Risk Parity - No Free Lunch
The poor performance of traditional portfolios over the last decade or so has increased the desire to find alternatives that can deliver better risk adjusted outcomes to investors. The key reason why traditional portfolios have performed poorly is the high exposure to growth assets, particularly equities. One of the alternatives that has been gaining traction is the concept of risk parity portfolios.
While there are considerable variants on the theme, in general risk parity portfolios are constructed on the basis that the contribution to risk from different asset classes is equal (which is different to the actual exposure to those asset classes). While the number of asset classes included and the names of those can vary, at its simplest a typical risk parity strategy involves a substantially lower equity exposure than a traditional portfolio and a much higher exposure to bonds and sometimes other alternatives such as commodities. Such portfolios are then leveraged to increase the expected return on the portfolio which would otherwise be lower than a traditional portfolio (given the higher exposure to lower return asset classes).
In this paper we compare the long term historical performance of a naïve risk parity strategy with that of a more traditional portfolio. While there are a number of theoretical arguments that have been proposed to suggest improved risk adjusted performance of risk parity portfolios, investors should be mindful of the underlying exposures of such strategies, especially the use of leverage and its potential implications.
In our analysis we find that while US centric naïve risk parity portfolios have outperformed naïve traditional portfolios (at the same level of risk), the same result does not necessarily hold true for Australian portfolios. In addition, while recent performance of naïve risk parity strategies has been good, this is not universally true and historically risk parity portfolios have underperformed traditional portfolios for very long periods of time, sometimes significantly.
Moreover, we find that naïve risk parity portfolios have not materially increased the likelihood of delivering better real outcomes to individuals over time frames that matter. Ultimately, it is the price you pay for an asset which drives the expected return and risk of that asset, and this should be the principal driver of investment exposures. In our view any strategy which ignores this in its portfolio construction methodology, runs the risk of underperforming objectives. Risk parity is not immune to this.
Traditional balanced portfolios have been shown to suffer from long periods of poor returns, offset by long periods of relatively good returns. This long cycle volatility is particularly detrimental to those who don’t have constant invested capital (i.e. pretty much everyone)1. This is quite evident in the chart below which shows rolling 10 year returns to a traditional balanced portfolio.
In particular, over the last decade returns of a balanced strategy have fallen considerably short of long run expectations. A significant driver of this real return behaviour is the concentration of equity risk in a traditional balanced portfolio and the long term volatility of real equity returns (which is largely a function of valuation). The chart below shows the asset allocation breakdown of a traditional balanced portfolio and the contribution to risk from the assets of that same portfolio.
It is clear from the above that while the asset allocation of a traditional portfolio may appear somewhat diversified, the contribution to portfolio risk is anything but. In simple terms, the returns of a traditional balanced portfolio are largely determined by the return on the equity market. This is good when equity returns are good, but less so when equity returns are poor.
In any case, the dominance of equity risk in a traditional portfolio construct and the relatively poor performance of equities over the last decade or so (particularly for US and European based investors) has meant that traditional portfolio returns have been relatively poor. In this environment it is not surprising that investors look for alternatives. The holy grail of investing, particularly for the average investor who accumulates assets through time then decumulates, would be a portfolio that offered a reasonable positive real return with very low volatility – i.e. a portfolio that produced consistently positive real returns irrespective of the underlying economic environment and the subsequent asset class returns.
The Risk Parity Solution
One of the “solutions” to this quest that has been gaining popularity is the concept of risk parity portfolios. There are quite a number of different variants to the risk parity concept but at its simplest, the point of a risk parity portfolio is to set the asset class weights such that the contribution to risk of each asset class is the same.
Keeping things simple, this invariably means that low risk asset classes (e.g. fixed income) have a significantly higher weight in the portfolio than high risk asset classes (e.g. equities). Not surprisingly, in an environment where equity returns have been quite poor against other assets (especially say, commodities and bonds) any portfolio structure that biases itself away from equities and towards these other alternatives will have performed better. A sample risk parity portfolio is set out below.
A sample risk parity portfolio is set out below.
The problem with such a portfolio is that by having a large exposure to low risk assets, over time the overall portfolio return is likely to be lower than a traditional portfolio. Consequently, the second stage of a risk parity approach is to leverage the risk parity portfolio to the degree required to make the expected return (or risk) broadly equivalent to the traditional portfolio.
The result is that the leveraged risk parity portfolio should offer returns similar to the balanced portfolio but with lower volatility (or higher returns with the same volatility) – i.e. it is a much more efficient portfolio.
