All-ETF Portfolios vs. Strategic Mix of Stocks

One of the interesting issues in portfolio planning is as to the relative merits of investing in a portfolio of all mutual funds or ETFs as compared to including a strategic mix of individual stocks. This is a particularly important issue because many retirement plans do not allow their participants to invest in individual stocks.

This is an issue that has long bothered me—so much so that when I set up the 401k plan for my company, I looked until I found one that supported investment in individual stocks. Plans that allow investment in individual stocks are few and far between, however. What this means is that participants in fund-only plans have to maintain separate accounts for individual equity holdings. There are a number of issues that have resulted in retirement plans not giving their participants the ability to choose their individual investments and I will not delve into these issues here. For a wide range of reasons, but with retirement plans playing a big role, the direct ownership of securities by individuals is quite low and dropping – as cited below from a talk by Vanguard founder, John Bogle:

“…direct holdings of stocks by individual investors have plummeted from 92 percent of all stocks in 1950 to only 32 percent today, as corporate control fell into the hands of giant financial institutions—largely pension funds and mutual funds—whose share soared commensurately, from 8 percent to 68 percent, a virtual revolution in ownership.”

There are, of course, real costs to investors when investments are handled by ‘intermediaries’—the fund managers. Aside from the known costs, made painfully clear by the eminent Mr. Bogle (see link above), there are lesser-known portfolio impacts of building a portfolio from a set of funds.

The risks associated with investing in individual securities (as opposed to baskets of securities in the form of mutual funds or ETFs) are understood. If you invest in a small number of individual stocks, you are exposed to substantially higher risk of company failure or malfeasance. Further, many individuals tend to overweight the proportion of their assets in their employer’s stock when this option is available—particularly when the employer stock is doing well. On the other hand, when you invest in individual stocks you pay no “annual fees” and you can control your distributions of gains. Mr. Bogle has long focused attention on the long-term costs of fund fees and these are a major drag on many investors’ portfolios. ETFs charge lower fees than mutual funds and can help investors to mitigate the long-term impact of fees on their portfolios.

There is another more subtle (but substantial) cost of building your portfolio out of funds—even low-cost ETFs—that is due to portfolio effects. This issue is not at all appreciated and I have yet to see an article that really details the problem with portfolios that are made up entirely of ‘baskets’ of securities (i.e. funds). To grasp the negative portfolio impacts of building a portfolio out of a series of funds, you must understand a bit of portfolio theory. What Harry Markowitz taught us with the invention of portfolio theory is that you can get a higher return relative to the risk that you bear if you combine assets in your portfolio that are not well correlated to one another.

When you combine individual securities to build your portfolios, you are in the best position to exploit the positive portfolio effects of combining assets with low correlation. A mutual fund or ETF is typically made up of enough individual issues that the fund has captured some portion of the benefits of diversification internally. When you combine two funds in your portfolio, there is less potential for exploiting diversification effects than if you combined two securities with the same individual risk/return statistics.

To start to appreciate these effects, you need to understand that the diversification benefits of combining assets in a portfolio effectively top out once you hit about twenty securities in your portfolio. This effect is well known. See, for example, this article from Investopedia on over-diversification.

The diversification benefits of combining ten individual stocks together in a portfolio are typically far greater than the diversification benefits of combining ten individual funds together in a portfolio (having accounted for any differences in risk in individual stocks and funds). This effect is present even without the impacts of fees and can easily be costing you 1%-2% in return per year.

This is all sounding pretty abstract, so let’s get down to an example. I have started by creating a portfolio of ETFs and a couple of good Vanguard bond funds. This portfolio will look pretty ‘diverse’ by the standards of most investors and financial advisors.

