Investor psychology Lesson 13 of 15
Articles · Behavioral finance
Hindsight bias: why past markets look easier than they were
How knowing the outcome rewrites what you thought before it, what the experiments measured, why warnings do not fix it, and how to check your own record against it.
On this page
Hindsight bias is the tendency, once you know how something turned out, to believe you saw it coming, or could easily have. It is sometimes called the "knew-it-all-along" effect. Knowing the outcome changes your memory of what you thought before, and it makes the outcome look more inevitable than it was.
Investors meet it every time they look at a chart. A finished chart has a bottom, clearly marked. On the day of that bottom there was only a price that had been falling for a year and a half. This lesson explains what the experiments found, shows the difference on a real chart, and covers why hindsight matters for money: it corrupts the one source every investor learns from, which is their own past.
What the experiments found
Baruch Fischhoff ran the founding experiments in the early 1970s. In one, published in 1975, people read a short account of a little-known 19th-century war between the British and the Gurkhas of Nepal and were asked how likely each of four outcomes had been: a British victory, a Gurkha victory, or a stalemate with or without a peace settlement. One group was told nothing about the result. Four other groups were each told that a different one of the outcomes was what had "really" happened, and asked to answer as if they did not know.
They could not do it. Whichever outcome a group had been told, that group rated it as substantially more likely than the uninformed group did, and found the facts supporting it more relevant. Fischhoff called this "creeping determinism": the known ending reorganizes everything that came before it into a story that leads there.
A second study, with Ruth Beyth, tested memory directly. Before President Nixon's 1972 trips to China and the Soviet Union, students estimated the probability of various possible events (Nixon would meet Mao, the US and USSR would announce a joint space program, and so on). Some time after the trips, they were asked to recall their own estimates. For events they believed had happened, they remembered having given higher probabilities than they really had. For events that had not happened, they remembered lower ones.
Psychologists Neal Roese and Kathleen Vohs later sorted the effect into three levels, which are useful because each does different damage.
| Level | What it sounds like | What it damages |
|---|---|---|
| Memory | "I said it would fall." | Your record of your own forecasts |
| Inevitability | "It had to happen." | Your sense of how uncertain things are |
| Foreseeability | "I knew it. Anyone could have." | Your estimate of your own skill |
The finding is robust. A 2004 meta-analysis by Rebecca Guilbault and colleagues combined 95 studies with 252 separate effect sizes and found a consistent effect of moderate size (a standardized difference of 0.39).
One low, two charts
The figure shows the same price history twice. On the left is what an investor in the SPDR S&P 500 ETF could see at the close of February 27, 2009. On the right is the chart as it looked at the end of 2013.
On the right-hand chart the low is unmistakable, and the eye draws the conclusion for you: that was the time to buy, and it was obvious. Here is what was actually available on the day.
- SPY had closed October 2007 at $154.65 and February 2009 at $73.93, a fall of 52% at month-ends.
- It had fallen in 12 of the previous 16 months. February itself was down 10.7%, after January's 8.2%.
- There had already been one convincing "bottom". In April and May 2008 the fund rose 6.4% in two months. From that May close it then fell another 47%.
- Nothing on the chart distinguished February 2009 from November 2008, when the fund was 22% higher and also looked like it had fallen far enough.
March 2009 turned out to be up 7.6%, and the fund did not close a month above its October 2007 level until March 2013. An investor who bought at the end of February 2009 made 150% by the end of 2013, before dividends. That is a fact about what happened. "It was obvious" is a fact about the chart on the right, which nobody had.
The same exercise works for every famous turning point, in both directions. The top in October 2007 looked, at the time, like one more month in a five-year rise. History of stock market crashes has the dates and depths of the major US declines.
Why it costs investors money
It inflates confidence. If you remember predicting the past, you expect to predict the future, and you size positions accordingly. Bruno Biais and Martin Weber measured hindsight bias in 85 investment bankers in London and Frankfurt and compared it with their performance records. The more biased bankers had lower performance. In a companion experiment with students, hindsight bias led people to underestimate volatility: if every past move "made sense", the range of future moves seems narrower than it is. The link to overconfidence and overtrading is direct.
