White Papers

Avoiding the valuation traps in SAA

18/10/2012

Chris Durack

Chris Durack

Country Head of Hong Kong and Head of Institutional Asia Pacific

Background

Common wisdom suggests that an average investor aged in their 20s or 30s might be expected to tolerate higher levels of volatility in exchange for higher long run expected returns. The risk reward trade-off is typically expressed in terms of a strategic asset allocation (SAA). For example, portfolios with higher expected returns (and higher expected volatility) might be invested 70% in growth assets and 30% in defensive assets while portfolios with more moderate expected returns (and lower levels of expected volatility) might be invested 30% in growth assets and 70% in defensive assets.

Investors are ultimately interested in the (money weighted) returns they earn rather than the long term averages often used to justify a particular SAA. To this end, we argue that relatively static portfolios can expose cohorts of investors to asset classes with systematically unattractive features. As an investor ages, and is exposed to increasingly higher proportions of defensive assets, this can (and does) occur in periods where growth assets display more favourable risk-return characteristics relative to defensive assets and their own history. The same can be said of defensive assets. This situation comes about due to a lack of incorporation of the longer term valuation and cyclical characteristics of asset classes. Essentially, the risk reward relationship of asset classes is non-stationary, so relatively static SAA portfolios, by definition, cannot offer stable risk-return outcomes. In summary, if these issues are not recognised in the design features of portfolios, they risk working against the objectives of investors.

In this paper, we argue that incorporating strategies that address the valuation characteristics of asset classes over the medium term (so called “objective based”) can significantly improve the behaviour of portfolios that continue to operate their overall structure driven by an SAA framework. We consider from first principles how objective based strategies should be treated in broader investment frameworks, including whether they possess growth and income asset class characteristics and consequently how such allocations should be funded.

Evaluating objective based investment portfolios from first principles

In our view there are two key considerations for investors when evaluating objective based investment strategies. The first is to make an assessment of the robustness of the conceptual investment framework, people and process applied to achieve the stated investment objectives. The second is to determine when such a portfolio might properly be allocated within a broader (say multi-manager) investment framework. The answer to the second of these considerations will, in our view, depend on an evaluation of the following:

- Can the characteristics of an objective based strategy be sufficiently identified?
- Are they fundamentally different to other asset classes?

If the answers to the above questions are yes, then there are strong arguments for an objective based strategy to be treated as a separate “asset class” for the purposes of determining an allocation. The funding of such an allocation would then be determined as part of an overall portfolio that is most efficient in meeting the defined objective for the overall portfolio. We look at these issues in detail below to draw some conclusions.

What are the investment characteristics of an objective based portfolio?

In the general case, an objective based fund may be defined as a portfolio designed to deliver on an explicit investment objective(s). The objective may be to deliver a target rate of return with the least risk of underperforming it over a nominated investment horizon. In the case of the Schroder Real Return Fund, the objective is to deliver a target rate of return of CPI+5% p.a. over rolling three year periods while minimising the risk of underperforming this target. Most importantly, the fund is explicitly managed to target this outcome, it is not a function of underlying assumptions.

When it comes to building portfolios, in the same way that assumptions are formed about the behaviour and expected returns of the major asset classes, similar assumptions can be made in relation to objective based portfolios. It is common to observe researchers articulate assumptions for asset classes such as equities, fixed income and cash on the basis of their expectations for economic growth, inflation and risk premia (as a function of things like liquidity, volatility and default risks). These assumptions are used to inform the expected return, volatility and correlation of each asset class but also tend to be quite long term in nature given the assumptions are based on equilibrium expectations.

For objective based portfolios, when forming expectations about returns, volatility and correlation behaviour, the same information is also available but in addition to the unconditional behaviour of asset classes which may be representative of the long run expected behaviour of capital markets, the conditional behaviour of asset classes is also explicitly taken into account. The conditional behaviour of an asset class may be seen as the expected return, risk and correlation assumptions of an asset class given that a pre-existing condition has been observed. For objective based strategies it is not simply the expected behaviour of an asset class over all periods that matters most - the unconditional distribution – rather, it is the behaviour of an asset class conditional on some observable (and causal) factor(s) that really matters when building objective based portfolios – the conditional distribution.

