I decided to revisit the tactical bond Low Volatility Strategy [LVS] that I discussed in this original article on LVS and this article on an enhanced version of LVS. The enhanced LVS features the use of five low volatility mutual funds including a senior floating rate fund that should do well in a rising rate environment (e.g. see this Vanguard article). LVS has different versions based on how many assets are selected each month. LVS-1 is the designation if one asset is selected each month. LVS-2 corresponds to two assets being selected each month.
It has been about a year since I originally published the articles about this strategy, and the enhanced LVS has done quite well since then (enhanced LVS-1 is up 11.8% YTD and enhanced LVS-2 is up 10.5% YTD). However, I decided I needed to try and make the strategy more robust and easier to implement in real life. So this current article will discuss some improvements to the momentum parameters to improve robustness and usability, and then I will present backtest results with the new set of momentum parameters.
Universe of Funds and Their Correlations
To review, the universe of the enhanced LVS consists of five diversified bond assets listed below:
1. Nuveen High Yield Municipal Bond Fund (MUTF:NHMAX);
2. Principal High Yield Corporate Fund (MUTF:CPHYX);
3. PIMCO Mortgage Backed Securities Fund (MUTF:PTMDX);
4. Oppenheimer Senior Floating Rate Fund (MUTF:OOSAX); and
5. Loomis Sayles Limited Term Government and Agency Fund (MUTF:NEFLX).
These funds are NTF funds (no loads/no fees) on Schwab and Fidelity platforms. The only cost is a $50 early redemption fee if a fund is sold within 60 days at Fidelity or 90 days at Schwab. There will be approximately 6 such trades per year at Schwab and five such trades per year at Fidelity; this will result in annual fees of $300 at Schwab and $250 at Fidelity. This should not be too problematic as long as an account value of $100K or more is held (0.30% annual loss at Schwab and 0.25% annual loss at Fidelity).
The overall correlations of the funds between (09/02/1999 - 11/03/2016) are shown below using the commercially-free Portfolio Visualizer [PV]. It is well known that correlations between assets are dynamic in behavior, i.e. they vary over time, but overall correlations give a basic feel for the interactions between assets.
The funds generally exhibit low correlations to each other, as desired. Also note that the Daily Standard Deviation [DSD] for each fund is 0.34% or less, signifying that all of the funds are low volatility. For reference, Vanguard S&P 500 Index Fund (MUTF:VFINX) has a DSD of 1.25% (Annualized Return of 4.54%) while Vanguard Total Bond Market Index Fund (MUTF:VBMFX) has a DSD of 0.26% (Annualized Return of 5.13%).
In order to backtest the enhanced LVS to 1990 using PV, proxies are substituted for most of the funds. These proxies should not be used in live trading on Fidelity or Schwab because of loads or excessive early redemption fees.
1. MFS Municipal High Yield Bond Fund (MUTF:MMHYX) is substituted for NHMAX;
2. Eaton Vance Income Fund (MUTF:EVIBX) is substituted for CPHYX;
3. Vanguard GNMA Fund (MUTF:VFIIX) is substituted for PTMDX;
4. BlackRock Floating Rate Income Fund (MUTF:BFRAX) is substituted for OOSAX; and
5. No substitute for NEFLX.
Previous Momentum Rules for Enhanced LVS
The original momentum rules for enhanced LVS are as follows:
1. At the end-of-month [EOM], the funds are ranked based on 10-day total returns. For LVS-1, only the top-ranked fund is selected each month. For LVS-2, the two top-ranked funds are selected each month.
2. The top-ranked fund(s) is purchased if it passes its 10-day Simple Moving Average [SMA] filter, i.e. if its current price is greater than its 10-day SMA. If not, the money goes into cash (money market).
3. Selections occur at EOM and transactions occur at EOM prices. One year ago, transactions could only be performed at EOM prices using PV. Today, PV has the option of making selections at EOM and transactions at EOM+1 prices. It turns out that the difference between making transactions at EOM prices versus EOM+1 prices is insignificant for LVS.