It ruins learning. Experience teaches only if you compare what you expected with what happened. Hindsight edits the first half of that comparison. An investor who was unsure in March and remembers being confident in December learns that her instincts are good, when the real lesson was that she got an uncertain call right.
It makes judgments of others unfair. A decision that turned out badly looks, afterwards, like one that anyone could have seen was wrong. That is how a fund manager, an adviser or a spouse gets blamed for a reasonable decision. Judging the decision by its result is a related error with its own lesson: process over outcome.
It makes backtests and charts look like forecasts. Reading a historical chart with the ending visible is hindsight by construction. Trading rules designed while looking at the full history fit turning points that were not identifiable in real time; see overfitting and backtest bias.
A worked example in an account
An illustrative investor, Lee, holds $50,000 in an index fund through a 20% decline. Near the low Lee thinks "this could get much worse or it could turn; I have no idea", and does nothing. The fund recovers and a year later is 15% above its old high.
Two years on, Lee's memory of the episode is: "I knew it was a buying opportunity. I should have put the other $20,000 in." Acting on that remembered confidence, Lee meets the next 12% decline by moving the whole $20,000 cash reserve into the fund at once, certain this is the same thing again. This time the decline continues to 35%.
On $20,000 the further fall costs about $5,200 at the low (from 88 to 65 on a scale where the old high is 100, a 26% drop from the purchase price). The loss on paper may well be recovered. The point is where the decision came from: not from anything Lee knew, but from the memory of having known something the first time. Lee's own note from the first decline, "I have no idea", would have prevented it.
What helps, and what does not
Warnings do not do much. In the 2004 meta-analysis, manipulations designed to reduce hindsight bias did not, on average, lower it. Knowing the name of the bias does not protect you, and reading this page will not either.
Considering the opposite helps somewhat. Paul Slovic and Baruch Fischhoff found that asking people to explain how the result could have come out the other way reduced the bias, without eliminating it. For an investor that means asking, of any past event: what would the story have been if the market had fallen another 30% from here, and what facts of the time would I now be citing as the obvious warning?
A written record sidesteps the problem. Memory is the thing that fails, so the reliable fix is to stop depending on it. A dated note of what you expected, how sure you were and why, written before the outcome, cannot be revised by later knowledge. Investment decision journal shows a format that takes two minutes per decision.
Practice on hidden charts. Covering the right-hand side of a chart and committing to a guess before revealing it restores a little of the uncertainty that the finished chart removes. It is a demonstration, not training: nobody has shown it improves returns.
How Finabase helps
The chart can be set to any range, so you can look at a past period first and widen the range only after writing down what you would have expected. Each symbol page has a Notes tab where a signed-in user can keep a note on that symbol: a view written there today is the foresight record that hindsight will later try to replace. The News tab shows what was being reported as it happened, a useful reminder that the coverage on any given day pointed in several directions at once.
Frequently asked questions
What is hindsight bias in simple terms?
It is the feeling, after something has happened, that it was predictable all along. Knowing the result changes both how likely it seems and what you remember having thought beforehand.
What is an example of hindsight bias in investing?
Looking at a chart of the 2008–2009 decline and thinking the March 2009 low was an obvious time to buy. At the time, the market had fallen in 12 of the previous 16 months and an earlier rebound had failed.
How is hindsight bias different from outcome bias?
Hindsight bias distorts what you think you knew: the outcome seems to have been predictable. Outcome bias distorts how you grade a decision: a good result makes the decision look good. They often occur together, and both are countered by records made before the result.
Can you get rid of hindsight bias?
Not by willpower or awareness. Studies find warnings have little effect. Explaining how the opposite outcome could have occurred reduces it somewhat, and a written forecast made in advance replaces the faulty memory with a document.
Summary
Hindsight bias makes known outcomes seem predictable and rewrites the memory of what you expected. Fischhoff's experiments showed both effects in the 1970s, 95 later studies confirm them, and among investment bankers more hindsight bias went with worse performance. A chart shows the bias at work: SPY's February 2009 month-end is an obvious low on the finished chart and was, on the day, the twelfth down month in sixteen after one failed recovery already. Because the bias corrupts what investors learn from their own history, and because warnings do little, the dependable remedy is a dated record of forecasts made before the results are known.
