Quarterly Tactical Strategy Using Fidelity Fixed-Income Mutual Funds
Summary This strategy consists of ranking four fixed-income mutual funds based on 3-month returns, and then selecting the top-ranked fund at the end of each quarter. The top-ranked fund must pass a 3-month simple moving average filter in order to be purchased. Otherwise, the money goes into a money market asset. Backtested to 1986, the CAGR is 11.1%, the MaxDD is 5.5%, the worst year is +0.73%, and the return-to-risk ratio [CAGR/MaxDD] is 2.03. The monthly win rate is 79%. The strategy appears to be very robust in terms of relative momentum look-back period length and moving average duration. The strategy can be traded between the end of quarter [EOQ] and the next four trade days without any significant detrimental effect on performance or risk. I have recently been developing monthly tactical strategies that employ mutual funds instead of ETFs (see here and here ). There are a number of benefits in trading mutual funds instead of ETFs. First, mutual funds of a certain class tend to have much less volatility than ETFs in the same class. This permits the use of shorter duration look-back periods and moving averages in a tactical strategy without as much whipsaw. Second, there is the benefit of trading at one closing price, thus avoiding slippage losses (bid/ask losses) associated with trading ETFs. Third, mutual funds tend to have higher liquidity than ETFs. This avoids sudden price changes caused by lack of asset liquidity. Fourth, there are no fees/loads at all if Vanguard funds are traded in a Vanguard account, or Fidelity funds in a Fidelity account. And fifth, using mutual funds with long histories enables backtesting of almost 30 years, back to the mid-1980s. This is in contrast to ETFs that have very short histories, especially bond ETFs, that limit the timeframe of backtests. One of the negatives against mutual funds is the higher management expenses, but in some cases mutual funds actually have similar expenses as ETFs (e.g. Vanguard Admiral funds versus corresponding ETFs). And then there are the practical issues of trading mutual funds. These practical issues are challenging, but can be solved. Until recently, mutual funds did not permit monthly trading; severe short-term redemption penalties were charged or frequent-trading restrictions were imposed. But these penalties/restrictions have been lifted so that monthly trading is now permissible on some platforms, most notably Vanguard and Fidelity. This is the case as long as trades are made at least 30 calendar days apart. So a strict trading schedule must be followed that I have discussed previously (see here ). However, most of the trading issues are eliminated if a quarterly strategy is implemented. In past articles, I have presented monthly strategies using Vanguard mutual funds. But in this article, I am proposing a fixed-income asset allocation strategy that uses Fidelity mutual funds and trades on a quarterly basis. So the trading issues are greatly reduced. Four asset classes are used in the strategy: High yield corporate bonds: Fidelity Capital and Income Fund (MUTF: FAGIX ) High yield municipal bonds: Fidelity California Municipal Income Fund (MUTF: FCTFX ) Mortgaged-backed bonds: Fidelity Mortgage Securities Fund (MUTF: FMSFX ) Money market: CASHX (in Portfolio Visualizer). The overall objectives of this moderate growth/low risk strategy are: To attain a 10% compounded annualized growth rate [CAGR]; To achieve a maximum drawdown [MaxDD] of -5.0% (based on monthly returns); To produce a return-to-risk MAR [CAGR/MaxDD] of 2 or greater; To have positive returns every year in backtesting; and To attain a monthly win rate over 75%. The correlations between these funds are shown below, taken from Portfolio Visualizer [PV]. It can be seen that the funds have low correlation to each other, as desired. (click to enlarge) The strategy consists of ranking the 3-month total returns of each fund, and selecting the top-ranked fund at the end-of-the-quarter [EOQ]. The top-ranked fund must then pass a 3-month simple moving average [SMA] screen in order to be purchased. Otherwise, the money goes to the money market fund. It’s a pretty simple set of rules. What seems to make this strategy work is the relatively high return of FAGIX and its low correlation to FCTFX and FMSFX that have moderate return. CASHX is included as an absolute momentum filter to control risk. Backtest Results Using Portfolio Visualizer The strategy was first backtested using the PV software. All of the funds have histories that date back to at least 1985, so the backtesting went from Jan 1986 to Nov 2015. By using only Fidelity funds and trading on a quarterly basis, there are no trading costs, loads or restrictions if a Fidelity platform is used. The backtest results are shown below. Trading is done at the EOQ. (click to enlarge) (click to enlarge) (click to enlarge) (click to enlarge) The tactical strategy is compared with a buy & hold strategy in which the four funds are held continuously and rebalanced annually. The thing that jumps out at you is the large annual returns in 2003 and 2009; the rest of the time the tactical strategy has more modest returns as expected. The overall results show that the tactical strategy has a much higher CAGR (11.1% to 6.5%) and much lower MaxDD (-5.5% to -10.0%) than the buy & hold strategy. The worst year for the tactical strategy is a positive 0.7% (in 2008), while the buy and hold strategy has a worst year of negative 8.6%. It can be seen that the tactical strategy matches the buy & hold strategy over much of the timeframe, but in times of market stress, the tactical strategy performs much better than buy & hold. Backtest Results Using the Haynes’ Backtester The next step in backtesting was to assess the effects of look-back period length, SMA length, number of assets held, and trade day on the performance and risk of the tactical strategy. These calculations were performed using Herbert Haynes’ backtester. We first made sure that the Haynes’ backtester matched PV’s results for EOQ calculations. The comparative results are: PV’s Summary Results, CAGR = 11.1%, MaxDD = 5.5% (monthly basis); Haynes’ Summary Results, CAGR = 11.2%, MaxDD = 5.5% (monthly