The recent Wall Street Journal article "A Portfolio That's as Simple as One, Two, Three" got us thinking about the possibilities of using just three ETFs to build a portfolio. But the article was sparse on examples. It provided only two specific portfolio allocation ideas: 1) 40% U.S. stocks, 20% international stocks, 40% total bond market and 2) an allocation using an all-country equity fund (e.g., ACWI) plus a U.S. bond fund and an international bond fund.
That left us hungry for more actionable asset allocation ideas using three ETFs. We also wanted to explore portfolio allocations using other types of assets. This article explains how we looked at a full range of portfolios using just three ETFs, and which portfolios performed best.
Our approach to finding the best portfolios using just three ETFs
1. Choose a list of global asset classes from which to pick the three ETFs.
2. Create portfolios by choosing every unique combination of three ETFs from the asset class list.
3. Apply an allocation (% weighting) to each ETF within each portfolio.
4. Calculate risk and return metrics for each portfolio.
5. Identify winners and losers based on risk and return.
Choosing the asset classes
We chose eight asset classes and applied a data set with 40 years of history through 2012. Since the ETFs listed below are not 40 years old, we've used an appropriate mutual fund or proxy to complete each data series.
Are there asset classes we've missed? Sure, but these eight should give us some idea of where to look first. If you can't imagine yourself owning some of the above asset classes, then we hope there will still be some portfolio combinations worth considering.
We've applied an equal weight of one-third (33.33%) to each item in each portfolio. This equal weighting is a simplification; there are many other allocation possibilities (e.g., 60%/20%/20%, 20%/20%/60%) but this allows us to explore a manageable number of portfolios and should give us an idea of what looks most promising.
Then we calculated the risk and return for each portfolio using annual data and the assumption of annual rebalancing and no transaction costs or taxes.
Which three-fund portfolios had lower risk and higher return? Here's what we found. The top left of the chart contains the portfolios with the lowest risk and highest return.
Portfolio #30 (Real Estate, Short Term Treasuries, International Equities) has an annual return of 10.99% with a standard deviation of 9.82%
Portfolio #11, the highest return portfolio, would have increased 82-fold (8,200%) over the 40-year period.
We're aware of the particular behavior of the gold market in the early 1970's. Portfolios #1 through #21 contain gold, and Portfolios #22 through #56 have no gold.
Looking at the poorer performing portfolios,
Portfolio #40 offers significantly higher risk without much extra return.
Portfolio #19 offers significantly lower return compared to other portfolios with similar risk (e.g., Portfolios #9, #39).
The lowest return (and lowest risk) portfolio, #42, would have increased 21-fold (2,100%) over the 40-year period.
Several portfolios, such as #30, have offered solid returns of 11% per year at below-market risk.
Looking at volatility, all 56 of the 3-ETF portfolios in this analysis were less volatile than holding the single asset class of U.S. equities, which has a standard deviation of 18.1% (!) over the 40-year period.
Looking at total annual returns, most of the portfolios in this analysis had a higher return than the U.S. equities asset class, which had an annual return of 9.5%.
The top-performing portfolios are worth further analysis to see if additional optimizations can be made with asset selection, factor tilts, or other variations.
Our recent article "3 Portfolios Inspired By Graham, Swensen, Browne" showed Harry Browne's Permanent Portfolio to be a solid performer, returning 6.6% over the past five years with less risk than an S&P 500 portfolio or a balanced portfolio (60% stocks / 40% bonds). But some unanswered questions emerged from that analysis.
The New Questions
Over a longer time horizon, how has the Permanent Portfolio performed?
During a period of rising interest rates, how has the Permanent Portfolio performed?
We intend to answer these questions in this article.
As a recap, the Permanent Portfolio can be constructed as follows:
25% Cash or Short-Term Treasuries (using SHV or VUSXX)
The Permanent Portfolio's design is intended to weather varied market conditions and provide a reasonable risk-adjusted return. Our model of the Permanent Portfolio rebalances yearly and we've used annual mutual fund returns as proxies for each portfolio component, plus historical indexes to complete the data series.
Looking at 39 years of history
Due to structural changes in the gold market from late 1971 through 1973, and due to the importance of gold in the Permanent Portfolio, we start our long-term analysis in 1974. So we are looking at the 39 years beginning with 1974 and ending with 2012.
