Sector Rotation: Adding a “Hold” Range

A couple days ago, I looked at how buying the best performing sectors can outperform the market. In this post, I try to reduce false signals.

From the results of many research papers on momentum, is should be safe to assume that high momentum stocks outperform low momentum stocks. If we transfer this assumption to the sector rotation strategy, we would also assume that high momentum sectors outperform low momentum sectors. This has, for the most part, been the case with the Top 1, Top 2, and Top 3 portfolios outperforming the market. However, when we look at the performance of those three portfolios relative to each other, it no longer seems to hold true. The holdings of the Top 1 portfolio have, on average, higher momentum than the Top 2 and Top 3 portfolios and therefore we would expect higher returns. In the Fidelity sector version, this has been the case, but with the ETF version there was not much of a difference in performance. If I apply the same momentum strategy to the iShares Dow Jones sector ETFs, than the results are even worse, with Top 1 significantly underperforming Top 2 and Top 3.

So how can we explain the results?

I think this has to do with trades that are quickly reversed such as buying a sector one month and selling it the next or selling a top performing sector that is temporarily ranked second or third. In a portfolio with multiple holdings, this occurs much less frequently than with one holding because there is more room for a small change in rankings. When it does happen, then the effect on performance is smaller because that holding is only a part of the portfolio.

So the simple way to reduce false signals is to increase the number of holdings, but it also increases the number of trades. The Top 1 portfolio has 49 trades, while the Top 2 has 81.

The best solution I have found so far is to add a range of rankings where the sector is a “hold”. Sectors are bought if they are above the hold range, and sold only if they drop below it. With the Top 1 portfolio, we can specify the second place ranking as the “hold” range. The top 1 sector is bought if another sector has been sold, and held until it drops to third place or below. This improves the performance and almost cuts in half the number of trades to 27.

I think that this method is a good idea when several of the potential holdings have high correlation. Sectors of the stock market tend to have a high correlation and so this improves returns by avoiding some trades that are quickly reversed.

Sector Rotation Overview

Our recent website outage and subsequent switching of web hosts completely distracted me from what I like to focus on, quantitative strategies. Hopefully, I’ll be able to post more frequently than I have in the last two weeks. In this post, I am returning to the overview and backtest of the sector rotation strategy.

The sector rotation strategy, like the market rotation strategy, is based purely on momentum. I started trading it at the same time as I did the market rotation strategy, and the method I used evolved in the same way also. The only difference is the group of funds it selects from. Instead of multiple asset classes, it picks from sectors of the stock market.

I update two variations of this strategy, one for the Vanguard sector ETFs and one for the Fidelity funds that correspond to the 10 major stock market sectors. The Fidelity sector funds have been around for a while and have gained a following among some investors due to their tolerance of sales 30 days after the purchase.

Choosing the Vanguard sector ETFs over the other sector ETFs was not so clear. The other major sector ETF sponsors are iShares, State Street, Rydex, and PowerShares. PowerShares uses a quantitative methodology and the ETFs are not as liquid. While I actually like the quantitative approach to sector investing, I can’t update too many variations of the sector rotation strategy and the PowerShares sector ETFs aren’t . The Rydex ETFs use an equal weighting method, which makes some sense, but is far from industry standard resulting in lower volume for the ETFs. The iShares sector ETFs have relatively high expense ratios and use a less common sector classification system. Ruling out those three families of sector ETFs leaves Vanguard and State Street. The price action of Vanguard and State Street’s sector ETFs are almost identical, but Vanguard’s sector ETFs have more holdings than State Street’s, so I thought they were the best for this strategy. All of the sector ETF families have their strong points, so I may add other variations based on them in the future.

I think that this strategy could also be useful for determining which sectors to buy individual stocks in. If an investor can earn extra returns from this strategy and also some alpha from picking stocks, than he may be able to earn higher returns by picking stocks withing the sectors that have high momentum.

Vanguard ETFs Backtest

	Top 1	Top 2	Top 3	S&P 500
Annualized Return	6.8355437%	8.0792%	5.3031%	2.3343%
Standard Deviation of Monthly Returns	6.0940864%	5.2256%	4.9961%	5.0326%
Sharpe Ratio	1.121668311	1.546091	1.061451	0.463829

The Vanguard ETFs have a limited amount of data, and the backtest does not tell us much. Since the State Street sector ETFs have almost identical performance, we can use them instead. However, State Street does not offer a telecommunications ETF, so the next backtest only selects from the other 9 sectors.

