Follow me on Twitter!
When using tactical asset allocation and ETF rotation systems, does the number of ETFs an investor chooses from impact returns? In other words, will rotating between 5 ETFs under-perform a system that rotates among 20? The answer seems apparent, more options should lead to higher returns. I decided to perform a small backtest of a relative strength ETF system. This test is not meant to be definitive and has many limitations. The primary limitation being that many ETFs have a trading history of less than five years so the sampling size is small.
While much of my interest in tactical asset allocation was initially inspired by Mebane Faber’s The Ivy Portfolio: How to Invest Like the Top Endowments and Avoid Bear Markets, this test was done using a different method available for free using etfreplay.com. I have written about this method on several occasions on my site and also track a free portfolio on a monthly basis. The primary difference between the Ivy Portfolio strategy is the etfreplay rotation system purchases the top ETFs based on a combination of returns and volatility and does not require the ETF to be trading above a moving average.
This test used a combination of 3 month returns, 20 day returns, and 20 day volatility to rank ETFs. The ETF is not required to be trading above any moving average to be available for purchase. The test purchased the top ETF and rebalanced semi-monthly (twice per month), which leads to higher turnover although with the introduction of free ETF trades and free equity trades, the added cost of commissions could be minimal.
The first portfolio examined is Faber’s Ivy Portfolio of 5 ETFs with one exception, I used AGG instead of BND because AGG has a longer trading history and the two are highly correlated. Thus, the first portfolio tested was AGG, DBC, VEU, VNQ, and VTI. The returns starting in 2007:
Total Changes (trades) 46
Total Return 110.9%
Benchmark (SPY) Return -9.7%
Benchmark Volatility 27.7%
Benchmark (SPY) CAGR -2.6%
Strategy Drawdown -10.5 %
Benchmark Drawdown -50.9%
Next, I tested the 10 ETF portfolio suggested in Faber’s book with again the substitution of AGG for BND. Eight of the 10 ETFs were trading at the start of 2007, the remaining 2 began trading in mid 2007. The ETFs are AGG, DBC, GSG, RWX, TIP, VB, VEU, VNQ, VTI, and VWO. The results:
Finally, I tested the “broad mix” portfolio I track for free on my site consisting of 25 ETFs. Of the 25 on the list, only 15 were trading at the start of 2007 which limits the tests to an extent. The 25 ETFs are BWX, DBA, DBB, DBC, DBV, EEM, EFA, GLD, HYG, IEF, LQD, LSC, PCY, PFF, RWX, SCZ, SHY, SPY, TIP, TLT, VBR, VNQ, WIP, XLE, and XLU.
By the start of 2008 20 of the 25 ETFs were trading. The results:
Given the relative short trading histories of many ETFs, it is difficult to draw many conclusions from a small sampling. However, a few things are evident. All three of the ETF portfolios/strategies clearly outperformed the benchmark, SPY. There was a slight increase in returns as the investment pool increased but the increase was not as large as I had expected.
Finally, this test purchased the top 1 ETF which may explain the similar returns among the different portfolios. I think if the top 3, for example, in each portfolio were purchased semi-monthly we would see a widening of the performance gap between the small (5) ETF portfolio and the large (25) ETF portfolio. If we assume that relative strength impacts performance and rotating into the strongest ETFs positively impacts returns, then buying 3 out of 5 ETFs, or 60%, of the available investments versus 3 out of 25, or just 12%, could reduce returns. In a strategy that seeks only the strongest ETFs, having to purchase 60% of the available investments should, in theory, be a drag on overall returns versus a strategy that purchases a smaller percentage of available investments. This will be tested in a follow-up article.
For those interested, below is a spreadsheet summarizing returns side by side:
|Ivy 5 Portfolio||Ivy 10 Portfolio||Broad Mix||Benchmark (SPY)|
|Total Changes (trades)||46||50||64||None|
Data courtesy of ETF Replay