Covid-19 naturally raises the issue of how investors can incorporate the importance of such significant shocks into their risk management processes. Practically speaking, this issue comes down to three questions. First, is this environment conducive to my bond portfolio acting as a hedge or not? Second, what does the evolution of public health policy mean for my factor overweights (e.g. small cap and value stock)? Finally, can I still expect a premium from my allocation to illiquid alternatives? Answering these questions is made more difficult by the additional uncertainty of public health policy.

Unfortunately, existing data-driven tools are unlikely to give investors immediately meaningful and useful answers to these questions. It is not because the models underlying these tools are bad, or that data-driven analytics are inappropriate. Indeed, standard risk models should form a part of every investor’s analytic tool kit. It is just that these models were designed for different business end-goals. To put this differently - when installing a screw, we usually look for a screwdriver, not a wrench.  

The question for risk management in the time of Covid-19 is identifying the real source of risk, and then looking for the correct tools. Three connected ideas are helpful to framing a discussion of why standard risk models should be complemented with other analytic tools. These ideas are time horizon, structure and uncertainty.  

A straightforward way to understand why time horizon matters is to link it to the underlying business decision. For example, trading desk risk managers are tasked with analyzing whether there is sufficient capital to withstand shocks to the daily valuations. By contrast, risk managers at investment management firms (e.g. mutual funds) are more worried about the impact of shocks to quarterly performance (benchmark sensitive or otherwise). Finally, the main consideration for institutional investors such as pension funds and sovereign wealth funds is whether their asset portfolios will achieve long-term objectives such as paying pension benefits. Each of these business purposes carries with it a different time horizon and potentially different sources of risk. And, because the sources of risk could be different, the demands for data-driven analytic tools will be different.

The second key idea is structure - a theoretical model of some sort embedded in the data analysis. Simply put, the analyst proposes a theoretical model that is assumed to be able to generate the observed data. These models always require that a parsimoniously chosen set of parameters be estimated from the existing data. And, the estimated parameters are evaluated in terms of how they match against the proposed theoretical model. Adding structure would seem to be relatively less important for very short horizon risk management problems, but incredibly relevant for long-term risk management problems.

Finally, we get to the idea of uncertainty. This idea traces back to the economist Frank Knight, among others. Essentially, it says that there are some phenomena where it is impossible to define, a priori, a distribution. Uncertainty stands in contrast with risk, where estimating a distribution is par for the course. It stands to reason that solutions to problems involving uncertainty cannot be addressed with standard data-driven techniques. Similarly, it makes sense that risk managers, and the decision makers they support, will need to impose some kind of structure if they want to incorporate uncertainty into their analysis.

The Global Financial Crisis of 2008-9 provided a good illustration of how uncertainty entered into asset pricing and portfolio decisions. The uncertainty issues revolved around how long it would take to return to a positive long-term trend, and whether the trend would equal the pre-crisis trend. In fact, this was the most tangible example for investors of uncertainty in action until Covid-19.  

As distinct from the GFC, Covid-19 has multiple sources of uncertainty. Investors, and others, need to worry about the epidemiology of Covid-19 as well as the policy responses. Each of these has consequences for underlying economic growth and hence asset prices. Since we have no real data to represent the policy responses to pandemics, risk managers and decision makers will need to rely on other tools. And, they will need to use structural models to fully incorporate the results of these tools into their analysis and decisions.

What kinds of additional tools could investors and risk managers use? There are three complementary places to look. First, investors and risk managers should exploit models that explicitly link asset prices to macroeconomic events. Modern asset pricing models do this in a way that fully incorporates the role of uncertainty into a structure. Accounting for macro uncertainty help investors identify the systematic sources of long-term risk and return to factor-based strategies (i.e. value, small cap, quality) and illiquid alternatives. Second, investors and risk managers should make use of collective intelligence to better understand the dimensions surrounding policy uncertainty. Analysis from collective intelligence can be fully incorporated into modern asset pricing models. Finally, investors and risk managers should make use of timely economic and market data, possibly beyond what is available in the usual data releases.

Investors know that every decision carries risk. The three questions posed at the beginning are illustrations of risk-type questions where additional analytic tools are required in order to provide meaningful and useful analysis. In our view, the use of modern macro asset pricing models together with more timely publicly available data and crowdsourced information provides a step forward in addressing the uncertainties posed by Covid-19. By doing so, risk managers can provide richer analysis for investors and decision makers.