How do you define your clients’ risk tolerance in conjunction with their risk capacity?
Michael Kitces has written about the three factors that should, ideally, go into determining the ‘right’ asset allocation for a client portfolio (https://www.kitces.com/blog/risk-tolerance-questionnaire-and-risk-profiling-problems-for-financial-advisors-planplus-study/), and noted that most advisors tend to focus on only one of them. There are a lot of risk tolerance tools in today’s marketplace; the still-collecting-data T3/Inside Information software survey (https://www.surveymonkey.com/r/2021SoftwareSurvey) lists 14 (and counting).
The second factor is risk perception, which might be described as the variability factor in risk tolerance. In the relaxed atmosphere of an advisor’s office, when the markets are chugging along as normal, clients might give answers that would suggest they’re totally comfortable with a risk-on mix of assets. But in February of 2009 or March of 2020, when their net worth is dropping like a stone, they might give a very different profile—because they expect more downside than they did back when they took the risk tolerance instrument in the first place. The solution there is to give clients a risk tolerance assessment every once in a while—which, of course, most advisors don’t do.
But what about the third factor that Kitces says that advisors should consider: risk capacity? It’s obvious that if a client has $100,000 set aside for the down payment on a house that will be sent to the escrow agent in a couple of months, that part of the portfolio should be invested in cash rather than oil futures. But what about a client who has that upcoming down payment, and in a few years the first child goes to college, and a couple of years after that the second child needs a tuition payment, and the couple plans to retire in 25 years and has a certain monthly lifestyle cost that will fluctuate depending on changes in the inflation rate, and there’s a vacation home in the future?
Add in the regular fluctuations in the market (and the value of the client portfolio) and, well, most advisors use a Monte Carlo analysis on their recommended portfolio that matches a client’s risk tolerance and, if the sufficiency odds seem low, they advise the client to save more or retire later.
Mark Friedenthal, proprietor of a software program called Tolerisk Pro (https://www.tolerisk.com/), finds this Monte Carlo solution to the risk capacity problem to be very unsophisticated. “A lot of people don’t need to take as much risk as they’re willing to accept,” he says. “With many clients, there’s a clear mismatch between how they score on a risk tolerance instrument and the risk they need to take to achieve their goals.”
He offers the simple example of a 30-year-old client whose risk tolerance score might indicate a very low-volatility portfolio, so the advisor puts her in an all-bond portfolio. “At age 90, that person might have a 32% chance of having retirement assets, because she’s not protected from inflation,” says Friedenthal. “Her primary risk is not a market contraction. It is that stuff is going to cost more, and she’s not earning enough over time to compensate for it.”
This same result might also show up in planning software. “In most cases with that situation,” says Friedenthal, “the advisor would say, well, you need to take on more risk. And he shows that with a higher equity portfolio there is a higher expected return over the long run, and that magically diminishes the probability of failure,” he adds,—though it may increase the variability of outcomes.
But then, Friedenthal points out, a portfolio outside the client’s comfort zone makes behavioral risk much more likely. The client might bail out of the allocation when the market goes down.
Moreover, the traditional Monte Carlo approach doesn’t take into account the timing of monies that will need to be withdrawn. “There are really two independent dimensions to the risk problem,” says Friedenthal. “One of them is completely mathematical. We can mathematically measure someone’s ability to take risk using an objective scale that relates to the ratio of stock and bond allocations, where we show that a mix of these assets and these volatilities have a probability associated with achieving a particular mix of future withdrawals.”
The other dimension, of course, is the clients’ willingness to take risk. Friedenthal doesn’t use the word “tolerance” because ‘willingness’ is the term the SEC uses in its quantitative descriptions.
If there is a third dimension, it is how the willingness and the ability to take risk interact, mathematically, to produce an appropriate portfolio. As it turns out, this last problem is the easiest of all.
