Asset allocation - We can't afford to be passive in asset allocation
Countless shots have been fired in the active versus passive management debate, with most academic research landing on the side of passive management - even the odd government enquiry has questioned the relevance or need for active management. The debate around active management is usually framed around the security selection part of the investment decision. Yet countless studies (and more importantly observation) suggest that the most influential decision on long-term investment outcomes is not security selection but asset allocation. The degree to which investors adopt a passive approach on asset allocation has a much greater bearing on investment outcomes than security selection.
The purpose of this paper is to provide an alternative view on the active versus passive debate from this much more important perspective of asset allocation that also takes into account investors’ objectives. In summary, we show that if investors adopt a completely passive approach to investment they will require substantially higher contribution rates in order to achieve the same degree of certainty of outcomes. Put simply, a purely passive approach is too expensive in aggregate to be a consideration for investors with real (non-index) liabilities. An active approach is the only viable alternative.
Typically the debate starts (and finishes) with the mathematical understanding that by definition all the active participants must collectively equal the passive market, and given the higher fees associated with active (trading and management), active must underperform passive. Sharpe’s "The Arithmetic of Active Management"1 hit this on the head over 25 years ago. There is no debate here.
However, while true, these mathematics are premised on a number of important assumptions:
The mathematics are only correct if the particular opportunity sub-set that is defined as the passive portfolio – "off index" does not fall into this calculation;
It only applies to market cap weighted indices (i.e. listed equities) – there is no capital-asset-pricing model for bonds and a passive approach to bonds is counterintuitive (greatest exposure to the entity with the most debt). The growth in smart-beta and "fundamental" index-based strategies are an indication of investors views on the relevancy of cap-weighted indices to achieving objectives;
It only applies in a time-weighted construct, not money-weighted (in fact any decision to invest or divest in a non-systematic way is an active decision);
It applies only when all active market participants are included – e.g. professional money managers, retail/individual investors, corporates, foreigners and governments or other public institutions etc.
It is framed around total returns only – no account is taken of real portfolio risks or underlying investor liabilities, liquidity preferences, income preferences, time horizons and the differences in these (i.e. an index or less than index return over time with lower risk is absolutely worth paying for – indeed, it is part of modern portfolio theory).
It is relevant to liquid, listed markets only. Illiquid and private assets cannot be invested passively.
Warren and Ezra2 outline these and other points in considerably more detail, in particular that "a passive approach no longer becomes optimal under a number of conditions. These conditions extend beyond the issue of market efficiency and beyond the choice between passive and active management as traditionally defined. They include questions about the efficacy of the index itself and how well it meets investor objectives."
In addition, all of this is only relevant to the extent that there is, in fact, a passive solution to the investment problem. In particular we note that most investors’ liabilities are NOT defined by market-cap-weighted, asset-class based indices, but are more likely:
a. Real liabilities (e.g. inflation plus);
b. Absolute liabilities (e.g. cash plus);
c. Bond valuation based (e.g. NPV driven liabilities such as pension liabilities); or
d. Cash-flow or liquidity based (e.g immunized insurance portfolios).
This latter point changes the dynamics of active management completely – in fact in respect of a), b) and c) above, where some level of risk is required to be taken over and above the liability portfolio, active management of the asset allocation decision is the ONLY feasible approach to take. A passive approach will result in a guaranteed poor outcome a significant proportion of the time.
While individual active security-selection strategies may be able to demonstrate persistent added value, it may be that the implementation of certain parts of the management can (and indeed should) be undertaken passively (or quasi-passively). In most cases, however, (and certainly in the case of retirement income funding), it is not possible to construct a completely passive portfolio that will meet the required liability hurdle over the required time frame. More to the point, investors who do passively construct portfolios to meet these liabilities, as is effectively the case with strategic asset allocation (SAA) portfolios, are taking considerable liability risk in building their portfolios. It may be cheaper to implement, but it is also quite likely not to work.