Therefore, when making assumptions about an objective based portfolio in a comparable manner to traditional asset classes, we can set clear expectations based on the behaviour of such investment portfolios using more information than simply the sum of unconditional expectations and instead use conditional expectations about the behaviour of asset classes used to build the portfolio. In the case of the Schroder Real Return strategy, this approach enables a robust framework for setting expectations of a real rate of return over the long run (and over rolling three year periods) of CPI+5%, expected volatility of around 5%, and expected covariances with other asset classes.

Are the characteristics of objective based portfolios fundamentally different to other asset classes?

Using the analysis outlined above, we can show that it must be the case that the characteristics of an objective based portfolio will fundamentally differ to other asset classes. The key reason for this is the portfolio’s use of conditional (as opposed to unconditional) distributions to determine its composition. Rather than assume the long run equilibrium expected return and risk characteristics, the objective based portfolio will assume that markets are always in disequilibrium – or in an adjustment process – relative to long run return expectations. When viewed this way, the views on long run are conditioned on important observable information such as valuation, cycle and liquidity variables to determine expectations for the forward three year period.

While the approach to determining the behaviour of objective based portfolios may be significant in theory, a simple example can assist in illustrating the difference in practice. In the table below, we show the behaviour of US equity market on an unconditional basis versus a conditional basis over the period from 31 December 1899 through to 31 July 2012. In showing the conditional distribution for equities over this period, the factor used to condition the data was a simple valuation measure, price earnings (PE) multiples using reported earnings. Over the period we examined observed monthly PE ratios and then examined the three year equity market return that immediately followed in each case. By doing this we were able to disaggregate the unconditional equity distribution (entire sample period), into three conditional equity distributions:

- one distribution of returns based on conditional “cheap” valuations (i.e. when PE ratios were less than 10 times),
- one distribution of returns when equities were in a range of “mid” valuation (i.e. when PE ratios ranged between 10 times and 20 times), and then on a distribution of returns conditional on valuations that were “expensive”.

  Unconditional Conditional
PE<10x 10x<PE<20x PE>20x
Number of observations 1316 278 880 158
Average 3 year return 9.8% p.a. 16.4% p.a. 8.8% p.a. 4.3% p.a.
Max 3 year return 42.4% p.a. 36.5% p.a. 42.4% p.a. 29.2
Min 3 year return -42.7% p.a. 0.5% p.a. -42.7% p.a. -34.7% p.a.
Frequency < 0% p.a. 16% 0% 17% 35%
Frequency between 0% p.a. and 10% p.a. 34% 28% 37% 25%
Frequency > 10% p.a. 51% 72% 46% 40%
Return range 85.1% 36.1% 85.1% 63.9%

*Maximum 3 year return p.a. less minimum 3 year return p.a. based on monthly data
Source: GFD, Schroders

Key observations

- The characteristics of the conditional distributions are fundamentally different when compared with the entire (unconditional sample). Most notably the average 3 year return in the “expensive” market was 4.3% p.a.. When PE ratios were in the “mid-range” the distribution of rolling 3 year returns showed an average return of around double the expensive sample at 8.8%. This average return then roughly doubled again when conditioned on “cheap” market valuations.

- Again, in terms of range of return outcomes, each distribution showed marked difference. Of note here was the absence of a negative three year return when the starting PE ratio for the market was less than 10 times. The overall range of returns for both the “cheap” and “expensive” conditional distributions were both less than the unconditional and “mid-range” conditional distributions indicating a higher degree of consistency in return outcomes in both cases. Positive for the cheap market, but negative for the expensive market.

The main reason for showing the analysis above is to demonstrate that a conditional distribution will not have the same characteristics as an unconditional distribution. While the underlying drivers of an asset class may be common to both distributions, the act of conditioning on an explanatory variable (such as valuations) will alter the behaviour of that asset class distribution from a risk and return perspective. This in turn leads to some portfolio construction insights which can be used to ensure that the incorporation of objective based portfolios in a broader investment framework is dealt with in a robust manner and makes sense from first principles.

Is growth always growth and defensive always defensive?

