The Affect Heuristic: Feeling as a Substitute for Analysis

The affect heuristic answers a hard question about risk and reward with an easy one about how something feels. One feeling supplies both answers, so a liked holding looks better and safer at the same time. The facts rarely arrange themselves that way. Fear and greed are ordinary names for that single feeling running in each direction.

The clearest starting point is well away from any decision about money. Two glasses of water are put in front of a person. One came from a tap in a kitchen that person has eaten in for years; the other came from a tap in a corridor never walked down. Two separate questions are then put about the corridor glass. How good is this likely to taste? How likely is it to cause illness? Both answers were already waiting, they arrived together rather than one at a time, and nothing that could be called a measurement took place in between. The single arrival of both answers together is the whole of the affect heuristic. One feeling turned up first, and two different judgements were then read off it.

What does the affect heuristic actually substitute?

The shape of a heuristic is familiar from representativeness: a hard question gets quietly swapped for an easy one, the easy question is answered, and the answer is handed over as though it belonged to the hard question. The affect heuristicLetting how something feels answer questions about its risk and its reward. differs from the substitutions already covered in one respect. The heuristic does the swap for two hard questions at once, using a single easy one.

The two hard questions attached to any decision about money are these. How much might this gain? And how much might this lose? Answering either properly takes work. An answer needs something about the holding's business, its cost, what has to go right, and how often that sort of thing has gone right before. The easy question is one word long. How does it feel?

AffectThe quick good-or-bad feeling attached to something before any analysis of it. is the technical name for that quick good-or-bad reading. Affect is not an emotion in the everyday sense of joy or anger. The feeling is thinner than an emotion, faster than an emotion, and usually not noticed at all: a faint pleasantness or unpleasantness that attaches to a name, a logo, a phrase, a person on a screen, before anything has been worked out. Tversky and Kahneman set out the heuristics programme in Science in 1974, and the general move was theirs. The affect heuristic is that same move made with a feeling as the easy question, and with two hard questions replaced rather than one.

How are risk and reward arranged in the world itself?

Before the judged version can be surprising, the real version has to be on the table. In the world, risk and reward usually run together in the same direction. The things that pay more generally carry more that can go wrong, and the things that carry very little that can go wrong generally pay very little.

The pairing is felt without any finance at all. A scooter ridden faster gets a rider to work sooner and is more likely to hurt them. A stall in the busiest lane takes more money and pays a rent that will break its holder in a slow month. A cousin who offers a share of a new business is offering more than the post office would, and is also offering a real chance of getting nothing back. Nobody has to teach that pairing; a person has been living inside it since being old enough to be sent out for milk.

The pairing is a rough regularity rather than a law. A regularity can be broken in individual cases and still be the way to bet in general. There are badly arranged offers where a great deal can go wrong and very little is paid for bearing it. There are well arranged ones the other way. Out in the world there is no systematic supply of things that pay a lot and risk almost nothing. Anybody able to make one would rather keep it.

How are risk and reward related once they are judged by feeling?

Now the finding. When people are asked to rate a long list of things on two separate scales, one for how much benefit each brings and one for how much risk each carries, the two ratings come out running in opposite directions. The things rated as bringing more benefit are rated as carrying less risk. Finucane and others, writing in the Journal of Behavioral Decision Making in 2000, reported the measurement and went further than simply observing the pattern: when the information given to people was changed so that one of the two judgements moved, the other judgement moved as well, in the opposite direction, even though nothing supplied had said anything about it.

The second finding, that moving one judgement moves the other, is the part worth sitting with. If judged benefitHow rewarding something seems, as against what it has actually paid. and judged riskHow risky something seems, as against how risky it is. were two independent readings of two independent properties, telling somebody something flattering about the reward side would have left the risk side alone. It does not. Push one and the other slides. A slide in the untouched reading is the signature of a shared input rather than two separate assessments.

So the world arranges risk and reward positively and judgement arranges them inversely, and the affect heuristic lives in the gap between those two sentences. The inverse relationshipTwo things moving in opposite directions, so that when one reading goes up the other goes down. is not a small error of degree. The judged relationship has the sign wrong. A person judging by feeling is not merely mis-sizing the trade-off between risk and reward; they are, at the moment of deciding, not seeing a trade-off at all.

Why can one feeling not answer two questions without bending one?

The reason is structural rather than psychological, and it is worth taking slowly. Structure is why the effect is not simply a mistake that a careful person could avoid by concentrating harder.

Suppose two readings are produced from a single input. Whatever else is true, those two readings now carry information about the input and nothing else. If the input rises they both move; if it falls they both move. Because there is nothing for the two readings to disagree about, they cannot disagree. Their agreement is a property of the wiring, not a discovery about the thing being judged.

