Friday, November 20, 2015

The Current Market Environment and Implications for Tactical Asset Allocation

The past two years have been difficult ones for tactical asset allocation (TAA).  The continual V shaped corrections and a lack of trend in any asset class has caused TAA strategies to lag the US market.  On the plus side, TAA strategies have outperformed global asset classes and since we are probably closer to the top of the market than the bottom we are probably very close to a point where TAA will really show its value when we hit a bear market.  Any investment strategy will have a market environment it doesn't work as well in.  For traditional asset allocation it is a bear market and markets where most asset classes are not doing well.  For TAA it is a choppy market.  I would much rather talk to clients about underperformance in a choppy market than try to justify being down 30% while the market is down 40%.  Putting all that aside we need to consider some important issues about the past two years as it relates to the future of TAA. 

The questions that practioners should be asking now are:

1. Is the choppiness of the past couple of years a fluke that is not likely to persist
2. Is the choppiness of the past couple of years the "new normal"

The answer to both of those questions is probably no, but we need to be better prepared for years like this.  In thinking through this problem a number of important thoughts come to mind.

1. We need to take backtests with a grain of salt.  We can do walk forward testing, 3D analysis, etc to make sure we are not curve fitting but any long term backtest that shows great results doesn't have a lot of relevance in today's markets.  On the flip side, any backtest that handles these markets and struggles in trending markets is just curve fitting and also doomed to failure.  Backtesting systems is still important but we need to apply a lot of forward looking common sense.

2. All in or all out models might not be optimal.  Traditionally, TAA models were all in or all  out.  For example, a system that bought the S&P 500 when it is above its 200 day moving average would move 100% of its portfolio to stocks or cash on one fixed day a month.  That works fine when you have strong trends but fails miserably in  a choppy market. 

3. Rebalance date risk must be taken into account.  In the example above the moving average system would rebalance on one fixed day.  Sometimes it will be a perfect day to rebalance, other times it won't.  Over time this should even out, but in a choppy market with large moves missing something by a day or two can be devastating. 

4. TAA strategies need to include multiple methodologies.  In a trending market just about everything works all the time.  In a choppy market different methodologies will cycle in and out of favor.  Just like asset allocators diversify by asset class, TAA practioners need to diversify by tactical methodology.  Just like Modern Portfolio Theory (MPT) optimizes asset classes, TAA can optimize tactical methodologies.  With MPT asset classes that could never stand alone in a portfolio can be a strong part of a diversified portfolio.  The same approach can work with TAA, methodologies that might not be ideal for an entire portfolio can be part of a well balanced strategy.

5. TAA strategies need to include some sort of counter trend methodology.  A TAA strategy is never going to be nimble enough to turn on a dime in a V shaped recovery, momentum and trend following models take time to move from safe havens back into the market.  Counter trend methodologies have no such limitations and can get in at short term market bottoms and out at short term market tops.  Having these methodologies as part of a TAA strategy will allow the portfolio to get invested once a recovery has started.

We can either put our heads in the sand and hope that the choppiness of the past couple of years is just a fluke or we can adapt our models so that they not only work in trending markets but they can also do fine in a choppy market.

Monday, November 16, 2015

Using Reblance Date Diversification to Reduce Rebalancing Risk

Tactical methodologies have a number of different factors which influence their results.  Making a small change to any factor can drastically change the results of the methodology.  For example, most trend following methodologies have a lookback period and a rebalancing frequency.  The lookback period can either be long or short and the rebalancing frequency could be monthly, weekly quarterly, daily, etc.  In a previous post we talked about the concept of tactical duration where longer lookback periods and rebalancing frequency models will be less sensitive to market movements and shorter term models will be more sensitive.  We also recommened that multiple lookbacks and rebalancing frequencies be combined, in either an approach that allocates to all of them or in an approach that dynamically shifts based on the market environment.  Doing this smoothes out the return stream but it does very little to reduce the risk of when you actually rebalance.

Any strategy that rebalances must deal with rebalancing risk.  Take two identical 60% stock and 40% bond strategies that both rebalance annually.  Strategy 1 rebalances on the first trading day of the year and strategy 2 rebalances on the second trading day of the year.  Assume that stocks had a good year and bonds had a horrible year so both strategies end the year 80/20 and must rebalance back to 60/40.  Strategy 1 sells 20% stocks and buys 20% bonds on the first trading day of the year.  If on the second trading day of the year the market crashes then strategy 1 will substantially outperform strategy 2 solely based on the rebalance date.  Since tactical strategies often have larger and more frequent rebalancing this risk is especially important to address.

This risk can be reduced by using rebalance date diversification.  The concept is fairly simple, lets take a tactical model that rebalances monthly on the last trading day of the month.  This type of model obviously has a great degree of rebalancing risk as it has 12 fixed rebalance periods per year.  A way to smooth this out would be to split the model into 4 equally weighted portfolios.  Portfolio 1 would rebalance after the first week of the month, portfolio 2 would rebalance on the second week, and so on.  The final portfolio would be the average of these weekly rebalances. 