Funds in the Portfolio

The portfolio components (above) include a broad market index as well as focused ETFs for utilities, basic materials, and REITS. There are two bond funds, and I have included an emerging markets fund and a European large-cap fund. I am not saying that this is an optimal mix of funds. This is an example case. We are going to look at the relative merits of adding some individual securities (XOM, C, EXC, and BA) to this mix. I selected these individual securities to keep default risk fairly low—these are large firms—and also to allow me to create an overall portfolio with similar total risk to the original (all fund) portfolio.

To build and analyze the portfolio, I have used our Quantext Portfolio Planner [QPP]. This is a portfolio analysis tool that generates forward-looking projections of how a portfolio will perform, as well as analyzing historical performance. QPP uses historical data as inputs and then generates projections based on long-term relationships between risk and return in capital markets combined with the historical data. I have assumed a long-term average return of 8.3% per year for the S&P500, with a standard deviation of 15.07% per year—my standard baseline assumptions. To understand where these assumptions come from, see http://www.quantext.com/EquityRiskPremium.pdf [pdf file].

My initial all-fund portfolio is a fairly generic allocation, with 30% in the ‘all market’ fund, IYY, and 20% in bonds (below). The portfolio has a 20% allocation to international funds via ADRE and ADRU and some concentrations in utilities (IDU), basic materials (IYH), and REITS (ICF).

All-Fund Portfolio

This portfolio has performed rather well over the past three years, the period from which we draw out historical data, with an average return of 15.27% per year and a standard deviation in annual return of 7.98% (Historical Data above). This portfolio has beaten the S&P500 in terms of average return by almost 6% per year over this period, with about the same level of total volatility as S&P500 (as measured by standard deviation in return). The projected future performance (Portfolio Stats) of this portfolio is also pretty good, with an average annual return of 10.94% per year—2.6% per year better than the future performance of the S&P500 with only slightly more risk than the market as a whole.

Note that this portfolio is projected to have a long-term volatility that is about twice what it has generated over the past several years—but this is a consequence of our standard assumption that market volatility will revert to its long-term average as opposed to the very low levels that we have seen over the past few years. For more discussion of where we get assumptions about total market volatility, see this article.

The all-fund portfolio has a Beta of 89% and the mix of stocks has allowed us to increase the return relative to total risk in the portfolio by combining high Beta and low Beta assets together. To really make the most of portfolio effects, you need to exploit low correlation between the non-systematic returns of the portfolio components. The degree to which we have accomplished this is measured by the Diversification Metric [DM] in QPP. If the non-Beta components of returns (i.e. the fraction of return that is not driven by the market as a whole) are perfectly correlated, the Diversification Metric has a value of 0%. As the non-systematic components of return are less correlated, the value of the Diversification Metric increases. The value of 38% for the DM for this portfolio is quite good but not optimal. For realistic portfolios with this level of risk tolerance, the maximum levels of the DM that I have been able to generate are between 50% and 60%. Note that DM is an indicative metric. The higher the value of DM, the higher the average return relative to total portfolio risk that you can achieve.

The total portfolio return for a given level of risk is constrained because this portfolio consists of assets that are already somewhat diversified—they have already exploited some potential diversification effects internally and these are already reflected in the risk/return balance of the individual assets. If we were to replace one of these funds with a single stock that had the same average return and standard deviation, we have the potential for a far more significant portfolio diversification impact. Obviously this is not true for all possible stocks, but lowering the weight in a fund and raising the weight in an individual equity has good potential to improve the overall portfolio diversification benefits which equates to raising the average return for a given total portfolio risk.

After a bit of experimentation, I came up with a mixed portfolio that includes the funds along with allocations into some of the aforementioned individual stocks (below).

Portfolio Mix of Stocks and Funds

The mixed portfolio that I ended up with has markedly higher historical and projected average return than the all-fund portfolio, with less risk (i.e. lower standard deviation in returns). The inclusion of a significant concentration into Citigroup [C], Excelon (EXC), and Boeing (BA) allowed the portfolio to substantially increase the benefits of diversification. The portfolio is projected to generate 2.6% more in average return with slightly less risk than the all fund portfolio. The difference in historical performance is even more impressive, but we all know that using trailing performance as a forecast of future performance is a recipe for problems. While XOM (Exxon Mobil) was a candidate for inclusion, it did not help in building a portfolio with the risk constraints that I set forth, largely because XOM is extremely volatile and the portfolio effects just weren’t too impressive.

