Home Bias, Local Bias and Familiarity Bias Compared

Familiarity bias is a preference for what is recognised. Home bias is that same preference with a national border drawn round it, and local bias the same preference with a shorter radius. All three cut down what is considered before anything is compared, and outcome bias is the error that stops the cut ever being noticed or tested.

Every choice anybody has ever made began with a list somebody else drew up, or with a list nobody drew up at all. By the time a decision feels like a decision, something has already removed almost everything from consideration, and that something is almost never examined. Familiarity is one of the commonest things doing the removing. Familiarity is not a preference for the sound, the cheap or the well run. The preference is for the recognised, and it does its work on the list rather than on the choice.

What is familiarity bias, and what is it a bias in favour of?

Take it out of money first. A household in a large city has forty schools within reach. Asked which ones are being weighed, the household names three or four: the one a neighbour's child attends, the one on the route to work, the one whose board is painted on the wall by the bus stop. The other thirty six are not rejected. The thirty six never became candidates in the first place. Nobody struck them out, no reason was written against any of them, and asked whether the whole set had been looked at, the household would honestly feel the answer was yes.

The options that survived that process make up the consideration setThe options a person actually weighs against each other, which is whatever survived everything that ran before the weighing began., and the surviving names have one thing in common that is worth holding on to. Not quality, not price, not record. The survivors are the recognised ones. RecognitionKnowing a name has been met before. The feeling of having encountered something arrives without any accompanying content. is a far cheaper thing than knowledge: it needs only that the name has crossed a person's path once, and it arrives with no attached account of why. Gur Huberman, writing in the Review of Financial Studies in 2001, set this out in a finance setting and showed that a preference for the familiar operates on its own, separately from what a person actually knows about what is being chosen. The bias is in favour of recognition itself, so a hoarding at a bus stop can produce it.

What happens when a national border is drawn round the same rule?

Home bias is the identical mechanism with one boundary specified. Instead of the question being do I recognise this name, it becomes was this issued inside my own country, and the second question is answered faster than the first because a whole set of ordinary things settle it: the language of the statement, the currency of the price, the name of the street the office sits on, the fact that news about it arrives without being sought.

Kenneth French and James Poterba, writing in the American Economic Review in 1991, documented the pattern across several countries and gave it its name. French and Poterba measured how much of what people held had been issued at home, set against how much of what was available in the world had been issued there. The two numbers were a long way apart, and they were a long way apart everywhere they looked. The gap is a description of what people held, not a statement about what anybody should have held instead. How a holding ought to be spread across anything is a separate subject entirely.

What changes when the boundary shrinks to a radius?

Shortening the line from a border to a distance gives local bias. Joshua Coval and Tobias Moskowitz, in the Journal of Finance in 1999, documented exactly that: a measurable preference for what is based near, measured in miles rather than by any frontier. The finding carried extra weight because the people it was measured on were professionals whose whole occupation was choosing, working with research budgets and a duty to somebody else. The preference did not need an amateur to show up.

The everyday version is easy to feel. A street vendor sourcing vegetables has forty possible suppliers in the district and buys from four, all within a ten minute walk. Asked why, he gives real answers: he can see the stock, he can walk over when a crate is short, he hears within an hour when somebody has been shorting weights. Those answers are not superstitions. The advantages of a short radius are genuine, and they have to be granted before anything critical about local bias can be said. Proximity is not worthless. The bias is that the radius gets applied before anything is weighed and is never itself stated as a reason.

Familiarity vs Home vs Local Bias: where does each line get drawn?

Set side by side, the three show the same machinery. Each one takes a large set, applies a test that is not about the merits of anything in it, and passes forward whatever satisfies the test. The test is the only thing that differs. Familiarity asks whether the name has been met before. Home asks which side of a national line it sits on. Local asks how many miles away it is. Everything downstream of the test is the same, including the part where the person applying it believes they looked at everything.

The three overlap heavily without being the same thing. Something issued at home is more likely to be recognised, and something nearby is more likely still, so the three filters keep a great deal of the same material. But they come apart at the edges, and the edges are where the distinction earns its keep. A heavily advertised name from far away passes the familiarity test and fails both the others. A small concern two streets away, whose name has never once been heard, passes the local test and fails the familiarity one. Three tests that usually agree are still three different tests, and naming which one ran is the first step to being able to inspect it.

Why does recognising a name feel like knowing something?

Because most of the time it is. Skipping that answer makes the whole subject sound like a list of ways people are foolish. A name that has been met before has usually been around a while. Things that have been around a while have usually survived something. So in ordinary life, recognition and durability travel together closely enough that treating one as evidence of the other is a sensible way to save effort.

