In this lesson · 8 sections
Two investors hold the same stock at CHF 100. One bought at CHF 60, the other at CHF 140. The company is identical, the outlook is identical, the price is identical. Ask each whether to sell and you will reliably get different answers — and the difference has nothing to do with the company.
That is the whole subject of this lesson. Standard finance says only the future matters: what you paid is a sunk fact, and a rational holder evaluates the position against its expected return from here. Real investors do not work that way, and the deviation is systematic enough to have a name, a measured size, and a predictable cost.
Reference dependence: the mechanism underneath
Prospect theory, developed by Daniel Kahneman and Amos Tversky, replaces the standard assumption that people evaluate states of wealth with the observation that they evaluate changes relative to a reference point.
Three properties of that value function do the work.
It is defined over gains and losses, not levels. CHF 100,000 does not feel like CHF 100,000. It feels like +CHF 20,000 or −CHF 40,000, depending on where you started.
It is steeper for losses than for gains. A loss of a given size hurts more than an equivalent gain pleases. In Tversky and Kahneman's later cumulative formulation, the estimated loss aversion coefficient was around 2.25 — meaning a CHF 1,000 loss carried roughly the emotional weight of a CHF 2,250 gain. Later work has found values commonly in the range of about 1.5 to 2.5 depending on domain and method. The exact number varies; the asymmetry does not.
It is concave in gains and convex in losses. This is the property most people skip, and it is the one that actually produces bad selling. Concavity in the gain region means diminishing pleasure from further gains, which makes people risk-averse when ahead — they take the certain gain rather than gamble for more. Convexity in the loss region means diminishing additional pain from further losses, which makes people risk-seeking when behind — they prefer a gamble that might get them back to even over a certain, smaller loss.
Put those together with a reference point set at the purchase price and the prediction is immediate: investors sell winners too early and hold losers too long.
The disposition effect, and what it costs
Hersh Shefrin and Meir Statman named this the disposition effect in 1985. Terrance Odean tested it on a large sample of individual brokerage accounts and found what the theory predicts: investors realised gains at a substantially higher rate than losses, holding losing positions far longer than winning ones. He also examined what happened next — and the positions investors sold went on, on average, to outperform the positions they kept. The behaviour was not merely psychologically costly. It was financially costly.
The effect has since been documented across retail investors, professional fund managers, real estate sellers and even experienced traders, in many countries. It weakens with experience and sophistication but rarely disappears.
Notice what the reference point does. It is arbitrary — the purchase price is a historical accident of when you happened to place an order — but it partitions an otherwise continuous decision into two psychologically different regimes. Above it, you are in the concave region and want certainty. Below it, you are in the convex region and want the gamble.
Two extensions make the effect stronger in practice. Realisation utility, described formally by Nicholas Barberis and Wei Xiong, is the observation that the pleasure or pain is felt at the moment the trade is executed, not while the position drifts. An unrealised loss is, emotionally, not yet a loss — so not selling is a way of not having lost. Mental accounting, in Richard Thaler's sense, means each position is tracked in its own ledger rather than netted at portfolio level, so the decision is made stock by stock instead of on the portfolio as a whole.
Myopic loss aversion: why checking often makes you hold less risk
Loss aversion interacts with how often you look, and the arithmetic here is worth doing explicitly.
Assume, purely for illustration, a diversified equity portfolio with an expected return of 10% per year and volatility of 20% per year, with returns approximately normal.
Over one year, the expected return is half a standard deviation above zero, so the probability of a positive year is about 69%.
Over one month, the expected return is 0.83% and the standard deviation is 20/√12 ≈ 5.77%. The probability of a positive month is about 56%.
Over one day, the expected return is roughly 0.04% and the standard deviation is 20/√252 ≈ 1.26%. The probability of a positive day is about 51%.
Now apply a loss aversion coefficient of 2 — losses weighted twice as heavily as gains. The emotional balance of an observation is favourable only when the probability of a gain exceeds λ/(1 + λ), which for λ = 2 is two-thirds.
Checking daily: 51% pleasant, 49% unpleasant at double weight. Net emotional experience is clearly negative.
Checking monthly: 56% versus 44% at double weight. Still negative.
Checking annually: 69% versus 31% at double weight — 0.69 against 0.62. Just positive.
The same portfolio, with the same distribution of returns, is emotionally intolerable at daily frequency and tolerable at annual frequency. Nothing about the asset changed. Shlomo Benartzi and Richard Thaler used this argument — myopic loss aversion — to explain why investors demand such a large premium to hold equities at all, and Thaler and colleagues confirmed experimentally that subjects shown returns more frequently allocate less to the risky asset.
The practical implication is not "ignore your portfolio". It is that evaluation frequency is a decision variable you control, and setting it too high converts a good long-horizon strategy into one you will abandon.
Where this bites specifically in Switzerland
The wealth tax makes one evaluation event compulsory. The section above treated evaluation frequency as something you control. In Switzerland one observation per year is imposed on you by statute, and it is worth understanding what kind of observation it is.
There is no federal wealth tax, but the cantons and communes levy one on worldwide net assets, with rates and allowances that differ substantially from one place to another. To assess it, a Swiss resident files a securities and credit balances schedule — the Wertschriftenverzeichnis — with the annual tax return, listing every holding at its taxable value as at 31 December; the Federal Tax Administration publishes a price list from which those values are taken. Once a year, therefore, you sit down and write out every position you own at a fresh mark.