The example below demonstrates how a simple risk parity approach would work.
|Traditional Portfolio||Risk Parity Portfolio(Unlevered)||Risk Parity Portfolio(Leveraged)|
Assuming borrowing costs at the cash rate of 3.25%, we can lever the risk parity portfolio by 167% (ie. borrow $1.67 for every $1 of capital) to get a portfolio that will have an expected return of 2.67 x 5.4% less borrowing costs of 1.67 x 3.25% = 9.0%. On these assumptions the volatility of the levered risk parity portfolio would rise to 6.8%.
There are a number of theoretical explanations as to why risk parity portfolios should deliver better risk adjusted outcomes2, which we would broadly summarise as:
- In a Capital Asset Pricing Model (CAPM) world, the risk parity portfolio moves us up the capital market line; and
- Investors as a rule are “leverage averse” and so there is an arbitrage opportunity that can be captured by being prepared to accept leverage (obviously these academics haven’t met Australian households).
We do not propose to consider the theoretical arguments for or against CAPM and in any case we are not great believers in CAPM (or more accurately the assumptions that underlie CAPM upon which the theory is based). Rather, the purpose of this paper is to consider the practical implications of a risk parity portfolio and examine the consequences of following a risk parity approach through time.
Our concerns about risk parity and the issues that investors in such strategies need to be aware are principally:
- Risk is not volatility. “Risk” means different things to different people, however, should most accurately be described as the “probability and magnitude by which objectives are not met”. The riskiness of asset classes to us is a function of the degree of leverage (financial and operating), liquidity and price risk. However, this makes it difficult to formulate a risk neutral portfolio and consequently risk is often taken to be volatility or a volatility based measure. In addition, volatility is generally taken to be historical volatility. This is not necessarily a good indicator of real risk.
- Leverage. At a basic level, the risk parity portfolio represents a leveraged portfolio of bonds. In environments where there is an upward sloping yield curve (and borrowing is typically short vs investing which is longer duration) there is a positive carry to the portfolio. The same doesn’t necessarily apply when the yield curve is downward sloping. (e.g. Australia today). More importantly, this approach will generate positive returns when interest rates are trending down. A sharp move up in interest rates could be highly detrimental to such a portfolio. As such, the leverage can introduce a significant negative skew to the return profile. This negative skew could be exacerbated in times of stress by the unavailability of leverage, a significant increase in the cost of leverage or a lack of liquidity in the underlying assets which is required to fund the leverage.
A look at the longer term movement in yields suggests that the last few decades should have been relatively kind to leveraged bond portfolios. This environment is unlikely to be repeated over the next decade or two.
- Geography. Much of the academic analysis on risk parity portfolios is centred on the US markets and most risk parity portfolios are constructed around US (or to a lesser extent European) assets. This may not always be appropriate for an Australian centric liability portfolio.
- Ignorance of valuation. At its heart, a risk parity portfolio represents a structural bias away from equities and to everything else relative to a traditional portfolio. We should bear in mind that such a structure may be appropriate when equities are expensive, but what about when equities are cheap, or worse, when the alternatives to which one is biasing the portfolio are expensive? It is our strong view that any portfolio construct which ignores valuation (and hence expected return) will only work episodically. Of course one could argue that the same applies to a traditional portfolio being structurally biased towards equities. We would agree with this sentiment – any portfolio structure which is ignorant of valuation is unlikely to deliver optimal results for investor’s.
In other research we have reviewed the performance of a naïve strategic asset allocation portfolio back to January 19003. For the purposes of this analysis we have taken the same historical data set and constructed a naïve risk parity portfolio utilising those same asset classes. We have analysed a US risk parity portfolio vis a vis a US balanced portfolio (upon which much of the academic research has been based) but then also constructed the same analysis from an Australian context. Note however, there is no universally accepted method when it comes to constructing risk parity portfolios. As such different methods of constructing a portfolio will lead inevitably to different results. Broadly, our approach has been to:
- Construct the relevant balanced portfolio and calculate the 36 month trailing volatility of that portfolio;
- Determine the unlevered risk parity portfolio by analysing the 36 month trailing volatility of each asset class and setting the portfolio weights in the unlevered portfolio such that the contribution to risk of the total portfolio is the same from each asset class and the asset class weights sum to 100%;
- Calculate the 36 month trailing volatility of the unlevered risk parity portfolio and set the leverage rate on this portfolio such that the levered risk parity portfolio has the same trailing 36 month volatility as the traditional balanced portfolio.
- Borrowing costs for any leverage in the portfolio are optimally assumed at the cash rate.