Substituting recognition for knowledge is a heuristicA short rule that answers an easy question in place of a hard one, usually well enough and always cheaply. in the exact sense Amos Tversky and Daniel Kahneman set out in Judgment under Uncertainty: Heuristics and Biases, in Science in 1974. The hard question is what do I know about this. The easy question that gets answered instead is do I recognise this. The rule is not stupid; it is a rule that works in the setting it grew up in, applied in a setting it did not grow up in. Every bias worth the name has that shape, and none of them can be understood by starting from the assumption that people are careless.

Where does the rule break, and what produced the recognition?

The break happens at exactly one joint. Recognition works as a signal only when its source is also connected to whatever the decision turns on. Long establishment produces recognition and is connected. Advertising produces recognition and is not. Proximity produces recognition and is connected only sometimes, and only for particular kinds of knowledge. Language produces recognition and is not connected at all.

Put a date on it. On 19 February a television segment named Suvarna Chemicals Limited, and Meera Sundaram, an investor in the invented Palash decision log, added Rs 1,00,000/- to that position the same evening, taking its cost to Rs 4,00,000/- and the total cost of the holding to Rs 13,00,000/-. Whatever else was true of Suvarna Chemicals that morning was equally true of it that evening. The strength of one name in one person's head changed in those hours, and a broadcast slot produced the change. The wider log says this is not one person's habit. Of the 96 buys recorded, 41 followed a media mention within three days, being 42.7 per cent. In an ordinary week only about 11.0 per cent of the eligible list got mentioned anywhere at all. Recognition manufactured by coverage is the same feeling as recognition earned by forty years of trading, and nothing inside the feeling tells them apart.

Outcome Bias: what is it, and why does it sit beside these?

Outcome bias is judging the quality of a decision, or of a rule for making decisions, by how the result turned out. Jonathan Baron and John Hershey, in the Journal of Personality and Social Psychology in 1988, set the effect out cleanly: they held the information available at the moment of the decision completely constant, varied only what happened afterwards, and found that judgements of how good the decision had been moved with the outcome. Same facts at the time, same reasoning available at the time. The result came out differently, and so did the verdict.

The everyday case takes a second. Two people leave the same office at the same hour and take the same road home. One arrives in twenty minutes and one sits in an accident queue for ninety. Nobody thinks the second person decided worse. Move the same structure to a decision with money in it and the intuition collapses immediately: the person whose holding rose decided well, and the person whose holding fell decided badly. Outcome bias is not a failure to look at results; it is treating the result as the whole of the evidence about a decision that was taken before the result existed.

Why does outcome bias keep a familiarity rule alive?

Here the familiarity filter and outcome bias lock together, and the argument is structural rather than psychological. Suppose somebody buys what they recognise and it does well. The conclusion available to them is that the rule works. Suppose instead it does badly. The conclusion available to them is that this particular holding was poor. Neither conclusion is about the rule. The first credits the rule for something a surviving option did, and the second blames the option and leaves the rule untouched. Both feel exactly like learning from experience.

The reason the rule cannot be assessed this way is a selection effectWhat happens when the thing being examined was chosen by a rule that also decided which cases would never be seen., and it is not a matter of not having reviewed carefully enough. The 203 options the filter removed have outcomes. The outcomes exist and they happened. Removal ran before anything was tracked, so the person who did the removing never observes one of them. A rule that decides what will never be seen cannot be tested against the results of what was seen, however many of those there are and however honestly they are reviewed. More discipline about reviewing holdings does not touch this. Reviewing harder examines the survivors more thoroughly, and the survivors were never the question.

What does the filter do to a list of 214 options?

Now put a number on it. The platform Meera Sundaram uses through Palash Advisory Services Private Limited carries 214 options that she is eligible to choose from, and the Palash decision log records the list as well as the choices. Her holding opened on 4 January at Rs 12,00,000/-, four positions of Rs 3,00,000/- each: the Vindhya index scheme, the Nilgiri mid-cap scheme, Suvarna Chemicals Limited and Kesari Logistics Limited. Four names came out of it. The interesting question is not why those four beat the others, but how many of the others were ever in the running.

Work the filterAnything that removes options from a set before they are judged on their merits, whether or not anybody meant to apply it. at four settings. Recognise a twentieth of the list, five per cent, and 214 times 0.05 is 10.7, so 11 options survive. Recognise a quarter and 214 times 0.25 is 53.5, so 54 survive. Recognise half and 107 survive. Recognise every single name and all 214 survive, the only setting at which the filter does nothing at all. At the setting most people are actually at, the filter throws away 203 of 214 options, and the entire remaining process of comparing, weighing and choosing operates on the 11 it left.