Two things follow. The first is mildly encouraging: the mandated cadence is annual, which is roughly the frequency at which Benartzi and Thaler found a diversified equity portfolio becomes psychologically tolerable. The statutory rhythm is not the one that does the damage — the brokerage app checked at breakfast is.
The second is that this event is deliberately indifferent to your reference point. The wealth tax is charged on the level of your assets, not on the change in them, so it asks what a holding is worth and never asks what you paid. A winner and a loser of equal current value are taxed identically. That is the opposite of a jurisdiction with capital gains tax and loss offsets, where tax-loss harvesting supplies a real fiscal reason to realise losses and push directly against the disposition effect. Swiss private investors generally have no such counterweight, so whatever bias exists in a Swiss private portfolio shows up undiluted, and the discipline has to be supplied by process rather than by the tax code.
There is an uncomfortable corollary. Because the base is the value held, a bad year mechanically lowers next year's bill. Nobody has ever experienced that as compensation, which is a small demonstration of the asymmetry this whole lesson is about: the loss is felt at full weight and the offsetting benefit is not felt at all.
Currency creates phantom reference points. A Swiss investor holding foreign assets sees a CHF-translated number. Suppose a US holding rises 8% in dollars while the dollar falls 10% against the franc. The CHF return is 1.08 × 0.90 − 1 = −2.8%. The asset did well; the account shows red.
Loss aversion responds to the red, not to the decomposition. Over the decades of franc strength, Swiss investors have repeatedly experienced foreign holdings that performed adequately in local currency while displaying CHF losses — and have sold them for that reason. This is one of the mechanisms behind Swiss home bias: unhedged foreign exposure produces more frequent apparent losses in the reporting currency, which loss-averse investors then avoid. Whether to hedge is a separate question with real costs on both sides; the point here is that the currency layer manufactures loss experiences that trigger the bias independently of investment merit.
Being fair to loss aversion
Two honest caveats.
First, cutting losses is not always a bias. Where there is leverage, a path to ruin, or a genuine thesis break — the reason you bought is now known to be false — selling at a loss is correct risk management, not weakness. The bias is specifically the tendency to hold because of the loss, using the purchase price rather than the current outlook as the criterion. The diagnostic question is not "am I down?" but "at today's price, with what I now know, would I open this position?" If the answer is no, the only thing keeping you in it is the reference point.
Second, the empirical robustness of loss aversion as a universal law has been challenged — David Gal and Derek Rucker argued that many demonstrations are better explained by inertia and status quo preference than by an asymmetric value function. This does not rescue anyone's portfolio: whether the mechanism is asymmetric pain or plain inaction, the observed behaviour of holding losers and selling winners is well documented. But intellectual honesty requires saying that the size and universality of the coefficient are contested.
Countermeasures that actually work
The bias does not respond to knowing about it. It responds to structure that removes the decision from the moment.
Write the exit condition when you open the position, in terms of the business or the thesis rather than the price. Rebalance on a calendar or a band rule, which mechanically sells winners and buys losers and never asks how you feel. Evaluate at portfolio level rather than position level, which is what mental accounting resists. Reduce evaluation frequency to something matched to your horizon. And when you are tempted to hold a loser, force the reframe: if you would not buy it today, holding it is the same decision as buying it.
What to take away
- Loss aversion works through a reference point, usually the purchase price: above it you become risk-averse and take gains early, below it you become risk-seeking and hold losses hoping to get back to even.
- The disposition effect is measurable and costly — investors realise gains at a much higher rate than losses, and the positions they sell have tended to outperform the ones they keep.
- Evaluation frequency is a lever you control: with losses weighted twice as heavily as gains, a portfolio needs roughly a two-thirds chance of a gain per observation to feel tolerable, which a diversified equity portfolio achieves at annual frequency and not at monthly or daily.
- The Swiss cantonal wealth tax imposes one compulsory annual mark on every position through the Wertschriftenverzeichnis, and it taxes the level of your wealth rather than the change in it — so it is blind to your purchase price and offers no fiscal counterweight to the bias, while franc strength separately manufactures CHF-denominated losses on foreign assets that performed well in local currency.
- The operative test on any losing position is whether you would open it today at today's price with today's information; if not, the purchase price is the only thing holding you in.
Sources
- Daniel Kahneman and Amos Tversky, "Prospect Theory: An Analysis of Decision under Risk", Econometrica, 1979.
- Amos Tversky and Daniel Kahneman, "Advances in Prospect Theory: Cumulative Representation of Uncertainty", Journal of Risk and Uncertainty, 1992 — source of the commonly cited loss aversion coefficient near 2.25.
- Hersh Shefrin and Meir Statman, "The Disposition to Sell Winners Too Early and Ride Losers Too Long", Journal of Finance, 1985.
- Terrance Odean, "Are Investors Reluctant to Realize Their Losses?", Journal of Finance, 1998.
- Nicholas Barberis and Wei Xiong, "Realization Utility", Journal of Financial Economics, 2012.
- Shlomo Benartzi and Richard Thaler, "Myopic Loss Aversion and the Equity Premium Puzzle", Quarterly Journal of Economics, 1995; and Thaler, Tversky, Kahneman and Schwartz, "The Effect of Myopia and Loss Aversion on Risk Taking: An Experimental Test", Quarterly Journal of Economics, 1997.
- David Gal and Derek Rucker, "The Loss of Loss Aversion: Will It Loom Larger Than Its Gain?", Journal of Consumer Psychology, 2018 — the principal critique.
- Swiss Federal Tax Administration, guidance on the Wertschriftenverzeichnis and the annual securities price list (Kursliste) used to set taxable values for cantonal and communal wealth tax.