Note from our CIO: Expected Investment Returns Going Forward

In November 2011 our Investment Committee gathered for a two-day off-site retreat, specifically for “big-picture” discussion, and to review the specific investment vehicles we use in every asset class we recommend to clients. We also considered whether to include new asset classes. We surveyed a vast amount of academic literature in advance, and invited four guest speakers on specific topics. The 29 pages of written minutes generated many actions we’re taking now, actions we may pursue as we investigate the practicalities, and several other areas for further investigation. We devoted considerable attention to coming up with long term expectations for future investment rates of return.

This quarter, I’d like to focus the considerable attention we devoted to our long-term expectations for future investment rates of return. Our focus here is on decades into the future, not just the next five or ten years. This is a primary driver of our investment advice to you, and this drives the financial planning which affects your income and lifestyle.

The bottom line is that we lowered our expectations for the real (inflation-adjusted) rate of return for a globally diversified portfolio of stocks which includes substantial allocations to small and value stocks, from 6.7% to 6.0%, after assumed fees. For those investors who don’t own small or value stocks, we expect real returns of about 4%. Of course, any allocations to bonds reduce these estimates further.

For this exercise we’ll set aside some of the major potential changes that could occur in the world in the next several decades, such as depletion of oil or other natural resources, huge unfunded pensions and social programs, escalating health care costs, and ongoing transfer of intellectual property and technology dissipating developed nations’ wealth. All of these may affect investment returns, but we feel they are best handled by considering those individual scenarios, rather than trying to average them all into an overall discussion.

We will use a building block approach to estimating future global stock market returns for US investors. It will be the sum of the following elements:

1. Inflation: the rate at which the price of goods increases from year to year.
2. Real risk-free rate: the extent to which a “risk-free” asset (historically, US Treasury bills) provides return in excess of inflation.
3. Equity premium: the extent to which stocks outperform the “risk-free” asset.
4. Factor premia: the extent to which increased exposure to small or value stocks provides additional return.

The result will give us an “expected” return, before taxes, and before investment fees. Let’s examine each of these components in turn.

Inflation, as measured by the US Consumer Price Index, has averaged almost exactly 3% per year from 1926 through 2011. While it has had significantly higher levels for some sub-periods (e.g. 9.3% from 1973 to 1980), it was under 3% for 14 out of 19 years from 1983 to 2011. We feel 3% is a good estimate for long-term future inflation.

The “risk-free” rate is a notion created by academics for convenience. There is always some risk the US could default on its debt. Also, after taxes and inflation, there is a risk of reduced purchasing power for money invested in short-term Treasury bills. Perhaps it should be called the “least-risk” rate. Nonetheless, T-bills have beaten inflation by about 0.6% on average since 1926 (before taxes). From 1987 to today, T-bills have provided about 1.2% above inflation. We feel 1% is a good estimate for the long-term pre-tax “risk-free” rate.

The equity premium is a perennial subject of discussion among investment advisers, not to mention academics. Here some points from a few of the many papers we considered in discussing this point.

a) The US equity premium of 6% from 1889 to 1978 was inexplicably high to economists. (paper #57, 1998)

b) The US equity premium was 6.6% from 1880 to 2004. The potential for disasters helps explain high equity premia. Even a low probability of a recession “has a major effect on stock prices, risk-free rates, and the spread between risky and risk-free yields, even though the disasters that occur have only moderate effects on long-term averages of consumption growth rates and rates of return on equity.” That is, people hate losing money in crashes, so they demand a higher premium, and pay less for stocks. Evidence is given that people accept negative real rates of return on T-bills in wartime. That is, people hate losing money in risky assets during risky times, and are willing to pay too much and accept a small loss in the safest asset rather than suffer potentially catastrophic losses in other assets. (paper #115)

c) The US equity premium was 7.43% from 1951 to 2000, far higher than the first half of the century, and far higher than the 4.3% they would predict based on earnings growth. This has set the stage for mean reversion i.e. lower returns in the future. (paper #18)

d) The US equity risk premium was 7.5% from 1900-2001 (arithmetic mean, with a standard error of 1.9%). Since 1950, “stock markets rose for reasons that are unlikely to be repeated” i.e. increased productivity through technology and enhanced management, a lower required rate of return due to diminished business and investment risk, and lower transaction and management costs leading to large inflows of cash to stock investment. Adjusting for these, the authors indicate the past 7.5% should be adjusted to a future 5.3%. A similar analysis for other major markets leads them to posit a forward-looking 3.5% world equity premium. (“Global evidence” 2002)

e) From 1921 to 1996, the appreciation on US stocks “is rather exceptional.” “Major disruptions have afflicted nearly all the markets in our sample, with the exception of a few such as the United States.” (#56)