This is best demonstrated by our prior work on "Why Strategic Asset Allocation is flawed".3 In summary, we showed that the popular approach to investing for longer-term inflation-related liability profiles of using an SAA to represent the benchmark (and by definition the liability portfolio) does not work. The SAA varies dramatically from its liability portfolio for long periods of time. This creates large asset and liability mismatches and, together with cash-flow timing constraints from most investors, makes achieving objectives unrealisable a considerable proportion of the time.
Chart 1: Rolling 10-year real returns on a balanced portfolio
With a 20-year time horizon, the best that can be achieved on this history with 80% probability is CPI+2.7% - this is from a strategy that has a long-term time-weighted return of CPI+5.3% p.a. The medium- to long-term volatility of the completely passive approach to asset allocation is highly destructive to achieving objectives.
In a subsequent paper "Anchoring asset allocation to valuation rather than averages",4 we showed how a simple, rules-based approach to changing the asset allocation that was focused around valuation would significantly improve investment outcomes.
Overall, based on data from 1929 to 2014, the conditional SAA model outperformed the traditional SAA model by an average of 1.5% p.a. (which relative to the total average return of the 70/30 SAA portfolio is an improvement of 29.7%). In addition, the performance was generated more efficiently in risk-adjusted terms with a Sharpe ratio of 0.83 versus 0.47. The following chart shows the relative performance of the conditional SAA model versus the traditional unconditional SAA model on a cumulative basis.
Chart 2: Cumulative portfolio performance – SAA versus conditional SAA
Source: Schroders, "Anchoring strategic asset allocation to valuation rather than averages"
Defined contribution (DC) schemes
The objective of retirement scheme funding is generally to provide a level of income in retirement that will support the retiree in whole or in part with a reasonable degree of certainty. There are two critical risks to achieving this for the typical retiree. Firstly, that there is insufficient money accumulated to provide a reasonable retirement income stream (largely a function of investment risk and return) and, secondly, that the retiree lives longer than expected and the income stream is exhausted too early (longevity risk). As the world shifts from defined benefit (DB) schemes where the employer bore much of the longevity and investment risk to defined contribution (DC) schemes, where risks are borne by individuals. We concentrate in this paper on the investment part of these two risks.
By virtue of DC members having their own individual accounts, the "opportunity" to smooth outcomes across equivalent member cohorts or inter-generationally with a guarantor (as with DB schemes) is removed. This makes the provision of a stable set of outcomes at the individual level most paramount. Essentially the investment risk for DC members can be decomposed into two parts:
Overall returns through time ("generational returns") are insufficient to meet the required outcomes; and/or
The pattern of returns through time when overlaid on the accumulation and decumulation process results in a poor overall outcome (sequencing risk)
Consequently, to investigate the particular circumstances of DC members, we have to investigate scenarios that capture the full spectrum of accumulation and decumulation. In order to accomplish this, we have to utilise in the order of 70 years of returns (40 years accumulation and 30 years decumulation). While we could create a "simulated" history of returns, this obviously lacks real market relevance and, in any case, will be biased by the methodology used to simulate returns. Real investment market data (even though life expectancy and retirement funding systems will have changed) would be far more appropriate.
Unfortunately, in many markets, market data is only available since the 1900s. What this means is that in most cases we don’t have two independent, non-overlapping 70-year periods to consider. While using yearly data we can sometimes create a number of sample periods from this history, there are going to be considerable overlaps in the data set.
Analysing long-term outcomes
However, there are a few data sets that can provide a significantly richer history of returns, particularly in the UK. To this end, we have obtained UK equity, bond/interest rate and inflation data going back to 1694. The period covers the industrial revolution, numerous regional and global wars, and the invention of almost everything that today underpins modern society. We have had the South Sea Bubble (think tech bubble), the invention and use of electricity, the telephone and modern transportation systems and a variety of conflicts, all particularly disruptive events akin to today’s disruptions brought on by the internet and biotechnology.
