Some months ago, I wrote a piece on retirement risk and what really matters to people. It was part of our ongoing desire to build an evidence-based approach to fill the gaps in the retirement planning process and enable people to make more informed choices. This reflected our belief that most risk profiling tools haven’t been well calibrated to the way people see risk and uncertainty and more reflect the industry’s idealised views.
This process is an ongoing one at OnTrack. The next phase was conducting analysis on a statistically significant sample of mass market investors and their attitudes towards uncertainty. Building a psychologically valid process to ascertain attitudes to uncertainty and then analysing the results was a non-trivial exercise. It required some excellent assistance from a group of PhD candidates we have had consulting to us for quite some time now - much of the credit goes to the great work they did.
What have we learned? There are many dimensions to uncertainty; from ambiguity to order and decisiveness, predictability, risk aversion and the actual act of making investment decisions. Ultimately, we found that some dominant characteristics describe the multiple dimensions to uncertainty in a consistent and differentiated way. The key factors appear to be assessing a person’s attitude to predictability and their risk aversion level.
To quantify some of the results, a low score on our latent factor of predictability, means the individual is 5 times more likely than someone who scores at the high end of the scale to be focused on investment gains rather than loses. That person is also substantially more likely (72% vs 44%) to take a gamble to receive a risky payoff than someone who scores highly.
Risk aversion reveals similar preferences.
A low score investor is 15 times more likely to be interested in gains than a high score person. The high score person is also twice as likely to want a more certain investment outcome than a low score investor and three times more likely to take a gamble for a risky payoff. Only half the number of people with a low score versus a high score say they invest heavily in extremely safe assets - and the evidence suggests they are much happier with the decision.
Is all this intuitively obvious? Perhaps, but quantifying the intuitively obvious is the critical first step in building a behavioural based risk profile.
This research – and a lot more besides – informs us as to the latent desire our users have for dealing with uncertainty, and how happy they have been when they adhere to that desire. Planning for retirement involves making complex choices with uncertain outcomes at an unknown point in the future. The better we understand people’s ability to deal with uncertainty, the more personalised the plan can be.
When we combine the work on behaviour with the modelling, we do around capacity to take risk, we get a much better picture as to the type of investments that best match a person’s actual risk appetite and capacity.
Pigeon holing people into risk buckets is convenient but the evidence to support it is slim and it has little relevance to actual people. When these models were developed in the mid-1990s, they were premised more on anecdote than on evidence. Adding rigour to this key component of the retirement planning process reduces the sub-optimal outcomes that are not in the persons best interests. Our work demonstrates that you can remediate these problems by assessing individual preferences as part of building a fuller picture of their appetite for uncertainty.
As much as I would love to include terms like “artificial intelligence” and “machine learning” – this analysis comes from real experts, doing structured analysis of real-world data. And there is nothing artificial about it.