.

Behavioral view: indicators are popular with investors because they turn feelings into signals you can measure.

9. The big picture: what momentum says about markets

The human optimism cycle

Rising markets lift confidence, and with it momentum. The more investors join in, the "stronger" the trend looks. Once the mood turns, a steep fall, crash or reversal follows. Shiller reads this as mass psychology, alongside fundamentals, driving bull and bear phases.

Adaptive Markets Hypothesis (Lo, 2004)

In Andrew Lo's view, markets are "adaptively efficient". Investors learn and adjust; a strategy like momentum keeps working until the market "learns" how to cancel it out, and then fresh opportunities arise. Momentum is a "product of evolution": it lasts as long as human frictions remain. That matches the profits that change over time in the rolling chart.

Momentum and limits to arbitrage

If momentum is a known mispricing, why don't arbitrageurs remove it? Because arbitrage is limited: transaction and shorting costs, timing risk (the mispricing can grow before it shrinks) and the limited capital of arbitrageurs. There is also a behavioral element: fear of joining "crowded trades". This explains why anomalies persist even when widely known.

How momentum and the EMH coexist

The EMH says prices reflect information; momentum shows that prices also reflect investor behavior. Many researchers treat the EMH as a "first approximation" and momentum as a "behavioral correction". The EMH is not refuted but complemented by the behavioral view.

Lessons momentum offers about investors: they (1) are slow to learn (underreaction), (2) put too much faith in trends (representativeness) and (3) copy one another (herding). Taken together, these behaviors are predictable. The human element of markets produces statistical regularities, not because people are irrational, but simply because they are human.

Practical lessons for portfolio management

• Apply momentum in a diversified framework (equities, FX, futures).
• Scale positions by volatility and add value to limit the risk of a crash.
• Avoid overconcentration in crowded trades.
• Recognize that success often comes from psychology, not superior information.

Summary of the momentum part: momentum is strong and persistent; markets are "adaptively rational" rather than fully efficient; psychology explains it and limits to arbitrage let it survive.

10. Mean reversion and reversal

Theme two: when do markets "correct themselves"? If momentum is about trends continuing, why do those trends so often turn around over longer horizons? The core idea is that prices drift away from fundamentals for a while but tend to come back in the long run.

The "Reversal Graph" shows the Nasdaq 100 from 1998 to 2002 with moving averages and RSI: after the 2000 peak the index drops below its averages and falls to about 1,000.

De Bondt & Thaler (1985): "Does the Stock Market Overreact?"

• Sample: NYSE stocks, 1926 to 1982.
• Method: sort stocks into "winners" and "losers" based on their past 3-year performance, then hold these portfolios for the next 3 years.
• Result: former losers outperformed former winners by about 25% over the following three years.
• Interpretation: investors overreact. They extrapolate trends too far, overvaluing winners and undervaluing losers, and later correct.

The visualization shows cumulative returns after ranking: the loser portfolio climbs steadily to above 20% after about two to three years, while the winner portfolio reaches only about 5%. Yesterday's losers become tomorrow's winners.

Short-term reversal: Jegadeesh (1990)

Weekly returns are negatively autocorrelated: a stock that dropped hard in one week usually bounces back in the following one. Possible causes: liquidity shocks (a seller who must sell drives the price down for a while), noise trading, and the bid-ask bounce (trades switch between the bid and the ask, which by itself produces negative autocorrelation in measured returns). The slides' conclusion is that short-term reversal is microstructural rather than behavioral.

Long-term reversal

This shows up over 3 to 5 years and can be read in two ways: (1) behavioral: overreaction combined with representativeness bias; (2) risk-based: loser portfolios carry more risk and therefore earn a contrarian premium. Long-term reversal is tightly connected to the value effect, since cheap stocks tend to recover. Too much pessimism opens the door for contrarians.

Behavioral explanations of reversal

• Disposition effect: losers are held too long because investors are reluctant to realize losses; later these losers bounce back.
• Anchoring: clinging to earlier prices, so updates come too late.
• Overreaction: responding too strongly to past events.
• Representativeness: perceiving patterns that are not there (after a few bad years a firm looks like a "permanent loser").

Reversal is the "psychological restoration" of an earlier overreaction.

The value effect as long-term reversal

High book-to-market (B/M) stocks, called value stocks, are typically former losers: their price has fallen relative to the accounting value of equity. Low B/M stocks, called growth stocks, are former winners. Value's outperformance over growth is therefore a form of long-term reversal. In the illustrative chart value's cumulative return reaches about 22% after four periods versus about 9% for growth.