Let's first compare the Permanent Portfolio to two benchmark portfolios for the 39-year period.
Return and Risk: 1974-2012
Permanent Portfolio: Total return is 8.9%, Risk (as measured by standard deviation) is 7.7%
S&P 500 Portfolio: Total return is 10.2%, Standard deviation is 17.8%
60/40 Balanced Portfolio: Total return is 9.9%, Standard deviation is 11.6%
Looking at a period of rising interest rates
Now let's focus on the period 1977-1981 when interest rates were rising. The Fed began raising the Fed Funds rate in February 1977, and this continued through 1981 under Paul Volcker. Since current rates have nowhere to go but up, we also wanted to look at this time span as a comparison.
Return and Risk: 1977-1981
Permanent Portfolio: Total return is 11.9%, Standard deviation is 14.5%.
S&P 500 Portfolio: Total return is 7.6%, Standard deviation is 16.1%.
60/40 Balanced Portfolio: Total return is 8.6%, Standard deviation is 10.1%
Risk vs. Return Scatterplot
Below is a view of the three portfolios during both time spans. Looking at the two Permanent Portfolio data points, we see that its performance during the difficult 1977-1981 period exceeded its overall performance during the 39-year period. The two benchmark portfolios had a rougher time during the period of rising interest rates.
1. Over a longer time horizon, how has the Permanent Portfolio performed?
Over the 39-year period, the Permanent Portfolio has lagged the benchmark portfolios in total return, but the Permanent Portfolio has exhibited less risk.
2. During a period of rising interest rates, how has the Permanent Portfolio performed?
Over the rising interest rate period, the Permanent Portfolio outperformed the S&P 500 with less risk.
Our recent article "3 Portfolios Inspired By Graham, Swensen, Browne" elicited the suggestion that we compare a do-it-yourself ETF portfolio of 60% stocks and 40% bonds to Vanguard's Balanced Fund (VBIAX). That also made us think about the other balanced funds from the leading fund companies.
Specifically, here are the two questions we want to answer:
Which "balanced fund" has the best risk-adjusted returns?
Can I do better creating my own balanced ETF portfolio using 60% stocks (VTI) and 40% bonds (BND)?
Choosing the Portfolios
First, we identified balanced funds from the major fund companies. Here's the list with each ticker and annual expense according to the prospectus.
Dodge & Cox Balanced (DODBX) 0.53%
Fidelity Balanced (FBALX) 0.60%
Fidelity Global Balanced (FGBLX) 1.04%
Ibbotson Balanced (BETFX) 0.92%
Janus Balanced (JANBX) 0.72%
Schwab Balanced (SWOBX) 0.65%
Vanguard Balanced (VBIAX) 0.10%
Wells Fargo Advantage Index Asset Allocation (WFAIX) 0.91%
We also created a benchmark portfolio of 60% VTI and 40% BND.
This static 60-40 portfolio has a 0.07% blended annual expense based on VTI expenses of 0.05% and BND expenses of 0.10%.
Looking at the numbers
In the table below, we see that Ibbotson has lagged the group, Dodge & Cox has performed the best over the past year, and Janus leads over the past 5 years.
Note the varying amount of equity that each fund can hold, as shown in the descriptions above. Each manager's discretion varies between these funds. While "balanced fund" usually implies a fixed or stable stock-bond allocation, such as 60% stocks and 40% bonds, some of these fund managers have more discretion to change the level of stock holdings. For example, looking at the prospectus for Janus Balanced, we see that the fund manager can hold between 35% and 65% in equities. In contrast, Vanguard sticks to a firmer 60% equity guideline.
Turning to the topic of a do-it-yourself ETF portfolio, a balanced ETF portfolio using a combination of 60% VTI and 40% BND has seen these results:
1-year return: 11.84%
3-year return: 12.60%
5-year return: 6.97%, with a maximum drawdown of -26.7% over the past 5 years.
Meanwhile, VBIAX has seen the following returns:
1-year return: 12.18%
3-year return: 12.70%
5-year return: 7.07%, with a maximum drawdown of -27.0% over the past 5 years.
So VBIAX edges out the do-it-yourself ETF combo in all three time periods, but with a slightly higher maximum drawdown.