Check your understanding
5 questions on this lesson. Each answer comes with the reason, and nothing is scored.
In Fischhoff’s war experiment, groups were told different “true” outcomes and asked to judge as if they did not know. What happened?
- Each group rated the outcome it had been told as more likely than the uninformed group did
- All groups gave the same probabilities
- Informed groups rated their outcome as less likely, to compensate
Show the answer
A. Each group rated the outcome it had been told as more likely than the uninformed group did Outcome knowledge could not be set aside. Whatever ending people were given, the earlier facts rearranged themselves into a story leading to it.
An investor looks at a finished chart and says the 2009 low “was obviously the time to buy”. Which fact about that month-end contradicts “obviously”?
- The fund paid a dividend that month
- It had fallen in 12 of the previous 16 months, and a two-month rebound in 2008 had been followed by a further 47% fall
- Trading volume was low
Show the answer
B. It had fallen in 12 of the previous 16 months, and a two-month rebound in 2008 had been followed by a further 47% fall The low is visible only because the later recovery is on the chart. On the day, nothing separated it from earlier points that had also looked like a bottom.
Why does hindsight bias make it hard to learn from your own investing experience?
- It makes you forget which stocks you owned
- It changes your memory of what you expected, so the comparison between forecast and result is made against a forecast you never held
- It only affects professionals
Show the answer
B. It changes your memory of what you expected, so the comparison between forecast and result is made against a forecast you never held Learning needs the original expectation. If memory moves it toward the outcome, every result seems to confirm your judgment.
According to the research, which is most likely to reduce hindsight bias?
- Being told that hindsight bias exists
- Trying harder to remember accurately
- Explaining how the opposite outcome could have happened, and keeping forecasts written before the outcome
Show the answer
C. Explaining how the opposite outcome could have happened, and keeping forecasts written before the outcome Warnings had little effect in the meta-analysis. Considering the opposite reduced the bias in experiments, and a written record does not depend on memory at all.
After a decline and recovery, an investor remembers having been “sure it was a buying opportunity”, though at the time she was undecided. What is the likely consequence?
- She will take smaller risks next time
- She will meet the next decline with more confidence, and more money, than her real record justifies
- No consequence
Show the answer
B. She will meet the next decline with more confidence, and more money, than her real record justifies Remembered foresight inflates confidence. The study of investment bankers found that more hindsight bias went with lower performance.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open the SPY chart and choose the 1Y range. Cover the right half of the chart with your hand or a sheet of paper. Looking only at the left half, write down one sentence: where you would have expected the price to be at the right edge, and how confident you would have been from 50% to 100%. Then uncover it. Do the same once more on the 5Y range.
Open the SPY chartWhat you should find
Notice two things. First, how hard it is to un-know the right half once you have glimpsed it: that is the bias. Second, how wide the range of plausible endings looked from the left half alone. If your written confidence was above 70%, ask what you could actually see that justified it. Keep the note: it is your first journal entry.
Sources and further reading
- Fischhoff — Hindsight ≠ foresight: the effect of outcome knowledge on judgment under uncertainty, Journal of Experimental Psychology: Human Perception and Performance 1(3), 1975
- Fischhoff and Beyth — “I knew it would happen”: remembered probabilities of once-future things, Organizational Behavior and Human Performance 13(1), 1975
- Slovic and Fischhoff — On the psychology of experimental surprises, Journal of Experimental Psychology: Human Perception and Performance 3(4), 1977
- Guilbault, Bryant, Brockway and Posavac — A meta-analysis of research on hindsight bias, Basic and Applied Social Psychology 26(2–3), 2004
- Biais and Weber — Hindsight bias, risk perception, and investment performance, Management Science 55(6), 2009
- Roese and Vohs — Hindsight bias, Perspectives on Psychological Science 7(5), 2012
- Finabase chart data — SPDR S&P 500 ETF (SPY) monthly closes, January 2007 to December 2013
