TAA Investing

TAA portfolio performance: theory vs reality

One of the biggest challenges in implementing TAA portfolios is coming as close to the theoretical returns as possible. Theoretical returns are based on index returns which are not available in the real world. In this post I’ll explore the major items that keep investors from achieving published theoretical returns of TAA strategies and discuss some ways to minimize the gap between theory and reality. This is definitely an advanced topic but a critical one that I really never seen addressed in the financial blogosphere.

First, let’s look at the three big reasons for the gap between theoretical and real returns for TAA portfolios.

Poor index replication: TAA portfolio returns are based on indexes, e.g. small cap momentum, for some which no reasonable investable ETF exists. This is becoming less and less for an issue, e.g. the DWAS ETF is a potential candidate for small cap momentum, but many of these new ETFs are still quite small. Even if an investable ETF exists there is will be some tracking error between it’s index and the ETF.

Fees: there are two sources of fees – trading fees and management fees. Many of the ETFs in TAA portfolios are available as commission-free ETFs but some are not. And of course, every ETF has a management fee which detracts directly from the index returns.

Slippage: this is the largest source of the gap between theoretical TAA returns and real TAA returns. TAA portfolios are based on monthly investment signals. Monthly investment signals are based on closing ETF prices. Actions based on those signals are done on the following trading day. Any difference between the closing ETF price and your trade price the following day constitutes slippage. For example, a sell signal was generated on Aug 31, 2015 when VBK closed at $124.77. The following day, Sep 1, VBK traded in a range from $123.53 to $121.06. Selling VBK in that range would generate a difference from the theoretical sell price (the previous close) of 1% to 3% depending on where you sold during the day. And this does not even account for the bid-ask spread. Needless to say that would impact your returns. Usually, it is not as bad as this example and the slippage can even go in your favor but in general it detracts significantly from theoretical returns.

Now, lets put these reasons into context. I ran some backtests with the AGG3 and AGG6 strategies with some different slippage numbers. Since these backtests use real ETFs, all management fees are taken into account. The results from March 2007 through Mid Sept 2015 are below.

TAA theory vs reality Oct 2015

If you were able to trade at the theoretical closing price of the ETFs then with AGG3 the return would have been 13.77% annualized over the period. With just 0.25% negative slippage on every trade that would return would have decreased by 2% annualized to 11.77%. And with 0.5% negative slippage per trade that annualized return would have been only 9.85% annualized. As I like to say, slippage kills!

BTW, any portfolio strategy has the exact same issues – even buy and hold. The issues are exacerbated when a strategy is more active and thus trades more often.

OK, so what can we do about this. Lets address each reason individually. For poor index replication, we can always be on the look out for better constructed ETFs that more closely match the indexes and do so at reasonable costs. As I said earlier, this is less and less of an issue today.

As far as fees, we can look for the lowest cost commission free ETFs that best implement the index. Sometimes this can conflict with the first goal of good index replication. For example, is DWAS a better choice for small cap momentum at 0.6% per year in fees, vs VBK which is really a small cap growth ETF (not momentum), but is only 0.09% per year in expenses. In other words, the better index replication may not be worth the extra fees.

And then there is the big one, slippage. In theory, the solution is easy. Trade as close to the theoretical model price as possible. At the minimum, this ideally means the use of high volume, low bid-ask spread ETFs. I’ll give you my favorite example. VGLT trades 50K shares per day at an avg bid/ask spread of 0.2%. TLT trades 9M shares a day at an avg bid/ask of 0.02%. Which one would best minimize slippage? TLT by a long shot, despite the slightly higher management fee (0.15% vs 0.12%). So, to minimize slippage we may sometimes actually end up using different ETFs…You may also want to change from end of  month portfolio signals to some other day during the month to avoid volatile month end periods due to options expirations, portfolio window dressing, etc…

Then, at the advanced end of the spectrum, you can actually ‘trade the close’, i.e. execute your trades as near the end of trading as possible on the last day of the month (the day that generates your portfolio signals). The use of conditional orders and MOC (market on close) orders greatly simplifies this strategy.

I’ve been working on this strategy most of the year and have found it quite effective in minimizing slippage. Also, even choice of brokerage can impact slippage. I have stopped using TAA strategies at certain brokers due to poor execution prices. 

