Now is when investors need to be especially vigilant to avoid recency bias, which occurs when we give greater weight to events that have happened recently than those that occurred further in the past. Investors are guilty of this, for example, when they get rid of a financial adviser with a great long-term record because of recent lagging performance. Such behavior is especially prevalent in January.
While I have written many times about the dangers of recency bias, I was prompted to write about it now by an analysis from one of the newsletters I monitor. The newsletter’s headline read “Don’t overvalue ancient history.” The history in question occurred only 20 years ago, which is hardly “ancient.” In fact, from a statistical point of view, what happened two decades ago is no less relevant when picking an adviser or strategy than the more recent past.
To illustrate this, I analyzed the five dozen investment newsletter model portfolios that my auditing firm tracks. Specifically, I imagined that, five years ago, we were choosing an adviser to follow for the subsequent five years. In this thought experiment, the only pieces of information we had to inform our choice were those advisers’ returns over various past five-year periods extending back to the mid-1990s.
Which of those five-year periods would have been the most helpful? To understand what I found, it’s helpful to focus on a statistic known as the r-squared. This statistic measures the extent to which one data series (in this case newsletters’ margins of victory over the market, or alphas, during a certain five-year period in the past) explains or predicts another data series (in this case those same newsletters’ alphas over the most recent five-year period).
The r-squareds are reported in the table below.
Newsletters’ alphas over this five-year period | r-squared of the correlation with newsletters’ alphas over the five-year period from 1/1/2015 to 12/31/2019 |
1/1/1995 to 12/31/1999 | 5.6% |
1/1/2000 to 12/31/2004 | 14.1% |
1/1/2005 to 12/31/2009 | 1.8% |
1/1/2010 to 12/31/2014 | 1.2% |
Notice that the highest r-squared emerged when using the five-year period from 2000 through 2004 — almost 20 years ago. In contrast, the most recent five-year period did the worst job of helping us to pick a newsletter.
These results do not mean you should now de-emphasize recent performance when picking an adviser. It’s just the luck of the draw that, in picking an adviser for the 2015-2019 period, performance between 10- and 15 years previously would have been most helpful. On some other occasions the most helpful performance period will be even further in the past, while in still others it will be more recent.
Bear this in mind as you read the 2020 performance scoreboards. Returns in 2020 are no more relevant to your choice of adviser than returns, say, in 2010 or 2000.
What the low r-squareds mean
Notice also that the r-squared values in the table are strikingly low. This is the statistical basis for the oft-repeated mantra that past performance is no guarantee of future performance. Not only is it no guarantee, it provides depressingly modest guidance.
That’s why you can’t afford to ignore past performance, no matter how far back in the past it was produced. You need every bit of statistical help when choosing an adviser. Call it ancient history if you want, but your odds of success go up when you pay attention to it.
Mark Hulbert is a regular contributor to MarketWatch. His Hulbert Ratings tracks investment newsletters that pay a flat fee to be audited. He can be reached at mark@hulbertratings.com.
More: Bulls again are putting stock in whatever valuation measure validates their market view
Plus: How to win the ‘loser’s game’ of investing and still keep the excitement alive
