Articles · Behavioral finance
Representativeness and overreaction: why glamour stocks disappoint
How the representativeness heuristic makes investors extrapolate winners and losers, what De Bondt-Thaler and LSV found, and a real glamour-stock case.
On this page
Representativeness is the habit of judging how likely something is by how much it resembles a familiar pattern. A company with five years of soaring sales looks like a "great company," and a great company looks like a great stock. A company that has disappointed for years looks like a loser, and losers look like bad investments. Both judgments feel natural, and both ignore the question that decides an investment's return: what is already in the price?
The psychologists Daniel Kahneman and Amos Tversky described representativeness in the early 1970s. A decade later, finance researchers found its footprint in stock prices: portfolios of past losers went on to beat portfolios of past winners, and cheap "value" stocks beat expensive "glamour" stocks by wide margins. This article explains the heuristic, the evidence, the debate about what causes it, and a real case of a glamour stock after the cheering stopped.
The heuristic
Kahneman and Tversky's classic experiments described a person and asked how likely it was that he or she belonged to a group. People judged mainly by resemblance to a stereotype and largely ignored the base rate, how common each group actually is. In another line of work they found what they called the "law of small numbers": people expect a short sequence to look like the process that generated it, so they read too much into a few observations.
Both errors carry straight into investing:
- Base-rate neglect. Few companies sustain rapid growth for a decade; most fast growers slow down. An investor who judges by resemblance ("it looks like the next great tech company") ignores how rarely that story comes true.
- Small samples. Three good years of sales or returns are a small sample, yet they are often treated as evidence of a permanent trait.
- Good company, good stock. A company's quality and its stock's future return are different things. The price can already reflect the quality, and more.
The cousin of this bias, chasing recent fund or stock performance, is covered in Recency bias and performance chasing: buying what just went up. Here the focus is on what representativeness does to prices over years.
The overreaction study
In 1985 Werner De Bondt and Richard Thaler asked whether investors overreact to a run of good or bad news. Using monthly returns for New York Stock Exchange stocks from 1926 to 1982, they repeatedly ranked stocks by their returns over the previous three years. They formed a "loser" portfolio of the 35 worst performers and a "winner" portfolio of the 35 best, and tracked each for the next three years.
The results reversed the past. On average over the three years after formation, the losers beat the market by about 19.6% and the winners trailed it by about 5.0%, a gap of roughly 25 percentage points. Much of the losers' gain came in January, and the effect was stronger for losers than for winners. De Bondt and Thaler read it as evidence that prices overshoot: investors become too pessimistic about companies that have done badly and too optimistic about those that have done well, and the correction comes slowly.
Glamour versus value
Josef Lakonishok, Andrei Shleifer and Robert Vishny took the idea further in 1994. They sorted US stocks each April from 1968 to 1989 by measures of how expensive they were, and by how fast their sales had grown, and followed each group for five years.
The cheapest tenth of stocks by book-to-market earned 19.8% a year on average over the next five years; the most expensive tenth earned 9.3%. Sorting by cash flow-to-price gave a gap of about 11 points, and by earnings-to-price about 7.6. Companies with the slowest past sales growth beat those with the fastest by almost 7 points a year. Combining the two ideas, low price relative to cash flow and low past growth, gave a gap of about 10 to 11 points a year.
The authors' explanation was extrapolation. Investors projected the past growth of glamour companies too far into the future and paid too much for them; they did the opposite for value companies. When later growth turned out closer to average for both, glamour stocks disappointed and value stocks surprised. LSV tested whether value stocks were simply riskier and found that they did not do worse in bad markets or recessions, which made a pure risk story harder to sustain.
The debate and the momentum puzzle
Not everyone accepted the behavioral reading. Eugene Fama and Kenneth French argued in 1996 that the long-run reversal and the value premium are largely captured by their size and value factors, which they interpreted as compensation for risk. Other critics showed that part of De Bondt and Thaler's result came from small stocks and January.
A second complication is momentum. Narasimhan Jegadeesh and Sheridan Titman found in 1993 that stocks that did best over the past three to twelve months tend to keep doing well for several more months. So prices seem to underreact in the short run and overreact over three to five years. A 1998 model by Barberis, Shleifer and Vishny combined two biases to explain both: investors are slow to update on a single piece of news (conservatism), then, after a long enough string of similar news, decide they are seeing a pattern and extrapolate it (representativeness). The momentum side is covered in Time-series momentum: the 12-month return rule, tested and 52-week high momentum: buying what is near its high, tested.
Whatever the cause, the premium has not been steady. Value stocks trailed growth stocks for most of the period from 2007 to 2020, a long stretch that tested anyone who relied on the pattern (see Growth vs value investing: styles, P/E and the long record). The history shows a tendency, not a law.
A glamour stock after the peak: Cisco
In early 2000 Cisco Systems, the maker of the routers that ran the internet, was the archetype of a great company. Its sales and profits had grown rapidly for years, and for a short time in March 2000 it was the most valuable company in the world by market capitalization. The price assumed that growth would continue at a pace few companies have ever sustained.
Cisco remained a large, profitable business and kept growing for years. Its stock did not. From a month-end close of $77.31 in March 2000, the shares fell to $10.48 by September 2002, a decline of about 86%. They did not close a month above the March 2000 level until January 2026, almost 26 years later. Shareholders who held throughout collected dividends from 2011, which this price-only chart leaves out, but the lesson is plain: the company's quality was real, and the price had run far ahead of it.
Working against representativeness
- Start from the base rate. Before judging a company's prospects, ask how often companies of its size and growth rate have kept growing at that pace. The answer is usually: rarely.