Results (1990 - Present) of Enhanced LVS-1 Using Proxy Funds
Overall backtest results to 1990 for the enhanced LVS-1 using proxy mutual funds are shown below. The LVS-1 results are designated Timing Portfolio in the figures. Also presented are the results of an Equal Weight Portfolio and VBMFX for reference. The results are obtained using PV.
It can be seen that the Compounded Annual Growth Rate [CAGR] is 12.2%, the Maximum DrawDown [MaxDD] is -4.5%, the annualized Standard Deviation [SD] is 5.4%, and the Worst Year is +0.9%. The risk-adjusted return ratio MAR (CAGR/MaxDD) is 2.7. The monthly win rate is 83%.
Compared to the Equal Weight Portfolio, LVS-1 has twice the growth (12.2% versus 6.0%) with 1/3 the drawdown (-4.5% versus -13.5%). Compared to VBMFX, LVS-1 has twice the growth (12.2% versus 6.1%) and the same drawdown (-4.5% versus -5.0%).
Usability and Robustness Improvements
These results are very encouraging, but there is a usability challenge and a robustness challenge that I now want to address. The usability challenge is the calculation of the 10-day SMA based on adjusted prices. The difficulty occurs because of the large EOM distribution; the EOM distribution is not usually included in free commercial adjusted price data (such as Yahoo) in a timely manner. The lack of EOM distribution information causes a challenge in calculating the 10-day SMA for the average investor.
The robustness challenge involves the use of the 10-day look-back period for ranking funds based on total returns; this duration is the short duration limit on PV. In other words, I cannot backtest a strategy with a 9-day look-back period using PV. Thus, the robustness of the 10-day look-back period cannot be truly assessed on the short duration side.
Here are my suggestions to improve the usability and robustness of the enhanced LVS. First, rather than use a 10-day SMA, I will use a two calendar month SMA as the risk-off cash filter. The two-month SMA filter becomes very easy to assess by looking at one-month total returns. If the one-month total return is positive for the top-ranked fund, then the fund passes its two-month SMA filter and the top ranked asset is bought. If not, then the fund fails its two-month SMA filter, and money for that asset goes to money market.
Second, I looked at the backtest results over a range of days and months. It turns out that the CAGR and MaxDD do not change significantly if the look-back period is changed from 10 days to 19 days (remember, I couldn't check less than 10 days). So, to improve robustness, I changed the look-back period to 15 days, i.e. midway between 10 days and 19 days.
Backtest Results (1991 - Present) of LVS-1 Using Proxy Funds and Improved Momentum Parameters
Using a 15-day look-back period of total returns for ranking funds, and a 2-month SMA as a cash filter, I reran the LVS-1 strategy. The backtest only goes back to 1991 (instead of 1990) because the data of BFRAX combined with the 2-month SMA filter limits the backtest. Here are the 1991-present overall backtest results.
It can be seen that CAGR between the original momentum parameters and new momentum parameters drops by 1.6% (12.2% to 10.6%), but the MaxDD improves (-4.5% to -3.5%). The Worst Year is also a little lower, i.e. +0.9% to -1.5%. So although the performance and risk numbers are still quite good, there is some performance degradation when the modifications are employed. Once again, these calculations are performed with proxy funds, so these results are not our final outcome with the improved momentum parameters.
Backtest Results (2000 - Present) of Improved LVS-1 Using NTF Funds
Next, I ran LVS-1 with the improved momentum parameters and the Schwab and Fidelity NTF funds (no fee/no load funds). Backtesting can only be done from 2000 - present because of the limited histories of the funds. PV results are shown below.
It can be seen that CAGR = 12.0%, MaxDD = -3.8%, SD = 5.7%, and Worst Year = +2.9%. The MAR = 3.2 and the monthly win rate = 82%. I think any conservative retired investor would be glad with this performance and risk in the years ahead.
Backtest Results (2000 - Present) of Improved LVS-2 Using NTF Funds
But for the really conservative investor, I ran the LVS-2 with the improved momentum parameters. Here are the 2000 - October 2016 PV backtest results.
