Kendrick Wakeman, CFA, is founder and CEO of FinMason, a Boston-based financial technology and investment analytics firm.
Traditionally, when advisers think about portfolio construction and drivers of risk, they think in terms of asset classes. However, in the institutional investment space, more and more asset allocators and risk managers are thinking in terms of factors instead of asset classes. This begs the question: if institutional investors such as CalPERS find factor analysis useful, could it benefit retail wealth management as well?
First, let’s start with a discussion of the difference between factors and asset classes. Strictly speaking, factors are underlying drivers of risk and return while asset classes are groupings of similar types of investments. Put another way, an investment’s asset class is usually part of a family of investments, such as equities, debt, or cash. Factors are different in that they are drivers of return instead of groups of assets, examples being inflation, the U.S. dollar, or interest rates.
In his 2013 article on factor investing (“Risk Factors as Building Blocks for Portfolio Diversification: The Chemistry of Asset Allocation,” published by the CFA Institute), Callahan Associates’ Eugene Podkaminer described factors as atoms and assets as molecules. Being a foodie, I prefer CalPERS’ approach, where they describe the difference between factors and asset classes using two bowls of soup. One soup, a clam chowder, has 220 calories, 10 grams of fat, 23 grams of carbohydrates, and 9 grams of protein. The other soup, a tomato basil, has 49 calories, 0 grams fat, 10 grams carbohydrates, and 2 grams of protein. In their example, asset classes are like bowls of soup with the nutritional information being the factors. A key point of differentiation is that factors are a part of asset class returns, but asset class returns are not a part of factors.
Unfortunately, this last point speaks to one of the main sources of confusion since many factors sound very similar to asset classes, or actually are considered asset classes. For example, some people might consider “U.S. equity market” a factor, but “U.S. equities” might also be considered an asset class. It’s a very subtle difference, but consider this: each stock in the asset class “U.S. equities” has a different exposure to the U.S. equity market (beta). Google has a beta of 0.93x, meaning it generally returns 93 percent of the return of the S&P 500. So, if you are asking what the factor exposure of Google is, the answer might be that it has a 0.93 exposure to the U.S. equity market factor. But, if you asked what Google’s asset class was, the answer might be “U.S. large cap equity.”
One of the advantages of factor analysis is we can break investments down into their drivers of return to get a better picture of the risk and return drivers in the portfolio.
Consider a portfolio of two investments: a 50 percent allocation to a global bond fund and a 50 percent allocation to a utility fund. A typical asset allocation pie chart might show a 50/50 split between bonds and equity. If we look at the same portfolio using factor analysis, the pie chart may show the portfolio is 32 percent U.S. equity markets, 46 percent interest rate exposure, and 22 percent foreign exchange risk.
An expert might know that utility funds carry a lot of interest rate exposure or that global bond funds carry a lot of foreign exchange risk, but they might not know how much. Needless to say, a less-informed investor would not likely know about these hidden risks without factor analysis.
What factor analysis reveals in this instance is a higher exposure to U.S. interest rates than might be suggested by asset class analysis alone. This is because the utilities fund is considered 100 percent equity from an asset class perspective, but shows a fairly strong sensitivity to interest rates from a factor perspective.
In 2008, many investors looked at asset allocation pie charts showing that their investments were distributed across a great many asset classes, giving them the impression that they were well diversified. What the pie charts did not show was that underlying many of their asset classes were common risk factors that actually lead to less diversification than they had thought. Factor analysis uncovers and quantifies these common, and sometimes hidden, risk factors to give investors and advisers a better view on the true diversification of their portfolios.
Another potential benefit to factor analysis is that it can be easier to interpret for people with no financial training. While many non-professional investors might struggle to form a mental picture of a portfolio described as large-cap growth, international large-cap core, mid-cap value, small-cap value, and U.S. Treasury bonds, they might find it easier to grasp a portfolio defined as equity, U.S. dollar, and U.S. interest rates. This might be of interest for advisers who are trying to improve their communication strategies with respect to increasing the education and understanding of their clients.
