Stocks for the Long Run? Setting the Record Straight

Editor’s Note: This is the final article in a three-part series that challenges the conventional wisdom that stocks always outperform bonds over the long term and that a negative correlation between bonds and stocks leads to effective diversification. In it, Edward McQuarrie draws from his research analyzing US stock and bond records dating back to 1792.

CFA Institute Research and Policy Center recently hosted a panel discussion comprising McQuarrie, Rob Arnott, Elroy Dimson, Roger Ibbotson, and Jeremy Siegel. Laurence B. Siegel moderated and Marg Franklin, CFA, president and CEO of CFA Institute introduced the debate.

Edward McQuarrie:

In my first two blog posts, I reviewed the new historical findings presented in my Financial Analysts Journal paper. Relative to when Jeremy Siegel first formulated the Stocks for the Long Run thesis 30 years ago, better and more complete information on 19^th century US stock and bond returns has emerged. Likewise, courtesy of the work of Dimson and others, a far richer and more complete understanding of international returns is now in hand.

I summarized the new historical findings in my paper’s title: “Stocks for the Long Run? Sometimes Yes, Sometimes No.”

In this concluding post, I will highlight the implications of these new findings for investors today. I will address several misconceptions that I’ve encountered interacting with readers of the paper.

Misconception #1: McQuarrie doubts whether stocks are a good investment over the long term.

Nope. Rather, I want you to adjust your expectations for the long-term wealth accumulation that you can expect from holding stocks, especially a 100% stock portfolio, over your idiosyncratic personal time horizon.

Here’s why I think some adjustment of expectations is necessary.

Let me first acknowledge that no author is responsible for what readers do with their work once published and diffused, so what follows is not a criticism of Siegel or his research.

That said, some readers of Siegel’s Stocks for the Long Run conclude: “If I can hold for decades, stocks are a sure thing, a no-lose proposition. It could be a wild ride over the short-term, but not over the long-term, where buying and holding a broad stock index essentially guarantees a strong return.”

Siegel never said any such thing. But I can assure you, more than a few investors drew the conclusion that for holding periods of 20 years or more, stocks are like certificates of deposits with above-market interest rates.

The inference that my paper attempts to refute is that stocks somehow cease to be a risky investment once they are held for decades. I presented numerous cases where investors in other nations had lost money in stocks over holding periods of 20 years or more. And to make the demonstration more compelling, I first excluded war-torn nations and periods.

My point is: Stocks are NOT guaranteed to make you money over the long term.

In fact, stocks have often rewarded investors over the long term, despite large fluctuations in the short term. Patient investors have reaped huge rewards, especially US investors fortunate enough to be active during the “American Century.”

• Over the 20 years from the end of 1948 to the end of 1968, an investment in US stocks would have turned $10,000 into almost $170,000.
• Over the 18 years from the end of 1981, that investment would have turned $10,000 into almost $175,000
• And over the 36 years from 1922 to 1958, that investment would have turned $10,000 into almost $340,000, despite the, ahem, hiccup that occurred after 1929.

Huge rewards can be reaped from stocks. But there is no guarantee of any reward.

You make a wager when you invest in stocks. It remains a wager when you invest in a broadly diversified index such as the S&P 500. And it is still a bet even when you hold it for 20 years.

Odds are good that your bet will pay, especially if you are investing in a globally dominant nation, such as the US in the 20^th century, or the UK in the 19^th century.

But the odds never approach 100%.

Misconception #2: McQuarrie wants me to own more bonds.

It would be more correct to say that I wish to rehabilitate bonds from the disrepute in which they fell after their terrible, horrible, no good, very bad performance in the decades from 1946 to 1981. Those years dominated the record in the Stocks, Bonds, Bills & Inflation yearbook compiled by Roger Ibbotson and colleagues when Siegel first formulated his thesis.

The new historical record reveals that the divergent performance of stocks and bonds from 1946 to1981 was unique. Nothing like it had ever occurred in the century-and-a-half before. The most recent four decades look quite different, with stock and bond performance again approximating parity.

Here is where it becomes important to tread very carefully in constructing a forward-looking interpretation of the historical record with respect to the equity premium, i.e., the advantage of owning stocks instead of bonds.

If you calculate the mean or average stock performance relative to bond performance over the entire two-century US record, you get an equity premium of about 300 to 400 bp annualized. That’s huge. Compound that for 20 or 30 years and you’ll find yourself chanting “Stocks for the Long Run.”

Outcomes by Century

Note. Reproduced from “Stocks for the Long Run? Sometimes Yes, Sometimes No.” Arithmetic mean of real total returns. Wealth is the value of $1.00 invested for 100 years (compounded returns can be extracted by taking the 100^th root). Equity premium is the mean of the annual subtractions. Standard deviations are in parentheses. Means with superscript ^a are different across periods and those with superscript ^b are different within period (t-tests with heterogenous variance, all p-values < .01).

If you separate out the 19^th century from the 20^th century, as I did in the table, you find: The equity premium for the 19^th century was just under zero, while the equity premium for the 20^th century was just over 600 bp.

Average those two together, along with the omitted years from the 18^th and 21^st centuries to get a complete record, and you get the expected result: an historical equity premium of 300+ bp, in the new historical record, which is consistent with the old record.

But can you be confident that stocks will outperform bonds by 300 bp per year over your decade or two or three, over your personal horizon?

Of course not. The equity premium has exhibited too much variance even over very long intervals.

Let’s return to Misconception #2 and drill down. In the old historical record, first compiled by Ibbotson back to 1926 and then extended by Siegel back to 1802, a long-term investor had no good reason to own any bonds. At all.

