Research Affiliates: Forecasting Returns: Simple Is Not Simplistic

January 12, 2016

Key Points

  1. The first and most important step in constructing portfolios is selecting the best available model to forecast asset class returns.
  2. The performance of a modeling system is measured by how accurately it identifies undervalued asset classes, which when combined in a portfolio, consistently generate alpha over the investor's time horizon.
  3. A model's value is not deter­mined by its level of complexity, but by its forecasting ability.

"It is far better to foresee even without certainty than not to foresee at all."

–Henri Poincaré1

Another year, another body blow delivered by the market to "cheap" investments. One popular definition of cheap (i.e., value) has now underperformed growth on a total return basis for six of the last nine years. Can we blame the investor who is considering throwing in the towel, dropping to the canvas, and taking a 10 count on value strategies? Is it now time to leave the ring, sell value, and pick up the growth gloves, or is a better option to stay in the ring and buy even cheaper cheap assets? To make this important determination, a reliable expected returns model is a good referee.

The choice of model is important. After all, a model's forecasted return for an asset class is only as good as its structure, assumptions, and inputs allow it to be. In this article, we compare three models. Each can be classified as simple in contrast to the quite complex models used by many institutional investors. One of the three is the model used by Research Affiliates, which although simple has performed well, not only in terms of making long-term asset class forecasts, but in combining undervalued asset classes to build alpha-generating portfolios. This latter consideration is a prime attribute of a successful model.

The Rational Return Expectation

Let's begin our analysis with the return we should rationally expect from the investments we make. Whether an investor practices top-down asset allocation or bottom-up security selection, investing is about nothing more than securing cash flows at a reasonable price. After all, the price of an asset is simply the sum of its discounted cash flows, which can be affected by two forces: 1) changes in the cash flows and/or 2) changes in the discount rate. If the cash flows and discount rate remain constant over the holding period, the asset's value will remain the same throughout its life as on the day it was purchased. Therefore, it is a change in the cash flows and/or the discount rate that ultimately drives an asset's realized return over time, and the possibility of such changes that drives an asset's expected return over time.

As mentioned in the introduction, the implementer of a value strategy would have experienced a long string of annual negative returns over the past several years. Figure 1 illustrates quite vividly the disappointing returns associated with a U.S. equity value strategy compared with a U.S. equity growth strategy since 2007.

Simple Is Not Simplistic

For a larger view, please click on the image above.

Although this period of underperformance may be disheartening for many value investors, the precepts of finding, and then investing in, undervalued assets will, tautologically,2 be rewarded with outperformance in the long run. The question then becomes, does "cheap" mean undervalued?

To aid in answering this question, a variety of expected return models are available in the marketplace, including the model on the Research Affiliates website.3 From the first day we published our long-term expected returns on the site, we have received questions from clients and peers on the efficacy of our model. The question usually posed is: "What's the R2 of your expected return model for [insert favorite asset class here]?"4 Granted, it seems like a pretty obvious question, but we would argue it is actually not all that relevant. A better question, and the one we address here, is how our model compares with other commonly used models. Because investors need some method or modeling system to estimate forward returns, the issue is not just a matter of how "good" a single model is, but also how it compares to available alternatives; simply improving on the alternatives can be quite beneficial.


Find your next ETF

Reset All