TAA Investing


There is a lot of interest in Tactical Asset Allocation (TAA) portfolios these days. The big TAA models are the various versions of the IVY portfolios (GTAA5, GTAA13, GTAA AGG3/6) and the Antonacci GEM/GBM portfolios. See here for a recent comparison. There are many others. The promise of higher than equity-like returns with low risk and drawdowns would be appealing to any investor. But often an investor’s actual real world experience with TAA portfolios can be a lot different than what the historical backtests or what investors’ expectations would suggest. In this post I’m going to list what I think are the biggest problems with TAA portfolios and what, if any, the alternatives or solutions are to those problems. This is going to turn into a series of posts. Here, I’ll mainly present an overview of the problems and possible solutions. In future posts I’ll dive into the nitty gritty of some of these problems, discuss some of the research in these areas, and present data on potential solutions. I’m still in the process of doing some of this research but I wanted to start the discussion before I have everything wrapped up. As usual, I’m sure I’ll get some great suggestions from my readers. Here is my list of the top 5 problems with TAA portfolios .

  1. Poor replication of the asset classes . This one is pretty fundamental. And it is also an issue with many buy and hold portfolios as well but more so with TAA. If you look at the 13 asset classes in the GTAA 13 list of assets there are a couple that have no true or good real world implementations. The best examples are US large cap momentum and US small cap momentum. In the past couple of years there are a few ETFs that are attempting to track large cap momentum (MTUM, PDP). But we don’t know how well they will track the large cap momentum index and whether it is worth the extra fees. In small caps we need to use growth ETFs, like VBK, to try and get close to the momentum index but growth and momentum are not quite the same thing. These discrepancies in replication of the asset classes will lead to differences in returns mostly to the downside. This issue is not a huge one in my opinion. Most of the asset classes in the popular TAA models, e.g large cap value, are pretty well represented by the ETFs and the coverage will probably improve over time but it is a discrepancy that will lead to tracking error and needs to be accounted for.
  2. Asset class coverage . Any TAA portfolio will leave out some asset classes. There is no point to including them all. But there are some pretty big holes in some of the models. This is even more true with some of the buy and hold portfolios. For example, the traditional 60/40 portfolio and the Permanent Portfolio leave out all international asset classes. That’s a huge problem. But even in the more modern TAA portfolios there are some big holes; international credit bonds, international real estate, international small cap stocks are just a few examples. This omission could lead to opportunity costs in terms of returns or risk adjusted performance. We should at least use a Global Market Portfolio as a benchmark. We can improve this by intelligently adding asset classes to the TAA models.
  3. Slippage, fees, taxes . This one is a biggie. Any TAA system will have higher splippage, fees, and taxes than a buy and hold system. All these three will lead to lower than theoretical returns. Taxes are an issue for any portfolio in a taxable account but more of an issue for portfolios with higher turnover as TAA portfolios tend to be. Some of this can be mitigated with tax management strategies but it is an issue that lowers returns. Management fees are a direct subtraction from the published theoretical returns can could possibly destroy any advantage a particular strategy is trying to exploit. My favorite example here is the atrocious fees and implementation of commodity ETFs. Next, trading fees nowadays are not a huge impact to TAA systems with the rise if discount brokers and commission-free ETF lists. Slippage, however, is the big issue in this category. The difference between the theoretical price the model is buying at, usually the previous closing price, is the not the price the investor will buy at. The difference from the previous close plus bid/ask spreads can make a huge difference in returns when combined with the higher turnover of TAA systems. The good thing is that modern portfolio testers let you account for these factors in the TAA system but no matter, there will be significant differences between theoretical performance and real-world returns.
  4. Moving average rules are timing the market. This issue is the single biggest discussion point for TAA portfolios. Most TAA portfolios use trend following with a moving average to lower risk and drawdowns. Many investors think this is just trying to time the market and will fail despite what the backtest results say. There is a lot of research that has been done in this area. The basic conclusion is the moving average rules work over the long term but they have significant periods where they don’t work at all and lead to lower returns than not using them. Also, the win rate for many moving average rules is barely above 50%. But the research shows there are some rules that are much more robust than others. Hint: the 200-day SMA is not the best rule. And that in some cases there is not a big impact to leaving out moving average rules all together. I’ll discuss this research in a coming post.
  5. Markets have changed, the systems can be gamed . This is an issue with any portfolio or system that is different than the market portfolio made up of purely passive indexes. The argument here goes that TAA portfolios worked in the past but since markets have changed they are not working any more. Or that now that the systems are becoming more popular that they will stop working. There is some truth to these. And this has been happening forever. For example, Price to Book (P/B) used to be the most powerful value metric but has become much less effective over time. But there are some fundamentals that have stood the test of time: diversification, value,  and momentum are the key ones. As long as the TAA systems stay true to fundamentals then they stand a good chance of outperforming in the future. Also, there are ways to get around some of the more tactical issues. For example, a system can be implemented to take action during the middle of a month, rather than end of month to avoid some of market gaming concerns.