I was able to generate the higher total portfolio returns with less risk because of improved portfolio diversification effects between the non-systematic returns of the portfolio components. This is exactly the effect described earlier. The individual equity components have more potential for adding diversification effects to the portfolio than the funds because the funds’ risk / return balance reflects a considerable fraction of the available diversification effect that has already been exploited between the stocks that make up the fund. We see the impact of the inclusion of individual equities very clearly when we look at the Diversification Metric [DM] value of 56%. This is much higher than the 36% from the original portfolio. The higher value of DM means that the non-systematic volatility in returns (the fraction of returns that are not explained by moves in the broader market) offset each other better between these assets.

The benefits of adding individual equities to a portfolio made up mostly of funds are easily explained from portfolio theory (for standard presentations see the Investopedia article cited above or A Random Walk Down Wall Street), but most investors don’t grasp the implications of this issue for their own portfolios. Investing in a number of funds has the benefit of lowering your exposure to the default risk of any individual firm, but an all-fund portfolio diminished your ability to fully exploit the only real ‘free lunch’ available to investors: diversification effects.

While most individual investors feel more comfortable investing in mutual funds or ETFs for their simplicity and the desire to minimize stock-specific risk, it does not take a lot of work with QPP (or with a careful analysis of historical data for that matter) to demonstrate the value to the overall portfolio of having a substantial allocation into a small number of well-chosen individual stocks. Three individual stocks (as shown here) will typically be too few for a real application, but five to ten individual stocks should be plenty to effectively maximize diversification effects in a portfolio with broad exposure to asset classes via ETFs or low-cost mutual funds.

My analysis suggests that many investors can generate an average of 1% to 2% (or more in the case shown here) per year in additional return (for a given level of total portfolio risk) by investing in perhaps 1/4 to 1/3 of their portfolios in a small number of high-quality individual stocks. To really calculate the effects, you must run the numbers of course.

Note: The Quantext Portfolio Planner [QPP] and Quantext Retirement Planner [QRP] are the only tools available to individual investors and financial advisors that I am aware of that are capable of effectively accounting for total diversification effects (systematic and non-systematic) and the ability to capture these effects is very important in building your best portfolio to capture the benefits discussed in this article. For more information on this issue, see our recent article on this topic.

Comments

[RogerNusbaum]

From a big picture standpoint, I agree all ETF is not ideal. Investors should seek the best tool for each part of the market they want to own.

Might be an ETF, OEF, whatever.

[GeoffConsidine]

By the way, this title was added by an SA Editor--not my title.

[XTF Group]

It is our view that a properly allocated portfolio of ETFs can have a better return on investment than a portfolio of individual stocks. Moreover, even a single broad-based ETF generally is a better strategy than a portfolio of individual stocks.

The out-performance results shown in the above article are a good example of “data mining” – plugging in past performance and volatility of individual stocks to predict future performance. Numerous studies have shown that this is not a repeatable process.

Also, the C, BA and EXC stocks mentioned in the stock portfolio are included in the Total Market Index, and as such do not add diversification. Again, this study appears to be a statistical data mining exercise, where components are specifically selected based on past-performance to demonstrate out-performance of the market.

We see little justification for the ETF asset-allocation shown here. It is our belief that simply adding utilities and basic materials to a portfolio does not necessarily add diversification. These sectors are already included in the Total Market Index; adding them simply results in over-weighting these sectors.

Additionally, the study does confirm the fact that there are asset allocation models whose performance can be matched by a few individual stocks. It is essential that an optimal asset allocation model is in place to form any portfolio. ETFs may offer the best available option for efficient asset allocation without negative-alpha risk and with low expenses.