- Both the risk parity strategy and traditional portfolio are rebalanced monthly (which may influence the results compared to daily, quarterly or annual rebalancing). Transactions costs and tax effects are ignored in both portfolios (albeit these may be higher in risk parity portfolios given the higher degree of portfolio turnover).
We note that throughout this paper we use the construct of a naïve risk parity portfolio as described above. However, it is only fair to note that there are multiple variations on the risk parity theme. Our purpose here is to determine the generalised characteristics of risk parity and consequently better understand the real risks of such an approach.
The chart below shows the decade by decade relative performance of the 60/40 traditional portfolio versus a (leveraged) risk parity portfolio constructed from only US equities and bonds. Over the entire sample period (1903 to 2009) the risk parity portfolio outperforms by 0.71% p.a., however when we break the relative performance down across decades risk parity outperformed in only 4 of the 10 decades, and one of those was by only 0.17% p.a.
Consequently, while the risk parity portfolio has outperformed over the entire sample period, this outperformance has been episodic and is heavily biased to the 1930-1950 period and the most recent decade of 2000-2009.
Another important consideration that is often overlooked when analysing historical portfolio outcomes is the effective level of exposures that would have been required in such a portfolio to generate these outcomes. Consequently, we can also examine the degree of leverage the risk parity portfolio construction methodology above would have resulted in through time. This is shown in the chart below.
Clearly at times the degree of leverage required has been historically very high. In particular in the two prior decades when the risk parity approach outperformed - the 1930’s and 1940’s – average leverage rates were 420% and 270% respectively for the decade. It is probably somewhat academic to expect that during the great depression it would have been possible to construct a portfolio that had borrowed $4.20 for every $1 of capital (and have borrowing costs at the risk free rate), likewise with $2.70 of debt for every $1 of capital during World War II.
Since the 1970’s leverage rates have been somewhat more realistic at less than 100%. As of today, the portfolio would be approximately 70% exposed to bonds and leveraged 60%. i.e. for every $100 of capital another $60 would be borrowed and $48 would be invested in equities and $112 in bonds.
By way of example, if we limited the maximum level of leverage to say 200% ($2 of borrowing for every $1 of capital) the average outperformance falls from 0.71% p.a. to -0.13% p.a.
We note that our results differ somewhat from the historical results obtained in the study by Asness et al. In that analysis, the risk parity portfolio outperformed the traditional 60/40 portfolio by approximately 3.34% p.a. over the period 1926 to 2009. The same period in our analysis resulted in an outperformance of only 1.15% p.a., a substantial difference. While we were unable to replicate the Asness results exactly given the use of a different data set and a different basis for setting the volatility of the risk parity portfolio, we did manage to approximate these results. In particular, we noted that that study set the risk parity portfolio leverage such that the volatility of the portfolio was in general about 30% higher than the 60/40 portfolio (having been set to equate the volatility of a different “value weight portfolio”). This increase in leverage has the impact of substantially improving the returns to risk parity (not surprising if you increase the risk of the portfolio, returns should rise). For example a 30% increase in leverage would have resulted in the risk parity portfolio outperforming by 2.66% and a 40% increase in leverage would have resulted in an outperformance of 3.13%.
However an increase in leverage of 30% would have implied leverage levels of 600% to 700% during the great depression and 400% to 500% during World War II. In any case, that portfolio still underperformed a traditional (less risky) portfolio in the 1910’s, 1950’s, 1960’s and 1970’s – all notable as periods when interest rates were generally rising.
Having examined the results from a US perspective, we now turn to an Australian equivalent portfolio. Obviously the structure of market returns in Australia has been somewhat different to that of the US, most noticeably equity returns have generally been better than the US and in particular the serious underperformance of equities in the US in the 1930’s was not as evident in Australia (likewise the most recent decade of the 2000’s).
The chart below shows the same 2 asset (equity and bond) portfolio comparison between a risk parity portfolio and a 60/40 traditional portfolio on a decade by decade basis.
We now observe that not only does the risk parity portfolio underperform by just over 1%p.a. over the entire period, but the very strong performance in the 1930’s and 1940’s is no longer evident. We can also see that there was a 40 year period of underperformance from the 1950’s to the 1990’s. Not surprisingly, the leverage profile of this portfolio is entirely different to that of the US example as set out below.
If we were to broaden the Australian portfolio to include global stocks rather than just local stocks and bonds, the overall results improve somewhat as set out in the table below, however, only to the point that the risk parity portfolio now outperforms the traditional portfolio marginally by 0.1% p.a. While the degree of underperformance in the 1950’s to 1990’s has reduced, it is still in the order of approximately 2% for this entire period.