Interestingly (but not surprisingly), this return history matches up reasonably well with the long-term history from some major western world developed markets (Australia, Canada and the US). However this is not the case for markets that have suffered severe political dislocations (e.g. Argentina, China, Japan, Germany and Russia etc). That presents a strong argument in favour of geographical diversification and non-home country bias.
While we might question the veracity of some of the older equity, bond and inflation data, the table below summarises the key outcomes over time and suggests that it is not out of line with more recent data and expectations.
*Denotes a portfolio that is 60% equities, 40% bonds. All return data in this paper is sourced from Global Financial Data. Bond returns have been proxied by using the Bank of England Base Rate (BOE) from 1695-1700, 60% Consols and 40% BOE from 1700-1932, and the UK 10 year government bond total return index for 1932-2015. Equity data is the UK-FTSE All Share Index. Inflation is the UK Retail Price Index.
Given likely changes in data collection and reporting over the time frame, we would be careful with looking beyond nominal and real returns in this dataset. However, it is interesting to note that the volatility of the equity-market return (standard deviation of annual results) was 11.9% in the 1700s, only 6.9% in the 1800s but surged to 21.6% since 1900. Given that the real return is essentially unchanged over these time frames, investors now get more risk for their return.
We note that positive long-term inflation has been a phenomenon of the 1900s and beyond (with the advent of fiat money). However, irrespective of inflation the real return on equities has been remarkably consistent at circa 5% p.a, while bond holders have been the real sufferers from inflation with average real bond returns declining from 4% to 5% p.a. real in the 1700s and 1800s to 2% p.a. real in the post 1900s (this may be partly due to the data veracity in bonds with that prior to 1932 being a proxy based on Consols and the Bank of England Base Rate).
Interestingly the "extremes" of returns have been well dispersed through the 300-year history, with the possible exception of the outlier high inflation years of the mid-1970s. The table below shows the 20 best and worst returns on UK equities and the corresponding year.
The beauty of such a long history (for actuaries) of relatively robust return data is that we can consider the impact of alternative accumulation and decumulation policies using real data. By going back over 300 years, we extend the range of possibilities significantly.
So what does all this data tell us?
Firstly, the rolling real outcomes from this long-term data series (chart 3) are just as volatile as those outlined in our shorter (Australian) data set in Chart 1. Even on 10-year view, the strategic passive portfolio is extremely volatile.
Chart 3: Rolling 10-year real returns on a passive balanced portfolio
Source: Schroders, GFD, Balanced Portfolio is 60% UK equities, 40% UK bonds
However, far more important than time-weighted returns, is the impact on the members' potential retirement income stream. We analyse this by calculating the inflation-adjusted final salary pension that could be paid for 30 years after a 40-year accumulation at 10% of salary (with no taxes, fees etc). In particular, we look at the median final salary pension that could be paid, but also the range of outcomes that would have been experienced using this 300-year historical data set.
The chart below shows the 90th percentile, median and 10th percentile final salary pension outcomes from a long-term DC accumulation over the past 300 years from alternative passive investment strategies. For those concerned about the veracity of data from the 1700s, we have also calculated and shown the results using data from January 1800 only.
Chart 4: Range of final salary pension outcomes from different passive asset-allocation strategies
Source: Schroders, based on GFD data, 10% of salary accumulation for 40 years with a final salary pension for 30 years. Salary inflation at CPI+1%. Period 1694-2015 and 1800-2015.
The key result from this is that, irrespective of the passive approach adopted, the range of outcomes varies enormously. For a standard passive 60/40 investor, you could have ended up with a final salary pension of between 27% and 73% with an average outcome of 44%. For an investor who chose 100% in equities, the range was 28% to 80%. With the exception of the pure bond strategy, it made little difference if we utilised the full 300-year period or just the period from 1800 onwards.
In summary, a passive approach to asset allocation in the investment strategy leads to an enormously volatile set of retirement outcomes.
What is the alternative?