[GeoffConsidine]

Reply to XTF:

I am afraid that you have missed the point here. This is not a data mining exercise--this effect shows up commonly and is a natural consequence of portfolio theory. You can find this effect with many possible combinations of portfolio assets because of the non-linear diversification benefit and the number of holdings in a portfolio increases. If you have the time to read other articles you have written, you will find that our portfolio tools are remarkable resistant to 'over fitting' to historical data and this is well documented.

Further, you will note that I specifically stated that this sample portfolio is not ideal or suggested--it is just there for purposes of illustration--as I said, this effect will show up in almost any portfolio.

[John Doe]

It appears both remarks on this article missed the issue with the allocation. As a journalist, if you are going to quote Harry Markowitz you should apply his theories as they are being used in portfolio theory like at the Yale Endowment, which is to use mean variance optimization to come up with a portfolio that contains investments which are not correlated to each other. The investments that should not be correlated to each other should come from different asset classes. Your first allocation appeared to be a negligent shotgun approach to using ETFs especially if you compare the portfolios in the big picture. Portfolio 1, has 50% in domestic equity IYY, IDU, IYM, and 20% in foreign equity for a total of 60% in one asset class, the stock market with only 20% in bonds and 10% in real estate. You forgot a major asset class which is commodities. However, you illustrate the beauty of Markowitz work by reducing the stock market exposure by 20% (helpful when trying to reduce beta to the market) and you add 13% EXC which is another way to get exposure to the energy or "commodity" market, then you say here is a great example of why only ETFs should not be used? Try adding a commodity ETN like the new DJP which tracks the Dow Jones commodity index and you might be shocked to find about a 13% return with a 9% standard deviation of the last 10 years. If anything your article was an excellent example of why journalists should not be creating ETF allocations for the public to invest in. This was some of the most irresponsible journalism that I have read in a long time and it is sad to see reports like this over an investment vehicle which is still a bit confusing to the public with no help from your original article.

[CrossProfit]

John Doe-

Geoff is making two points and <b>NOT</b> trying to set up a portfolio.

Point one: Use mean variance optimization thus achieving proper diversification.
Point two: ETFs or low-cost mutual funds have a cost.

You agree with point one and disagree with point two as the aprox. 1% difference between ETN and Geoff's method is negligable in your opinion (the ETF cost).

Disclosure: Comment written by a CrossProfit non journalist professional analyst. This is not an opinion on the article.
www.crossprofit.com

[John Doe]

It appears anyone can call themself a professional. Is there a reason why you do not have any bio information on the "about us" section of your crossprofit.com website? I was curious especially since it says that you have "teams are composed of many of the best and brightest people."

[CrossProfit]

Just curious but I have to ask being that you choose to maintain anonymity. Do you appreciate that there may be reasons that others like the heads of our teams and analysts are required to maintain anonymity. As for our CEO, Saul Sterman’s picture appears in his SA Bio! You have me at a disadvantage being that I have no idea who you are or what you do for a living.

The reason that I didn’t state an opinion on Geoff’s article is that I am not the ETF expert at CrossProfit. The CrossProfit ETF expert is David Kimche who happens to have 30 years of banking and investment experience. Until recently he held one of the top positions at Israel’s fourth largest bank. Being that Mr. Kimche has his own website and is no longer with Mizrahi Bank anonymity is not required. Mr. Kimche specializes in all ETF portfolios for himself and his clients (as well as other types of investments for some larger family funds).

www.kimfo-fs.com/found...

It appears to me that you are striking out at the wrong people. Both the "journalist" comment and the "about us" slur were superfluous.

Should you wish to apologize, please do so at faq [at] crossprofit.com, Attn:BBK

[GeoffConsidine]

To John Doe:

I would like to address several of John's points. The essence of portfolio theory is not to invest in different asset classes because they are different. To the contrary, it is about correlation. Utilities, for example, have low correlation to the S&amp;P500 and thus are very useful in exploiting 'portfolio effects'. My articles emphasize exploiting portfolio effects--correlations-... than simply thinking that you can diversify by buying 'different' asset classes. It is unwise to think of 'stocks' as an asset class and to thereby constrain your thinking.