The reward answer and the risk answer sit in exactly that position under the affect heuristic. The feeling supplies both. So when they line up neatly, and a holding looks like a good deal of upside with very little that could go wrong, the neatness is not evidence. The two judgements were never in a position to contradict each other, so their agreement reports on the feeling and nothing whatever on the holding.

And the distortion is not fixed in size. The distortion scales with the strength of the feeling. A faint liking pulls the two readings slightly apart. A strong feeling pulls them a long way apart. The further apart they go, the more the picture in somebody's head describes a thing that pays well and risks little, and that is the one arrangement the world does not reliably supply.

Greed: what does the word name, and why is it a poor explanation?

Now to the word everybody reaches for. When a person puts a large amount into something that then falls a long way, the explanation offered afterwards, by the person and by everyone around them, is greedAn ordinary word for wanting a large gain, which describes a want rather than explaining a judgement.. Greed is a satisfying word. Greed is also, as an explanation, nearly empty, and it is worth being precise about why.

Greed names two things. The first is a want: somebody wanted a large gain. The second is an observation: they accepted a large risk in order to have a chance at it. Both of those are true descriptions of what happened. Neither is an explanation. Neither answers the only question that was ever interesting: why did the risk look acceptable at the time it was accepted?

A person who wants a large gain and correctly sees a large risk beside it is not being reckless if they take it; they are making a trade they can see. The case that needs explaining is the one where the trade was not visible. And here the affect account says something quite specific, and quite different from the word greed. The affect account does not say the risk looked acceptable relative to the reward. The account says the risk looked small. The same feeling that raised the reward reading lowered the risk reading in one movement, so at the moment of deciding there was no trade to weigh.

The affect claim is a different claim, and it comes with a different correction. The difference in correction is the practical reason to care. Telling somebody to be less greedy asks them to want less. Wanting less is not a thing anybody knows how to do on request, and it is a moral instruction wearing the coat of an analysis. The mechanism asks for something else entirely: put the risk question back on its own feet. Ask it separately, in different words, at a different sitting. Asked that way, it is not being answered off the same feeling that just answered the reward question. A change of that kind is a change to a procedure, and procedures can be changed.

Is fear a different mechanism, or the same one reversed?

The same one, with the sign flipped, and this is the cleanest test that the account is doing real work rather than restating a mood in longer words. If one feeling answers both questions, then a bad feeling should do to the two readings exactly what a good one does, only backwards. The disliked thing should look less rewarding and more dangerous at once, from one feeling rather than two judgements.

The reversal is what turns up. A lane somebody was once frightened on serves as the example, with the two questions put separately: how quick is that lane as a route, and how safe is it? Most people discover that the frightening lane has also become, in their head, the longer way round, even where a map says otherwise. Nothing was learned about distance. One feeling answered a question about danger and a question about convenience in the same movement.

Back to a decision. The Nilgiri mid-cap scheme in this worked case stood at Rs 2,55,000/- on 30 September against a cost of Rs 3,00,000/-, down 15.0 per cent. A fall of that size does something to how a holding feels quite apart from anything it reports. If the fall has attached an unpleasantness to the name, then on the next reading the scheme will look both more dangerous than before and less rewarding than before, and the person doing the reading will experience that as having thought about it. Fear and greed are therefore not two effects requiring two explanations; they are one substitution running in two directions. An account of greed is therefore also an account of fear.

How does a feeling get attached to a holding in the first place?

If the feeling is doing all this work, the interesting question becomes where it came from. There are three common routes, and the trouble is that the tag they deposit is identical whichever route delivered it.

The first route is one vivid encounter. A segment on a screen, a story from somebody at a wedding, one bad experience with a similar-sounding name years ago. One encounter takes seconds and it sticks. The second route is long repeated experience of the same situation, where the outcome was seen each time and fed back into the feeling. Repeated experience takes years and it also sticks. The third route is attachmentHow a feeling becomes fixed to a particular thing, often by one vivid encounter rather than by anything the thing did. borrowed from something standing next to the thing: a name that sounds established, an office that looks solid, a presenter who is calm and well dressed. The feeling attaches to what was in the room, not to what was being judged.

The Palash decision log, an invented record of 240 decisions taken by 60 investors over eight quarters, shows that 41 of the 96 buys followed a media mention within three days, being 42.7 per cent, against only about 11.0 per cent of the eligible list being mentioned at all in a given week. The gap between those two shares is usually read as an attention finding, and it is one. But it is also a supply figure for the mechanism described here: it counts how often a decision was taken shortly after an encounter that was well placed to leave a tag.

The outside route is precisely what makes the effect manipulable, and manipulable is a stronger word than biased. A bias that comes from inside a person can only be worked on from inside. An input that arrives from outside can be arranged by whoever controls the encounter, and anybody selling anything to anybody has always known it. The music under a segment, the order of the two numbers on a printed sheet, the confidence of the voice: none of those is information about risk or reward, and all of them move the feeling that is about to answer both questions.

