Q and A: Estimating Long-Term Market Returns

Each year, Charles Schwab Investment Advisory, Inc. (CSIA) calculates long-term return estimates for stock, bond and cash investments. Here, we'll answer common client questions concerning this research, including an explanation of the methodology behind our estimates.

• Why are long-term return estimates important?
• How do you define "long term"? 
• How do short- and long-term forecasts differ? Is one better than the other?
• What are your long-term return estimates for stocks, bonds and cash investments?
• How do you calculate your estimates?

Why are long-term return estimates important? 

Timing market returns from year to year is difficult, but over a long period of time, research shows that risky asset classes such as U.S. equities can offer a positive risk premium for a patient investor. Having a sound financial plan serves as a road map to help investors reach long-term financial goals, but to get there, you need reasonable estimates of what long-term stock- and bond-market returns might be.

For example, if your return estimates are too optimistic, you run the risk of not being able to retire on time or pay for a child's education. If they're too pessimistic, you may needlessly sacrifice some of your current lifestyle by over-saving for retirement.

Similar to the axiom "garbage in, garbage out," you can't use unrealistic assumptions to determine realistic outcomes, and this is especially true when developing your long-term financial plan.  

How do you define "long term"? 

When it comes to return forecasts, there's no specific definition of "long term," though a widely accepted rule of thumb is a time period of more than 10 years. A balance is struck when you consider both shorter-term market fluctuations (think 2008) and extremely long periods of time when your confidence in making predictions greatly diminishes. Accordingly, CSIA used a 20-year time horizon for the estimates provided here, though calculations using a time horizon between 10 and 30 years should produce similar results.

How do short- and long-term forecasts differ? Is one better than the other? 

For some investors, the strategic asset allocation can serve as a starting point to make shorter-term tactical changes to their asset allocation. For example, an investor may target a long-term, strategic allocation of 50% stocks and 50% bonds. Depending on the market environment, the investor may want to temporarily favor stocks over bonds, or vice versa.

Continuing with the example, suppose the investor thinks that the stock market is currently undervalued. The investor may choose to act on this belief by temporarily adjusting her current allocation, possibly to 60% stocks and 40% bonds.

The process of making these shorter-term changes is called tactical asset allocation. These temporary shifts generally occur when estimates of short-term returns deviate from long-term estimates. Short-term return estimates are typically based on current economic and market conditions, whereas current conditions are not as relevant for estimating long-term returns.

When it comes to meeting your long-term goals, however, choosing an appropriate long-term, strategic asset allocation is more important than making short-term, tactical bets.

Some people argue that investors should focus exclusively on short-term returns and tactical asset allocation because it's difficult to accurately estimate long-term returns. The problem is that it's even more difficult to accurately estimate short-term returns!

And because most investors have at least one long-term goal—retirement—they need reasonable long-term return estimates to help determine how much money they'll need to fund their retirement lifestyle, and in turn, how much they'll need to save to get there.

For this reason, the focus of this study is on long-term returns.

What are your long-term return estimates for stocks, bonds and cash investments?  

These estimates are significantly below the historical annual compound returns on large-cap stocks and bonds of 10.5% and 7.9%, respectively, during the 1970-2014 time period. Of course, these are estimates of average returns—in any given year, stocks and bonds may return far more or far less and may even be negative.

Why are the estimates below historical averages? There are two reasons:

• Our estimate of long-run inflation is 1.8%, just shy of two percentage points below the actual inflation rate during the 1970-2014 time period of 4.2%.
• Current and expected interest rates are much lower than what has transpired historically, especially compared to the high-interest-rate environment of the 1980s.

What you can do now

So, what can you do in a single-digit-return environment? Thanks to the power of compound returns, what you do (or don't do) today can have big implications for your ability to meet your long-term goals.

When faced with expected returns that are lower than you may have anticipated, try to resist the temptation to simply wait in the hope that the market will provide higher returns in the future that will allow you to "catch up" on your financial plan. If it does, that will be a great bonus. But it's far better to plan for a more realistic scenario.

Here are a couple things you can do. First, try to avoid unnecessary fees and taxes, particularly in a lower-return environment. Second, if you don't have a long-term financial plan, it's a good time to put one together.  

How do you calculate your estimates? 

Our return estimates contain two parts: a current risk-free rate component that's the same for all asset classes and an asset-class premium that varies by each asset class because of differences in expected risk.