It may be readily seen that the drivers of asset class returns over time are a function of the income produced by the asset class, the growth in income of the asset class over time and the amount that is paid (the valuation) for the income and growth for the asset class. At Schroders, each asset class, is decomposed into these three elements and can be combined to produce the following formula:

R = Y + G + V

Where:
Y is the current investment yield
G is the annualised growth in income or earnings from the asset
V is the valuation effect

All asset classes, whether growth or defensive, produce income . However, the common definition of a growth asset using the framework outlined above is where G does not equal zero and is over longer periods expected to be positive. In other words, there is a long run expectation that the income produced by the asset class will grow over time. It is readily apparent that both equities and property possess this characteristic. On the other hand, fixed income and cash asset classes do not have annualised growth in their earnings and therefore do not meet the definition of a growth asset.

On an analysis of income and its growth path over time for each asset class, the definition of growth and defensive is not problematic. However, what about the impact of valuation on returns? If objectives are set over the medium term, say on a rolling three to five year outlook, the impact of valuations on returns can be large and experience has borne this out. If, at the start of a three year period, the set of valuation indicators show a highly overvalued market, all other things being equal, this may lead to a view that the “V” effect would be negative over the following three years as valuations revert to more normal levels. If this is the case then the unconditional growth asset may be held to be conditionally likely to produce negative returns.

By similar analysis, defensive assets may be held to be conditionally not defensive if their valuation effect is expected to outweigh the investment yield. Current yields, on sovereign bonds may well serve to highlight these issues.

Overall, when asset classes are decomposed into their major return drivers, it can be seen that growth is not always growth, nor defensive always defensive in a conditional sense. If these issues are of concern to investors in seeking to address the impact of valuation in their overall portfolios, the question as to how to fund an investment in an objective based portfolio where valuation is always conditionally assessed should be in reference to the best portfolio composition relative to the objectives set for the portfolio, not based on a preservation of the existing asset allocation of the strategic asset allocation on a look through basis.

Portfolio construction implications using the Schroder Real Return strategy

As discussed above, the Schroder Real Return strategy is constructed using the conditional distributions of each asset class by first understanding the drivers of the relevant return distribution over long run periods, then conditioning these distributions on valuation, cycle and liquidity variables. Given that conditioning on relevant and causal variables fundamentally alters each asset class distribution relative to its unconditional counterpart, this means that the overall real return fund characteristics should be used to determine the appropriate allocation in a broader multi-manager framework. This is in contrast to “looking through” to the underlying asset allocations of the objective based portfolio and adjusting the overall multi manager asset allocation weights on that basis.

Consequently, for example a 10% allocation to equities in the Schroder Real Return strategy should not be funded by a directly proportional reduction in equities from the broader framework given that the former allocation is based on the conditional distribution and the second on the unconditional distribution. As has been demonstrated, the two distributions are fundamentally different. In this context, an allocation to an objective based strategy with a long run return of CPI+5%p.a and volatility of between 5% to 7% p.a. should be the unconditional assumptions used and these characteristics can compete with other unconditional asset class distributions to be optimised to meet overall portfolio objectives.

In practice, we find that when the unconditional characteristics of an objective based portfolio such as the Schroder Real Return strategy are used as inputs in a mean-variance optimisation framework to determine an overall multi-asset class (and multi-manager) portfolio, the optimisation process will converge on a very high weighting to the objective based portfolio. This is particularly the case where similar overall real return objectives are targeted to those of the objective based portfolio. The reason for this is that through the use of conditional asset class distributions to build an objective based portfolio, the risk characteristics (expressed as say a standard deviation) will generally be superior in the case of the objective based portfolio compared with the long run (unconditional) strategic asset allocation derived from the optimisation process. This comes about through the process outlined above in the example of US equity markets where a superior conditional distribution is selected in preference to the unconditional distribution.

While an optimiser will therefore prefer a very high allocation to objective based strategies, there may be reasons to constrain the allocation to these strategies in a broader framework. Three major reasons are analysed below.