In summary, there will always be a difference in model returns vs real world returns. The question is how can we minimize these gaps. With attention to detail in choosing the best, liquid, low bid-ask, low cost ETFs and some smart trading strategies you can keep the gap down to a minimum.





Full Disclaimer - Nothing on this site should ever be considered advice, research or the invitation to buy or sell securities. These are my personal opinions only.

Tagged , , , , , , , ,

About paul.novell@gmail.com

12 thoughts on “TAA portfolio performance: theory vs reality

  1. Paul,
    What are your thoughts on swapping out VBK with DWAS? Is the better index replication worth the higher expense ratio?

    1. Bryce,

      DWAS is pretty new, 3.5 yrs or so, so I’d say jury is still out. Since inception it has only slightly outperformed VBK. However, it trading volume is lower and bid/ask spread higher than VBK. So slippage could be worse. I’d choose to stay with VBK for now.


  2. Hi Paul,

    Great topic and enjoy your website. I am constantly analyzing and monitoring those 3 items with my own TAA strategy (mainly based on Faber and Antonacci). It almost feels like a juggling act sometimes.

    There is another item that is sort of an “elephant-in-the-room” with me since I started using TAA back in March. I will lead into it by using a quote from your article:

    “You may also want to change from end of month portfolio signals to some other day during the month to avoid volatile month end periods due to options expirations, portfolio window dressing, etc…”

    Here (http://www.thinknewfound.com/wp-content/uploads/2014/11/Minimizing-Timing-Luck-with-Portfolio-Tranching.pdf) is a research article I found on the internet that addresses the topic about the day of the month that is used, etc., and the differences in returns, etc. and a suggested solution to address the issue.

    I track the calculated allocations daily for my own TAA model. It is amazing how much the allocations can change in just a few days. For example, back in August SPY went from a 20% (max) allocation to a 4% allocation in about 3 days, which starts the differences in returns.

    Do you have any thoughts on this issue?


    1. Yes, I think it primarily is a slippage issue. Any difference in TAA returns based on the day of month of the signal I think is purely statistical noise. I think it’s mainly about avoiding potential large volatile periods at the end of months.


  3. Hi Paul,

    Awesome Article! Well done.

    Which broker is better/worse with the slippage issue from experience?

    How’s interactive brokers if you have used them?


    1. Hey Fred,

      Interactive Brokers has the best execution of any broker I’ve tried.


  4. Hi Paul,

    I have always thought the setup of the GTAA portfolios was a bit weird, with the buy/sell signals being based on closing prices, but the actual buying/selling taking place at the open. Some of the bigger gap up/down days seem to happen on the first of the month too, which only further magnifies this problem (just my limited personal experience, not sure if the data actually support this).

    I didn’t have the guts on my own to change the system, but I am going to give it a shot from now on, so thanks!

    You mentioned using Interactive Brokers to further minimize slippage. Have you found that the reduction in slippage compensates for the increased commissions compared to a broker with more commission-free ETFs?

    1. Hey Tony,

      Yes, I have found that IB’s better execution more than compensates for trading fees. I would say account size needs to be over $100K to make it worthwhile. Just my experience.


  5. Paul,

    How does .5% slippage become a nearly 4% drag on the trading system? How did you build your model and what assumptions are you making?


    1. Adam, I use P123 in my backtests. Slippage is applied per trade. I just wanted to see what a careless implementation of the model would do to performance.


  6. Paul..

    Update: I just did a year-to-date audit of all my trades at Interactive Brokers where I’ve been trading a TAA system. Audit runs from 1/1/2015 to 10/23/2015. My system generates signals at the end of each month and I typically run my spreadsheets during the last trading day and make my trades a few hours before the close.

    Results: Slippage (difference between actual fill price and ideal end of month signals) = -0.1% of account value year to date. Also noteworthy is that commissions are -0.16%.

    1. Thanks for the data Adam. It is similar to what I’m seeing. Since I switched to a better broker and run my system before the close on the signal day, like you’re doing, my slippage has improved a lot, similar to what you’re seeing.

      Implementation details of these systems matter a lot. Which almost no one talks about.


Comments are closed.