All together these problems will lead to lower than theoretical and backtested returns. The question then becomes whether the TAA portfolio in question has enough of a performance advantage to make it a viable investment choice versus the buy and hold methods.

That’s my list of top 5 problems with tactical asset allocation portfolios of any variety. What are your concerns, issues with tactical asset allocation portfolios?

Tagged , , , , , , ,

17 thoughts on “ The top 5 problems with tactical asset allocation portfolios

  1. Good post, thank you. I’m implementing the TAA5 since taking back my portfolio from a 1% money manager (100% large cap value stock picker) this past January. I have to get out of my own way though. I stalled on each part of buying, selling and buying back VEU during its whipsaw these last few months – it tested my commitment). They analyzed the GTAA5 to see if the monthly trigger occurring on another day of a month rather the end of the month close and I think found the end of the month fared well but other days were ok too. I’m layering on other thinking around the simple 200 MA but that’s probably getting “too cute by a half” as a friend says and might contribute to more tripping over my own feet. I’m very interested in your future posts on your line above: “Hint: the 200-day SMA is not the best rule. And that in some cases there is not a big impact to leaving out moving average rules all together.”

    1. Thanks Douglas. Yes, sticking with the systems just like any other investment approach is key.


  2. Good post. I pretty much agree with the problems you list for TAA. You are fair: you now list the limitations to go along with the Positives (returns, SWR, sharpe, etc) and Tips that you have posted in the past for TAA.

    I suppose my biggest concern is that TAA just stops “working as well” (lower returns and/or higher draw-downs). It may be caused by any combination of the 5 problems you list. I’m thinking reason 5 may be the biggest concern (markets have changed).

    I suppose if we continue to use TAA, we are presuming the rule-based, non-emotional, diversified approach of TAA will overcome the problems to a high degree and allow TAA to achieve similar results in the future.

    PS – I am curious about the research to other options for the 200 day SMA you will cover in a later post … something to look forward too! Thank you.

  3. I enjoy reading your thoughtful and well researched posts and this topic is especially timely, and you are off to a great start. I have used the GTAA AGG 6 for some time now and despite the issues referenced have been satisfied with the results. I feel confident that drawdowns will continue to compare favorably with raw indices due to volatility clustering. But I wouldn’t be surprised if we have many years of underperformance due to many of the factors you reference. Three or four years of negative returns in a row, or supposed uncorrelated assets all trading in the same direction will test our faith in these tactical systems- and I think it is likely given the current CAPE. The most frustrating thing for me is in trend less markets we are often forced into frequent trades that go nowhere. But for most of us even a less than perfect system is way better than seat of the pants trading- and I think the research is pretty clear on that point.

  4. Apologies — bungled the formatting in the last reply. Please read this one instead:

    A concern I have with GTAA — across all of its permutations — is that Meb seems to have been pretty flip regarding his selection of asset classes. Here’s how he describes his selection of the 13 asset classes used in GTAA13 (and, by extension, in GTAA AGG3 and AGG6):

    “We also have the challenge that many asset classes and indexes simply have not existed for a very long time. For example, we do not include TIPs, junk or high yield bonds, emerging bonds, foreign REITs, fundamental indexes, managed futures, currencies, or other asset classes we might otherwise consider. However, thirteen asset class subgroups will likely cover the majority of the world that we would like to allocate to.”

    Perhaps this is informed by significant thought on his part, but there’s no evidence of that. The “will likely cover the majority of the world” comment seems pretty casual and consistent with the view espoused in his latest book, Global Asset Allocation, that the individual ingredients of a portfolio don’t matter that much because even if you get them wrong, you’ll “still be able to bake a pretty good cookie.” (If you read the book, this will make more sense.)

    My concern is that he was more worried with “covering the world” in GTAA13, in a pretty loose fashion, than in intelligently selecting a good set of uncorrelated asset classes that would perform better in GTAA AGG6 or GTAA AGG3.

    (With this, I don’t mean to criticize Faber … he’s obviously done a lot for those who want to try out tactical approaches. He just seems a bit flip sometimes with regard to asset selection, and it’s not clear that a lot of thought went into the asset classes he chose in GTAA13.)

    1. Daminan, I’ve read the book. I was one of the early draft reviewers. I think he put plenty of thought into the choices he made despite the way his comments come across sometimes. He started from a bottoms up approach, using the 60/40 US stock US govt bond model, which the the most popular diversified portfolio benchmark, and added to it from there, to reach a more global diversified portfolio of asset classes. If you compare the GTAA13 asset classes to all the other portfolios in the book or the ones I’ve reviewed here, it is pretty much better than all of the others. Also, he was limited by data availability. It is not possible to add many modern sub asset classes and do a robust test of the portfolio. They simply did not exist. Correlations are also not a panacea for asset class choice as they change constantly. So, given the constraints I think it’s pretty good.