In all of my backtesting, I have never seen a tactical strategy having greater than 10% CAGR and MaxDD of -1.3%. The risk-adjusted return ratio MAR (CAGR/MaxDD) is 7.7. This is a remarkably high number; good tactical strategies typically have MARs between 1.5 and 2.5. The monthly win rate is 85%, and the Worst Year is +2.2%.
Including TIP Fund in Basket of Funds
Another possible improvement to LVS is to include a Treasury Inflation Protection [TIP] fund. However, in the past (e.g. 2013) when rates increased, TIP funds did not do well (whereas the senior floating rate fund OOSAX did quite well). So a TIP fund may be a hedge against inflation, but not necessarily against rates increasing. The other issue is that TIP funds only go back to just before 2000, so backtesting is somewhat limited. There are very few instances when rates actually increased after 2000.
However, I included VIPSX in the basket of funds as a proxy for NTF funds such as Schwab Treasury Inflation Protected Bond Index Fund (MUTF:SWRSX) or Fidelity Inflation Protected Bond Index Fund (MUTF:FSIYX). We observe that including a TIP fund might reduce CAGR somewhat, but overall the strategy's results are not compromised. We can only backtest to 2001 because of the history of VIPSX. The PV results for the improved LVS-1 with VIPSX included are shown below.
Comparing LVS-1 with VIPSX to LVS-1 without VIPSX indicates that including VIPSX slightly degrades CAGR and MaxDD. But the degradation is rather minor, and an investor may choose to include VIPSX if he desires.
Frequent Trading Restrictions and Other Practical Issues
As far as I can tell, excessive trading (per prospectus) for each of the mutual funds is as follows:
- NHMAX: Limited to two round-trips in a 60-day period.
- CPHYX: Must hold the fund for 30 days before selling.
- PTMDX: Nothing specific stated.
- NEFLX: Limited to two round-trips within a rolling 90-day period.
- OOSAX: 30-day exchange limit. Fund is blocked for 30 days.
In order to avoid excessive trading rules, it is advisable to follow a strict 30-day trading schedule that moves between EOM transactions and EOM+1 transactions depending on the month. If you are interested, I have a trading schedule through 2020 that can be followed to avoid excessive trading.
There are two other practical issues to discuss. First, I would advise holding two separate accounts if you want to use LVS-2, holding one fund in one account and the other fund in the second account. This permits flexibility in avoiding excessive trading rules.
Second, it is important to sell one fund and buy another fund on the same day, i.e. not waiting a day between selling one fund and buying the other fund. If you just transfer funds, there will be a one day lag between selling one fund and buying the other fund. But if you place two transactions on the same day (one to sell and one to buy), there is no lag between selling and buying as long as there is enough money in the account to cover the buy transaction.
In conclusion, the improved LVS is a very conservative tactical strategy that selects one (LVS-1) or two funds (LVS-2) from a basket of five low volatility NTF bond funds. The strategy can be executed on either Fidelity or Schwab accounts. Trades are executed at the end of each month, either EOM or EOM+1 depending on the month. For LVS-2, the backtest results from January 2000 - October 2016 show that CAGR is 10.2%, MaxDD is -1.3%, and risk-adjusted return ratio MAR (CAGR/MaxDD) is 7.7.
An investor needs to be aware of the frequent trading rules of each fund, and be sure to avoid breaking them. Trading restrictions should never occur if a strict 30-day trading policy is followed and/or if multiple accounts are utilized (for LVS-2).
Hopefully, the modifications I have implemented in this article will improve the usability and robustness of the strategy. Here is the PV link that gives the parameters for the Schwab/Fidelity version of LVS-2. For LVS-1, just change the assets to hold from 2 to 1.
But please remember that the selections provided by PV at EOM are not correct until EOM distributions are included in the data, and this doesn't happen until many days after EOM. Hence, an investor needs to use other means to arrive at EOM selections. My suggestion is to use the commercially-free StockCharts to arrive at the selection(s) each month.
Disclosure: I am/we are long OOSAX. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.