Monte Carlo limitations
On his LinkedIn page, Friedenthal describes himself as a ‘math nerd,’ and his background suggests that he has fully lived up to this self-characterization. After graduating from Emory University in Atlanta, Friedenthal spent several years at the Atlanta Fed, building mathematical models in asset/liability management and modeling related to the banking industry. That was when he began exploring the limitations of Monte Carlo mathematics—specifically, that they are not optimal for long-term projections.
Friedenthal took his mathematical expertise to a management position at GE Capital, where he was tasked with creating risk management models, and then later he was hired away to manage billions of dollars in company assets at a proprietary trading desk at Citigroup. In 2009, he left to start his own asset management firm, Friedenthal Financial, as a way to offer the same institutional asset management models to Main Street investors. “I thought that people with $250,000 to invest would want the same math, data-oriented, technology-driven processes that we had at Citigroup,” he says.
Working with individual investors brought Friedenthal back to pondering the limitations of Monte Carlo analysis, and how to provide his clients with appropriate portfolios using math instead of instinct.
“If I were measuring some complex derivative instrument that has an expiration of six months, I’m going to want to use the most current information, the most current relationships, of whatever could be contributing to it,” he says. “And I’m going to simulate a whole bunch of different paths for those six months, and I’ll get a value for that instrument, as well as the sensitivity.”
But with a financial plan, he says, you might be simulating three, four, five or more decades. “That’s a very long time to implicitly make assumptions about a constant volatility or a constant correlation,” says Friedenthal. “I’ve asked advisors: do you think that volatility in the stock market will generally remain constant? Invariably, they will say no; sometimes it is very volatile, sometimes not volatile at all. So then I say, what about the correlations between the broad stock market and the broad bond market? Or the stock market and inflation? Or the bond market and inflation? Do those correlations remain at a constant level? And then,” says Friedenthal, “I’ll ask them: does your planning process account for that?”
What’s better? Start with the first dimension, Tolerisk Pro’s risk tolerance (willingness) assessment, which is structured as a psychometric profile like FinaMetrica rather than a comparison of different portfolios where clients pick the one they prefer over and over again, like Riskalyze. I didn’t see anything much different about the 20 questions that Tolerisk Pro asks clients than you would see in the FinaMetrica assessment tool; both probe whether someone is a natural risk-taker (or not) in general life decisions, with frequent references to investing. (Sample questions: I would rather earn half the return on an investment if I knew I could not lose any money. I feel anxious after I make an investment decision.)
This leads to a score ranging from 0 to 100, and that number becomes the percentage of client portfolio assets which, in this first dimension, at least, could be allocated to stocks. If the Tolerisk Pro willingness score is a 34, that would imply a 35/65 portfolio, or maybe a 40/60. Advisors can obviously diversify this stock allocation among domestic and international, small cap, midcap and large cap, but the overall percentage of stocks is largely determined.
There are some checks and balances built into this. Friedenthal included some redundant questions in the instrument, to see if the client actually understood what was being asked. In one instance, a client answered ‘rarely’ to the question “I tend to overreact when financial decisions turn out poorly.” But then answered ‘never’ to the proposal “I usually adapt easily when things go wrong financially.”
“Those are pretty dichotomous,” says Friedenthal. “So the system would flag them, so the advisor could have a deeper discussion.” He adds that for most advisors, both spouses should take the risk willingness assessment, and then the advisor should ask whether they want to meet halfway or use the lower score as the baseline for building a common portfolio.
What about that trickier second dimension, the risk capacity? Here, the system gathers much the same client data that you might input into a financial planning program: ages of both clients; the value of their existing portfolios and which are taxable and which are tax-deferred; the amount being saved each month; monthly expenses; and (very importantly) the timing and amount of future big ticket items like college, wedding, buying a vacation home, funding a business—and, of course, retirement. You would add Social Security benefits, annuities or pensions or any other retirement income streams.
Then Tolerisk Pro applies a mathematical formulation to these various inputs, and what comes out is a graph similar to what you see here (above), which shows both the willingness and capacity score evolving over time. As a simple example, if somebody has to make that $200,000 down payment on a second home, then the risk capacity score would be lower before that payment was made (because you’d naturally not want to take a lot of risk on money that will be needed soon) than it is afterwards (when there’s less need to hold an asset in or near cash equivalents).