Over the past year, Dodge & Cox has shown a 13.4% Modigliani risk-adjusted return, which is the best in this group. Over the past 5 years, Janus Balanced has shown the best risk-adjusted return, with a 12.5% Modigliani risk-adjusted measure.
Now back to our questions at the beginning of the article:
Question 1: Which balanced fund is best on a risk-adjusted basis?
Answer: Janus Balanced is the clear winner on a 5-year risk-adjusted basis. If you have an appetite for more risk, Dodge & Cox Balanced has performed better over the past year, but slightly lags Janus Balanced over the 5-year period.
Question 2: Can I do better creating my own balanced ETF portfolio using 60% VTI and 40% BND?
Answer: Not likely. If you are interested in an easy 60-40 portfolio, you are probably better off (in terms of overall risk and performance) by simply buying VBIAX. The ETF portfolio does offer a 0.03% cost advantage over VBIAX, but for that extra $30 (if you have $100K invested) Vanguard's team will do the rebalancing for you.
As part of our ongoing effort to identify the best and worst asset allocation portfolios, we compare some ETF portfolios inspired by investing gurus: Benjamin Graham, David Swensen, and Harry Browne.
Here's our plan of attack:
1. Create an ETF portfolio that resembles each guru's target allocation, based on their investment philosophy, book, or other statements. These are static allocations, rebalanced monthly.
2. Crunch the numbers to get risk adjusted returns for the past 5 years for each portfolio.
3. Look at the risk/return metrics to see if there are any clear winners. Compare to some benchmarks, too.
Creating the Portfolios
First up, Benjamin Graham, the "Father of Value Investing." We have modeled a Graham-inspired Portfolio using 50% equity and 50% fixed income, implemented with a value and dividend tilt to the equity selections, as follows:
Second, David Swensen. He is the Chief Investment Officer of Yale University and author of Unconventional Success: A Fundamental Approach to Personal Investment. His target portfolio described in the book uses total-market funds, real estate, and inflation-protected securities (TIPS). We have modeled a Swensen-inspired portfolio as follows:
Third, Harry Browne. He suggests a simple, "permanent" allocation for weathering various market conditions. His Permanent Portfolio concept is designed to provide some diversification across several different market scenarios. We have create a Browne-inspired Portfolio as follows:
Fourth, let's add a 50-50 portfolio. This is similar to the "Couch Potato Portfolio" espoused by the Dallas Morning News columnist Scott Burns. We've used TIP instead of a broader-based bond fund like BND.
50% VTI (Vanguard Total Stock Market)
50% TIP (iShares TIPS Bond)
Finally, for we'll add two more portfolios as benchmarks:
S&P 500 (represented by 100% allocation to SPY) plus a balanced 60-40 portfolio of 60% equities and 40% bonds (BND).
Crunching the numbers
Below is the annualized performance data through June. Looking at the 5-year column ('5y' on the chart), we see the annualized performance over five years is quite similar for these portfolios. However there have been striking differences over the past 12 months. The portfolios with more equity exposure have clearly outperformed. But this outperformance comes with extra risk, as we will see next.
(Click to enlarge)
The 5-year risk-adjusted winner is the Browne portfolio with a 6% risk-adjusted annual return using the Modigliani-Modigliani metric. We prefer using the Modigliani-Modigliani ("M-squared") measure instead of the unitless Sharpe ratio. The Browne portfolio's alpha is 14.2% over the past 5 years, which is the best in this comparison. Also notable is the maximum drawdown that the Browne portfolio has experienced over the past five years: only 11%.
This compares favorably to the 42% drawdown that SPY has seen over the same period.
But over the past 1-year and 3-year periods, the 60-40 portfolio has offered the best risk-adjusted returns. The maximum drawdown for the 60-40 portfolio is 27% over the past 5 years.
To see a graphical comparison of risk and risk-adjusted returns visit Inspired Portfolios: Risk Metrics. The analysis uses standard deviation, drawdown, beta, Modigliani, and alpha over the past 1, 3, and 5 years.
Risk vs. Return Scatterplot
The 5-year risk vs. return scatterplot (below) shows these portfolios have similar returns but different risk levels.