Estimating current risk-free rates

The current risk-free rate is estimated by directly observing Treasury yields in the marketplace. Because we're estimating returns for a 20-year time horizon, the risk-free rate is measured as the yield of a 20-year U.S. Treasury bond, which was 2.6% as of December 31, 2014. Keep in mind that no investment is entirely free of risk, but because U.S. Treasuries are generally considered to be the least risky asset class (aside from cash), Treasury rates are typically used as a "risk-free" benchmark.

Estimating asset-class premiums

The asset-class premium measures the incremental return (generally higher for stock asset classes and lower for fixed-income asset classes) demanded by investors for investing in that asset class as opposed to a risk-free bond.

Stocks: The asset-class premium for large-cap stocks is called the equity risk premium (ERP), which measures the relative attractiveness of large-capitalization stocks versus a risk-free bond. It also serves as the foundation for estimating asset-class premiums for mid/small-cap stocks and international stocks. There are two primary ways of estimating the ERP:

• The historical long-term approach takes the historical difference in returns between stocks and risk-free bonds and assumes that the future will look like the past.
• The valuation approach relies on fundamental data, such as dividends, earnings, gross domestic product (GDP) growth and valuation levels and then uses well-established financial theory to estimate an ERP.

Valuation approach vs. historical long-term approach

The primary criticism of the valuation approach is that it’s very difficult to forecast variables such as dividends, earnings or GDP growth over the short-run, let alone over long horizons. As such, we forgo this approach in forming our long-term estimates.

The historical-return approach is based on the realization that it's difficult, if not impossible, to forecast long-run stock-market returns using current market or economic conditions. Since current market information is generally not a useful predictor of long-run ERP, the basis of the historical-return approach is that the best estimate of the future ERP is the historical average ERP calculated over a long history.

The primary criticism of the historical-return approach is that realized returns over a particular time period can differ, sometimes dramatically, from what's expected. As such, blindly extrapolating these returns into the future can result in unreasonable estimates.

The approach adopted in this study addresses this criticism.¹ To better understand it, we first break down the sources of average returns for large-cap stocks. In doing so, we look "under the hood" to help determine which components of average returns may be expected to repeat in the future and, more importantly, which ones may not.

In researching the sources of historical returns for large capitalization stocks as represented by the S&P 500 total return index, we don't expect the growth in the P/E ratio from 1926 to 2014—amounting to a roughly 0.9% per year average return—to repeat in the future, as this return did not come from earnings growth.   Instead, it represents what the market was willing to pay for every dollar in earnings during the 1926-2014 time period.

There are a number of possible reasons why the P/E ratio expanded during this time, including higher expectations for future earnings and less return demanded by investors for holding stocks. Regardless, it's not realistic to think that such an expansion will occur again.

As a result, we do not include the 0.9% attributed to P/E growth when estimating future returns, which results in an adjusted historical return on large-cap stocks equal to the following components:

Inflation + growth in inflation-adjusted EPS + dividends
Or A + E + C = historical average return

3% + 2.0% + 4.0% ≈ 9.0%²

The adjusted historical return of 9.0% is not our estimate of future returns because it reflects historical interest rates and inflation. It's used to estimate the ERP. Specifically, we take the adjusted historical return on large-cap stocks and subtract from it the historical income return provided by the risk-free asset (proxied by the Ibbotson Long-term Government Bond Index)³:

ERP ≈ 9.0% - 5.2% ≈ 3.8% (compounded annually)

Mid-/small-cap stocks: When estimating the asset-class premium for mid-/small-cap stocks, we use the ERP of 3.8% as the starting point, and then make adjustments based on the unique risk level for the mid-/small-cap asset class relative to large-cap stocks.

To do this, we estimate the premium of mid/small–cap stocks relative to large-cap stocks from 1926 to 2014, which equals 0.7%. To arrive at our final estimate, we simply add the premium of 0.8% to our large-cap estimate of 3.8%.

International stocks: Data limitations prevent us from analyzing the sources of historical returns for international stocks. As such, we use the domestic stock market asset-class premium as an anchor in developing the international equity premium.

This approach estimates the international risk premium as measured by the return demanded by investors holding an international stock portfolio.  The domestic ERP of 3.8% is multiplied by the historical sensitivity of international market returns (excluding U.S. stocks) to U.S. large-cap stocks.  The sensitivity or beta of 0.94 is then multiplied by the domestic ERP of 3.8%, resulting in an international asset premium of 3.6%.  

This approach assumes that domestic and international stock markets are integrated, meaning there are no barriers to financial flows and that assets with the same levels of risk command the same return no matter the country. In addition, the approach relies heavily on sensitivities between domestic and international returns that prove to be relatively stable since 1990.

