The Covid-19 shock has brought sharply into focus the inherent uncertainty of life. Such uncertainty extended into financial markets as investors grappled with the unprecedented nature of the shock, policy response and reaction of asset prices. The experience of the past two years has reinforced the benefits of incorporating uncertainty into portfolio construction. Standard investment models only consider risk – situations where we do not the know the outcome but can measure the odds of different outcomes. But they often ignore uncertainty – situations where we may not have enough information to even define the odds.
Arguably, the coming years present a scenario that remains highly uncertain. New central bank frameworks are being put to the test, a rewiring of global supply chains is taking place and major shifts such as the transition to net zero are underway. Such a backdrop warrants some humility around economic projections and estimates of asset price returns – the building blocks of strategic asset allocation. Importantly, we need to acknowledge that there is no single portfolio that will be optimal for the wide range of significantly divergent yet plausible economic outcomes.
Yet traditional portfolio techniques, such as mean variance optimisation, take the approach of seeking to achieve an ‘optimal’ asset allocation. They arrive at such a portfolio by placing too much confidence in a set of economic and market estimates that leaves little room for uncertainty. The risk is that such an approach typically leads to concentrated – or less diversified – portfolios that are ill-prepared for outcomes that differ materially from a base case.
How can uncertainty be embraced in portfolio construction? A recent joint paper from GIC and BlackRock discusses how uncertainty can be better incorporated in strategic asset allocation to overcome shortcomings in traditional portfolio construction methods.
One approach is to seek to minimise ‘regret risk’. Here, a handful of reasonably probable macroeconomic scenarios could be modelled explicitly using key variables, such as inflation, growth and valuations, which would then lead to a set of capital market assumptions – or long-run expectations for asset class returns. Optimal portfolios for each scenario are identified and then blended using scenario-probability weights. The resulting blended portfolio may not be perfectly optimal for the outcome that unfolds, but it seeks to minimise the lost returns from reality being too different from the scenario-optimal portfolios.
Another approach seeks to ‘minimise downside risk’. Instead of explicitly specifying alternative scenarios, thousands of possible market outcomes are simulated within a set macroeconomic scenario and used to reflect both the uncertainty and volatility of future asset returns. When constructing portfolios, we seek to identify the portfolio with the ‘best worst outcome’, defined as the best average outcome in a given proportion of the simulated outcomes with the biggest downside. The choice of what proportion to select ought to reflect the investor’s aversion to uncertainty and risk. This approach accounts for the behavioural finance finding that the pain felt from losses exceeds the joy felt from (similarly sized) gains.
Both approaches seek to achieve a desired portfolio outcome: a more diversified asset allocation than under the traditional approach and one that is likely to perform better in adverse market environments. Both approaches also provide investors with the flexibility to reflect their specific aversion to uncertainty. While the first approach explicitly specifies a few alternative scenarios and their likelihoods, the second approach defines a full range of plausible outcomes. These two approaches are not mutually exclusive: for each alternative macroeconomic scenario, a wide range of market scenarios could be calculated and used as an input for portfolio construction.
Incorporating uncertainty serves two important purposes. First, it acknowledges that most investors might not have full conviction on a specific value for expected returns. More plausible is that asset prices may follow a few potential pathways. Accounting for these pathways can either be done by adding uncertainty to the input variables – that is the expected return itself – or by employing bespoke capital market assumptions for different macroeconomic scenarios. Second, the variability in levels of uncertainty across time and asset classes can be captured by attaching different probabilities to certain macroeconomic environments or varying the level of uncertainty by asset class. Why is this important? A lower ability to estimate returns for one asset class – for instance when an asset’s returns are poorly explained by well-known public market factors – should warrant higher uncertainty around its expected returns.
Most of us deal with uncertainty in our daily lives by considering various scenarios and staying flexible and adaptable. As investors, we should look to mirror this approach when designing long-term portfolios in today’s financial markets. There is no guarantee that this alone will lead to better returns, but it could help in making more deliberate investment decisions and making communication with key stakeholders more effective.
Stephan Meschenmoser is Senior Portfolio Strategist at BlackRock’s Official Institutions Group and Anthony Chan is Researcher in the Portfolio Research Group at the BlackRock Investment Institute