Clear and effective communication of portfolios can be facilitated by scenario analysis in some less obvious ways. For example, it is generally easier to estimate how a portfolio would behave under different economic scenarios when you use factor analysis. This is because factors tend to relate more directly to macroeconomic factors than asset classes.
For example, if your clients are suddenly concerned about an oil war developing in the Middle East, a factor analysis would show your exposure to oil directly while an asset allocation would not. Even if you looked at how many oil stocks you owned from an asset class perspective, you would still need to take into account other industries that are also sensitive to oil, like airlines. Building that sort of analysis is relatively easy with factor analysis. The net result is a clearer communication about that particular risk.
Factor analysis can help inform your expected return assumptions for a portfolio and, subsequently, help you to explain how you developed your expected return assumptions in a way more people can understand. For example, if you have a gold fund in the portfolio, you could find its sensitivity to gold and its sensitivity to stock prices, two factors that generally effect gold stocks. Now you are in the comparatively luxurious position of being able to explain your expected return assumption for the fund as a certain assumption for the long-range returns of stocks and a certain assumption for the long-range return on gold. That is much more intuitive and probably the exact way a CIO would approach it.
One sometimes frustrating aspect of factor-based analysis, which is shared somewhat with asset class analysis, is that there is not much agreement on a universal set of factors to consider. The right factors to consider depends a great deal on what questions you are trying to answer.
For example, from a high-level risk standpoint, CalPERS likes to think in terms of their overall exposure to interest rates, inflation, and growth. For more precise risk, or for more precise predictions, an investor might want to consider an expanded factor set, such as U.S. equity, U.S. dollar, international developed equity, emerging markets equity, U.S. interest rates, corporate bond spreads, oil prices, and gold prices. There is no need to stop there and you can get as detailed as you choose.
However, asset classes have one important advantage over factors: they are easy to produce. If you want to find out what asset class a fund is in, you probably just need to read the name of the fund and, failing that, read the investment objectives. If you want to find out the factors of a fund, you generally need to perform a regression analysis or use some other statistical technique. This explains the popularity of asset class reporting over factor-based reporting. But, as computational power and the Internet pervade the wealth management space, such computations are becoming more available at a cheaper cost. At some point, cost and ease will approach parity and, as incredible as it sounds, we may see an end to the ubiquitous asset allocation pie chart.
No modern discussion about factor-based investing can be complete without discussing one particular class of factors due to their fairly significant popularity in the equity markets and the rise of smart beta: the Fama French Factors.
Eugene Fama and Kenneth French developed an extension of the CAPM model that basically stated that a stock’s systemic return is not just based upon the movement of the stock market, but an influence of two other factors: (1) how well small-cap stocks were performing relative to large-cap stocks; and (2) how growth stocks were performing relative to value stocks. This is referred to as the “Fama-French Three-Factor Model” and adds a size factor and a growth factor to the usual equity market factor. This not only spawned the popular “style boxes” at places like Morningstar, it became the early basis for the smart beta movement. Several other academics have built upon that basic model, including Fama and French themselves, who are now up to five factors: stock market, size factor, growth factor, momentum factor, and quality factor.
Fama-French factors and other forms of smart beta have arisen out of a realization that asset classes are somewhat limited, being a somewhat arbitrary construct. By building portfolios of factor exposures instead of asset classes, many academics and practitioners feel that diversification can be increased, thereby increasing expected return and/or reducing risk.
The smart beta movement is an attempt to not only give investors the ability to see the factor exposures in their portfolios, but to sculpt those exposures to their exact specification by providing investments that closely mimic the factors themselves. Whether or not smart beta funds are used, using factor analysis in portfolio construction can give advisers another, perhaps more flexible and inclusive, tool to inform their investment decisions.