In the old record, stocks always outperformed bonds, and the outperformance became increasingly dependable and grew larger in magnitude as the holding period stretched out to 20 years, 30 years, and longer.

The only justification for holding any bonds was if the investor lacked the stomach for the short-term volatility of stocks. Bonds were for the pusillanimous investor who didn’t have the spine to harvest the magnificent long-term returns on stocks.

Spineless investors had to settle for the much lower returns offered by a bond allocation because of their urgent need to dampen the intolerable short-term volatility of stocks.

Any financial adviser will confirm that many clients can’t abide the short-term volatility of a 100% stock portfolio. One of several contributions of Siegel’s work was to stiffen the spines of investors who were prey to such fears but who could be persuaded by evidence.

Such risk-averse investors could only maximize utility, net of return and risk, by including bonds in their portfolios, sacrificing return to reduce risk to a tolerable level.

Using the Ibbotson-Siegel historical data, the investor with a cast iron stomach would be inclined to invest 100% of their long-term funds in stocks. Given their high tolerance for risk, it would be irrational to do otherwise.

On the new historical record, in which stocks do not always beat bonds, the choice is less clear. A balanced portfolio, such as the 60/40 portfolio popularized by Peter Bernstein, might not produce any less return than a 100% stock portfolio. It might even produce somewhat more wealth if stocks go through a bad stretch.

Conversely, a 60/40 portfolio will almost certainly be less volatile than a 100% stock portfolio for reasons explained by the late Harry Markowitz: the expected lack of correlation between stocks and bonds and bonds’ historically lower volatility.

In the absence of certainty that stocks will outperform bonds, combined with the near certainty that a balanced portfolio of stocks and bonds will be less volatile than a 100% stock portfolio and subject to more shallow drawdowns, a balanced portfolio becomes a viable option for any investor.

That’s the gist of the new historical record.

Misconception #3: McQuarrie steers US investors away from owning international stocks.

This one surprised me when I first heard it. I never dreamed that the tables in my preceding post would be interpreted that way.

In my paper, I tabulated bad periods for stocks — periods showing equity deficits where stocks underperformed bonds — across 19 nations outside of the US. I showed multiple instances of losses on stocks over 20 years, 30 years, and more rarely, 50 years.

But that doesn’t mean that international stocks are a bad bet for US investors going forward. I would expect that sometimes international stocks will outperform US stocks and sometimes US stocks will outperform. It varies by regime and can’t be predicted any more than the future performance of US stocks can be known in advance.

How then to interpret the woeful episodes of underperformance by international stocks tabulated in the paper?

First, each of those international results was cherry-picked. I had a 300-year record of UK stock performance available, courtesy of Bryan Taylor at Global Financial Data. That means I had 281 twenty-year rolls to choose from: 1700 to 1719, 1701 to 1720, etc.

I picked the very worst one for the UK entry in the 20-year column in my table. For the other 18 countries, I typically had between 150 and 200 years from which to cherry-pick the very worst episode.

The purpose of the exercise was to expand the sample size of stock market histories beyond a one-market, one-century record: the period from 1926 in the United States, which has dominated most investors’ historical understanding ever since Ibbotson first assembled the Stocks, Bonds, Bills & Inflation record in 1976.

In that one-market, one-century record, stocks always do well if you hold on long enough, and stocks always beat bonds over those long periods.

But that result was obtained in one market over one century. The 19^th century US data I compiled gave me a second century, but still only for that one market.

Paul Samuelson among others famously observed that “history has a sample size of one.” That’s true if you confine attention to one national history and one century. When the only historical record available covers but one nation, and only during the period when it rose to world dominance with the largest economy — the United States post-1926 — generalization is fraught indeed.

Would stock investors fare just as well in a nation less favorably situated, over a less sunny period? There was no way to know, decades ago, when Siegel first assembled the Stocks for the Long Run thesis. The international record was very sparse back then.

In my thinking, the newly emerged international record, launched initially by William Goetzmann and Philippe Jorion in 1999, takes the historical record from a sample size of one to a sample of about 40 (20 nations across two centuries). Or, if you will, from 100 market years to 4,000 market years.

As a rule, expanding the sample size helps to refine the estimate of the range of potential outcomes. If you walk down Fifth Avenue in Manhattan with a surveyor’s laser sight and measure the height of the first 100 adults you pass, you will likely infer that most US adults are between five and six feet tall. You might find a few individuals shorter than five feet, and you will probably find a few taller than six feet.

If it is a reasonably co-ed sample, you might formulate the hypothesis that US males are taller than females on average, but you wouldn’t have much confidence in that generalization if the sample consisted of only 23 women and 77 men.

To continue the metaphor, suppose you added to the sample by walking down the main street of Stockholm. Your estimate of the maximum adult height to be found in a sample of 100 people would probably increase.

Switching up the metaphor, suppose the first sample of 100 was taken outside the largest high school in Los Angeles just after Winter sports practice let out, and that you confined the sample to female students. As those basketball and volleyball players streamed past, how good an estimate would you get of the average female height globally?

That’s how I think of both the 19^th century US data I collected and the international data I drew from others: as expanding the sample size of stock and bond returns beyond what could be glimpsed from Ibbotson’s Stocks, Bonds, Bills & Inflation yearbook.

The expansion in size is greatest for longer holding periods. There are, after all, only 10 separate decade samples in a century, and only five independent two-decade samples. Once you have two centuries and 20 markets, there are 400 separate market-decades, and 200 distinct 20-year cases.

It should come as no surprise that the international sample included measurably worse stock market outcomes than anything seen in the post-1926 United States.

That’s an expected outcome from expanding the sample size. It says nothing about the future results that might be obtained from an investment in international stocks.