What does the 19 February decision look like read this way?

Take the decision the log records for 19 February. A television segment named Suvarna Chemicals Limited, and the same evening Meera Sundaram added Rs 1,00,000/- to the position she already held. Its cost went from Rs 3,00,000/- to Rs 4,00,000/-, and the cost of everything she held went from Rs 12,00,000/- to Rs 13,00,000/-.

The attention account, set out under limited attention, explains one part of that evening well: it explains why Suvarna Chemicals Limited was in the running at all, out of everything she could have added to, and it explains why the decision came that evening rather than at any other time. The one thing attention does not explain is the size. Attention is a filter, and a filter reports what got through, never how much got committed.

The affect account reaches the size. Imagine putting the two hard questions to her separately that evening, in the plainest words available. How much might this gain? A good deal. How much might this lose? Not much. Take those two answers seriously for a moment and ask what sort of thing they jointly describe: something that pays well and puts very little at stake. Held together, the two answers describe a holding no market reliably supplies, and the reason they could be held together comfortably is that they were never two answers in the first place.

By 30 September, eight quarters in, the position stood at Rs 4,60,000/- against its cost of Rs 4,00,000/-, a gain of Rs 60,000/- and 15.0 per cent. Kesari Logistics Limited, bought at the same time by the same person on the same sort of reasoning, stood at Rs 1,95,000/- against Rs 3,00,000/-, a loss of Rs 1,05,000/- and 35.0 per cent. Two decisions, one procedure, opposite directions. Two positions are not a measurement of anything and nobody should treat them as one, but they make the point available to look at: the feeling did not know which of the two it was attached to.

Where is deciding on feeling exactly the right move?

Here is the part that gets left out, and leaving it out turns a mechanism into a scolding. Deciding on feeling is frequently correct, and sometimes it is the only thing that could possibly work. Feeling is how a judgement gets made when there is no time, no data and no second opinion available. A great many real situations are like that, and every emergency that has ever happened.

The test is not how strong the feeling is. Strength is worthless as a signal. A tag deposited by a thirty second encounter can be every bit as strong as one deposited by twenty years. The test is where the feeling was built. A feeling assembled from many repetitions of the same situation, where the outcome was visible each time and could correct the feeling afterwards, is compressed evidence: real information, held in a form that can be read instantly. A feeling assembled from one vivid encounter is a tag with no evidence behind it at all.

A woman who has sorted vegetables at the same market for twenty years can say by feel which crate will be turning by Thursday, and she is right, and she cannot say how she knows. Her feeling has been corrected thousands of times by Thursdays. Somebody who has bought vegetables twice has a feeling of the same strength and none of the correction. Both experience the feeling identically, and only the history behind it decides whether it is worth acting on. The useful question is therefore never how sure am I but how did I come to be sure. Herbert Simon, in the Quarterly Journal of Economics in 1955, put the general case for judgement built to fit the situation rather than to fit an ideal, and the compressed sort of feeling is one of the best examples of it.

Applied to a decision about money, this is not encouraging. The situations that repeat often enough with visible outcomes to build a corrected feeling are the everyday ones: what a household actually spends, how a particular sort of person behaves under pressure, whether a shop is busier than last year. The situations that do not repeat that way are precisely the ones that involve committing a large amount to something encountered recently, where the outcome arrives years later, arrives once, and is never clearly attributable to the decision anyway.

What does the affect account not explain?

A mechanism is only useful if it has edges, and this one has them. The affect account does not explain which holding came into view in the first place. Selection is attention, set out by Kahneman in Attention and Effort in 1973 and taken into decisions about money by Barber and Odean in the Review of Financial Studies in 2008, and it is set out under limited attention. Attention decides what is on the list; affect decides how large the commitment is once something is on it.

Affect does not explain why a thing that resembles a story gets treated as probable. Judgement by resemblance rather than by feeling was measured by Kahneman and Tversky in Cognitive Psychology in 1972 and is set out under representativeness. Affect does not explain why a particular number stays stuck in somebody's head and drags every later estimate towards itself. Anchoring, from Tversky and Kahneman in Science in 1974, is set out under anchoring and adjustment. And affect does not explain a background state that colours everything on a given afternoon regardless of what is being judged. Mood was separated out by Schwarz and Clore in the Journal of Personality and Social Psychology in 1983.

Affect is a feeling about a specific thing, arriving with the thing and attached to it. Mood is a state already in place before the thing showed up, and the difference between the two is what keeps the distinction honest. Affect and mood can look identical in a decision log and are corrected in completely different ways, so they are taught separately.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log, the Vindhya index scheme, the Nilgiri mid-cap scheme, Suvarna Chemicals Limited and Kesari Logistics Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.