Disposition effect

Picture an investor with two holdings in a general investment account. One has risen 40% since purchase. The other has fallen 30%. No cash is needed. The investor still sells the winner and keeps the loser, with a quiet plan to sell once it “gets back to what I paid”.

Nothing in the numbers says the winner should go first. The purchase price is history and the market does not know it. What matters is each holding’s prospects from here. The tax position, if anything, points the other way: gains above the annual exempt amount attract capital gains tax, while a realised loss can be set against gains. Yet the pattern is common enough to have a name.

The disposition effect is the tendency of investors to sell assets that have gained in value too soon, and to hold assets that have lost value too long.

What the research shows

Shefrin and Statman (1985) gave the effect its name. They drew on prospect theory, mental accounting, regret aversion and self-control to explain why investors sell winners too early and ride losers too long. They also argued that tax considerations alone cannot account for the pattern.

The best-known empirical test is Odean (1998). He analysed trading records for 10,000 accounts at a large US discount brokerage, covering 1987 to 1993. His method compared how often investors sold winners, relative to the winners they could have sold, with the same ratio for losers. On days when an investor sold something, they realised about 14.8% of their available paper gains and about 9.8% of their available paper losses. That is roughly a 50% higher propensity to sell winners.

Odean then tested the innocent explanations. The pattern did not appear to be driven by portfolio rebalancing or by the higher trading costs of low-priced stocks. Nor was it justified by what followed: the winners investors sold went on to outperform the losers they kept by about 3.4 percentage points over the next year. Odean concluded that for taxable accounts the behaviour is suboptimal.

The finding has travelled. Grinblatt and Keloharju (2001) used daily buy, sell and hold data for individuals and institutions in the Finnish stock market over a two-year period. They found evidence that investors are reluctant to realise losses, and that they also engage in tax-loss selling. The two behaviours coexist, which is part of why the tax explanation is incomplete.

The effect also appears in the laboratory, where purchase prices and information can be controlled. In Weber and Camerer’s (1998) experiment, participants traded artificial stocks over a series of rounds. They sold risen stocks more readily than fallen ones.

Where the evidence is contested

The behaviour is well documented. Its cause is less settled.

The textbook explanation uses prospect theory. Kaustia (2010) argues that prospect theory is unlikely to explain the disposition effect. Prospect theory predicts that the propensity to sell falls as the price moves away from the purchase price in either direction. In trading data, the propensity to sell jumps at zero return, but is roughly constant over a wide range of losses. The purchase price appears to act as a threshold in its own right.

Barberis and Xiong (2009) reach a related conclusion by modelling. They tested two versions of prospect theory. One defines preferences over annual gains and losses, and often fails to predict a disposition effect. The other defines them over realised gains and losses, and predicts the effect more reliably. In other words, the result depends on whether investors feel the gain or loss when the price moves or only when they sell. That is a modelling argument, not direct proof of what investors feel, so it is better read as a hypothesis than a settled mechanism.

How it shows up in portfolios

• Selling the winner to bank the gain. A holding crosses a round number or a previous high, and the sale feels like a profit secured. The proceeds may then sit in cash or move into something with weaker prospects than the holding sold.

• Waiting to break even. A loser stays because selling would make the loss “real”, and the stated plan is to sell once the price recovers to cost. Kaustia’s finding that the propensity to sell jumps at zero return is consistent with the purchase price acting as a finish line.

• Treating each position as its own account. Each holding is judged by its gain or loss against cost, not by how the whole portfolio sits against its goals. The position-by-position view tends to crowd out the portfolio-level one.

• Giving cost basis weight where it has no tax meaning. Inside an ISA or SIPP, gains and losses carry no capital gains tax consequence, so the purchase price has no tax significance there. In a general investment account it does matter, and Odean’s finding is that the usual pattern runs against tax efficiency. In either setting, the cost basis is a poor guide to what a holding is likely to do next.

• Letting tax-loss selling mask the pattern. Realising losses before the tax year end can look like discipline. Grinblatt and Keloharju’s results suggest tax-loss selling and loss reluctance can occur together, so a burst of loss-selling does not show the bias is absent the rest of the year.

Why it happens

Kahneman and Tversky (1979) provide the starting point. Prospect theory assigns value to gains and losses relative to a reference point, not to final wealth. The value function is normally concave for gains, commonly convex for losses, and steeper for losses than for gains. This contributes to risk aversion in choices involving sure gains and to risk seeking in choices involving sure losses. For an investor, the purchase price is a natural reference point. Selling a winner locks in a sure gain. Selling a loser makes the loss certain, while holding keeps a chance of recovery alive.

Mental accounting adds a bookkeeping habit. Each purchase opens a mental account that feels unfinished until it is closed. Closing it at a loss means admitting the decision did not work, which most people avoid.

Regret and pride supply the emotional payoff. Selling a winner earns a small sense of having been right. Selling a loser triggers regret. Because the feeling attaches to the sale and not to the price move, holding feels like deferring the verdict. Barberis and Xiong’s realisation-utility framing captures this idea: the pleasure and pain arrive when a position is closed.

Researchers still debate how much each ingredient contributes. The evidence supports the pattern more firmly than any single explanation of it.

Ways to counter it

None of these removes the bias. They aim to make it harder for the bias to decide alone, and the evidence that any one of them works for individual investors is thin.

• One approach is to ask a forward-looking question before a sale: would this holding be bought today, at today’s price, ignoring what was paid? If the answer is no, the purchase price may be doing the work that the outlook should be doing.

• Some investors write sell rules in advance, such as a maximum portfolio weight or a review triggered when the original reasons for buying no longer hold. Rules set when no position is under pressure may reduce the scope for in-the-moment emotion. Rules built on weights and fundamentals avoid the purchase price altogether.

• Reviewing total allocation against goals, rather than position by position, may weaken the pull of individual mental accounts.

• A decision journal records why a holding was bought. Later, it separates the question “do those reasons still hold?” from “is the price above or below cost?”

• Investors who know that gains and losses inside an ISA or SIPP carry no tax consequence can remove one common rationalisation for holding or selling. In a general investment account, the capital gains tax position is a legitimate input, but it is a different input from the purchase price.

• Odean’s measure can be run on a spreadsheet. Listing a year of sales, split into gains and losses, and comparing them with the gains and losses that were available but not sold, shows whether the pattern exists in a particular portfolio.

• Automating contributions and rebalancing to fixed weights removes some discretionary decisions. By construction, that leaves fewer occasions for the bias to operate.

Related biases

• Loss aversion: losses weigh more heavily than equivalent gains, which makes closing a losing position feel worse than closing a winning one feels good.
• Mental accounting: treating each holding as a separate account is one of the mechanisms Shefrin and Statman proposed.
• Sunk cost fallacy: past spending, such as the purchase price, is allowed to influence a decision that should look only forward.
• Regret aversion: avoiding the feeling of having been wrong pushes investors to defer a sale of a loser, and to sell winners early for the satisfaction of having been right.