Suitability and Appropriateness: Risk Tolerance and Capacity

Risk tolerance is a preference: how much variation somebody is willing to bear. Risk capacity is arithmetic: how much they could bear before the plan itself fails. The two are measured in different units, so they cannot sensibly be averaged, and a questionnaire that returns one score has quietly merged a calculation with a feeling and lost the use of both.

Begin with something easy to picture, with no finance in it. A cousin says he is comfortable driving at speed on a hill road, and he means it, and there is no reason to doubt him. The cousin's comfort is a statement about him. Now the vehicle: the tyres are worn, there is no spare, and the brakes were last checked a long time ago. The vehicle's condition is a statement about the situation, and anybody with a torch can check it. Comfort and condition are two different readings, taken from two different places, and nobody would think of averaging them into one number out of five. Suitability is exactly that pair of readings, and averaging them is the mistake.

What is the difference between risk tolerance and risk capacity?

Two different questions get asked about the same person, and they only sound alike. The first is risk toleranceHow much variation somebody is willing to bear, obtained by asking them.: how much movement in the value of a holding is this person willing to sit through without stopping. The second is risk capacityHow much variation somebody could bear before the plan itself stops working, worked out from facts.: how much movement could this person absorb before something in the plan actually breaks. Willing, and could. One is about the person, the other is about the arrangement the person is standing in.

The gap between them is easiest to feel outside investing. A street vendor who has sold vegetables from the same corner for twenty years may be entirely willing to lose a week of takings to a monsoon, and say so without flinching. Willingness of that kind is tolerance, and it is real. But if his stall rent is due on the fifth and he has nothing set aside, a lost week means borrowing at a punishing rate from whoever will lend on the day. Capacity is what the due rent and the empty reserve describe, and capacity does not care in the least how relaxed he feels. His willingness and his position are both facts, but only one of them can be checked by somebody standing next to him with a notebook.

Meera Sundaram, an invented investor whose decisions run through the case below, sits in exactly that shape. On 6 June she completed a risk questionnaire of 12 questions scored 1 to 5, so a maximum of 60, and scored 44. Palash Advisory Services Private Limited, where Devika Rao is the adviser, reads a 44 as growth. Meanwhile she holds Rs 1,10,000/- back against interruption, against a monthly outgo of Rs 55,000/-, and two months is what that covers. The questionnaire says one thing. The bank record says another. Neither is wrong, and neither is a version of the other.

The last row of that comparison does the most work. Capacity comes out in months of outgo, or in rupees, or in years to a date. Tolerance comes out as a ratio, or a score, or a place on a five point scale. There is no conversion between the two, in the way there is no conversion between the weight of a suitcase and how much its owner wants to go on the trip. Both readings can be perfectly accurate. Neither can be turned into the other.

Why does a difference in units matter this much?

Units are not pedantry. A unit states which operations are allowed on a number. Two lengths add to a length. A distance divided by a time gives a speed, a new and meaningful thing. A distance cannot be added to a temperature, and the reason is not that the arithmetic is hard: it is that the result would not refer to anything in the world. Nothing in the calculator stops the buttons being pressed. The number that comes out is simply not about anything.

Capacity is a quantity in months, or in rupees, and it comes out of a division anybody can redo on paper. Two months of holding back is Rs 1,10,000/- divided by Rs 55,000/-, and a disagreement about the answer means one side has the wrong input. Tolerance is a quantity in a unit that only exists inside the measuring instrument: a score of 44 out of 60 means something only relative to the same twelve questions scored the same way. Change the questionnaire and the tolerance number changes while the person has not moved at all. The capacity figure never behaves that way.

The asymmetry between a checkable figure and an elicited one is the whole of the subject. The same asymmetry is why a profile is worth filling in a particular order, which client profiling returns to below. Computing capacity gives a fixed point. Asking about tolerance gives a reading whose meaning depends on the instrument that happened to be used.

What is risk capacity actually made of?

Capacity is not one number waiting to be looked up. Capacity is assembled from four separable things, and each of them is a fact somebody could produce a document for. The first is the money held back against interruption, usually stated as how many months of outgo it would cover. The second is what is going in and for how long, set against what it has to reach. The third is the time remaining to the date the money is wanted. The fourth is what has to be paid whatever else happens. People forget to write that one down.

Think of a household running on one salary with a wedding to pay for in eighteen months. The capacity question is not whether they feel brave. The question is how many months of expenses sit in the account, how much goes in each month, how far that gets them by the date, and what the existing repayments already take out before anything else moves. Every one of those four is a subtraction or a division somebody else could redo, and none of them requires knowing how the household feels about anything.

What does this case add up to when the capacity is worked out?

The four parts in order, with the arithmetic done on the case figures. Meera holds a reserveMoney held back against an interruption to income, usually counted in months of ordinary spending. of Rs 1,10,000/- and her monthly outgo is Rs 55,000/-. Rs 1,10,000/- divided by Rs 55,000/- is 2.0, so the reserve covers exactly two months. Not roughly two, not two and a bit. Two. Nobody had to interpret anything to get there, and if her outgo were Rs 44,000/- instead the same division would give 2.5 months, and that would be equally uncontestable.