Bonds and cash investments: In 2014, the methodology for estimating the bond asset class premium was changed. With our previous approach, the 20 year Treasury was the baseline estimate to which was added a default premium to reflect the additional amount of compensation an investor requires for holding credit risk.  It was measured as historical difference in monthly total returns between the Barclays U.S. Aggregate Bond Index and a government bond maturity-matched to the Barclays U.S. Aggregate Bond Index.

Using historical data to form a forward-looking estimate for bonds is currently problematic, however. With U.S. Treasuries at historically low yields, potentially a secular bottom, significant over-estimation of returns may occur if interest rates rise. Since 1982, interest rates have been falling, allowing bond investors to achieve additional returns in the form of capital gains. Counting on a similar, long, historical trend of falling interest rates may prove to be disappointing for investors.

A more reasonable view of forward returns an investor will receive for holding bonds, even if rates do rise at some time in the future, is captured with yield-to-maturity. Yield-to-maturity is flexible enough to capture the ups and downs of interest rate movements, while avoiding the problem of depending on historical data. 

Research shows that yield-to-maturity also provides a better forward-looking indication of total compensation an investor will receive. In the chart below, the yield-to-maturity of the Barclays U.S. Aggregate is plotted from 1976 to 2014 against the next 10 year total returns of the Barclays U.S. Aggregate Index. The ten year forward returns are calculated by annualizing returns of the Barclays U.S. Aggregate Index over the next 10 years. As the chart shows, when the yield-to-maturity today is plotted next to the forward 10 year returns, they tightly track each other. 

Barclays U.S. Aggregate Bond Index

The foundation of our new method for estimating returns is the yield-to-maturity on the Barclays U.S. Aggregate Bond Index, which was 2.3% as of December 31, 2014.

We do not stop here, however. Our final bond return estimate contains an additional adjustment to reflect the fact that, as of December 31, 2014, the average maturity of the Barclays U.S. Aggregate Index was 7.7 years versus our long-term investment horizon of 20 years. 

Investors should receive additional compensation (a horizon premium) for the purchase of longer maturity bonds, so using the yield-to-maturity of the bond index alone under-estimates potential returns. For instance, rolling over three month U.S. Treasury Bills for 20 years will produce less return than buying a 20-year U.S. Treasury bond. 

To calculate the horizon premium, we simply subtract the yield-to-maturity of a zero coupon security, matching the modified adjusted duration of the U.S. Barclays Aggregate Bond Index, from the yield-to-maturity of a 20-year zero coupon U.S. Treasury. Modified adjusted duration measures the price sensitivity of bonds to changes in market interest rates, so it provides an excellent way to approximate the additional risk investor assume when purchasing longer maturity bonds. As of December 31, 2014, the modified adjusted duration of the Barclays U.S. Aggregate Bond  Index was 5.0.⁴ The yield-to-maturity of the 20 year STRIP was 2.7% and the yield-to-maturity of the 5 year STRIP was 1.7%. To calculate the horizon premium, we take the yield-to-maturity of the 20 year STRIP and subtract the yield-to-maturity of the 5 year STRIP which equals 1.0%. Adding the 15 year horizon premium of 1.0% to the bond index yield-to-maturity of 2.3% produces a final return estimate of 3.3%.   

How we estimate long-term inflation

The 20-year inflation estimate is derived by comparing the yield of 20-year Treasury Inflation Protected Securities (TIPS) to the yield of U.S. Treasury bonds of the same maturity. The yield on a conventional Treasury bond must compensate the investor for the expected decrease in purchasing power associated with inflation. Buyers of inflation-protected securities require no such compensation because interest and principal payments are indexed to inflation. Treasury bonds and TIPS of the same maturity should offer the same inflation-adjusted return because the U.S. Treasury backs both of them.

If this were not the case, savvy bond-market investors would buy the security with the higher inflation-adjusted yield, causing its price to adjust, and resulting in both securities offering the same inflation-adjusted yield. Therefore, the yield difference between conventional Treasuries and TIPS of the same maturity represents an estimate of the inflation rate expected by market participants. Using the spread as of December 31st. 2014, this approach resulted in a long-term inflation estimate of roughly 1.8% per year for the next 20 years.⁶

Asset class benchmarks

The table below lists the benchmarks assigned to each asset class. In cases where the benchmark has a short history, it's extended by using a statistically similar longer-lived proxy.