Wednesday, July 27, 2016

Momentum on Individual Stocks vs Asset Classes


I had the pleasure of finally meeting Gary Antonacci earlier this year.

Gary is the creator of the momentum strategy that I follow and have been discussing on this blog.  I first came across his work in 2011 on the blog Abnormal Returns (which should be a daily read for investors). Gary and I have been e-mailing each other ever since. After over 4 years, it was nice to finally see him in person.

Gary gave an excellent 2+ hour presentation on momentum in Seattle. There were over 100 people in attendance and it was the local AAII chapter’s largest turnout. The presentation covered everything: what momentum is, why it works, its history and correct use.

Two main ideas on the correct use of momentum were discussed:
  1. Momentum applied on asset classes works much better than on individual stocks
  2. The role of absolute and relative momentum and the synergistic effect of combining the two
In this post, I will focus on point #1. When momentum is applied on individual stocks, there are 3 main drawbacks:
  1. Low scalability
  2. High volatility
  3. High transaction costs
Let’s discuss each of these in detail.


SCALABILITY

When applying momentum on individual stocks, trade execution can suffer for large trade sizes or thinly traded stocks. The very act of buying and selling changes the market price, hence reducing the alpha that can be achieved.

Contrast this with momentum applied at the asset-class level. You can enter/exit large positions in an ETF such as SPY (SPDR S&P 500 ETF) and the vast liquidity of this ETF would absorb the buying/selling pressure.

The chart below shows that momentum on individual stocks is theoretically best when both the number of stocks and holding period is small. However, a small number of stocks means low scalability and high volatility while a small holding period means high transaction costs.


VOLATILITY

The second drawback to trading individual stocks is that they are also more volatile. This is due to company-specific risk such as earnings reports or regulatory changes. It can also be due to stocks being thinly traded or belonging to a volatile industry such as mining.  

The table below shows a momentum strategy using individual stocks compared to the S&P 500. While the returns are higher for the momentum strategy, the volatility is also a lot higher. The Sharpe Ratio does not improve much (after trading costs, it may actually be lower for the momentum strategy).



TRANSACTION COSTS

In an individual-stock momentum strategy, you typically will hold a much larger number of stocks than you would ETFs in an asset-class momentum strategy. This results in a higher volume of trades and hence, transaction costs.

The table below shows that after transaction costs, momentum applied on individual stocks (whether large cap or small cap) actually performs worse than a buy-and-hold strategy.

We can alleviate the drawbacks of scalability, volatility and transaction costs by instead applying momentum at the asset-class level. But which asset classes should we use? 

The book “Stocks for the Long Run” by Jeremy Siegel shows that equities (especially US equities) have had the highest risk premium among all asset classes, so they should be the focus of any momentum model. Bonds should also be included since they tend to outperform equities during recessions. 

The GEM strategy I follow uses 3 asset classes: US stocks, non-US stocks and bonds. You can refer to this page to see how the strategy has fared over the past 40+ years compared to each of the 3 assets on their own. 

Despite all the evidence supporting asset-class momentum over individual-stock momentum, it is astonishing to see so many commercial momentum funds doing the opposite: AQR Mutual Funds (AMOMX, ASMOX, etc), PowerShares ETFs (PDP, PIZ, PIE, etc), iShares ETF (MTUM), Alpha Architect ETFs (QMOM, IMOM). 

Results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Additional information regarding the construction of these results is available upon request. Past performance is no assurance of future success. Please see our disclaimer page for more information.