A recent article for Money Marketing raised the question of whether risk profiling tools should have standardised scales
A recent article in Money Marketing raised the question of whether risk profiling tools should have standardised scales. At present, some profiling tools map investors onto a variety of scales: five-point and seven-point scales are common, while some produce an output alleging to discriminate between investors on 100 fine gradations.
Should profiling tools be standardised to map to the 7-point scale used in the Key Investor Information Document (KIID) prescribed by regulation?
On the first question, the answer is a clear ‘no’. There is a vital difference between the KIID scale and risk profiling outputs: the first is a measure of the risk of investments, the second of a personality trait of investors. There is absolutely no reason that the (very wide) range of risks that an investment can possibly exhibit should be the same as the range of risks that an investor can reasonably tolerate.
We need to always have in mind that, measured properly, risk tolerance is a feature of the individual. It signals the trade-off, over their total wealth, between the chance of getting better returns and the risk of ending up with bad returns that this investor finds most effective. The “total wealth” point is particularly important: investors may quite reasonably have a mix of very safe and very risky assets in their portfolio, as long as the combination of all their holdings reflects their suitable risk–return trade-off overall.
There are innumerable investments at the upper end of the KIID scale that would be criminally irresponsible to give to an investor for the entirety of their wealth (reverse triple leveraged bitcoin ETFs, anyone?). Thus, the range of possible risk profiles for investors should always be less than the total range needed to measure the riskiness of investments.
A further reason it is inappropriate to establish a fixed mapping between the risk profiles and the KIID is that the risk–return combinations that are suitable for an investor with a given profile are much more stable than the risk–return characteristics of investments. The latter may fluctuate in different market environments, over different time horizons, and with different assumptions used to forecast risk and return (put differently, the investment efficient frontier may vary for all these reasons).
The right level of risk for each investor depends not only on the risk of the portfolio, but also on the expected returns that are deemed achievable for this risk level. For example, when market conditions are poor and investors expect to receive lower returns for each level of risk taken (i.e., when the efficient frontier is lower), the rational response is for an investor to take less risk. Investors expect to be paid less for each unit of risk they take, and thus, for the same risk profile, should take less risk. A fixed mapping between investor risk tolerance and investment risk condemns investors to taking a fixed amount of risk, even when market conditions make it irrational to do so.
Is there a ‘best’ number of buckets into which to divide investors?
The answer is that there may be, but there is no strong theoretical reason to fix this at a particular number. Risk tolerance in the population undoubtedly falls on a continuum, but there are constraints on our measurement ability to finely discriminate between investors. Risk tolerance is a measure of an investor’s long-term willingness to trade-off risk for the chance of extra returns, and it would be disingenuous to pretend that we can measure an abstract personality trait like this on more than, say, 10 buckets in the extreme. To avoid such spurious precision, it would seem reasonable to say that scales of five or seven buckets come close to the ability of psychometric instruments to discern differences between people.
We also need to consider the degree to which it is possible to discriminate between adjacent portfolios on the efficient frontier. If we were to fix end points of a) the least-risky portfolio that would be appropriate to offer an investor for their total wealth (as close to zero as possible in the extreme), and b) the most risk that even an extreme investor should be willing to accept for their entire wealth (at a very rough guess perhaps at a standard deviation of 25%), we need to ask how closely we can position portfolios in between before the differences become indiscernible. Due to the inevitable noise in investing, it would seem almost impossible to divide the intervening space into more than about seven divisions (resulting in nine model portfolios including the end points) before the adjacent portfolios are perceived as effectively identical. Many investment managers quite reasonably maintain fewer model portfolios to span this space. There is little point to profiling investors with more granularity than you can provide solutions for.
Is there nonetheless a case for standardisation of risk profiling outputs amongst themselves?
The idea of standardising the outputs of different risk profilers is superficially appealing, as it would aid comparisons between tools and between advisers. However, here again our answer is no. Risk profiling methodology is not static, and whilst there are certainly many ways of doing it badly, there is not a single way that can be considered “right” to the exclusion of all others (see New Vistas in Risk Profiling for a discussion of where current approaches are flawed, how to do it better, and where the future lies for risk profiling). Enforcing a single standard is likely to stifle valuable innovation in profiling methodology and lead to a regulatory environment that remains flawed and dated.