This article is a follow-up to the article we wrote last week regarding our conservative model portfolio, which has triggered quite a lot of interest. We will try here to dig further into the mechanics of this strategy, which returned 16% per year on average (CAGR) since 2003 with limited volatility and drawdowns. We will also provide some insights on how it has performed during critical periods.
1. First of all, we would like to underline the dynamic nature of the strategy. The portfolio allocation is not static, as it is updated every month. On average, around two positions are changed every month. Here is a recent example through which it is very interesting to see how the dynamic allocation process adapts to market developments:
- In the early months of 2011, the portfolio was ideally positioned to benefit from the ongoing commodity boom, with exposures to agricultural commodities (DBA), base metals (DBB) and Energy stocks (XLE). We started 2011 with 50% in cash (SHY) and 5 equal positions in DBA, DBB, XLE, Japan Equities (EWJ), and U.S. small caps (VB), for a monthly return of 1.5% in January. Overall for 2011, as of Friday June 3, the portfolio was up 4.3% with a 7% volatility and a maximum drawdown (ongoing) of -1.6%.
- After the May correction, it has now switched gear and appears to be much more conservative. As mentioned in the previous article, the portfolio allocation is currently 60% in a mix of SHY, AGG and TLT, and 4 equal positions in Gold (GLD), U.S. REITs (VNQ), U.S. Preferred shares (PFF), and Chinese Equity (FXI). Note how the mix of fixed-income has also been altered: the initial 50% cash position suggested that the bond market was performing poorly in the middle of the stock bull market. With increasing risk aversion, bonds are temporarily back in favor over the recent months.
The following graph is probably more illustrative than a thousand words. It shows how the share of fixed-income (dark blue) - among which cash (lighter blue) - in the portfolio has evolved since 2003 (click to enlarge image):
2. Second, a closer look at the historic performance of our conservative ETF model portfolio helps understand further how the strategy works. Although 8 years of backtesting may be considered by some to be too limited, one can agree that when a strategy is robust to theory and practice, flexible enough to adapt to market developments, and builds upon studies that cover much longer periods (see Mebane Faber’s study on relative strength for instance), one can implement it with enough confidence. Furthermore, the 2003-2010 period has been representative of many different market conditions (bull, sideways and historic bear market).
- During the first half of the period (2003-2006), the conservative portfolio achieved a compounded annual gross return (CAGR) of 14.8%, roughly that of the U.S. Equity market (14.6% for the SPY, but with almost half the volatility (7.0% versus 12.3% for the SPY)!
- During the second half (2007- 2010), it has been “bear market proof”, with an even increasing CAGR (17.4%) with the volatility remaining moderate (11.2%). Most interestingly, the allocation switched to 90% cash and 10% short S&P500 (SH) as early as July 31, 2008 and remained unchanged until end-november where two positions in GLD and TLT were added for December. Overall, the last five months of 2008 were all positive and returned +9.6% while the SPY was tumbling (-27.9%).
One of the reasons why our conservative portfolio performed so well during the crash is that there had been several months of weaknesses and many “warning signs” before. The strategy is less effective to fully protect against corrections during strong trends. Not surprisingly, among the worst months are April 2004 (-4.5%), May 2006 (-3.3%), and January 2009 (-5.1%). But these months are usually followed by recoveries where the portfolio is positioned to rebound strongly:
- Between June 2006 and October 2007, the portfolio had its longest winning streak (17 months).
- In 2009, the model did not catch the first weeks of the rebound in March (-0.2%), but then recorded its best two-month return in April (+6.9%) and May (+10.5%)
3. A key factor for this kind of strategy to perform well is the definition of the ETF universe used. The universe needs to be comprehensive in its coverage of broad asset classes, limit the number of extremely correlated ETFs. One also needs to carefully consider the volatility and behavior of the ETFs you include in your universe. There is also a degree of subjectivity when creating a universe, as this implies some anticipation of what you expect the future to bring. This is why you will find in our universe an ETF like XLE.
We do believe, indeed, that we are likely to experience several periods of strong trends in energy stocks over the next decades. Supply constraints, booming world demography and ever increasing demand from emerging markets trying to catch up with advanced economies are among the factors behind such analysis. We want the model to be able to catch such trends when they occur in the future. The universe we have built tries to balance all these considerations. It can be broken-down between the following asset classes:
- Advanced Markets Equities: 8 ETFs, among which SPY, EWJ, XLE
- Emerging Markets Equities: 6 ETFs (both regional and large countries ETFs)
- Real Estate Investment Trusts (REITs) and Quasi-Equity: VNQ and PFF
- Commodities: 4 ETFs covering the main sub-classes: DBA (Agricultural commodities), DBB (Base Metals), DBO (Oil), GLD (Gold)
- Fixed-Income: 5 ETFs, among which SHY (Cash), TLT (Long Term Treasuries), and AGG (Aggregate bond market).
- Inverse ETF: SH (Short S&P500)
Of course, the benefits of keeping such a large share of cash can still be discussed. In a next article, we will discuss a more aggressive portfolio where the condition on the minimum share of fixed-income ETFs is removed.