- Separate the company from the price. Write down what growth the current P/E implies, then ask whether that is likely (see Discounted cash flow (DCF): a step-by-step valuation on Apple).
- Treat short records as noisy. Three years of results, good or bad, is a small sample.
- Look at the losers too. A list of stocks that have fallen most is uncomfortable reading, which is part of why such stocks can be cheap.
- Use rules. A diversified value or quality rule applied mechanically, as in Value investing as a strategy: the rules and a 26-year test, keeps a vivid story from overriding the numbers.
How Finabase helps
Finabase shows the numbers that separate a good company from a good price. A symbol's Valuation tab shows multiples such as P/E and price-to-book over time, so you can see how expensive a stock is compared with its own history. The Screener sorts stocks by past performance and by valuation, which makes the winner-versus-loser and glamour-versus-value comparisons in this article easy to set up. The full chart with range=max shows what happened after any peak.
Frequently asked questions
What is the representativeness heuristic in investing?
It is judging an investment by how much it resembles a familiar pattern, such as "great company equals great stock," instead of by probabilities, base rates and price. It leads investors to extrapolate recent success or failure too far.
What is the overreaction hypothesis?
It is De Bondt and Thaler's 1985 proposal that investors overreact to long runs of good or bad news. They found that the prior three years' biggest losers beat the biggest winners by about 25 percentage points over the following three years.
Do value stocks always beat glamour stocks?
No. Lakonishok, Shleifer and Vishny found a large gap from 1968 to 1990, but value trailed growth for most of 2007 to 2020. The premium appears over long periods and many stocks, with long stretches when it disappears.
How is representativeness different from recency bias?
Recency bias overweights the latest events because they come to mind easily. Representativeness judges by resemblance to a stereotype and ignores base rates. They often work together when investors extrapolate a recent run.
Summary
Representativeness makes a strong record look like proof of a strong future and a weak record look like proof of failure. In stock prices that tendency shows up as long-run overreaction: past losers beating past winners, and value stocks beating glamour stocks over the 1968–1990 period studied by Lakonishok, Shleifer and Vishny. Whether the cause is psychology, risk or both is still debated, and the value premium has had long dry spells. Cisco after March 2000 shows the mechanism in one company: a business that stayed strong, a share price that took a quarter of a century to recover.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
An investor reasons: “This company has grown sales 30% a year for five years, so it is a great stock.” What is the main error?
- Sales growth never matters
- Treating a short record as a permanent trait and confusing a good company with a good investment, without asking what the price already assumes
- Five years is too long a record
Show the answer
B. Treating a short record as a permanent trait and confusing a good company with a good investment, without asking what the price already assumes Few companies sustain such growth (the base rate), and if the price already assumes it continues, even good results may disappoint.
In De Bondt and Thaler’s 1985 study, what happened to the prior three years’ biggest losers over the next three years?
- They kept falling relative to the market
- They beat the market by about 20% while the biggest winners trailed it by about 5%
- They matched the market exactly
Show the answer
B. They beat the market by about 20% while the biggest winners trailed it by about 5% The reversal is the overreaction hypothesis: prices overshoot on long runs of good or bad news and later correct.
Lakonishok, Shleifer and Vishny found value stocks beat glamour stocks by about 10 points a year. How did they explain it?
- Value stocks are much riskier
- Investors extrapolated past growth too far, overpaying for glamour stocks and underpaying for value stocks
- Glamour stocks paid no dividends
Show the answer
B. Investors extrapolated past growth too far, overpaying for glamour stocks and underpaying for value stocks They found value stocks did not do worse in bad markets, which made a pure risk explanation harder to sustain. Fama and French disagreed.
How can short-term momentum and long-term reversal both exist?
- They cannot; one of them must be wrong
- Prices may underreact to news over months, then, after a long run of similar news, investors extrapolate and prices overshoot over years
- They occur in different countries only
Show the answer
B. Prices may underreact to news over months, then, after a long run of similar news, investors extrapolate and prices overshoot over years Barberis, Shleifer and Vishny (1998) modeled exactly this: conservatism in the short run and representativeness in the long run.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open the Screener and sort stocks by their 1-year performance. Write down the five best and the five worst. Open each one’s Valuation tab and note its P/E or price-to-sales ratio. Which group is priced for more future growth?
Open the ScreenerWhat you should find
The best performers usually trade at much higher multiples than the worst, which means their prices already assume that recent success continues. That is the extrapolation Lakonishok, Shleifer and Vishny describe. It does not tell you which group will do better next year; it tells you how much each price expects.
Sources and further reading
- Tversky and Kahneman — Belief in the law of small numbers, Psychological Bulletin 76(2), 1971
- Kahneman and Tversky — Subjective probability: a judgment of representativeness, Cognitive Psychology 3(3), 1972
- Tversky and Kahneman — Judgment under uncertainty: heuristics and biases, Science 185(4157), 1974
- De Bondt and Thaler — Does the stock market overreact? Journal of Finance 40(3), 1985, pp. 793–805
- Lakonishok, Shleifer and Vishny — Contrarian investment, extrapolation, and risk, Journal of Finance 49(5), 1994, pp. 1541–1578
- Jegadeesh and Titman — Returns to buying winners and selling losers, Journal of Finance 48(1), 1993
- Barberis, Shleifer and Vishny — A model of investor sentiment, Journal of Financial Economics 49(3), 1998
- Fama and French — Multifactor explanations of asset pricing anomalies, Journal of Finance 51(1), 1996
- Finabase chart data — Cisco Systems (CSCO) monthly closes, January 1995 to October 2026, price only