Of course rebalancing risk can hurt or help performance.  In the example above strategy 1 would not have taken issue with the choice or rebalance date.  However, not diversifying rebalance dates subjects portfolios to large swings that can go either way.  Using rebalance date diversification can smooth these swings out and provide for a better investor experience.

Thursday, November 5, 2015

Tactical Duration

Investors are aware of the term duration as applied to bonds.  Duration is the measure of how sensitive a bond will be to moves in interest rates.  Typically, the longer term the bond is, the more sensitive to changes in rates.  Shorter term bonds are less sensitive to changes in rates.  Tactical methodologies also have duration in relation to how sensitive they will be to market movements.  Unlike bonds however, short term strategies have a high tactical duration, while long term strategies have a lower tactical duration.

When implementing a tactical methodology you are often looking back over a period of time to judge the trend of an asset class.  You also have to decide how often you do this lookback and rebalance your methodology.  Short term lookbacks and/or short term rebalancing make tactical strategies very sensitive to short term market shifts, creating a high tactical duration.  Longer term lookbacks and/or longer term rebalancing make tactical strategies much less sensitive to day to day market movements, creating a low tactical duration. For example, assume we wanted to create a moving average crossing system to trade the S&P 500.  It would generate a buy signal when the S&P crossed a moving average to the upside and a sell signal when it crossed a moving average to the downside.  You have two variables to your model---which moving average to choose, and how often to do the analysis.  If you pick a longer term moving average and a longer term analysis the model will be less sensitive to market moves.  A short term moving average with a shorter term analysis will be more sensitive.

A 200 day moving average with a daily rebalance applied to the market moves we saw during August, September and October would have gotten you out sometime near the end of August and would have gotten you back in near the end of October.  Assuming you were good with your trading you would have missed much of the decline and also much of the rally, but you would have ended up right about where you started, either a small loss or a small gain.  If instead of a daily rebalance you rebalanced at month end you would have gotten out at the end of August and back in at the end of October.  That would create a much different outcome as you would have experienced most of the loss and none of the rebound.

If you had decided to use a 50 day moving average and a daily rebalance you would have gotten out sooner and missed more of the loss.  You also would have gotten back in much sooner and experienced a good chunk of the gain.  However, if you had decided to rebalance this model monthly then it would have generated the same results as the monthly 200 day moving average.

During the most recent market decline the 50 day moving average with a daily rebalance would have generated the best results by far. Does that mean it is the best methodology?  No, it worked the best in that situation, but in other scenarios different combinations would have worked best.  During the market decline we had in January and subsequent rally in February a 50 day moving average with a daily rebalance wouldn't have done as well.  It would have exited the market three times during January, only to get back in each time at higher prices.  A 200 day moving average would have been much less sensitive, it would have stayed in during the decline in January and then experienced the entire rebound in February.

Some tactical practioners use methodologies based on some sort of fundamentals.  For example, we have a tactical model that looks at S&P 500 earnings.  Because fundamental factors are much less noisy and move much slower than market prices these types of methodologies tend to have the least sensitivity to market moves, and therefore have the lowest tactical duration.

So methodologies that have lower tactical duration (longer term lookbacks and/or longer term rebalancing or based on fundamentals) work best in certain markets while methodologies with higher tactical duration (shorter term lookbacks and/or shorter term rebalancing) work best in others.  Which ones should practioners and investors use?  Since no tactical duration is optimal the answer is to use all of them.  There are two ways to implement this.  Just like investors use laddered or barbell bond portfolios to lessen the impact of any one duration, investors can ladder or barbell tactical methodologies, combining different tactical durations.  More sophisticated investors could use an optimization approach that determines which duration is working best in a certain market and skew allocations to that methodology.

Whatever investors choose they need to be aware that different tactical methodologies have different sensitivities to market moves and need to allocate their portfolios accordingly.

Tuesday, September 15, 2015

Beware Tactical Asset Allocation? recently posted an article entitled

Beware Tactical Asset Allocation

The article brought up a couple of great points:

1. Tactical Asset Allocation (TAA) is really just market timing
2. Very few TAA funds as measured by Morningstar beat the benchmark of a balanced fund (60% Stocks/40% Bonds)

Market timers try to predict the stock market.  Their goal will be to try to get in at the bottom and out at the top.  True practitioners of TAA know that nobody can predict the market, but you can react to it.  Instead of getting out at the top, you get out before bad turns into really bad.  So instead of losing 30% in a bear market you might lose 5%.  They also try to get in once a rally has started.  Who cares if you miss some of the initial upside if you have missed most of the downside.

Benchmarking tactical strategies is also difficult, a 60/40 buy and hold fund is not a suitable benchmark for a tactical strategy.  Furthermore, Morningstar puts funds that it can't really figure out what they do into the tactical benchmark.  It is hard to say how many of the funds they call tactical are truly tactical.  Finally, saying a fund is tactical is like saying a fund invests in equities---do they buy small cap, large cap, international, growth, value, etc?  There are a number of different tactical methodologies, you cannot simply lump them together under tactical.

