This article isn’t strictly about infrastructure, but is about the search for defensive yield – which is often one of the justifications for a strategic asset allocation to the infrastructure asset class. Defensive yield is a low interest rate world is challenging!

Standard Risk Measure Background

Since 2010, APRA has required superannuation fund trustees to provide disclosure on the investment risk of the various investment choices offered by funds. Core to this disclosure obligation has been to provide a Standard Risk Measure disclosure for each investment option.

This Standard Risk Measure defines risk as the probability of a negative return (not its size). Super funds have to disclose the expected number of negative annual returns over a 20 year period and then classify each option accordingly (see table of Risk Labels below). This disclosure obligation creates a firm link between the assessed probability of a negative return and the riskiness – as described to members in fund’s marketing material – of an investment strategy.

In a more normal interest rate world, cash/fixed income offered a high return, particularly relative to the standard deviation of returns. Return to risk ratio (RR) that is the ratio of the expected return to the standard deviation of returns.

RR ratios are directly analogous to probability of a negative returns. The higher the RR ratio the lower the chance of a negative return. The chart below shows return outcomes both assuming a normal distribution as well for a leptokurtic or fat tailed distribution. While negative returns are more probable with fat tails – the overall shape is very similar.

Compared to bonds, equities have an expected return lower than their standard deviation – this results in a reasonably low RR ratio. Roughly a third of the time equity returns are negative. For this reason, a pure equity portfolio will typically be expected to have 6 or greater negative returns in a 20 year period and score a “Very High” risk rating under the standard risk measure.

But all is not lost, in constructing diversified portfolios you could mix cash/bonds (a high RR ratio) with equities (low RR ratio) and get whatever blended ratio you wanted. This meant you could dial in whatever probability of a negative return you wanted.

To illustrate this, I have constructed a very simple asset allocation model of the form used by superannuation funds and their asset consultants to underpin their strategic asset allocation decisions and risk disclosures.

My model has only two assets classes – bonds (proxied by the Bloomberg Ausbond composite) and Shares (proxied by the ASX200). My model assumes normal distributions (no fat tails) and ignores taxes and fees. It has been calibrated on risk data from 2012 to 2020 (monthly).

While more sophisticated modelling and a broader range of asset classes would be interesting – I would argue that the illustrations from this simple two asset class model show the heart of the modelling that underpins the majority of superannuation fund’s strategic asset allocations and standard risk measure disclosures.

The main aspect that might significantly affect results is the inclusion of alternative assets (that is private market assets like infrastructure). Readers should note that the modelling of private market assets – where, by definition, there aren’t reliable benchmark indices on which to base statistical models – falls very much into the dark art (aka witchcraft) rather than the scientific end of risk modelling – but I will leave that question and debate for another article.

Using my simple two asset class model, the chart below shows the probability of a negative return at a portfolio level and how this varies based on allocation to bonds. At 100% bond allocations this risk is very slow (sub 5% or 1 year in 20 – enough to get a Low standard risk measure). It then rises sharply as the allocation to bonds falls.

Note capacity to tune probability of negative returns between high (25%) and very low (<5%) by choice of asset allocation.

But what happens if risk free rates go to near zero?

If interest rates are very low, the RR ratio of bonds collapse. The RR ratio of bonds goes from 1.8 to 0.36 (see table 1). Now the RR ratios of shares and bonds are very similar. This means all portfolios to have the similar RR ratios and probability of a negative return.

The chart below shows the transition in the probability of a negative return frontier as base rates have collapsed (blue is historic data from 2012 to 2020, orange is a prospective outlook based on 2017 risk free rates and yellow is based on today’s 1% bond rates).

Note that the curve is basically flat. All portfolios have much the same risk of a negative return (they would all in the High category under the standard risk measure).

Key conclusions:

· There are Significant challenges in maintaining differentiation of investment choices under standard risk measure. There should be a focus on communicating the size of potential negative returns/drawdowns to members which becomes much more meaningful in differentiating the risk profile of different investment choices.

· The most valuable assets classes are those that maintain relatively high RR ratios in low interest rate environment. Vanilla fixed income doesn’t deliver any more. Some areas within credit and alternatives probably do. The key challenge is to find assets that offer high yields but aren’t just equities in disguise. Key theme music – we think infrastructure debt can offer pretty attractive risk adjusted returns – but it is these types of assets, anything that can deliver reasonable income/yield/return with modest risk, that will be highly sought after by investors.

· Standard risk measure has always been pretty dumb. Principally because it’s not standard – each fund uses its own asset allocation and return modelling to assess risk and so results aren’t comparable between funds. This is going to get even worse as assumptions regarding the risk modelling of alternative assets are going to be the key driver of standard risk measure outcomes and this is the area where there is the greatest variation in modelling approaches and assumptions.