Sometimes we are tempted to believe that more recent history, such as the last 50 years, might provide a more relevant framework for making predictions for the next 50 years. However, we feel figures like 7.4% to 7.5% are unsustainably high, for the reasons discussed above. Abacus has used 5% as its estimate of the equity premium for many years. Despite the 100-year history of the US equity premium being 6% to 6.6%, we were sympathetic to the idea that it should be 5.3% going forward, as suggested by the “Global Evidence” paper. We decided to reduce this figure somewhat, to reflect several potential “headwinds” to stock investors. These could include: increased volatility (reducing future compound returns), stock prices declining to allow for higher future returns (as people increase their subjective assessment of the risk of future recessions or market disruptions), increased international tension (decreasing global trade, and increasing costs of international diversification), or a long period of global delever aging or austerity (inhibiting global economic growth and asset returns).

Therefore, Abacus is decreasing its estimate of the arithmetic average of the long-term global equity premium from 5% per year to 4.5%.

Factor premia quantify how much additional return one should expect from investing in smaller-than-average stocks, or deeper-value-than-average stocks. Here again the academics have a lot to say, and we can learn much from listening in.

In reading these figures, note that academics usually compare the very extreme ends of the market when discussing factor premia. So they compare the very largest firms against the very smallest, or the most pumped-up growth companies against the most beaten-down value stocks. Abacus might only capture about one quarter of these effects, as we are trying to measure the additional return we’d capture by moving our portfolio only somewhat in the direction of the extremes, starting from the market average portfolio, not leaping from one end to the other as these studies generally do.

a) From 1929-1997, in US markets, value beat growth by 6.2% per year, and small stocks beat large by 2.4% per year. (#93)

b) From 1975-1995, in non-US markets, value beat growth by 7.6% per year, in 12 of 13 major markets. (#17)

c) From 1989-2005, in emerging markets, the small and value premia were larger than in the US, and non-correlated to the US returns, so very attractive to US investors. (#94)

d) “International small caps [are the] asset class that is most likely to offset US stock performance, regardless of the direction of US returns.” This makes it an unusually good diversifier. (#95)

e) In this seminal exposition of the three-factor model, the authors measure premia in the US from 1963 to 1990 and find a value premium of 1% and small premium of 0.5%…per month. That’s over 12% and 6% per year, respectively. (Cross-Section of Expected Stock Returns)

f) Since 1983 the size premium seems weaker, and statistically insignificant. This is probably caused by the significant, ongoing underperformance of small growth stocks. Discussion of this paper with other experts led us to the conclusion that small value stocks can be expected to outperform large value stocks, but small growth stocks are the worst performing quadrant, and are not expected to outperform large growth stocks. One explanation for this is that individual investors may overpay for small growth stocks, hoping to find “the next Apple Computer” (and possibly with the active urging of the brokers and IPO managers). Lumping all small stocks together masks the significant premium achieved by small value stocks, before and after 1983. (Post-Banz Size Premiums)

We took the longest perspective, noted the 6.2% per year value premium from 1929 to 1997, and estimated that we would capture about one quarter of that (since our portfolio starts with a market-weight “average” portfolio and tilts somewhat toward value). Including the additional return from the size effect, and the potential for factor premia to be higher in emerging markets (and non-correlated), we feel our figure of 2.0% for factor premia is well supported by the evidence.

Here is how it all stacks up:

At the risk of beating a dead horse, we had another whole discussion about what standard deviation to assume in our financial planning projections. We’ll spare you all the details, but you may be interested to know that we agreed upon a standard deviation of 15.43%, which reflects the results of our current allocation from 1987-2010. Emerging markets data become available as of 1987, and we felt volatility starting from the era of the “triple witching hour” and the junk bond meltdown is more relevant than prior years.

We expect our actual bond returns to exceed the 1% above inflation we’ve assumed for the “risk-free” rate, as we actually invest in bonds with higher yield and longer durations than T-bills.

There are many hazards in forecasting returns, so please see our disclosures below.