The S&P 500 benchmark portfolio (#6 on the chart) occupies the high-risk/high-return corner. Note that the 60-40 portfolio (#5 on the chart) and the Browne portfolio (#4 on the chart) offer similar returns at significantly lower risk. To find the Browne portfolio on the chart, look for #4 at the top left, slightly obscured by the Y-axis.
The 60-40 portfolio has outperformed these "guru inspired" portfolios, and it's hard to beat the simplicity of a two-ETF portfolio. But the 60-40 portfolio has been riskier than the Browne portfolio over the last five years. The Browne portfolio is worth a look if you can stomach some of the recent underperformance due to its gold holdings.
As part of our ongoing analysis of 59 global asset classes, we've identified the asset class performance leaders for June, based on total percent return for the month. We use leading ETFs as proxies for our asset class analytics.
Given the volatility this past month, which asset classes came out on top? Master Limited Partnerships (AMLP), Oil (DBO), and U.S. Microcaps (IWC) lead the pack.
U.S. Master Limited Partnerships
This was the #1 asset class in June, using AMLP as the ETF proxy for U.S. MLPs. In addition to strong June performance, AMLP is up 15.4% year-to-date, which beats the S&P 500 and the Dow Jones Industrial Average. In terms of momentum, AMLP crossed its 50-day moving average in June and is now 7.6% above its 200-day moving average.
Looking at DBO, our ETF proxy for the oil asset class, we see that despite a strong June (up 2.2%) DBO is essentially flat year-to-date (down 0.1%). Looking at momentum, DBO is close to its 50-day moving average (currently 0.5% below) and its 200-day moving average (currently 0.1% below).
U.S. Micro Caps
IWC, our ETF proxy for U.S. micro cap stocks, rounds out the Top 3 asset classes for the month. IWC was up just 0.8% in June, but IWC has been the No. 1 asset class performer year-to-date among 59 global asset classes, with a return of 18% through June 2013. IWC is still well above its 200-day moving average, but it has slipped from 13.1% above the 200-day simple moving average in May to 11.5% above in June.
We're aware of structural differences between these closed-end funds and GLD: tax considerations, fees, premium vs. NAV, spreads, country of origin (U.S. or Canada), and trading volume. Some of these differences might pique our interest in the closed-end funds. But what about returns, correlation, and risk? Does choosing a closed-end fund sacrifice returns or increase risk?
Let's take a quantitative look at the risk and return differences between these products.
Correlation to GLD
First, how well do the closed-end funds correlate with GLD?
PHYS has a perfect 1.00 coefficient of correlation with GLD. This is based on monthly total return figures over the past three years. The chart below shows the coefficient of correlation (R) multiplied by 100.
GTU has a very strong correlation of 0.97.
CEF has a 0.87 correlation, which is still strong, but is not as tight as the others due to CEF's silver exposure which makes up a significant portion of this fund.
On a day-to-day basis, returns vary considerably as seen in the 1-day (1d) column in the chart below. Even over a recent 4-week period, returns vary from -1.9% for CEF to -3.6% for GTU. The year-to-date returns are also not as tightly clustered as we expected, ranging from -17.4% to -22.6% .
Now let's take a step back and look at the 1-year and 5-year returns.
For the 12 months ending May 31, GTU lost the least with a return of -10% compared to GLD's -12%.
For the 60 months ending May 31, the ranking is as follows:
1. GTU +9.21% over the 5 years
2. GLD +8.91%
3. CEF +7.64%.
Note that PHYS isn't old enough to have 5-year returns.
The standard deviations for GTU, PHYS, and GLD have been in lockstep over the past three years. But CEF shows higher variability with a significantly higher standard deviation over the 1-year, 3-year, and 5-year timeframes. Interestingly, GTU has a slightly better maximum drawdown over the past 60 months (-23%) compared to GLD's maximum drawdown of -25%. CEF, with its silver exposure, had the largest drawdown.
For a chart with more risk measures for these funds (including Beta, R-squared, Modigliani, and Alpha) see our Gold Funds: Risk Analysis.
GTU is worth a look as a complement or replacement for GLD. It compares favorably to GLD over the 5-year period with a slightly better return, lower drawdown, and similar standard deviation. PHYS has a shorter track record and has performed worse than GLD over the past three years. CEF is more volatile than GLD since it is not a pure gold fund, but includes silver exposure.