The limits of value-at-risk models

While markets performed well overall in 2021, the year ended on a down note. The unpleasant trend has continued into this year, with all the major indexes flirting with or descending into correction territory amid soaring inflation, the threat of rising interest rates and Russia’s invasion of Ukraine. 

Naturally, in an environment like this, financial advisers must weigh whether their clients' goals and risk tolerance remain aligned with their investments. In some cases, this is a daunting task, perhaps involving hundreds of portfolios. 

However, one way to do this scalably is to give each portfolio a risk score. Traditionally, this involves categorizing portfolios by asset class composition and examining a host of other characteristics like the market cap and liquidity of its components. While these types of risk scores can be useful, they are, at best, a simplistic way of doing things. 

VALUE-AT-RISK MODELS

That's why some firms leverage so-called value-at-risk models for portfolio risk scoring. These go a step further, measuring not only how a portfolio's underlying securities have performed historically but how each one is correlated, all of which can provide a window into future performance.

Still, even this more rigorous approach is not foolproof, but is susceptible to a range of faulty assumptions. For instance, it’s often plagued by recency bias, particularly when markets encounter turbulence. Because of this, when a model spits out information is just as important as what it reveals.

Yet the more significant issue with value-at-risk models is that they fail to recognize that firms, advisers and clients all have different definitions of what's risky at a given point in time — and, importantly, what's not. In other words, even as risk is relative to the context of the investing landscape, value-at-risk models treat it as if it's not context-specific.

To illustrate why this is so important, think about how model portfolios act as a proxy for a firm's view of risk for a range of tolerance levels, regardless of short-term turbulence. This dynamic has a far-reaching impact, including informing how its advisers speak to clients about the market.

In theory, this ensures a level of consistency: The outlook is clear, and everyone's perception of risk is similar, depending on which model portfolio best reflects the client's investment strategy. The reality, though, is more complex.

NORMALIZATION

Let's say a firm has a higher-than-average tolerance for short-term volatility. If, for instance, others are worried that higher interest rates, persistent inflation and geopolitical turmoil pose a lasting threat, it believes that markets will overcome these issues in short order and quickly bounce back.

Whether this viewpoint is correct is immaterial. If that's the firm’s outlook, its value-at-risk scores should get normalized accordingly, across its conservative, moderate and aggressive model portfolios. Naturally, each client portfolio would also get the same treatment. But this isn't usually what happens, which could leave gains on the table.

Indeed, because unnormalized value-at-risk models tend to consider risk in a vacuum — meaning they make no distinction between short-term turmoil and something more structural — an adviser could believe that a portfolio has become too adrift from its stated intention.

As a result, they may be compelled to make a series of across-the-board adjustments to client portfolios, even as the environment is no riskier than before, based on the firm's view of risk and, thus, their clients’ understanding of it.

Is anyone well-served by a context-free process like that? Hardly. In fact, it’s every bit as simplistic as assigning a crude risk score

Sure, for some professionals, it's reasonable, or even appropriate, to have a bias for action. Firefighters, for example, frequently deploy without knowing the severity of a fire because it's better to be on the scene for a minor incident than back at the firehouse for a catastrophic one.

But financial advisers have a different mandate. Sometimes action is warranted. But other times, it's not. It all comes down to having the right tools armed with the right information.

Dr. Andrew Aziz is the chief strategy office at d1g1t, a Toronto-based fintech platform