It is true that many tactical managers have underperformed a buy and hold benchmark during the bull market, especially throwing in last year's choppy market which is the worst type of market for tactical,  it would be interesting to see the same analysis during a full market cycle that contains a bull and a bear.  Most tactical strategies are designed to get a decent upside capture with very little downside capture.  You can't just use a bull market to evaluate this.

Nobody knows what the market will do from here on out but most could agree we are closer to the top than we are to the bottom.  Given where we are in the market cycle it would seem much more prudent for investors to utilize tactical funds for some, or all, of their portfolio instead of riding the market down with buy and hold investments.

Wednesday, August 19, 2015

How Counter Trend Models Can Adapt to Changing Market Environments

Trend following methodologies attempt to buy into market strength and sell into market weakness.  Counter trend methodologies do the opposite, buying into market weakness and selling into market strength.  Trend following works under the premise that investments tend to trend over time and works best over intermediate term time frames (1-6 months).  Counter trend models work under the premise that over shorter time periods (daily to weekly) markets are dominated by noise, fear, and greed, causing them to overshoot to the upside and downside before snapping back to equilibrium.

Counter trend models can add needed diversification to tactical portfolios dominated by trend following models.  Trend following works extremely well in straight up or down markets, but doesn't work is well in choppy markets or equity peaks.  Trend following models also can't participate in bear market rallies.  Counter trend models don't work as well in straight up or down markets but do very well in choppy markets and equity peaks.  They can also participate in bear market rallies.  

The basic idea behind counter trend models is to use some measure to determine whether a market is oversold, overbought, or at equilibrium.  These could be static models that perhaps buy on n day lows and sell on n day highs or they could use some sort of dynamic logic.  Because markets are dynamic I have not seen static counter trend models work well over long periods of time.  Dynamic models are usually a much better option.  

There are two market factors that will determine how a counter trend model will react and perform, noise and volatility.  Noise is a measure of whether the market has a direction or not, while volatility measures the extent of up and down moves.  Market noise will determine how well a counter trend model will perform while volatility will determine what types of market moves a counter trend model will need for it to determine whether a market is overbought or oversold.  Counter trend models tend to work best in markets that are noisy and volatile.  They tend to work worst in markets that have low noise and high volatility (trend following models work best in this environment).   

So far in 2015 we have had an extremely noisy market with no direction, but volatility has been low.   In this environment---high noise, low volatility---an adaptive counter trend model wouldn't need large moves to determine whether the market is overbought or oversold and it could move in and out of noisy markets very frequently.  This is what we have seen in our counter trend models this year.  In a high volatility market, adaptive models would need larger changes to move in and out of markets.  

The Smart Beta Dilemma

The debate about whether smart beta strategies add value vs. market cap weighted indices will probably never end.  Smart beta strategies continue to launch with impressive backtested results that show outperformance over market cap weight.  The main questions revolve around whether those backtests are curve fitting and/or anomalies and whether that outperformance can persist.

If you are a long term buy and hold asset allocator then this presents a problem.  You are benchmarked against market cap weighted indices so any allocation to smart beta involves risk of underperformance.  You have to be sold on the backtest and the idea that any outperformance will be persistent.  If you are right you might add alpha, if you are wrong then you have a problem.

Tactical asset allocators don't have that problem.  Whether any smart beta idea is an anomaly or not isn't as relevant if you are not going to buy and hold something for ever.  The only relevant issue is whether or not a smart beta idea has periods where it outperforms a relevant market cap weighted index and how volatile that performance is.  Using relative strength or absolute momentum models of smart beta along with market cap weighted products allows the tactical investor to switch back and forth between smart beta and market cap weight (or cash in the case of absolute momentum).  Whether the outperformance will persist over the long term isn't relevant as long as it persists long enough for the tactical investor to make money and isn't so volatile that it moves down faster than a tactical model can exit.  

For the tactical investor, smart beta is another tool in the toolbox.  Maybe smart beta strategies are better than  market cap weighted indices, maybe they are not, but they do give tactical investors another set of assets to use in their models.

Wednesday, July 29, 2015

Factor Return Dispersion

So far in 2015 we have seen much more dispersion in the return of different factors---low volatility, momentum, value, increasing dividends, etc.  As you can see from the chart below it didn't really matter what factor you chose in 2014, all of them did well, except for size (small cap).  This year has show much more dispersion, from momentum (MTUM ETF up 9%) to high beta (SPHB down 5.78%):

Factor ETF 2014 YTD 7/28/15
Value RPV 12.21% -3.17%
Momentum MTUM 14.62% 9.00%
Dividends NOBL 15.54% 0.81%
Quality QUAL 11.70% 4.41%
Low Vol USMV 16.33% 3.64%
High Beta SPHB 12.68% -5.78%
Size IJR 5.85% 1.68%
Standard Deviation of Returns 3.49% 4.91%

Source: Morningstar

We already have a factor rotation model in TUTT and will be expanding the universe and adding a bit of factor rotation to our Core Satellite Strategies to take advantage of this dispersion.  The Core Satellite Strategies will keep a fixed 60% allocation to factor/smart beta ETFs but now they will incorporate a rotation model that can take more advantage of dispersion among factors.