  1. Peer based objectives

    While a CPI+4% to 5% return objective is very common as a primary growth portfolio objective, risk objectives can often be defined in terms of a peer group. It also remains reasonably common to observe investor objectives for (say) growth portfolios to target returns above the median manager/fund in nominated peer surveys. For investors where these objectives are in place (or implied), the path by which returns are generated becomes highly sensitive. To diversify too far away from the peer group’s asset allocation will involve taking higher levels of risk relative to the peer objective. This “peer relative” problem in the industry has led to an incumbency bias towards the use of reasonably static portfolios predicated on unconditional asset class distributions at the expense of more efficient portfolios that process and incorporate more available information.
  2. Return maximisation

    In a number of cases, the objectives set for an investor’s portfolio may be to target a return in excess of a target real rate of return. For example, a portfolio objective may be set to achieve a return in excess of CPI+5% p.a.. However given the aggressive nature of the objective, this target would usually be associated with a defined tolerance for downside outcomes including negative returns. In such cases, for reasons of diversification, it is rare to see such portfolios allocated solely to equities asset classes to address the potential downside outcomes. Again, due to its fundamentally different return drivers, an objective based portfolio when included in such a configuration to meet a (constrained) return maximisation portfolio will increase the efficiency of these portfolios relative to their objectives.
  3. Diversification of approach

    In the case of a multi-manager portfolio framework, it is currently commonplace to observe “growth” portfolios with real return targets of CPI+ 4% to 5% invested (say) 70% in growth assets (shares, property and growth alternatives), and 30% in defensive assets (bonds, cash and defensive alternatives). If an objective based portfolio such as Schroder Real Return strategy competes for an allocation in such a growth portfolio, based on its long term characteristics, a mean-variance optimiser will allocate close to 100% of the preferred (i.e. risk adjusted) weight to the objective based portfolio. However, if a multi-manager structure is required it may simply not be possible to identify a suitable number of objective based managers to given the required level of “manager diversification” for a high allocation to objective based strategies. In any case, regardless of the unambiguous efficiency improvement based on the optimised results, even a constrained allocation of around 20% would add meaningfully to the efficiency of such an overall portfolio. This improvement may be seen as being incorporated without changing the overall portfolio construction approach and execution using a diversified panel of investment managers. Furthermore, the approach may be seen as diversified (rather than just the mix of asset classes) because the drivers of returns in the objective based portfolio are fundamentally differentiated for the reasons discussed above.

The evolution of portfolio weights

We have already seen examples of investors in the Australian marketplace allocating to objective based strategies as the majority and in some cases entirety of their holding. The reasons driving the change to this portfolio construction is the recognition of the issues raised in this paper in pursuit of a more targeted focus on real return outcomes over relevant time periods for investors.

Disclaimer

Opinions, estimates and projections in this article constitute the current judgement of the author as of the date of this article. They do not necessarily reflect the opinions of Schroder Investment Management Australia Limited, ABN 22 000 443 274, AFS Licence 226473 ("Schroders") or any member of the Schroders Group and are subject to change without notice. In preparing this document, we have relied upon and assumed, without independent verification, the accuracy and completeness of all information available from public sources or which was otherwise reviewed by us. Schroders does not give any warranty as to the accuracy, reliability or completeness of information which is contained in this article. Except insofar as liability under any statute cannot be excluded, Schroders and its directors, employees, consultants or any company in the Schroders Group do not accept any liability (whether arising in contract, in tort or negligence or otherwise) for any error or omission in this article or for any resulting loss or damage (whether direct, indirect, consequential or otherwise) suffered by the recipient of this article or any other person. This document does not contain, and should not be relied on as containing any investment, accounting, legal or tax advice.

Important Information:
Opinions, estimates and projections in this article constitute the current judgement of the author as of the date of this article. They do not necessarily reflect the opinions of Schroder Investment Management Australia Limited, ABN 22 000 443 274, AFS Licence 226473 ("Schroders") or any member of the Schroders Group and are subject to change without notice. In preparing this document, we have relied upon and assumed, without independent verification, the accuracy and completeness of all information available from public sources or which was otherwise reviewed by us. Schroders does not give any warranty as to the accuracy, reliability or completeness of information which is contained in this article. Except insofar as liability under any statute cannot be excluded, Schroders and its directors, employees, consultants or any company in the Schroders Group do not accept any liability (whether arising in contract, in tort or negligence or otherwise) for any error or omission in this article or for any resulting loss or damage (whether direct, indirect, consequential or otherwise) suffered by the recipient of this article or any other person. This document does not contain, and should not be relied on as containing any investment, accounting, legal or tax advice. Schroders may record and monitor telephone calls for security, training and compliance purposes.