Using Analytics in Wealth Management: The Good and Bad
Part 1 – Portfolio risk measurement
This is the first part of a new series on investment analytics in the wealth management industry and how advisors are using analytics to communicate with clients and prospects. We will look at some of the more popular analytics and go over what they are, their good points, and their potential pitfalls.
Since the financial crisis, wealth advisory clients have shifted their relationships from one of blind trust to a desire to internally understand and validate that a particular advisor recommendation is the right thing to do. Advisors have obliged, taking their conversation from “trust me” to educating and partnering with their clients around their finances and future. Ironically, many advisors have reported that this education and partnership approach builds trust.
The challenge becomes how to educate clients about their finances when they may have little or no financial training. Turning to portfolio analytics may be one part of the solution. But most financial analytics are confusing for people outside of the industry. Forward-thinking advisors have been searching for and employing analytics very carefully, choosing only those metrics that are simple, understandable, and meaningful for people with no financial training. This series will explore some of these metrics, along with their benefits and pitfalls.
Today we will be looking at the important area of portfolio risk. Part 2 will focus on capital market assumptions. And part 3 will focus on Monte Carlo analysis.
Portfolio risk analytics
One of the most difficult topics to discuss with clients is risk. Unfortunately, it is also one of the most important.
For financial professionals, there is a wide array of analytics to measure and describe portfolio risk. Metrics like standard deviation, value-at-risk, expected shortfall, and others are powerful tools for professional risk management. However, for non-financial professionals, these measures are not intuitive, confusing, and occasionally downright terrifying.
To help communicate this important point, advisors have started using a few different types of analytics in their conversations with clients. Here is my take on the ones that seem popular, why they are effective, and what sort of problems might they create.
As I share my thoughts on the good and bad of these metrics, please try to keep in mind that the use of financial analytics in client conversations is not a clean and tidy utopia. It’s complicated and there is no silver bullet. Everything has a negative of some sort, so whatever you are going to use will have a negative and please take that in stride. I hope that readers see this more of a user’s guide and not an Amazon review.
Risk metric 1: “Risk numbers”
What are they?
Risk numbers take the complicated notion of portfolio risk and distill it down to one, easy-to-understand number. This innovative solution was pioneered by Riskalyze, which basically named the genre with its flagship Risk Number®.
Typically, risk numbers are based on a 1 to 100 numerical scale with 100 being the riskiest, but other variations are out there with a similar concept. For example, Orion Advisor Solutions has a scale that can go above or below 100, with a score of 100 being the risk of a generic global equity portfolio. It’s a linear scale relative to the global equity benchmark, so a score of 125 would be 25% riskier than global equities and a score of 80 would be 20% less risky, and so on.
To produce a risk number for a portfolio all you need is a methodology for measuring portfolio risk and mapping that risk to a number on the scale. This can be achieved in any number of ways using academically sound risk measuring techniques. However, there are some considerations to be aware of, as we will discuss in the following sections.
Intuitive. As I mentioned before, portfolio risk is a very difficult concept, even for professionals. The huge innovation of the risk number is that it is intuitive to most people. If you understand speed limits on the highway (or school zone), you can now understand portfolio risk. Riskalyze even frames its Risk Number® in a speed limit sign when displayed. This is basically the essence of trying to explain complicated financial topics to people without any financial training. If you drive below the speed limit, it can take you a lot longer to get where you want to go. But, if you drive over the speed limit, you risk getting a ticket or worse, delaying you or preventing you from reaching your destination.
Calming. Many clients are anxious about their finances. This anxiety often extends to the recommendations an advisor might make regarding finances. By using understandable and quantified language, it serves as a point of validation for clients, removing some of the stress that might otherwise hold a client back. Depending on the client, this can be of paramount importance, particularly in times of market stress.
Quantitative. Some might argue that a 100-point numerical scale is not all that different from the traditional “risk buckets” employed at large financial institutions. Many large financial institutions employ a three- or five-bucket scale when describing risk. These are usually something like “conservative, moderately conservative, moderate, moderately aggressive, and aggressive.” One could argue that these are a risk scale using words instead of numbers. They would be correct in that respect. But clients react to these scales differently. The 100-point scale seems more intuitive to some people, particularly since “moderately conservative” does not have context without first understanding what the other buckets are.
There is also some resistance to the notion that there are only five types of investors in the world. Even if true, most advisor clients don’t like being told that they are part of a herd. This may be one of the very rare instances where increasing the granularity of something actually makes it more intuitive and acceptable.
A more granular, quantitative scale makes a compliance officer’s surveillance job easier, potentially allowing them to test every client on a daily basis to ensure suitability.
