Hiroo Onoda was sent to Lubang in the Philippines in December 1944 with instructions to keep fighting and not to surrender. Japan capitulated in August 1945. Onoda stayed in the jungle until March 1974, when his former commanding officer travelled to the island and formally relieved him of duty. That is nearly 29 years.
The popular reading is that he was a loyal soldier who followed an outdated order. The more useful reading is that he was a reasoner who could not be corrected. Leaflets, newspapers and letters from his family all arrived as evidence that the war was over, and he classified each as enemy propaganda. His framework absorbed every contradiction and left the framework itself untouched. Psychologists call this belief perseverance. In a 1979 study by Lord, Ross and Lepper, participants who read the same mixed evidence on capital punishment each left more convinced of their original position.
Investors are not under orders, but they do carry rules. The question is how those rules were written, and whether anything can revise them.
How markets write personal rules
In their 2011 paper “Depression Babies”, Ulrike Malmendier and Stefan Nagel used US Survey of Consumer Finances data from 1960 to 2007. They found that the stock returns people personally lived through predicted their later behaviour. Those who experienced weak returns were less likely to own equities and, if they did, held a smaller share. They also expected lower future returns. Those who experienced strong returns did the opposite. Recent experience carried more weight, but early experience left a measurable imprint decades later.
A related finding comes from Greenwood and Nagel (2009). Fund managers who were young during the dot-com bubble were more likely to buy technology stocks at the peak than older managers. Professional training did not neutralise the cohort effect.
Two points follow. First, the effect is about forecasts, not just feelings. Investors who lived through different markets hold different return expectations, and they hold them with confidence. Second, the effect is persistent, which means it does not reliably fade as data accumulates.
Two Indian cohorts, same market
Take an investor who began in 2007. The Sensex peaked near 21,200 in January 2008 and touched roughly 7,700 in October, a fall of about 64 percent. The lesson from that: equities can lose most of their value, and cheap can get cheaper.
Take another who began in April 2020. The Nifty 50 had fallen about 38 percent from its January high to its March low near 7,600. It then rose to roughly 18,600 by October 2021, about 2.4 times the low. The lesson from that: drawdowns are gifts, and recovery is fast.
Here is the part the original article missed. The 2008 investor’s lesson was also contradicted by what followed. The Sensex rose from its October 2008 low to roughly 17,500 by December 2009, more than doubling in about 14 months. Anyone who stayed invested, or bought through the panic, was rewarded. The market itself had already shown that “stocks are dangerous, stay out” was an incomplete rule, even for the cohort that formed it. The rebound was not unique to 2020. What differed was the investor’s position when it arrived.
So the two lessons are not equally supported. The 2008 rule fails on its own cohort’s evidence. The 2020 rule rests on a single episode in which a fast recovery was helped by unusually large liquidity support, which is not a repeatable condition.
The mechanisms doing the damage
Several well-documented biases explain why these lessons harden.
Availability. Tversky and Kahneman (1973) showed that people judge probability by how easily examples come to mind. A crash you lived through is vivid and personal, so it feels more likely than base rates imply.
Recency weighting. Malmendier and Nagel found that recent returns carry more weight in expectations. A cohort whose first years were a rally extrapolates the rally.
Loss aversion. In Kahneman and Tversky’s work, losses weigh roughly twice as heavily as equivalent gains. A drawdown suffered early becomes the reference point for every later decision, which pushes investors toward holding cash even when expected returns are positive.
Confirmation and perseverance. Once a rule exists, evidence is filtered through it. Onoda read the leaflets as propaganda. An investor who believes every dip is a buying opportunity reads a falling price as validation, and one who believes equities are unsafe reads a rally as a trap.
Overconfidence from early success. Barber and Odean (2000) found that households that traded most earned about 11.4 percent annually from 1991 to 1996, against 17.9 percent for the market. A cohort that learned in a rising market is more likely to trade heavily and to add leverage.
What each wrong rule costs
The costs are asymmetric, and they show up differently.
The 2008 rule costs money quietly. Cash and low yield deposits compound at lower rates for decades, and the loss never appears as a statement line. It appears as a retirement corpus that falls short.
The 2020 rule costs money loudly. SEBI’s 2024 study of individual derivatives traders found that roughly nine in ten lost money across FY22 to FY24, with aggregate losses of about ₹1.8 lakh crore. Many of those traders entered after 2020, in a market that had trained them to expect fast recoveries. Leverage turns a drawdown that a long term holder survives into one that forces a sale.
Updating, with a procedure
Onoda needed a trusted authority to end his mission. Investors have no such authority, so the correction has to be designed in advance. Four checks help.
- Write the rule and its origin. “I avoid small caps because of 2008” is testable. “I’m cautious” is not.
- Separate principles from period responses. Diversification, leverage limits and matching assets to time horizons hold across regimes. “Never buy equities” and “always buy the dip” are responses to one regime.
- Test against the full sample. Ask what the rule would have produced across every major drawdown since 1990, not just the one you remember. If the answer is mixed, the rule is a hypothesis.
- Set position limits that do not depend on the rule being right. Caps on leverage and concentration protect both cohorts from the cost of being wrong.
The aim is not to discard experience. It is to treat each lesson as a hypothesis with a sample size of one, and to ask what evidence would change it. Onoda could not answer that question. Investors still can.