Mean - variance optimization is not all of portfolio theory. In fact, the pitfalls of mean-variance optimization are well known. I would suggest that you have a look at The Intelligent Asset Allocator in which Bernstein identifies many problems with applying that concept in raw form. As Bernstein shows, if you run a mean variance optimizer over the recent past, you can always find some portfolios that dramatically outperform on a risk-adjusted basis. In recent years, commodities have been outperformers. I understand that.

This article is written to focus in a single issue--the portfolio was not intended as a suggested portfolio. I feel that this is clear in the text.

As an aside, and if it matters to you, I have worked as a quantitative analyst for more than seven years--with a focus on portfolio risk management. I am not a professional journalist.

Nothing that I said is designed in any way as a judgement on any asset class. It is an example. I have no beef with commodities as an asset class and they have dramatically outperformed as an asset class in recent years.

The simple point of this article is that combining individual stocks can provide a level of diversification that is not entirely available buy buying funds. This is a direct consequence of the statistics and how volatility scales with the number of portfolio components. This is a complex idea and very new for many people. I suggest reading A Random Walk Down Wall St for intro to this idea.

[John Doe]

Geoff,

If you read my first reply you will find that I did not merely say that a portfolio should have "different investments." For convenience I have copied it in this post.

"As a journalist, if you are going to quote Harry Markowitz you should apply his theories as they are being used in portfolio theory like at the Yale Endowment, which is to use mean variance optimization to come up with a portfolio that contains investments which are not correlated to each other."

Thus, I agree with you and often state the same remarks you made. In other words, different investments do not create a better portfolio, it is better to have ones that are noncorrelated with each other. On that note, I have run some correlation studies that go back 10 and 20 years with monthly data and must disagree with you regarding your view on "utilities" because the correlation between utility "stocks" and the S&amp;P500 stock market are often above .70 which illustrates a statistically significant correlation as I am sure you understand (given your work with statistics). I also have to disagree with you when you state that I shouldn't think of stocks as an asset class. I am not sure how you can make a statement that stocks also known as "equities" are not an asset class? I am sure that you can mine the stock market to find some stocks that are not significantly correlated to the S&amp;P500, however, in most instances you might find that they are closely related to another actual "asset class" like commodities or maybe a REIT?

I'm also confused that if you did not intend to set up a portfolio then why would you use a group of stocks and ETFS that appear to look like a potential portfolio to illustrate how another group of ETFs with out a few random stocks is lacking in "diversity." I believe this would appear very misleading to the average investor and appears to be a bit of a plug for purchasing your Quantext Portfolio Planner software. QPP. When in the alternative, if one would take your own advice and use at least 10 years of data and some mean variance optimization for say the Sortino ratio that they would easily come up with an all ETF portfolio that not only has a better return but a dramatically lower risk of loss of money i.e. downside deviation. Moreover, I am not making a "beef" with commodities either, I am merely stating that regardless of their recent tremendous performance that they should have always had in a portfolio because of there unique characteristic of being "non" correlated with stocks, bonds, and real estate, hence primarily why your second portfolio outperformed given you used 13% with one outstanding stock like EXC that has commodity like returns.

While you are correct that "individual stocks can provide a level of diversification that is not entirely available buy buying funds"...The question is what level of diversification and I would argue a much less valuable type of diversity that only increases event risk in the portfolio and is of much less value to the average investor..AND it is much less sophisticated given the nature of the retail investor. Oh, and I read "A Random Walk.." by Burton Malkiel as an intro to investing when getting my MBA and it was a good read although not very sophisticated and a bit outdated which is apparent when it makes fun of technical analysis and analysts as those that have "holes in their shoes." In all fairness, Mr. Malkiel may not have had the opportunity to truly grasp the power of technical analysis or basic technology given that even his 6th Ed. was written in 1996 before the public started even using email or the Internet.

