The greatest trick the stock market ever pulled was convincing investors that historical returns are predictive. They aren’t.

In fact, historical returns not only give you very little information about future returns, but they can also increase the odds you’ll make a bad decision.

We often see this bias in investors. Both reporters and prospective customers often ask us, “What are your returns?”

I cringe when I hear this. Out of all the questions you should be asking, this one should be low on the list. There are far more informative and useful questions to ask, once you know what’s in our portfolio.

To be fair, there are aspects of the answer that can be helpful. Returns can give you an idea of the size of upswings and drawdowns, and how the portfolio relates to other asset classes. But in a passive, index-tracking portfolio, such as Betterment’s, you shouldn’t expect to see market alpha in our performance. When properly benchmarked, we are the benchmark.

The other common mistake people make is comparing our portfolio to another over a short period of time. If, after six months, our portfolio has a lower return, they’ll often ask, “Why should I use you if your returns are worse?”

Far too often, investors put too much weight on small sample, recent historical performance, choosing the investment with the highest investment return. How deceptive can this be? Our interactive tool below shows that this method leads to astonishingly high odds that they’ll underperform both in absolute and risk-adjusted terms in the future.

How the Data Deceives

You might not realize it, but when you look at historical returns, you’re doing a statistical analysis. Any set of historical returns comprises a sample of behavior over a certain period. Any inferences you make about what they tell you of the future should be balanced by placing them into context of how variable they are. And when you do that, two clear issues arise.

Fooled by Randomness

The first is being “fooled by randomness,” a phrase coined by Nicholas Nassim Taleb, a risk analyst and statistician. When you choose the highest returning of two correlated investments using a small sample of historical data, the odds are incredibly high that you picked the wrong fund. The randomness of small samples overwhelms the truth.

Let’s work through some examples. We’ll use hypothetical portfolios with return probabilities we know for certain, because we’ve created them through simulation, and see how well the short-term data mimics the long-term truth. These are not Betterment portfolios.

Portfolio A will have a mean annual return of 6% and a volatility of 14%. Portfolio B has a mean return of 6.5% and annual volatility of 13%. The portfolios will also have a 0.90 correlation to each other—most stock funds have higher correlations. By both measures of absolute return and risk-adjusted return, Portfolio B is better. Yet over the first randomly simulated six-month period, Portfolio A came out ahead.

One 6-Month Simulation

a-b-one-sim-6-months

How often does the worse portfolio come out ahead over a short time period? In this case, we’ll call them C and D, with the same parameters. Let’s look at running 1,000 of such simulations over a six-month period. How often does Portfolio D, who should be the winner, come out ahead?

Many Simulations Over 6 Months

c-d-six-month-dist

The answer is so close to 50% as to be indistinguishable from it.

In fact, we can increase the differences in expected returns and this remains true. Let’s give Portfolio D a mean return of 8% and Portfolio C a mean return of 6%. Both have 14% volatility. The significantly higher return Portfolio D will still lose over 40% of the time over a six-month period.

Many Simulations Over 6 Months

c-d-six-months-bigger-return-gap

While the odds are just better than 50/50 in the short term, they have big consequences in the long term. Here are the distributions of 20-year outcomes for those same portfolios:

Many Simulations Over 20 Years

c-d-20-years

The randomness in half-year returns results in choosing the wrong portfolio about half the time, even with large difference in return. You might as well save yourself the time and expense and flip a coin.

Over long periods of time (20 years), and with large differences in average returns, the odds of picking the correct choice do increase. But you may be surprised how long it can take. For portfolios with a 1% return difference, by 20 years you still have about a one-in-four chance of picking the portfolio that will have worse underlying returns over even longer periods of time.

Chance of Choosing Worse Portfolio Based on Performance

Return Difference 3 months 6 months 1 Year 5 Years 10 Years 20 Years
0.50% 49% 48% 48% 42% 40% 37%
1.0% 47% 46% 44% 36% 32% 26%
2.0% 44% 43% 37% 26% 16% 9%

Each cell based on 3,000 simulated cumulative returns of better portfolio (8% return) versus a benchmark portfolio with a mean return of 6% and 14% volatility. Correlation of 0.90 between portfolios.

To be clear, there are statistical tools you can use to improve your odds of picking the right portfolio, but most investors aren’t professional statisticians. They just go by the cumulative returns over a short period of time.

Performance Chasing Is Worse Than Random

If the low odds of correctly choosing a better portfolio above didn’t convince you, it’s even worse than that. Empirically, choosing the best funds, a strategy called performance chasing, is likely to reduce your returns.

The graph below comes from an excellent research paper from Vanguard. It shows the returns achieved by investing in the best fund in each asset class, compared to a buy-and-hold strategy. Performance chasing—picking investment based on recent performance—produced worse returns of about -2% to -3.5%.

Buy-and-Hold Superior to Performance Chasing, 2004-2013

If every year, you picked the investment manager with above average returns over the past 12 months, you’d end up underperforming an investor who stuck with the passive index-tracking manager.

The Right Things to Consider

If recent investment performance is such a poor way to choose an investment manager, how should you select one? Use a set of clear principles that are likely to be true in the future:

    • Monetary Cost: A certain drag on returns, if the service doesn’t deliver value above cost. Consider commissions, trade fees, and assets under management (AUM) fees.
    • Non-Money Costs: How much time and and effort does it take for you to use it well? Does it have a high time or stress cost for you to get the most out of it?
    • Services Offered: Do the services offered make you better off? Does it do things for you which you wouldn’t do yourself? Does it help you make better decisions? Does it make some of those decisions for you, automatically?
    • Experience: Is it easy to use? Do you enjoy using it?
    • Philosophy Fit: Consider its investment philosophy, and if it is parallel to yours. Some funds seek to deviate from the index and cost more, some seek to track it passively.
    • Tax Management: Returns will likely not take into account actual value-adds, such as tax loss harvesting. You won’t have received a comparison tax bill that allows you to compare after-tax returns across services; it will be up to you to compare them.
    • Behavior Management: Does the service have a proven track record of reducing the behavior gap?

When choosing an investment manager, the key isn’t to focus on investment performance; it’s to focus on service, fit, and investor returns.

More from Betterment: