Cognitive errors vs emotional biases

The eleven biases are the ones that cost investors the most money. They aren't the whole standard taxonomy, though, and the full taxonomy is worth having for two reasons. It's what an examination and an institutional adviser will use. And its top-level split answers a question the catalog leaves open. Can a given bias be corrected at all, or only accommodated?

What is the difference between cognitive and emotional biases?

The split is between cognitive errors and emotional biases. A cognitive error is a fault in reasoning, in information processing, in memory, or in the statistics a person is using. It survives because the person hasn't noticed it. So it yields to information, to training, to a written process and to a structure that forces the missing step. An emotional bias comes from feeling, impulse and intuition, with no faulty reasoning behind it. It survives despite the person knowing about it, so it usually has to be moderated or built around. You can't argue it away.

One diagnostic question sorts almost any bias into its category. Would showing this person the correct calculation, clearly, change what they do? If yes, the bias is cognitive. If the person can state the correct answer and still not act on it, the bias is emotional. Loss aversion is the standard demonstration. Anyone who has read the prospect-theory section can compute that the disposition effect costs money, and a great many of them still ride losers.

Map the catalog onto the split. Don't replace it, because the eleven remain the ones that matter most in practice.

Notice what follows. Six of the eleven are cognitive, and the answer to them is process. A checklist, a journal, base rates and a written thesis all supply the missing step. The four emotional ones are why the behavioral edge carries two other instruments. Pre-commitment and environment design don't correct a feeling. They remove the moment in which the feeling gets to act.

The eight the catalog leaves out

Each of these appears in the standard taxonomy, and each gets the same four-part treatment the catalog uses. The first three are cognitive. The last five are emotional.

Conservatism is a cognitive error. Mechanism: having formed a view, we update too little when new evidence arrives, and least of all when the evidence is abstract, statistical or effortful to process. The prior sticks and the posterior barely moves. Think of it as anchoring applied to beliefs instead of numbers. It's what Bayes's rule describes people failing to do. Investing failure mode: holding a thesis through three consecutive results that contradict it, calling each one an exception. Under-reaction to a real change in competitive position is one documented driver of post-earnings-announcement drift.

Real example: an analyst models a slow margin recovery for six quarters. The entrant's pricing had already reset the industry, and the forecast gets trimmed a little each quarter when it should have been rebuilt once. Counter-practice: write the update rule before the evidence arrives. "If same-store growth is below 4% for two consecutive quarters, I rebuild the model from scratch" is a pre-registered trigger. It turns an update into a rule, so you're no longer fighting your own prior.

What are representativeness and illusion of control?

Representativeness is a cognitive error. Mechanism: we judge probability by similarity to a stereotype and neglect the base rate. A company that looks like a compounder gets a compounder's odds, however rare compounders are. Two sub-forms are worth naming. Base-rate neglect ignores the underlying frequency. The sample-size fallacy treats a short record as representative. Investing failure mode: paying a quality multiple for a business that resembles a franchise but has no tested moat. Or treating three years of outperformance as evidence of skill, when three years can't separate skill from luck.

Real example: every cycle produces a cohort of newly listed companies that share the surface features of the last cycle's winners, and they're priced accordingly. The transition matrices from the returns-on-capital work give the base rate that resemblance is worth. Counter-practice: state the base rate first, in writing. Then ask what specific evidence justifies departing from it. "What fraction of companies with this profile still earn above their cost of capital in ten years?" is the question representativeness skips.

Illusion of control is a cognitive error. Mechanism: we overestimate our ability to influence outcomes that are substantially or wholly outside our control, and the illusion grows with involvement, effort and familiarity. Building the model makes the future feel more controllable, which it isn't. Investing failure mode: overtrading, because acting feels like controlling. Position sizes justified by the depth of the work instead of the range of the outcomes. Over-precise forecasts of things nobody forecasts.

Real example: the investor who, after a really thorough deep dive, sizes a position at 25% because the work felt conclusive. The work reduced uncertainty about the business, not about the price or the world. Counter-practice: before sizing, write down what you control and what you don't. You control the price you pay, the size, the holding period and your process. You don't control the outcome. The position-sizing rules are a structural version of that separation, because they tie size to the range of outcomes and leave your confidence out of it.

What are mental accounting and framing?

Mental accounting is an emotional bias. Mechanism: money gets sorted into separate mental pots with different rules attached, even though money is fungible. Thaler's formulation covers separating income by source, wealth by purpose, and gains by how they were earned. Investing failure mode: treating "house money" from a winning position as free to gamble with. Spending dividend income while refusing to sell an identical amount of an appreciated holding, which is the same rupee with a different label. Running a "safe" bucket and a "speculative" bucket with no single view of total risk. Carrying expensive debt alongside low-yielding savings.

Real example: the arithmetic at the end of the catalog, where a 9% loan held next to a 4.55% after-tax debt fund costs 4.45% a year on the overlap. Nobody accepts that once the two pots are shown on one page. Counter-practice: one balance sheet, one risk view, one optimisation. Then, where the buckets serve a real behavioural purpose, keep them deliberately and price what they cost. The goals-based allocation below does just that.

Framing is an emotional bias. Mechanism: the same information, described differently, produces different decisions. A fund with a "90% chance of meeting the goal" and one with a "10% chance of missing it" attract different money. Investing failure mode: accepting the frame the seller chose, which is most of what a factsheet, a pitch deck and a headline are for. Reading "down 40% from the high" as cheapness, which is anchoring dressed up as a frame. Choosing a risk level from how a question was worded instead of from the underlying exposure.

Real example: take one annual return and present it as a peer-group rank, as an absolute number, and as a gap against a benchmark. It supports three different conclusions, and every marketing document chooses among them. Counter-practice: restate every important decision in at least two frames before deciding, and include the frame the person presenting it didn't use. Absolute rupees and percentages, gains and losses, distributions rather than point estimates.

What are self-control and status quo bias?

Self-control is an emotional bias. Mechanism: we discount the future steeply and take the immediate reward against our own stated long-run preference, then regret it. You know the right action. You just can't take it now. Investing failure mode: under-saving relative to a plan you agree with. Spending capital instead of income. Abandoning a systematic investment plan in the month it matters most. Reaching for yield to fund current spending instead of selling appreciated assets, which tilts a whole book toward credit risk.

Real example: an investor's written plan calls for a fixed monthly contribution. The contributions stop for the eight months when equity prices are lowest. Counter-practice: automate the decision away from the moment. A standing instruction, a transfer on salary day and a spending rule that names a percentage instead of an amount all move the decision from the weak moment to the strong one. It's the same pre-commitment you'll build for trading, applied to saving.

Status quo is an emotional bias. Mechanism: doing nothing is the default, and defaults win. Inertia costs no effort and produces no felt responsibility for the outcome, so an allocation persists past every reason for holding it. It overlaps with the endowment effect and with regret aversion but is separable from both. The status-quo investor isn't attached to the holding and doesn't fear the loss. They simply never revisit it. Investing failure mode: an allocation that has drifted far from target because it was never rebalanced. A portfolio that still reflects a job, a market or a life stage that ended years ago. Cash left idle for years in a low-interest sweep.

Real example: the retirement account still in the fund selected on the day of joining, split across a default menu nobody has looked at since. Counter-practice: set a review date. A review intention isn't enough. Make rebalancing rule-based instead of discretionary, too. A calendar rebalance and a band rebalance both turn inaction into a decision that has to be justified.

What is regret aversion?

Regret aversion is an emotional bias. Mechanism: we avoid decisions whose bad outcomes would produce sharp regret, and regret from action is felt more sharply than regret from inaction. So we do less, and we prefer to be wrong alongside everyone else. Investing failure mode: refusing to sell a loser, because selling makes the loss real and attributable to you. Refusing to buy a good business after a fall, because buying it and being wrong would be unbearable. Herding into consensus positions, because a consensus loss carries less regret. Holding cash after a bad experience far longer than any analysis supports.

Real example: an investor doesn't buy in April 2020, because a further fall would then be their own fault. They miss the recovery too. Counter-practice: pre-commit in writing, and keep the decision separate from the outcome when you review it. A journal that records the reasoning at the time makes "good decision, bad outcome" a verdict you can trust. Only that verdict lowers anticipated regret enough to let you act next time.

Moderate or adapt: the rule that decides what to do with a bias

The taxonomy is only useful if it changes the action, and it does. The standard rule follows directly from the split.

Cognitive errors are moderated, meaning corrected. Supply the information, add the missing step to a written process, force the calculation, and the error shrinks. There's no reason to build a portfolio around an error the person will stop making once they see the arithmetic.

Emotional biases are usually accommodated. The portfolio and the process get built so the bias does less damage, because argument doesn't remove it. Accommodation isn't indulgence, though. It has a price, and the price has to be paid knowingly.

Two conditions decide how far to go, and an adviser applies both.

The first is wealth relative to the essential goals. A client whose portfolio comfortably exceeds the capital required to fund the goals that must be funded can afford accommodation, because a suboptimal allocation still meets the objective. A client whose portfolio barely covers those goals can't. There the adviser has to moderate the bias even where it's emotional, because the alternative is a plan that fails on its own terms. Put the test in numbers, because a feeling won't do. Take a portfolio of ₹1,200 lakh against essential annual spending of ₹36 lakh. That's a 3.0% spending rate. Against a rough 25-times benchmark of ₹900 lakh, it carries a ₹300 lakh surplus, and that surplus is the accommodation budget.

The second is the size of the bias's effect on the plan. A small deviation that keeps a nervous client invested is cheap. A large one that materially lowers expected return is not, and the adviser has to be able to say what it costs.

When the two conditions conflict, the tie-breaker is that a plan the client will actually follow beats a better plan they'll abandon in a drawdown. That's a real argument. It's also the most abused sentence in the advisory business, so the cost has to be computed, never just asserted.

Behaviourally modified asset allocation, worked

A client has ₹400 lakh. A conventional risk-tolerance questionnaire and a mean-variance process both point to a 60/40 equity-debt split. But the client is a mental accountant, and forcing a single blended portfolio on them has twice produced a panic sale in past drawdowns. So the adviser builds the allocation goals-first, giving each goal its own account with its own horizon and its own risk.

Add the buckets and the aggregate allocation falls out. Equity comes to (90 × 0) + (210 × 0.60) + (100 × 0.90) = ₹216 lakh, which is 54% of the total, against 46% in debt. The client experiences three accounts with three purposes. The portfolio is a single 54/46 book.

Now price the accommodation, because that's the step advisers skip. Assume an expected 11% on equity and 6.5% on debt. The unmodified 60/40 expects 0.60 × 11 + 0.40 × 6.5 = 9.20% a year, and the modified 54/46 expects 0.54 × 11 + 0.46 × 6.5 = 8.93%. So the accommodation costs 0.27 percentage points a year. That sounds small. Compounded on ₹400 lakh for twenty years, it's 400 × 1.0920²⁰ = ₹2,325.48 lakh against 400 × 1.0893²⁰ = ₹2,213.15 lakh, a gap of ₹112.33 lakh, or 4.83% of terminal wealth.

Then set that against what it buys. Suppose the accommodation prevents one panic sale of the sort the client has made before. The 60% equity sleeve, ₹240 lakh, gets sold at ₹156 lakh after a 35% drawdown and bought back after a 25% recovery. The shortfall against simply holding is 195.00 − 156.00 = ₹39.00 lakh. So the accommodation pays for itself only if it prevents 112.33 ÷ 39.00 = 2.88 such episodes in twenty years, roughly one every seven years.

That number is the argument stated fairly, and it can come out either way. For a client who has panicked twice in a decade, it's an easy case. For a client who merely says they might, it isn't, and the adviser's job is to say so instead of selling the comfortable structure. Read the whole exercise as the moderate-or-adapt rule turned into arithmetic. The accommodation is a real cost, paid to buy adherence. It's defensible when, and only when, the behaviour it prevents is more expensive than the return it gives up.