Risk-based explanations

Losers may carry more systematic risk and therefore earn higher future returns. However, reversal persists after controlling for the Fama-French risk factors, and behavioral elements seem to explain it better. Crowd-driven excesses (herding) create mispricing that ultimately corrects.

The mathematics: an AR(1) model

r_t = μ + ρ r_t−1 + ε_t

Here r_t is the current return, μ a constant, r_t−1 the return one period earlier, ρ (rho) the autocorrelation coefficient and ε_t a random shock. With ρ < 0 returns reverse (negative autocorrelation). With 0 < ρ < 1 returns persist (momentum). A pure random walk has ρ = 0. Over time, reversal is the mirror image of momentum, and markets switch between a momentum regime and a correction regime.

Value anomaly: conclusions

By extrapolating from the past, investors get carried away by new technologies or stocks that seem to offer strong growth. Compared with fundamentals, low B/M stocks end up too expensive and high B/M stocks too cheap. Once earnings fall short of those lofty expectations, prices adjust and the value effect emerges. The behavior of investors and analysts accounts for the anomaly.

De Bondt & Thaler (1987)

In their later work they found that the reversal effect comes back in different decades and samples, and shows up in markets outside the U.S. as well. Losers do have higher betas, yet risk cannot account for all of the reversal. Even with 50 years of data, overreaction stays statistically strong.

Rolling analysis and the behavioral cycle

Reversal is cyclical: stronger after periods of intense momentum, weaker in some decades (for example during the 1990s bull market). In the illustrative chart the premium dips to about 1% in 1990 and peaks at about 4.5% in 2000, after the tech bubble. Reversal intensifies after excessive momentum.

The behavioral cycle diagram: Underreaction (initial momentum) → Overreaction (excess) → Correction (reversal), after which the cycle starts again. Markets move cyclically between overreaction and correction.

Psychological causes of overreaction

• Representativeness: seeing patterns in random data.
• Anchoring: using recent prices as reference points.
• Availability bias: overweighting recent, easily recalled news.
• Overconfidence: underestimating the probability of being wrong.

Each of these leads to mispricing that is temporary and later unwinds. Investors chase "success stories" and shun losers, even though the losers often contain sound firms that are simply cheap; once the excess pessimism wears off, their prices come back.

11. Applying reversal: contrarian and value investing

A contrarian strategy buys losers and sells winners to exploit crowd overreaction, with a holding period of 3 to 5 years. In the illustrative chart the contrarian portfolio grows to about 15% after three years against about 4% for the market. It is rewarded when the market corrects its overreaction.

Value investors exploit reversal without using the word: they purchase cheaply valued companies, which tend to be recent losers. The fact that value beats growth is long-term reversal expressed in fundamentals. The risk-based objection is from Fama and French (1996): high B/M firms run a greater chance of financial distress and so carry more risk than low B/M firms, which makes their higher return a fair reward. The behavioral response: investors lose sight of the fact that bad news seldom lasts.

Reversal, macro cycles and sentiment

Market excesses follow optimism and pessimism cycles. In crises assets fall below fundamentals; when the economy recovers a strong reversal follows. The illustrative chart shows a value index falling from 100 to about 70 and then recovering to about 110.

Reversal becomes more likely as sentiment gets more extreme, in either direction. Measures such as the AAII investor survey and the Baker-Wurgler index often anticipate reversal phases. If the crowd is far too optimistic, a reversal is near, so the contrarian move is to bet against extreme sentiment. The second macro chart has a negative slope: future 3-year returns drop from roughly 8% at sentiment level 1 to roughly −3% at level 5. At extremes the crowd is seldom right.

Contrarian evidence

• Jordan and Pettengill (1990): over 1962 to 1986 the earlier losers portfolio earned an average excess return roughly 16% higher than the earlier winners portfolio.
• De Bondt and Thaler (1985), as this slide summarises them: past winners do not turn into losers afterwards; they return roughly 4.7%, against roughly 21.5% for the losers. The reversal is therefore driven mostly by a strong rebound of the losers (a difference of 21.5 − 4.7 = 16.8 percentage points) rather than by a fall of the winners.

12. A unified view of momentum and reversal

The two anomalies fit together: one reflects inertia, the other correction. Markets are slow to learn and frequently "overlearn". Reversal is how the market heals itself. For investors the lessons are that patience pays, that extreme prices do not last and that the crowd gets it wrong at the extremes. Shiller sums up the history of investing as an endless swing back and forth between euphoria and regret.