      However, it is by no means perfect. As I’m sure he would agree.


      1. Thanks, Paul. I knew there was some risk of coming off as unappreciative of his work with that comment, but that wasn’t the intent. He’s done great things for individual investors. That quote just seemed really casual. I know he tries to keep things simple to avoid those who are casually interested from tuning out, but in some cases, I think he’d benefit from going into just a bit more detail — not a ton, just a bit — in his work to avoid the undeserved appearance of a lack of rigor.

        (While on the topic of appreciation, I want to thank you for your articles, and especially those on tactical approaches … with discretionary approaches having failed me previously, I’m a relatively new, but highly interested, convert to tactical / quant approaches. Thanks for all you’re doing here.)

        I’d be very interested in seeing what you have to say about intelligently adding asset classes. The lack of historical data creates challenges for backtesting, and I would expect narrower asset classes to increase potential DD and volatility, but I’m interested in broad asset class coverage, especially if trying to use something like GTAA AGG3 (and especially given my belief that there’s some potential for broad markets to enter a 1970s-style sideways period).

        1. Damian, I don’t think your comments were unappreciative. Critical is not unappreciative. I agree that GAA lacked a lot of rigor. There is so much more he could of added. It seemed rushed to me.


    1. No, not really. These systems don’t trade anywhere near on the timeframes of HFT algos.


  5. As a 3o year veteran market timer, the biggest challenge I have had with TAA portfolio’s is acting on a signal, especially if a loss is involved. To overcome this, I always place a protective stop loss at the time of purchase. This is more difficult using the Ivy Portfolio signals as it is based on the “close” of the last trading day of the month. However, there are techniques one could use to limit losses if they are concerned they wont act when the signal is given. I believe any market timing system is incomplete unless it limits catastrophic losses.

    As for asset classes, I use the following: US Stocks, Foreign Stocks, US Bonds, Foreign Bonds, Reits, Commodities, Currencies, Cash. This leads to the following ETF’s
    VTI, VXUS, BND, BNDX, VNQ, GSG, GLD, MINT. (I consider GLD(gold) as a currency not a precious metal).

    I also trade the inverse when appropriate: Inverse US Stocks, Inverse Foreign Stocks, Inverse US Bonds, Inverse Reits, Inverse Commodities, Inverse Gold. These are SH, EFZ, TBF, REK, DDP, DGZ.

    I believe these asset classes fully cover the major markets of the world. There are many, many sub classes but simplicity is really needed when using a rules based timing system.

  6. Hi Paul,

    I am curious about the best time to use the GTAA buy/sell signals. I know you mentioned in a recent post that buying within the first 15-20 minutes after open (1st day of month) results in significantly better returns than buying later in the day.

    Have you also looked at buying just before the close (or in after-hours) the previous day (last day of the month)? Maybe this is something you will cover in this series.

    Thanks as always for the info you supply.


  7. I had been studying Mebane Faber’s work prior to discovering your blog, and your articles have added welcome clarity to the subject, much thanks. I invest in several of these screens along with a few I’ve customized to create system diversity. It’s my effort to smooth out gains and limit draw-downs. Initially I was drawn to Meb’s work because of the returns, but knowing my limited tolerance to draw-downs, I appreciate those that have the least. Over time I’ve learned I would rather give up returns than have excessive drawn-downs, and thus far I’ve not been disappointed with the returns. We shall see about the draw-downs when the market finally corrects.

    Regarding your point #3 and taxes, I expect you have invested in some of these momentum based systems longer than myself so I’m curious about your selling experience. Thus far I have found that many times the ETF that has fallen out of favor and is being sold/replaced due to new signals, is sold for a small loss. So while the winner’s are still gaining, there is some tax harvesting that occurs by selling the losers. At first this was a bit surprising, but after experiencing it repeatedly it makes perfect since. Over time, does this have much of a positive impact on reducing portfolio tax liability or is it mostly inconsequential? Great post.

    1. SSS, I agree that one of the best strengths of TAA portfolios is lower drawdowns. Even if the returns were the same, or are the same going forward, the higher risk-adjusted returns, make them a better choice for most investors.

      On taxes, there are many factors involved and the time period makes a big difference. My main point is that investing in any system in a taxable account will generate taxes, if only from dividends and income. And those taxes will lower the published theoretical returns. How much depends on the individual tax situation, the portfolio, etc… In my experience I have found that Meb’s estimate of 1% lower returns per year from taxes is about right.


Comments are closed.