Investing in retirement is especially challenging, as you are typically operating within an uncertain timeframe and battling the return drag caused by regular withdrawals. While such challenges require all of the tools at the disposal of the investment manager, for the majority of investors saving for retirement, the use of fixed income is the most important consideration, as these assets typically provide greater certainty in the short term together with a useful income stream.
Traditionally, a simple retirement proposition may have comprised fixed income and equities to achieve a safe withdrawal rate as an income stream. However, as the investment universe has expanded and interest rates pushed lower, this "safe withdrawal rate" has seemingly declined. In a modern context, the portfolios used in retirement typically now include property, infrastructure, commodity-related assets, currency exposures and other alternatives.
So, how do these assets complement fixed income? While this causes a significant amount of confusion for everyday investors, it is worthwhile stepping through the benefits of smart diversification and the ways one can think about improving total portfolio outcomes.
The equity/fixed income split
To start with what is most important, the equity to fixed income split remains the single biggest contributor to portfolio returns and risk in the vast majority of instances. The reason for this isn't necessarily due to correlations, but rather due to sizing.
As a generalised statement, a retiree yearns for low correlations between their key exposures as it diversifies the performance drivers and can smooth outcomes. Interestingly, an investor that has been predominately focused on fixed income could seemingly do a better job of diversifying than a simple fixed income/equities split. For example, the correlations to property, private equity, commodities and hedge funds have also offered diversification benefits over the past 10-years.
So, does this mean a defensive investor would be better placed mixing government bonds, private equity, property, hedge funds and commodities? Not necessarily. Correlations are very useful in understanding relationships, but they can be grossly misleading in understanding the total portfolio impact. The problem here is that correlations only explain the way the assets move together on average. It fails to tell you that these relationships can break down for considerable periods in stress. It also fails to tell you anything about valuations and the contributions to risk or return.
Thinking in a valuation-driven context
A better way to think about diversification is to assess "valuation-conditional drawdowns." Specifically, assets with lower valuations typically deliver higher returns and suffer lower drawdowns. The deeper the drawdowns, the more capital the retired investor will have to sell to generate sufficient income. Furthermore, the more capital they sell, the less they will have to deliver future returns. This spiral effect can cause the safe withdrawal rate to decline.
To take this to the present, there is no point comparing bonds at a 2% yield and expecting them to act the same way they did when they offered a 10% yield. Similarly, one shouldn't compare an equity market trading at an all-time high and expect it to behave the same way as an equity market that has already fallen 50%. Valuations matter.
The complexity of interpreting such information in a total portfolio context should not be underestimated, and there are several problems in comprehensively assessing this. First, an investor needs to have a means of understanding how expensive or cheap an asset is relative to fair value. Second, they then need to understand the way an expensive bond may move relative to an expensive/fair/cheap asset. We devote a lot of time to producing these as critical inputs to a valuation-driven risk assessment.
Unfortunately, the problems don't end there either. Inherent within all of these assessments are data reliability issues and historical objectivity. Obtaining sufficient long-term track records can be practically impossible in an asset class such as the emerging markets. It is especially difficult to compare apples with apples when one asset such as U.S. equities has a historical record back to 1871 and another such as emerging market debt only back to 1996.
The second element is related to experience. For example, assessing valuations for the technology sector relative to the utilities sector can be difficult due to varied drawdown histories. Utilities simply haven't experienced the equivalent to a "tech wreck," even though they tend to carry debt levels that wouldn't rule out such a disaster. The same can be said for shifting markets. While emerging markets are the obvious candidate, even global fixed income has seen dramatic shifts over time as Japan raised more and more debt via government bond issuance.
So, is it possible to accurately assess "valuation-condition drawdowns" relative to the valuation-implied returns? Absolutely. The solution is to consider both historical and simulated outcomes. While the historical analysis is useful in checking how a portfolio would have reacted in a financial crisis, the simulated outcomes also allow one to consider the unknown. It is backward-looking meets forward-looking -- and while there are issues in the design, it is all about reducing the ignorance one carries and to improve the total portfolio outcome.
Bringing this back to the goal of investing
Portfolio construction is an underrated element of the investment process. Using risk analytics tools is a great way of reducing any ignorance while building robust solutions. However, it is not a silver bullet. Ultimately, a portfolio is there to achieve outcomes, but the future cannot be predicted with precision and it is not possible to create something out of nothing.
This reinforces the need for a total portfolio viewpoint. It also emphasises the requirement for advisors to help clients by ensuring that they are not trying to achieve the impossible and drawing too much income. Therefore, portfolio construction should start with sound limits and an understanding of how behavioural biases can upset returns.