If the objective is to achieve a certain minimum outcome in retirement with reasonable probability, we have only two choices:
- Contribute more; or
- Change the investment approach.
If the desire in our passive 60/40 portfolio was that we wanted an outcome of a 40% pension with 90% probability, we would have had to contribute not 10% of salary, but 16.5% of salary. (The cost of being passive is a contribution rate 65% higher.) This is unlikely to be palatable or sustainable.
However, if we were to change the investment approach and move from a passive approach to one that targeted smoother money-weighted outcomes (which by definition is active), a considerable improvement in these outcomes is possible (assuming some degree of success).
The chart below shows the same analysis as above but with an additional portfolio that delivered a real return of 4.65% p.a. consistently. The 4.65% p.a. real return is interestingly less than the 4.78% p.a. real return of our passive 60/40 strategy and substantially less than the 5.4% real return on a pure equity passive portfolio. (Note that we have considered the whole period in this analysis from 1694. However, as shown in Chart 4 the results are similar to that for the shorter period from 1800.)
Chart 5: Range of final salary pension outcomes from passive and active strategies
Source: Schroders, based on GFD data, 10% of salary accumulation for 40 years with a final salary pension for 30 years. Salary inflation at CPI+1%. Period 1694-2015.
Clearly, a portfolio that targets (and achieves) more stable real-return outcomes is worth a lot. A real return of 4.65% p.a. provides the same median outcome as our 100% equity passive portfolio and virtually no downside volatility. In fact, the 10th percentile outcome from the real return portfolio was a final salary pension of 49%. To achieve this with a passive 60/40 approach we would need a contribution rate of 18.4% (compared to our assumed 10%).
The "cost" of being passive is clearly prohibitive and substantial relative to being active.
Can active do better?
A key point in all of the above is of course can active do a better job than passive in achieving objectives? While we think it can, and indeed have addressed some of the ways in which it can in other papers,5 the principal point is that there is no mathematical reason as to why it cannot. Here we would circle back to the mathematical arguments in support of passive management. Perversely, active approaches that target stability of outcomes would, at the extreme, result in financial markets being significantly less volatile as they would act to dampen volatility introduced by valuation changes in markets.
They would support better long-term capital allocations not just within markets, but across asset classes. This compares to passive approaches which work the opposite way to promote valuation volatility. In fact, Xiong and Sullivan6 comment that "the rise in popularity of index investing contributes to higher systematic market risk. More indexed equity assets correspond to increased cross-sectional trading commonality, in turn precipitating higher return correlations among stocks".
More active funds management (and the total portfolio level) would serve to make markets more "efficient", not less so. At a macro level, the mathematics would support an active investment approach rather than a passive one.
There has been considerable research conducted on why passive asset management should hold sway over active. However, much of this analysis relies on a series of assumptions that are not representations of the real-world issues faced by investors.
For most investors, liabilities cannot be represented by a passive investment construct and, as we have outlined in this paper, to not manage these in an active manner would be prohibitively expensive.
Passive (and quasi-passive or systematic) approaches can be useful for parts of the implementation as with active approaches. But at the aggregate level, in order for investors to achieve their objectives with a reasonable degree of certainty an active approach is necessary.
1 William F Sharpe, "The Arithmetic of Active Management", Financial Analysts Journal, Jan-Feb 1991
2 Warren and Ezra, "When Should Investors Consider an Alternative to Passive Investing?" Investments and Wealth Monitor, March/April 2014, Investment Management Consultants Association
3 See Cooper, Doyle, Durack and Stevenson, "Why strategic asset allocation is flawed", Schroders, March 2012
4 "Anchoring asset allocation to valuation rather than averages", Schroder Investment Management Australia Ltd, Durack and Zhou, February 2015
5 For example, see Durack and Zhou, "Anchoring strategic asset allocation to valuation rather than averages", Schroders February 2015
6 Xiong and Sullivan, "How Passive Investing Increases Market Vulnerability", August 20, 2011.
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.