Amid signs correlations between asset classes could be shifting like tectonic plates, investors may need some portfolio-proofing.
This article is part of a regular series of thought leadership pieces from some of the more influential ETF strategists in the money management industry. Today's article is by Corey Hoffstein, co-founder and chief investment strategist for Boston-based Newfound Research LLC.
For the past decade, the correlation between stocks and bonds has been largely negative, allowing investors to justify holding lower-expected-returning bonds because they could hedge away stock volatility.
But with short-term correlations approaching zero, we turn our eye toward history to get a better understanding of this critical relationship and the impact correlations may have on our portfolios.
As we get a handle on what could be happening to correlations, we're also interested in identifying off-the-beaten-track ETF tools that could help investors negotiate what looks to be shaping up as a very different environment. Such ETFs include the PowerShares S&P 500 Low Volatility Portfolio (SPLV | A-45), and I'll mention others below.
But first a bit of history.
Two Distinct Regimes Of Stock-Bond Correlation
The historical relationship between stocks and bonds has largely been a story of two correlation regimes.
From 1964 through 1998, the stock-bond correlation was strongly positive, with an average rolling one-year correlation of 0.28 over the period. Since 1998, however, a strong negative correlation regime emerged, with a rolling one-year correlation average of -0.28.
Evidence suggests that once set, the stock-bond correlation regime tends to be long-dated in maturity. However, a sample set of two is not enough to draw a statistically significant conclusion.
And as 1998 demonstrated, the regimes can change rapidly. As a critical input to the asset allocation decision, it's important to understand the impact correlation has on a portfolio, what drives the stock-bond correlation, and expectations for the correlation level going forward.
The Impact On Portfolio Construction
The correlation between stocks and bonds can have a large impact on asset allocation decisions, especially in the context of risk management.
Using equity figures from the Fama/French research library and bond figures from Antti Ilmanen's book "Expected Returns," we will assume the following long-term dynamics:
|Expected Excess Return||Excess Return Volatility|
From these dynamics, we can solve for optimal Sharpe ratio portfolios based on varying correlation assumptions:
Moving from the pre-1998 average correlation of 0.28 to the post-1998 average correlation of -0.28 results in a 12.5 percentage point decrease in recommended equity allocation for the maximum Sharpe ratio portfolio.
However, investors don't invest in the Sharpe optimal portfolio, but rather along the efficient frontier, as they are either seeking a target return level or a specific risk level.
While expected excess return levels will be independent of correlation, portfolio volatility will be dependent upon it and therefore, for risk-sensitive investors, correlation is a key ingredient of the decision-making process.
Not surprisingly, higher correlation levels between stocks and bonds will lead to higher volatility levels for the same asset allocation.
But how does the correlation regime come into play when it comes to flight to safety?
In the scatter charts below, we plot daily stock and corresponding bond returns. We split the data into the two long-dated correlation regimes to see if the average correlation has an impact on short-cycle flight to safety. If bonds serve as a good hedge, we expect to see more points in the top-left quadrant than in the bottom-left.
Simply scanning the plot, we can see that the number of points in the top-left quadrant versus the bottom-left is significantly higher post-1999. While not conclusive, this historical sample suggests that bonds may not serve as a good short-term hedge against equity volatility in high-correlation regimes.
So What's Driving Correlation?
Why the sudden shift in 1998, and what can we learn from this going forward from today's market environment? Like with most assets, the fundamental drivers of return for stocks and bonds will be a combination of inflation expectations, economic growth rates, nominal interest rates and policy.
The sensitivity stocks and bonds have to each of these factors will be different—but when one factor overwhelms all others, correlations between the two asset classes can spike.
The common hypothesis is that this is exactly the regime we saw pre-1999: Inflation, being the primary concern, was the latent factor that led to increased correlation between stocks and bonds. With the moderation of inflation during the 1990s, sensitivities to other return factors took over, particularly those around economic growth.
We can see this break highlighted in the graph below, where historical annual bond yield and earnings yields are plotted. Pre-1998, we see that these yield numbers moved closely in sync with one another. Once the long-inflation cycle has come to an end and the late-90s equity bubble picks up, we see these figures diverge and begin to move in opposite directions.
This simple model of whether prevailing risk is based on inflation or economic growth solves both the correlation and flight to risk puzzles. In the latter case, when inflation risk prevails, investors may not necessarily jump from stocks to bonds in fear that they are jumping from the frying pan into the fire.
Of course, academic studies are done with the benefit of hindsight. If I were writing this article in 1997 and not 2014, it would be unlikely that anything but inflation proved to be statistically significant in econometric models of stock-bond correlation levels.
However, such models would ignore the fact that price volatility and high inflation levels are likely an indicator—and a manifestation—of latent economic instability. Certainly the incorporation of economic response variables is critical in modeling the stock-bond correlation, but we should by no means assume a perfect model from just these inputs.
Forecasting To 2015
Both Pimco and Deutsche Bank have released papers in the last year: "The Stock-Bond Correlation" and "Long Cycles In The Bond-Equity Correlation: Where Next?" each, respectively, propose econometric models for estimating and forecasting the correlation relationship.
While the implementation of each model is unique, their forecasts are identical: A 0.0 stock-bond correlation in 2015.
If we're moving from a -0.28 average regime to a 0.0-level regime, with fluctuations into both positive and negative territory, investors should consider the implications of this change on the appropriateness of their current asset allocation profile.
To achieve the same expected return profile, increased correlation levels imply that investors will have to stomach higher volatility levels, both in the aggregate and in short-term shocks.
The portfolios most affected in such a change in correlation levels will be those with equity in the 20-45 percent range, seeing increases to volatility of more than 1 percentage point.
New Modes Of Diversification
The easiest way to combat this potential increase in correlation is through further diversification beyond just stocks and bonds.
While many investors have adopted a global approach to asset allocation already, they may not be aware of newer, innovative ETFs like the PowerShares Multi-Strategy Alternative Porfolio (LALT). This fund can provide access to long/short factor, volatility risk premium, and carry strategies. Since inception in late May, realized daily correlation of this fund to the S&P 500 ETF (SPY | A-98) has been -0.07, and 0.19 to the iShares 7-10 Year Treasury Bond ETF (IEF | A-58).
For more traditional holdings, diversifying the sources of total return between price return and income can also be beneficial. Investors may also consider dividend equities using the PowerShares Dividend Achievers Portfolio (PFM | B-76) or covered-call strategies such as the one in the PowerShares S&P 500 BuyWrite Portfolio (PBP | C-58).
They might also consider implementing their core positions with factors more aligned to their risk profile. Such choices include lower-volatility equities for conservative investors found in the PowerShares S&P 500 Low Volatility (SPLV | A-45), a deep-value play such as the Guggenheim S&P 500 Pure Value ETF (RPV | A-64) or a small-cap fund like the iShares Core S&P Small-Cap ETF (IJR | A-92) for growth investors.
Whatever the choice, the takeaway is this: Investors need to carefully consider how correlations may be changing significantly, and what they're going to do about it.
At the time this article was published, Newfound held positions in LALT, SPY, SPLV, IJR and IEF on behalf of clients.
Newfound Research LLC is a Boston-based quantitative asset management firm focused on rules-based, outcome-oriented investment strategies. Newfound specializes in tactical asset allocation and risk management solutions. Founded in August 2008, Newfound offers a full suite of tactical ETF managed portfolios covering global equity, U.S. small-cap equity, multi-asset income, fixed-income and liquid alternative asset classes. For more information about Newfound Research LLC, call us at 617-531-9773, visit us at www.thinknewfound.com or email us at [email protected]. For a list of relevant disclosures, click here.