The actual formula that takes into account all these variables is proprietary, but it is important to realize that, even if the clients’ risk willingness doesn’t change (and it may), the risk capacity changes over time, as goals are achieved and the cash is disbursed, and also taking into account the ever-fluctuating investment ‘regime.’ This accounts for something else that Kitces has talked about: the fact that future return expectations are lower when market PEs are high and bond rates are low, than they are when stocks are dirt cheap and you can get real yield on Treasuries.
“We use what is really tantamount to traditional fixed income mathematics,” says Friedenthal, “similar to the way you would measure a bond or a bond portfolio’s sensitivity to interest rate changes and credit spread changes. We’re using the same kind of functionality to measure all of the cash flows sitting underneath that financial plan. We can say: this is the sensitivity that can be sustained and still meet these obligations with that very high probability.”
How does the system recalculate the appropriate portfolio when stock PEs and correlations shift? “There are concepts from institutional fixed income and derivatives math where you are measuring forward curves,” says Friedenthal. “You can measure an implicit forward curve of something like interest rates based on today’s interest rate curve. Conceptually, you could think about it like migrating towards and through the client’s cash flows year by year, and recomputing mathematically what that ability to take risk is.”
Even if Friedenthal is a bit cagey on the actual formulas (he said at one point in the interview that he’s nervous that one of the risk tolerance competitors would simply steal them) you can see that this is very different from Monte Carlo mathematics. What Tolerisk Pro is graphing is similar to what many advisors are doing by hand. When the markets go down, you run the financial plan all over again and show clients that their chances of financial success dropped from 94% to 92%.
But there’s another difference. In addition to variables like expected return and standard deviations, changes in yield curves and correlation coefficients, movements in market PEs etc., Friedenthal’s calculations incorporate inflation as an additional variable.
“Monte Carlo simulations are only simulating potential portfolio returns,” he says. “They are not simulating inflation separately. They are really good at capturing sequence of return risk under a constant volatility/constant correlation kind of environment,” Friedenthal adds. “We use a different mathematical process that allows us to capture evolving volatility, evolving correlations, we use a dynamic asset allocation that measures how the clients’ risk ability mathematically evolves over time, a dynamic custom glide-path that is consistent with each client’s very specific cash flow chronology, subject to their personality constraints, and the shifting inflation rate over time.”
The mathematics create separate paths for equities, fixed income and inflation, and captures the evolving volatility between them, and then calculates real sequence of return risk relative to inflation.
There’s one more variable that Tolerisk Pro takes into account: lifespan. The program uses unisex mortality tables (which advisors can customize based on client health data) to calculate the probability of a 50 year-old client living to age 51, 52 and so forth, all the way out to age 120, with an average weighted average lifespan for each client, and most importantly it calculates second-to-die mortality probabilities. The result is a calculated set of last survivor probabilities, year-by-year, which is incorporated back into the risk ability (capacity) calculations and the calculation of the odds that the portfolio will survive the longer-lived client in a paired relationship.
Avoiding behavioral mistakes
But that brings us (finally) back to that question that was raised earlier: how do you combine the risk willingness data on a client with the risk ability data?
If you look back on the projected Tolerisk Pro scores chart, you see that this involves mathematics that even a journalist could fathom. The risk ability score is the dotted line through most of the left side of the graph, while the solid line represents the risk capacity—which, as you can see, doesn’t change over time. The risk willingness line serves as a ceiling for the portfolio allocations; that is, clients may have more risk ability (capacity) than they are willing to take, but the system generally requires the portfolio to not exceed the lesser of the two.