basis). Overall, we see very good agreement. All of the quarterly selections were exactly the same. The very small difference between CAGRs is probably caused by small variations in the adjusted price data between the two calculations. We proceeded to look at the effects of SMA duration. Rather than looking at calendar months, the SMA duration was switched to trade days. Twenty-one trade days corresponds to one calendar month, forty-two trade days corresponds to two calendar months, etc. The SMA duration was varied from 20 trade days to 70 trade days, and it was seen that SMA length had little impact on the results. CAGR varied from 11.2% to 11.3%, and MaxDD remained fixed at 5.5% Next, we studied the effect of the relative momentum lookback period. The lookback period was varied between 20 trade days and 84 trade days while the SMA screen was varied between 20 trade days and 50 trade days. As long as the SMA duration was 40 trade days or greater, the lookback period could be 2-months (42 trade days), 3-months (63 trade days) or 4-months (84 trade days) without any significant difference in CAGR or MaxDD. A final matrix was run in which the number of assets (1 to 3) and trade day (EOQ-20 to EOQ+20) were independently varied. For this matrix, the lookback period was fixed at 3 calendar months and the SMA screen duration was maintained at 63 days. The tabulated values and heatmaps are shown below for CAGR, MaxDD, and MAR. The tabulated values have the trade day on the top line (EOQ-20, EOQ-18, etc.) and the number of assets (1 to 3) in the first column. CAGR Results: Range = 6.1% [red] to 11.2 [blue] (click to enlarge) (click to enlarge) MaxDD Results: Range = -16.7% [red] to -4.0% [blue] (click to enlarge) (click to enlarge) MAR Results: Range = 0.5 [red] to 2.0 [blue] (click to enlarge) (click to enlarge) As expected, increasing the number of assets results in lower performance and lower risk. In terms of the return-to-risk metric [MAR], the optimal number of assets is one. One asset also produces the highest CAGR. The optimal trade days for one asset is seen to be EOQ to EOQ+4. It should be noted that this is not the equivalent of making a selection using EOQ data and waiting up to four days before making the trade. The way the program assessed the effect of trade day was to determine the fund selection and make the trade on the same day. Conclusions from Backtesting The tactical strategy is very robust in terms of the lookback duration length and SMA duration length. Significant variation of these parameters does not seem to greatly affect the backtest results. The selection of one asset each quarter (versus two or three assets) produces the best overall performance and risk adjusted returns. When only one fund is selected each quarter, the optimal trade day is EOQ to EOQ+4. Other trade days produce inferior results based on backtesting. 30-years of backtest results (1986 – 2015) show a CAGR of 11.1%, a MaxDD of 5.5%, and a MAR of 2.03. There are no losing years, and the monthly win rate is 79%. Some Practical Considerations These funds distribute their dividends on a monthly basis at the end of the month [EOM]. The dividend distribution does not make its way into the daily data until a number of days later. Thus, the selection that PV makes at EOQ may be in error until the correct data is available. The problem is that we don’t know exactly when the latest distribution information has been added to the adjusted data in PV’s selections. So a quarterly fund selection made by PV at the latest EOQ might change a few days later. Thus, each investor cannot just blindly use PV’s selection at the EOQ. Rather, each investor needs to look at the 3-month returns of the funds based on data that include the latest dividend distribution. There are two ways to determine the correct 3-month returns. One way is to take adjusted data from Yahoo and correct it for the latest dividend distribution. A second way is to use stockcharts.com (after the dividend distribution has been added to their data). Either way will work. There is an added benefit that can be achieved from this strategy that I want to discuss. It turns out that high yield mutual funds have a unique characteristic that I do not totally understand: when distributions occur on ex-div day, the price of the fund doesn’t drop by the amount of the distribution. For most funds, ETFs and stocks, whenever a dividend distribution occurs on ex-div day, the price of the asset drops by that amount. However, this does not occur for high yield mutual funds. I’m not exactly sure why the actual price does not drop on ex-div day, but it doesn’t. We can use this aspect of high yield mutual funds to our benefit. Thus, it will be better to always move from money market to FAGIX or FCTFX on EOQ-1 rather than on EOQ or later. In this way, you will receive the dividend without any accompanying loss in price. It’s like getting a free dividend payment. Likewise, if you are moving from FAGIX or FCTFX to money market, it will always be better to sell on EOQ or later (after the distribution is given). Because FMSFX has a relatively small distribution, the same rules apply to it too, i.e. selling FMSFX and buying FAGIX or FCTFX should be done on EOQ-1, and selling FAGIX or FCTFX and buying FMSFX should be done on EOQ or later. An Alternate Basket of Funds for Schwab Accounts For those investors who have Schwab accounts, an alternative basket of funds is recommended. Although there will be small costs for trading some of these funds, the costs will not be excessive because only one fund is traded each quarter. The basket of funds I recommend for use on the Schwab platform are the following: FAGIX, the Nuveen High Yield Municipal Fund (MUTF: NHMAX ), the Vanguard GNMA Fund (MUTF: VFIIX ), and a Schwab money market fund [CASHX in PV]. These funds can only be backtested from 2000 – 2015, and the results using PV are shown below, compared to the Fidelity version over the same years. Schwab Version (2000 – 2015) (click to enlarge) Fidelity Version (2000 – 2015) (click to enlarge) It can be seen that the Schwab version gives superior results in terms of CAGR (13.2% to 12.0%) while maintaining the same MaxDD (-5.5%). This is mainly caused by the superior returns of NHMAX compared to FCTFX.