Deep analysis with academic rigour is our core competency
The Stackup Risk scoring methodology was built to take the best elements of other surveys, purify them, then exploit them to generate highly accurate risk scores.The Methodology
Stackup Risk Scores are the product of hundreds of hours of finance and psychology research while taking the best elements of current industry practice. Combined, this works to produce highly reliable analysis. Here's what we did:
We started by looking for a theoretical model which as holistically as possible describes risk tolerance.
The literature is broad and diverse, however Cordell (2001) provides a solid framework encompassing each relevant factor. Risk tolerance or attitude is the most common, which is the propensity for someone to engage in risk behaviour. However there are many other considerations such as financial literacy, loss capacity and then psychological biases, which combine to determine their actualk risk taking behaviour. Cordell's model encompasses each of these concepts, so we use this as the starting point to develop our risk score.
From the theoretical model we progressed to developing questions which would measure each of the relevant factors for risk profile.
We began by creating a database of sample questions which were reflective of common indsutry practice. Many were similar to each other, but the subtle differences were important for how they could elicit nuanced responses. From the large sample database we analysed each question in consultation with our psychology and finance team to correct and filter for in built biases and readability. We were then left with a 125 question set which we considered to be valid.
Our next task was to determine which questions from our large sample were best at eliciting relevant risk profile preferences from our respondents.
To do this, we had over 300 people who were representative of the population as a whole to complete the full 125 question set. We then had a large sample of answers to conduct analysis on. Through a statistical process called factor analysis, we grouped questions which had correlated responses together. These groups of correlated questions corresponded to the factors we set out to test, as per our initial theoretical model. Through this process we were also able to determine which questions were best at eliciting responses for individual factors. Thus we were able to purify our question list down to a more manageable set, while retaining the ability to measure what we needed to. We now had a set of 25 questions.
With our survey length more manageable, we next validated our chosen questions statistically for their ability to measure risk behaviours.
The endpoint of this process is of course to place clients into investments which are appopriate for them, thus the survey's ability to predict which investments would make each client feel comfortable throughout the cycle was key. The analysis into reliability via a stepwise regression found that our questionnaire was indeed good at explaining people's allocation to risk assets, and through a statistical test called Cronbach's alpha we also showed that the answer to each question within a factor group was highly internally consistent and thus likely to be measuring the same thing.