Set that beside a measurement from the same log about how people cope with long lists. Thirty people were shown the whole set of 214, and 11 of them chose anything at all, being 36.7 per cent. Thirty people shown a prepared list of 7 instead produced 21 choosers, being 70.0 per cent. So a shortlist of 7 and a familiarity filter at five per cent do roughly the same arithmetic work: both take 214 down to something a human being can actually act on. Doing that arithmetic work is not a criticism of either. A list of 214 is unusable and everybody knows it.

The difference is who did the cutting and on what grounds. A shortlist has a stated basisThe reason a cut was made, written down somewhere a second person can read it, disagree with it and ask for it to be changed.: somebody wrote down why these 7, the writing can be read, and a reader who thinks the grounds are wrong can say so and get the list rebuilt. No basis was ever formed, so a familiarity filter cuts just as hard and leaves no record of its own. Both cuts are severe; only one of them can be inspected, argued with or corrected by anybody, including the person who made it.

Is preferring what is known actually a mistake?

No. Familiarity buys three real things, and they should be stated without hedging before anything critical is said. The first is lower effort. Eleven options can be read about properly by an ordinary person with an ordinary evening. Nobody can read properly about 214, and the log's own measurement makes the point: most of the thirty people shown all 214 did not choose at all.

The second is easier monitoringKeeping track of a holding after buying it: reading what arrives about it, noticing what has changed and when.. Things that are near or national produce news that arrives without being sought, in a language that reads without effort. A holding that actually gets kept track of is different from one that does not, and familiarity makes the difference between them. The third is that proximity sometimes carries genuine knowledge. The vendor really does know which supplier short weighs, and no amount of distance research would have told him. All three are worth something real, and none of them is a claim about what the excluded options would have done. That last clause is the whole of the criticism. The benefits are about the cost and quality of the process; they say nothing about the 203 that were never looked at.

How would somebody detect this in their own decisions?

Not by looking at results. Results are the trap. The detection question is about the cut rather than about the outcome, and it has two words in it that make it answerable: what was excluded, and on what grounds? Put to a decision taken last month, it either has an answer that is written down or at least stateable, in which case the cut can be examined, disagreed with and rebuilt, or it has no answer at all. Having no answer is not a failure of memory. The absence means the cut ran without ever being formed as a decision, and that is the ordinary case rather than the unusual one.

The second question is a useful companion and takes about a minute. How many options were on the list being chosen from, and how many of them could be named? Most people have never asked, and the gap between the two numbers is the size of the filter. Neither question needs a result, a valuation, or any judgement about whether a holding turned out well. The detection question can be answered honestly on the day the decision is taken, and that is precisely why it works when reviewing outcomes does not.

How does an adviser work with this without telling anybody what to hold?

Devika Rao, the adviser at Palash Advisory Services Private Limited, does not raise this with a client by saying anything about what they hold. She raises it one step earlier, at the list. Before a decision is taken she asks how the set got down to the names in front of them, and she writes the answer in the file whether or not it is a good answer. For a client who says the four names were the ones she had heard of, that sentence goes in the file as it stands. Nothing has been criticised. Something has been recorded that was previously not recorded.

The log measures what recording does to recording. Twenty of the sixty investors adopted a written checklist on 4 November. Across quarters five to eight, that twenty recorded a written reason on 34 of 41 decisions, being 82.9 per cent, against 19 of 63 decisions, being 30.2 per cent, for the other forty. The checklist made a large difference to whether a basis existed on paper afterwards. Eight quarters and sixty people cannot support a claim about returns, and one case is never evidence that a rule works. What the checklist produced is a stated basis where there had been none, and that is the entire finding.

For somebody deciding alone, with no adviser and no committee, the same thing works with a notebook and two lines. Line one: how many options were on the list. Line two: which ones I could name before I started, and why those. Written before the choice, both lines take a minute, and they turn an invisible cut into an inspectable one. A lender, an analyst or a selection committee runs the identical discipline at a larger scale under the name of a documented screening basis: the criteria are written before the screen runs, so anybody can later ask why a candidate was never on the list.

What does naming a filter not settle?

Naming a filter explains the cut and what the cut removed. Naming a cut says nothing about what anybody should do afterwards, and the distance between those two things is larger than it looks. Naming a filter establishes that a cut happened and that it was not inspected; it does not establish that the cut was wrong.

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.