In reviewing the four decades from the 1950’s to the end of the 1980’s, the first three of these were characterised by rising interest rates and the latter an environment of very strong equity performance.
In summary this points to the principal issue behind risk parity portfolios in that ultimately they are agnostic to asset prices and broadly very biased to bonds over equities. Consequently, we should expect that these portfolios will underperform a traditional portfolio when either bond yields are rising or equity returns are strong (or both). Interestingly the correlation of annual returns on the risk parity portfolio with our Australian global balanced portfolio was 90%.
The Real Issue
While the above analysis looks at relative performance of risk parity to a traditional portfolio, the problem investors are typically trying to resolve is that of delivery of consistent real returns over time. In prior papers we have examined the degree to which a traditional balanced portfolio may meet investor’s real return objectives. Similarly we can analyse the degree to which a risk parity portfolio might also be able to generate positive real returns through time. We show in the chart below rolling 10 year performance of an Australian risk parity portfolio versus a traditional balanced portfolio.
It is clear from the above that in the context of delivering returns relative to objectives, it is difficult to argue that a risk parity portfolio is any better than a traditional portfolio (and neither are that good in our view). Note that we would acknowledge that many investors’ recent experience of risk parity portfolios may be better than shown in the above chart, however, there are very few strategies with 10 years of history and none to our knowledge built from an Australian asset class frame of reference.
In particular, the historical probability through time of a risk parity portfolio meeting a CPI+4.5% p.a. target on a rolling 10 year basis was 58% versus 57% for a traditional portfolio – hardly a materially better outcome. At the same time, the potential drawdown on a risk parity approach was a lot worse than that for a traditional approach, at -8.5% p.a. for the 10 years to 1974 versus a traditional portfolio of -3.7% p.a. for the same 10 year period.
But haven’t risk parity portfolios performed better than balanced portfolios in the last decade? Bearing in mind that most (probably all) risk parity portfolios on offer in Australia are in fact overseas risk parity portfolios offered in Australian dollars (hedged or unhedged) then it is likely that returns on these have been somewhat better than the returns from an Australian perspective. However, investors in these strategies should understand that they are taking quite different risks in these strategies – in particular one could be taking a significant geared bet on US interest rates and commodity prices.
Risk parity solutions have gained popularity following the relatively poor performance of traditional portfolios over the last decade. However, in the context of building portfolios that are better aligned with investors’ real return objectives it is difficult to see how naïve risk parity portfolios achieve this any better than a traditional portfolio.
As with any portfolio construct that does not take account of valuation (and hence expected risk and return) it is our view that naïve risk parity is unlikely to offer investors a panacea to the poor performance of traditional portfolios that they seek. We would suggest that where investors do consider risk parity portfolios that those portfolios reflect current risk and return expectations for asset classes rather than historical outcomes. Investors should have a clear understanding of the leverage used and how this will change.
Interestingly, there is considerable academic literature around setting forward estimates for returns across asset classes which has been shown to be relatively accurate. The same cannot be said for estimating future volatility. Consequently investors may be sceptical about a portfolio construct that has at its heart an estimate of future volatility.
In our analysis we find that while US centric naïve risk parity portfolios have outperformed traditional portfolios (at the same level of risk), the same result does not necessarily hold for Australian portfolios. In addition, while recent performance of risk parity as a strategy has been good, this is not universally true and naïve risk parity portfolios have underperformed traditional portfolios for very long periods of time, sometimes significantly.
Moreover, we find that naïve risk parity portfolios have not materially increased the likelihood of delivering better real outcomes to individuals over time frames that matter. Ultimately, as we showed previously for traditional portfolios, it is the price you pay for an asset which drives the expected return and risk of that asset, and this should be the principal driver of investment exposures. In our view any strategy which ignores this in its portfolio construction methodology, runs the risk of underperforming objectives. Risk parity is not immune to this.
Data sourced from Global Financial Data used in this analysis:
Australia Consumer Price Index
Australia Total Return Bills Index
Australia 10-year Government Bond Return Index
GFD World Return Index
Australia S&P/ASX 200 Accumulation Index
S&P 500 Total Return Index (w/GFD extension)
US Government Bonds
1For a more detailed analysis of this see “Why SAA is Flawed?” and “Understanding the Journey to Retirement”, Schroder Investment Management Australia Ltd, March and April 2012.
2For example, see “Leverage Aversion and Risk Parity”, Asness, Frazzini and Pedersen; FAJ Jan/Feb 2012.
3“Why SAA is Flawed?”, “Understanding the Journey to Retirement”, and “How flexible do we need to be”, Schroder Investment Management Australia Ltd, March and April 2012.
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.