Things to be cautious of
The advisor may need to build the context. While the 100-point scale is a simple and understandable concept in a relative sense, some clients may still not quite understand what the numbers mean in an absolute sense. They can understand that their portfolio is scored 56 and that is less risky than a score of 65, but what does that mean in terms of risk? Advisors need to have their antennae up for this situation during their conversation and be prepared to help the clients better understand what that number represents. One approach might be to walk through a stress test to put the risk in terms of dollars. Another might be to talk about how it compares to an “average portfolio” or some other portfolio that might provide perspective. But that would depend on the client.
Need to be calibrated to the market. Many risk scales rely on historical volatility in some way. This is understandable because it is a well-accepted risk measure, and it is easy to calculate (just make sure you use total return including dividends and not just price return). However, it needs to be calibrated to the market to avoid potential disasters. The risk number should not change simply because the market is going through some increased volatility.
Volatility moves up and down over time as the market changes its stripes. In particular, it tends to increase during times of market turmoil. If your scale is based on absolute volatility, that means you will be seeing higher volatility in the portfolio and reporting higher risk numbers during a market decline. This might cause your client to panic and sell at the bottom. If your scale is based on relative volatility to the stock market, the risk ranking should be more stable during declines since volatility will increase on both the portfolio and the market at the same time. I do see a lot of risk numbers calibrated to the market, so this is usually not a concern.
Risk metric 2: “Confidence bands”
What are they?
Confidence bands are a way of explaining risk in terms of what a “bad” situation or “good” situation might look like, with the focus usually being on the former. They are generally a statistical measure that uses the volatility of a portfolio to calculate a certain probability of a particular loss. The most popular version is the “95% confidence” band. A 95% confidence band is a statistical measure saying that there is a 95% probability that you will not lose more than “xx%” over a certain period (usually a year). For example: “There is a 95% probability that you would not lose more than 17% in any given year.”
Nothing is understood better by clients than dollars and cents. The big advantage to this statistic is that it translates the unclear concept of “risk” into a percentage loss, or even better, a dollar loss based on the size of the portfolio. Bringing things down to dollars and cents is something that everyone understands regardless of their financial training.
Things to be cautious of
Need to be calibrated to the market. Similar to the need to calibrate risk numbers to the market, it can be highly counter-productive to an advisor and their clients if the “95% confidence loss” increases as the market declines. Again, this is because volatility tends to increase during market declines and as volatility increases so does the “95% confidence loss.” During a market crash, the last thing an advisor wants to tell their client is that it now looks like their portfolio could lose more than originally thought.
To mitigate this, it is possible to select longer historical periods to measure volatility. The longer the period, the less impact a recent market movement will have on the measured volatility. But one can only go so far with that before you start to include data that might not be relevant to nature of the investment. For example, a mutual fund may have changed its underlying investments over the years or a company may have changed its business mix or leverage.
I do not see a lot of “95% confidence” statistics that are normalized to the market.
Can sometimes overestimate loss during market declines. The way most confidence bands are implemented, it is always a calculation from the current date forward. This can introduce a problem during severe market declines since the statistic usually ignores the fact that the client has already incurred a portion of the projected loss. For example, if an advisor tells a client that they should not lose more than 25% with a 95% confidence, the client may find that acceptable. However, if the market declines 20% over the next month, most calculations would still come up with a 25% loss with a 95% confidence since measured volatility has not changed much. The problem is that the client has already lost 20%, so to lose another 25% from there would mean an actual loss of 40% (0.80 x 0.75 = 0.60 = 40% loss).
There is a danger at this point because the client may wonder why they were told the confidence band was a loss of 25% but now you are telling them a potential loss of 40%. It does not take a great deal of imagination to come up with a few possible client responses to this, many of which would not be good for anyone, except perhaps competing advisors.
There are several ways to mitigate this. I encourage advisors to make sure clients understand these mitigation methods and agree to them in advance, preferably in a documented way.
Risk metric 3: “Scenario analysis”
What are they?
A scenario analysis is a risk measure where an advisor uses a computer model or some other method to predict the portfolio loss under certain adverse conditions. The most popular stress test is a market crash, but there are others frequently used, such as rapidly rising interest rates or commodity spikes.
If there is a particular concern in the market at a given time, there is a scenario that might give clients a way to understand what it might mean for them.
Intuitive and meaningful. Like confidence bands, the big benefit of scenario analysis is that it can describe risk in terms of dollars and cents, which all investors understand.
Additionally, depending on what scenario you select, it can give an investor some context around how likely or unlikely that loss might be. A good example would be, “If we were to have another market crash like 2008, we would estimate your temporary loss to be 23% or $85,000.” Assuming the client was invested or news-aware in 2008, they might be able to better understand not just the loss, but how rare that is expected to be and how severe compared to other losses. This might be an important point of context for them to put the loss in perspective.
An alternative would be to select some arbitrary shock, such as a generic market decline of 25%. This will also produce a result in dollars and cents that is easily understood. However, I favor the historical scenario approach as it can lend more context to the result. If you just use a generic loss, you should also follow-up by putting that loss in the context of historical losses in terms of frequency and severity.