	Top 1	Top 2	Top 3	S&P 500
Annualized Return	3.2828%	2.6287%	3.4902%	0.0693%
Standard Deviation of Monthly Returns	5.918%	4.9608%	4.589%	4.7167%
Sharpe Ratio	0.554711	0.529905	0.760565	0.014699

Fidelity Funds

	Top 1	Top 2	Top 3	S&P 500
Annualized Return	12.9347%	9.7354%	8.788%	0.2591%
Standard Deviation of Monthly Returns	6.8045%	5.4378%	5.2986%	4.7248%
Sharpe Ratio	1.900901	1.790319	1.658551	0.054843

 

Modifications

Because sectors of the stock market are all to some degree correlated, there can be a large number of useless and performance reducing trades. There are some methods for reducing this, but they also cause positions that will continue underperforming to be held longer. I’ll be digging deeper into these in a future post.

Market and Sector Rotation Strategies Overview

As with the bond rotation strategy, I’ve gotten a few emails recently asking for more information about the market rotation strategy and so I’ll be posting about it in the next few weeks. Since I haven’t detailed the sector rotation strategy and the currency rotation strategy either, I think its best not to wait for questions about them and will go over them also. I am not going to discuss these strategies in depth, as they are, for the most part, just implementations of research done by others.

The market rotation strategy is a pure momentum strategy, buying the asset classes that have performed best recently. I started using the market rotation strategy just over two years ago after reading some research on momentum.  There is quite a lot of research which indicates that momentum is a viable strategy when applied to just about anything. Particularly relevant to the market rotation strategy, is Mebane Faber’s paper, which has a strategy that is essentially the same.

Asset Classes

For the market rotation strategy, I selected eight (now nine) asset classes that it would pick from.

• Large Cap Stocks
• Small Cap Stocks
• Foreign Stocks
• Emerging Markets Stocks
• Long-Term Treasuries
• Short-Term Treasuries
• Gold
• Commodities (previously Oil)
• Real Estate (was added later)

I am actually not to sure why I originally selected what I did; I can’t imagine why I put oil in and not broad commodities (I really should have used a blog back then).

The Momentum Metric

At first, I used an exponentially weighted rate of change to measure momentum, but a little less than a year ago, I switched to the more common and simpler method of summing the performance over the most recent 52, 26, and 13 week periods. The signals were almost exactly the same, and its a lot easier to explain.

Backtests

Backtests are, of course, subject to curve fitting, so these should be taken with a grain of salt. Moreover, the following backtests do not take into account any trading fees or bid-ask spreads, so do not expect the same performance in the future.

There isn’t enough ETF data to make a backtest meaningful, so instead, I used Vanguard index mutual funds, an index, and a closed end fund instead. While the correlation between the ETFs and the substitutes isn’t perfect, I think it is high enough to be a good indicator of performance before ETFs were available. I should point out that this backtest is particularly prone to curve fitting because ten years ago commodity investing had hardly gained acceptance as an asset class and wasn’t easily investable.

• SPY = VFINX
• IWM ~= NAESX
• VWO = VEIEX
• EFA ~= VGTSX
• TLT ~= VUSTX
• SHY ~= VFISX
• DBC ~= ^DJC – Ok, this one isn’t as close a fit, but its the best I could find.
• GLD ~= CEF – Again, not that close, but probably close enough.
• ICF ~= VGSIX

	Top 1	Top 2	Top 3	S&P 500	Equal Weight
Annualized Return (%)	17.21	16.93	14.03	4.00	8.18
Standard Deviation of 12 Mo. Returns (%)	18.65	13.91	11.84	19.33	14.41
Sharpe Ratio	0.9227	1.2166	1.1847	0.2070	0.5677

Looking at just the Sharpe Ratio, the Top 2 version is the best, with Top 3 close behind it. While the performance of the Top 1 was the best, the volatility was much higher than the other version.

Just to be sure that performance with ETFs is similar to that of the longer backtest with substitutes, below is a shorter backtest with the ETFs… and, yes it looks quite similar.

One point worth considering, is that even though this strategy has lower volatility than the stock market, it could still be emotionally hard to trade. It is hard to imagine anyone actually using this strategy in the years leading up to 2000, when its underperformance relative to the S&P would have made it difficult to trade even though it outperformed a strategy of equally weighting the nine asset classes.