Now the second component. Rs 25,000/- a month goes in by standing instructionAn arrangement that moves a fixed amount on a fixed date without anybody deciding again., and the stated goal is Rs 40,00,000/- for education in 11 years. Contributions alone come to Rs 25,000/- times 12 months times 11 years, or Rs 33,00,000/-. Set against the goal, that is 82.5 per cent of it. The gap is Rs 40,00,000/- less Rs 33,00,000/-, or Rs 7,00,000/-. That gap of Rs 7,00,000/- is 17.5 per cent of the goal and has to come from growth rather than from contributions. Seventeen and a half per cent of the target is a much smaller thing to ask of growth than most people assume before they do the sum, and that shift alone changes what the whole conversation is about.

How is tolerance measured, and what does the measurement inherit?

Tolerance is elicitedObtained by asking somebody, so the answer is the only record of it and there is nothing else to compare it with., a careful word for asked. Somebody writes a set of questions, somebody else answers them, and the answers are scored. Writing, answering and scoring is the entire production process. Instruments of this kind have been built and tested seriously for a long time, and John Grable and Ruth Lytton, in Financial Risk Tolerance Revisited in Financial Services Review in 1999, set out the work of building one that behaves consistently when the same person answers twice. The care that goes into a good instrument is real. No amount of that care turns an answer into an external fact.

Here is what the resulting number carries with it, whether anybody wants it to or not. The score carries the instrument. A 44 means something only against those twelve questions scored that way. The score carries the wording. A question about losing 20 per cent reads differently from a question about a fall of Rs 2,60,000/- on Rs 13,00,000/-, though the two can be the same event. The score carries the day it was taken on. The same person answers differently in a calm week and a frightening one. The tolerance number is a joint reading of the person and the instrument, and no part of the printed score reveals how much of it is which.

The case holds a second elicited reading, taken from the same 60 people on the same afternoon, and it is the one used here as the tolerance figure. Asked what gain would make an even chance against a Rs 10,000/- loss worth taking, the median answer was Rs 22,000/-. Rs 22,000/- divided by Rs 10,000/- is 2.2, the measured coefficient of loss aversionHow much more a loss of a given size weighs than a gain of the same size, for the same person.. Daniel Kahneman and Amos Tversky set out the structure that number sits inside in Econometrica in 1979. The measurement is careful. The 2.2 is also, still, an answer to a question.

The case actually holds two elicited readings of the same person, taken on the same afternoon. One is a score of 44 out of 60, or 73.3 per cent of the maximum available. The other is a loss aversion of 2.2. Neither of those two elicited numbers converts into the other, so the unit problem is not a quirk of comparing feelings with money.

None of this makes tolerance useless. A person who abandons an arrangement in the first bad quarter has not been well served by an arrangement built as though they would not, however good the arithmetic was. Willingness matters because behaviour follows it. No external record exists to appeal to, so the elicited number cannot settle a dispute. Meera says 44. If somebody thought 44 was wrong, what exactly would they check?

What is Client Profiling, and what does a profile actually contain?

Client profilingAssembling in one place what is known about somebody's situation and what they say about their preferences. is the assembling step. Profiling puts in one place everything known about a person's situation and everything they have said about their preferences, and a decision taken later can then be checked against it. Assembling is all the word means. A profile is not a form, though it usually arrives as one, and not a score, though it usually ends in one.

A useful profile has three parts rather than two, and the third is the one that gets left off. The first part is the computed column: income, outgo, the reserve, obligations in place, contributions, the goal and the date. The second is the elicited column: what the person says about variation, what they say they would do in a fall, and the score that comes out of it. The third is the provenance of the whole thing: who took it, on what date, using which instrument, and what has changed since. A profile without its provenance is a set of readings with no idea how old they are. A temperature with no time attached to it has exactly the same problem.

The order the columns get filled in turns out to matter more than it sounds. Filling the computed column first means entering the conversation holding a fixed point: two months of reserve, Rs 33,00,000/- of contributions against a Rs 40,00,000/- goal, Rs 1,80,000/- of borrowing at 36.0 per cent. Anything the person then says sits beside those numbers and can be compared with them. Filling the elicited column first does something subtler: the arithmetic gets done afterwards, in the shadow of an answer already given, and it is remarkably easy to do arithmetic that agrees with a conclusion already reached.

Why does a single score out of five destroy both readings?

Now the two quantities side by side, treated the way a great many forms treat them at the end. Meera has 2.0 months of reserve and a loss aversion of 2.2. Averaging them gives 2.1. Nothing prevented that. The calculator did not object, the form printed it neatly, and the resulting profile now carries a single figure that somebody will act on. The only question that matters about the 2.1, namely 2.1 of what, meets silence. Half the input was a ratio, so the 2.1 is not 2.1 months. Half the input was months, so it is not a ratio of 2.1 either. A compromise requires a scale both readings sit on, and no such scale exists, so the 2.1 is not a compromise between the two.