Oh, although your remarks about the "spark spread" seem very interesting and appears very complicated it misses the point I was making which was merely to address that the only reason that your example of adding some of these stocks improved the 2nd portfolio was because the return data stream has similar "uncorrelated&quo... movements of commodities which you failed to add as even an ETF in the first ALL ETF portfolio. Thus, it appeared intellectually dishonest, somewhat self serving to purchase QPP and simply not a fair comparison especially if you have a background of an impressive 7 years with Statistics. And now that it is known that you understand statistics as a professional it is easily arguable that you were intentionally misleading and an unfair pitch to the average guy.

[GeoffConsidine]

Oh--one other thought. A company like EXC is NOT the an ideal way to get exposure to commodities. Utilities such as EXC provide some exposure to commodities but their fundamental exposure is the spark spread in commodities. The spark spread is the difference between the cost of generating electricity and the price at which you can sell it. The spark spread is widely traded in commodities markets and has its own long-term dynamics. I am personally rather bullish on having long-term exposure to the spark spread. That said, utilities also do have exposure to the direction of energy commodities, but you invest in utilities because you want exposure to the spark spread and commodities. For a direct commodity play, it is better to look elsewhere.

[GeoffConsidine]

To John Doe:

Apparently you have just not read my paper. I specifically state and have reiterated that the article is designed to show one thing: how individual equities can have better diversification properties than a fund that is already asymptotically close to its maximum diversification.

As an aside, I will note that you are apparently the only person who has read this who felt that it is "intellectually dishonest" or in any way misleading. I write articles on topics that I find interesting. QPP is intended to help investors and their advisors make better allocation decisions and we get regular feedback that it is doing just that.

Of course you are correct that I want people to buy our software--that is the nature of business.

[John Doe]

No, I did read your "paper." In your "paper" you compared two groups of securities that appear to be portfolios and try to assert that an ALL ETF portfolio is not enough and you try to prove this by comparing a portfolio of highly correlated ETFs to a portfolio with 20% less of the same ETFs and add a stock that has uncorrelated return properties (which you apparently knew very well after reading your fancy description of the "spark spread" of EXC). Which is just plain misleading and intellectually dishonest especially given your "knowledge" of statistics.

Now, with your most recent remark you are asserting that you were only comparing some individual stocks to a "asymptotically maximum diversified fund." Are you serious? If you really wanted to sell some of your software why don't you suggest to readers to use the software to put together uncorrelated portfolios with ETFs which would be safer for the average investor or for the more experienced investor who may want to specialize in single stocks or areas of the stock market they could use your software to mine for this type of stocks? But, instead you created a confusing hypothetical scenerio between two portfolios that fails to satisfy basic logic.

To your aside, it appears you have a habit of making poor assumptions but at least it illustrates you are consistent. Simply because I am the only one silly enough to respond to this article does not mean that others have not read the article and simply decided they have more important things to do than to reveal someone's self serving, misleading article written to the average investor.

[GeoffConsidine]

Dear John (yes, this is a 'dear John' letter):

In the spirit of productive debate, why don't you email me (or simply post) a very diversified ETF portolio to start with that meets your criteria. If my supposition is correct, I will be able to find a variety of individual stocks that will improve the risk/return profile of this portfolio. Granted, such a test will be anecdotal but lets give it a shot. You will have the chance to demonstrate what you feel to be a 'best in breed' ETF portfolio. for simplicity, lets make it one with at least three years of historical data.

To make this work, I would ask that you refrain from pejorative name calling and the like. I believe that the other readers anc contributors to Seeking Alpha might feel the same.

[goatfarmer]

I fee the same. Name calling detracts from the credibility of the name caller. Geoff, nice to see an article with a simple point.