The Case for an All-ETF Portfolio With No Tech or Emerging Market Stocks

The ETF Investor article about my "tuned ETF portfolio" elicited an interesting response from Roger Nusbaum. Roger raises a key issue for investors who want to use ETFs to build a portfolio, and in responding to his article I'll explain why you might want to build an all-ETF portfolio that entirely omits emerging market and technology stocks.

But before I discuss that key issue, let me quickly clear up two important misunderstandings. First, Roger incorrectly states that my analysis simply uses the last three years as the basis for projected performance of the portfolio. The Monte Carlo model uses the last three years of market data to estimate Beta and volatility, but uses long-term risk-return relationships for stocks and bonds to estimate the future expected return. You'll find, for example, that my estimates for future returns on energy and REIT’s are considerably lower than their recent performance.

Second, my paper does not state that the suggested portfolio is optimal. I firmly believe that there is no single optimal asset mix for all investors. I was simply showing that including these low-Beta ETF’s in the mix could provide an improved risk-return balance. I was constraining the historical and projected total volatility of the portfolio to be reasonably close to that of the Radical Guide to Investing Core ETF Portfolio. There are many different portfolios that could achieve these types of results. Given the risk constraints, I feel that this portfolio is a considerable improvement over the original portfolio. This is not the ideal portfolio for me because I prefer a portfolio which carries more risk and targets higher returns.

Now on to the fundamental issues for investors who want to use ETFs to build a portfolio. Roger's key criticism was that the sector weightings of my portfolio do not seem plausible. You'll see from the portfolio (below) that there's signficant concentration in energy, and no technology or emerging market stocks:

IVV 5%
IJH 5%
IWM 10%
EFA 10%
EEM 0%
SHY 0%
IEF 15%
TLT 15%
RWR 10%
IDU 10%
IXC 10%
IGE 10%

Yes, this portfolio has a major concentration in energy. But is this weighting contingent on energy stocks continuing to outperform as they have for the past several years? Nope. I outlined the projected annual average returns for all portfolio components in the paper. I take the further step in the analysis of discounting these returns even further to see how much of the contribution of these components is due to average return, as opposed to diversification effects due to the low Beta of these ETFs. If one believes that the future return on these ETFs / sectors will be even lower, the Monte Carlo tool allows you to adjust the expected returns manually and see the impacts on the portfolio.

The key point in Roger’s comments is that this portfolio does not follow a traditional ‘pie chart’ allocation scheme. This is totally true. Many people end up with poor asset allocation by using style analysis (the pie chart) as opposed to looking at how the portfolio pieces actually work together. I constructed a portfolio with a targeted Beta and volatility and tried to increase expected return but without worrying about whether my allocation to utilities was similar to their weighting in the index. My goal is to build a portfolio that will give me the best chance of drawing income at a targeted level through retirement. If you want to own the index, you should buy the index. I believe that building a strategic asset allocation is one of the ways to beat the index and Monte Carlo tools can help you to see how the pieces of your portfolio work together.

Are these Monte Carlo projections correct? This is a tough question. I find, as do a range of users of the Monte Carlo software, that the projected risk and return of the model are realistic, though it is always a good idea to stress test a portfolio. The model is, of course, just a model. Is the model better or worse than any given analyst? I do not know. I find that the projected risk and return makes sense. The user is free to adjust the projected returns. The projected volatility has been benchmarked against options prices for a range of sectors, so most investors will want to use the model estimates—though the professional version of the tool allows the user to adjust all parameters.

So the real question is whether the portfolio is lacking weighting in tech and foreign stocks. This paper was intended to show an analysis using the Monte Carlo tool and to show users how they may want to test portfolios. I invite Roger and other interested investors to trial the software if they wish to build a portfolio with other style considerations in mind. The results from these analyses are often interesting. A recent analysis by a user showed that the impact of a large allocation to certain international funds actually did not provide a nice diversification effect because the Beta against the S&P500 for this fund was greater than 100%. Buying stock overseas does not necessarily mean that you are ‘diversifying’ if the Beta of your portfolio actually increases! Higher Beta means that these positions actually amplify your exposure to U.S. market volatility.

The idea that buying international funds is guaranteed to diversify your portfolio is one of the abiding misconceptions of portfolio management. Okay, let’s just take a sample of international ETF’s and see their profiles (below):

These stats are calculated by the Monte Carlo model using historical data. The projected standard deviation is based on a future market volatility of 15% per year. Yahoo! Finance does not agree perfectly with my numbers for Beta, but in all cases they estimate Beta as greater than 100% by a substantial amount. The upshot is that while people have this idea that foreign stocks will provide ‘diversification,’ many foreign funds will actually make your portfolio more sensitive to the moves in the U.S. market via high Beta and these funds can also add a lot of volatility to your portfolio—such as those above. Does it make sense to have foreign stock just to have it, even if it actually makes your portfolio riskier and more sensitive to the S&P500? Now, you may have a personal view that you want foreign stocks at some percentage and your insight is probably at least as good as my own---I claim no special knowledge of foreign economies. I will note that I did not include any analysis of foreign ETF’s in the paper in question and that would be an interesting future topic.

How about Tech stocks? Tech stocks tend to be high Beta. Tech stocks may or may not have a place in any given portfolio. Roger notes that the average market cap is fairly low. It was not my intent in this portfolio to specifically under-weight tech stocks relative to the index or to focus on smaller market caps. This is just the way the portfolio turned out, given the constraints and my less than exhaustive approach. Is there any reason that any portfolio should look like the total market index? If you want to have weights like the total market, it makes more sense to me to simply buy a total market ETF.

In summary, Roger’s comments on my study have led to a number of interesting points. There is no doubt that I am using statistical measures to build the portfolio. I am interested in building a portfolio with the highest projected returns, given certain constraints on portfolio Beta and standard deviation. These projections are made using a Monte Carlo model. I am far less worried about my ‘style chart’ allocation than I am about the statistical profile of my portfolio and the way that these statistics impact my ability to sustain a future income at some level of certainty. These points of discussion all highlight the value of Monte Carlo analysis in providing insight into portfolio planning. I will not add international funds just because are a different group of firms. The Monte Carlo tool shows that many international funds will not ‘diversify’ a portfolio at all if what you are trying to do with diversification is to have risks that offset the U.S. market.

As I noted earlier, anyone who is interested in testing the Monte Carlo model for themselves need only sign up for a free trial at www.Quantext.com and I will welcome comments and questions.