It can help hold clients in the market during declines. One of the more interesting benefits of scenario analysis is that it can help hold clients in the market during large declines. I am a believer in the concept that surprise intensifies emotional reaction. Perhaps this is why surprise birthday parties are often well received. Similarly, a surprise tragedy seems a little more tragic, at least at first. By talking about the potential loss of a portfolio during a calm market, you are theoretically preparing the client for the day when they might see a large temporary loss in their portfolio. I believe this to be much more effective than trying to explain risk to a client while the market is crashing.
By showing the client a stress test before there is a stress event, it may help them to avoid jumping to one of two unhelpful conclusions:
- “You put me in a portfolio that was way too risky,” or
- “The market is completely broken beyond repair and everything is going to zero.”
Many complaints about advisors contain the phrase, “I had no idea I could lose that much money,” which would be a difficult argument to sustain if you have documented showing them an analysis of how much money they could lose.
Things to be cautious of
It can overestimate loss when the market is crashing. Just like with confidence bands or any other statistic that tries to describe what a “bad” loss looks like in absolute terms, there is a fairly serious hazard during periods of significant market decline if you don’t adjust your loss estimate in a way that acknowledges the already incurred losses.
Just like with confidence bands, there are several ways to address this. You just need to make sure that the clients are aware of and in agreement with them before the market crashes. One method might be to subtract the already incurred loss from the original estimate to acknowledge that we are in the middle of the scenario and not the beginning. For example, if you had originally estimated a crash loss to be 25%, and the market and portfolio are already halfway to the bottom of the scenario, you might now say that the remaining crash loss is only 12.5%.
It is helpful from the standpoint of explaining this to clients if you have chosen an actual historical scenario, such as the 2008 market crash. That way you can point to a historical chart and say, “If this was a crash like 2008, this is about where we would be in the decline and this is what would be remaining.” Choosing a documented historical scenario might also make it more palatable for your compliance team, and you will need their sign-off on that sort of communication.
Several ways to calculate it, but none of them “perfect”. There are two main methods for calculating scenario returns, but both have flaws.
The simplest approach is to take a historical scenario, such as the 2008 crash, and see how each of the investments performed during that period to generate a historical return based on your current holdings. This has the advantage of being relatively easy to calculate and is easy for clients to understand. However, the problem is that a particular investment might not behave the same way as it did back then. Companies change around their leverage and business mix. Funds buy and sell the underlying investments. And, of course, there is the problem of what to do with investments that simply did not exist back in 2008, like Facebook or Tesla?
As a solution, some firms will look to the investment’s asset class and see how that asset class did during the time in question. The more narrowly and well-defined the asset class, the more academically sound the results are, so long as that asset class existed during the historical period in question. Of course, the trade-off here is that the investment might perform differently than the asset class by nature and some investments may not fit neatly into any asset class at all.
An alternate approach would be to figure out the sensitivity of the investment to certain macroeconomic and market factors, such as the “beta” to the S&P500 and interest rates. Next you would note how those factors moved during the period in question. To develop your return estimate, you would combine the scenario’s factor movement with the investment’s factor sensitivity to estimate the investment return under those conditions. Some popular factors are the stock market, interest rates, credit spreads, foreign currency, gold, oil, foreign markets, and so on. Most practitioners who use this method use a regression model to determine the sensitivity to each factor, similar to how people calculate the ”beta” of a stock to the stock market.
The advantage of using a factor approach is that it can solve for the two biggest problems with a pure historical approach: (1) it can handle investments that did not exist back in 2008, or whatever period you are looking at; and (2) it takes into account how the investment tends to move currently, which may be different than in the far past if the investment has changed its stripes over the years.
In addition, it considers the unique behavior of each investment. For example, you might be tempted to classify Tesla as a large-cap-growth stock from an asset class standpoint, but it is much more volatile than your typical large-cap-growth stock, statistically speaking.
As a downside, some compliance departments are not 100% comfortable with the factor approach and prefer the asset-class approach better, so make sure you check with your compliance team first.
Using investment analytics to build client understanding and acceptance is a practice that is rapidly gaining in popularity and is good for clients and advisors.
But it is not without its challenges:
- The need to be intuitive and meaningful even for clients without financial training;
- The needs to be academically supportable; and
- The need to understand the strengths and weaknesses of each analytic.
This is particularly important when it comes to portfolio risk, which is often a key consideration for clients when deciding to approve an investment recommendation.
Kendrick Wakeman, CFA, is the founder and managing partner of qSpur, a consultancy operating at the intersection of finance, technology, and people. Prior to founding qSpur, he was the founder and CEO of FinMason, a financial technology company that built one of the world’s largest and most scaled investment analytics platforms. He has over 30 years of institutional investment and technology experience across a broad range of asset classes and has developed, built, and run institutional-level analytics and risk management programs at both large and small financial institutions.