“Everything is calculated over time,” Friedenthal explains. The blue line starts out at a 34, it bottoms out next year at a 27 because of the capital expenditure to buy a house.” Then, toward the right of the screen, the risk ability score suddenly drops below the risk willingness score, and that becomes the recommended portfolio allocation, and the risk willingness score becomes the dotted (over-ridden) line to the right side of the graph. “The dark line is where we set the cap,” says Friedenthal. “Even though, mathematically, they have the ability to put 86% in risk assets, we are not going to recommend that because that would put them in an undue likelihood of making a behavioral mistake. It is beyond what we think they would be willing to accept.”
When you combine everything, the overall Tolerisk Pro score in any given year is really the ability subject to the cap, the lesser of the two dimensions. And, in this particular case, when the second-to-die client reaches age 90, there is an 87% chance that, if the couple follows this glidepath, there will still be money in the retirement portfolio.
Alternatively, the system calculates that the money has a 95% chance of lasting to age 86 on this glidepath. And it calculates the probability of meeting each of the individual large expenditure goals as well. “If the chances of achieving Mary’s college goal crosses a threshold, the system will highlight that,” says Friedenthal.
“It all boils down, with all the fancy math, to simple outputs,” says Friedenthal. “These are designed on the dashboard to answer the most common questions that advisors could field from their clients. What do we recommend? That is the Tolerisk score, calibrated to be the numerator in the stocks to bonds benchmark. What is the likelihood that we will will run out of money with that initial portfolio? After all the fancy math and moving parts, it boils down nice and simply to, in this particular case, there is a 13% chance that one or the other spouse will be alive after their money is gone.”
At the moment, Tolerisk Pro doesn’t have nearly the market share of some of its risk tolerance competitors, and after looking at the risk capacity (ability) features, one might argue that the program is mischaracterized in this category, and might be viewed more as a financial planning program. But whatever you call it, Tolerisk Pro has a small but devoted band of users.
“What we have found, anecdotally from advisors, is that their clients have a lot more confidence in the decisions that they make after using our software,” says Friedenthal. “They see that the analysis is tailored to them, instead of just a personality questionnaire. It incorporates all their cash flows. Their mortality probability. College, some other goals.”
And how does that translate into better outcomes? “The confident client sticks with the plan,” says Friedenthal. “By measuring their ability, rather than just their willingness to take risk, it actually reduces the chances of a behavioral mistake.”
In addition, for advisors who set up small business 401(k) or other plans, there’s a Tolerisk Pro version that allows plan participants to enter information, where they would take a smaller psychometric test, and get a recommended allocation of the investment options offered by the plan and see online the odds of enjoying a sustainable retirement.
“We call that the Personal Asset Allocation Roadmap,” says Friedenthal. “Instead of a target date fund, you have seven or eight different inputs plus a personality constraint that goes into the glidepath.” The plan participants can play with the inputs—changing, for example, their retirement year from 65 to 68, changing the asset allocation, and the system will automatically calculate the employer match for different levels of monthly contributions.
“The 401(k) version is for advisors who are working with small business owners, who have set up a plan as an accommodation,” says Friedenthal, “but they’re finding that giving individualized advice to so many plan participants is crippling them because they have to spend so much time on it.”
Price? You can find a somewhat complicated pricing structure for the individual advisor version on the Tolerisk Pro website, but there is a free trial option, and then the highest price is $99 per user per month. (Firms with between 2 and 99 advisors would pay $89 per user, and there are discounts if you pay the full year in advance.) The program is cloud-based, and does not collect personally-identifiable investor information.
As I go through the Tolerisk Pro demo, I find myself wondering if it doesn’t define an entirely new category of software. The risk tolerance component is just a piece of the much larger “appropriate portfolio analysis” puzzle, and it’s interesting that the appropriate portfolio might evolve over time for clients who have different goals, different cash flow needs, and invest in evolving market “regimes.”
The financial planning space has many examples of institutional mathematics making their way into mainstream client advice—Monte Carlo analysis is a great example of this. Tolerisk Pro’s models might be the next example.
Is it time for the math-heavy portfolio analytical tools on Wall Street to enhance investment recommendations in the retail space? Time, as always, will tell.