The damage is not only that the combined number is empty. The worse damage is that both original readings are now hard to recover. Somebody reading the profile six months later sees a 3 out of 5, or a 2.1, and cannot tell whether the reserve was thin, or the willingness was high, or both, or neither. The two facts that could have driven a sensible conversation have been mixed into a single grey figure and cannot be separated again. A score is a lossy summary of things that were not summarisable in the first place.

Running the merging backwards shows more sharply what it costs. A printed 2.1 could have come from 2.0 months beside 2.2. A printed 2.1 could equally have come from 1.0 month beside 3.2, or 3.0 months beside 1.2, or half a month beside 3.7. The four pairs call for four completely different conversations, and the score cannot tell them apart afterwards.

Where does appropriateness sit beside suitability?

Suitability and appropriatenessWhether somebody has the knowledge and experience to understand what is being proposed to them. are often said in one breath, and they are two tests rather than one. Suitability asks whether the arrangement fits this person's situation and preferences. Appropriateness asks something narrower and prior: does this person have the knowledge and experience to understand what is being put in front of them. A structure can fit somebody's capacity beautifully and still be something they cannot follow, and something perfectly comprehensible can be entirely wrong for the reserve they hold.

Kept apart, the two give four situations rather than a spectrum. Both pass. Suitability fails on its own, meaning the person understands the thing perfectly and it does not fit their position. Appropriateness fails on its own, meaning the arrangement fits and the person cannot follow what has been described. Or both fail together. Merging the two tests into one impression loses exactly the information about which of them failed, and only that information shows what needs fixing. What any conduct duty requires in either case is set by the Securities and Exchange Board of India and must be confirmed at sebi.gov.in.

What does a questionnaire miss most often?

One pattern shows up again and again. The elicited column is filled in with real care. Four or five questions probe how the person would feel about a fall, what they think they would do, how they describe themselves, what they remember doing last time. Meanwhile the computed column has one line in it, or none, and the arithmetic that would have mattered most was never attempted. The form was thorough about the part that cannot be checked and casual about the part that can.

Why is the date a profile was taken part of the reading?

Every reading so far has a date on it, and one of them has a gap after it. The questionnaire was completed on 6 June. The decision the case turns on came on 12 October, 128 days later. In between, the eight quarter valuation was struck on 30 September, showing a cost of Rs 13,00,000/- against a value of Rs 12,46,000/-, down Rs 54,000/- or 4.2 per cent. Nothing in the log records whether anything about her situation moved between June and October, and that absence is itself a finding.

A thermometer reading with no time attached is not a fact about the room, it is a fact about a moment. Tolerance readings are the same, and worse. The willingness being measured is known to move with recent experience. A person answering in a calm week and the same person answering after watching a holding fall 35 per cent are not answering the same question in any useful sense. William Samuelson and Richard Zeckhauser showed in the Journal of Risk and Uncertainty in 1988 how strongly an arrangement already in place tends to stay in place. A profile taken once quietly becomes the permanent description of somebody for the same reason. The date is not administrative detail on a profile; it is the part that determines how much weight the rest of it can still carry.

How does anybody actually use the split in practice?

A lender has never been confused about this, so watch one for a minute. When somebody applies for a loan against a shop, nobody asks the applicant how they feel about repayment risk. The lender computes. Monthly takings, monthly outgoings, what is already committed to other repayments, and what is left over each month with a margin on top. The whole assessment is arithmetic, and the applicant's confidence, however genuine, does not enter it. Ignoring the applicant's confidence is not a judgement that feelings are unimportant. The lender is measuring the one quantity that will still be true in a bad month.

Read the case with a lender's eye and the sheet fills itself in. Monthly outgo Rs 55,000/-. Interest running on the borrowing at 36.0 per cent on Rs 1,80,000/-, or Rs 64,800/- a year, or Rs 5,400/- a month, being 9.8 per cent of the monthly outgo before anything else is paid. Rs 25,000/- a month going out by standing instruction. A reserve covering 2.0 months. Every line on that sheet is a number, and there is no column the tolerance reading of 2.2 belongs in, so the reading appears nowhere on the sheet. For a person deciding alone the discipline is the same and harder, because they have to be both the one asking and the one answering, and the only protection is doing the arithmetic before the conversation with themselves begins.

One line on that sheet surprises people most, and it is worth its own picture. Rs 64,800/- a year of interest is Rs 5,400/- a month, and Rs 5,400/- set against a monthly outgo of Rs 55,000/- is 9.8 per cent of everything leaving the account every month, committed before any decision is taken at all.

The practical shape of all this is undramatic. The four capacity components are computed and written down in their own units. The tolerance questions are asked and the answers written down in theirs, with the date and the instrument beside them. The two sit together in one record and are never added. The two disagreeing is the interesting case, and a 44 out of 60 beside a two month reserve is exactly that. The disagreement is the finding, and flattening it into an average is how the finding gets thrown away.

The boundary of the subject is worth setting out plainly. Everything on the left of the picture below is arithmetic anybody can redo, everything in the middle is recorded with its date and never merged, and everything on the right belongs to a source that has to be checked directly.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log, the Palash 100 index, 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.