Any investor who has spent time looking into benchmarking active managers has probably come across the notion of survivorship bias and how it makes the group look better than they really are. A less discussed, but apparently bigger problem could be backfill bias.
The active investment management business is a very profitable one. With management expense ratios higher in some cases by 2 per cent over passive alternatives, it’s understandable to see why. It’s also understandable that active managers would want to protect the status quo. So, a lot of money goes into building the case for why you should invest with manager X, Y and Z versus an index fund.
We already know that the returns for active managers can be bolstered when we don’t correct for survivorship bias. If during a 10-year period, we started with 500 funds in our universe but end up with only 300, chances are the 200 that didn’t survive were underperformers. If they no longer count towards the 10-year average, it’s easy to see that the group’s performance average is now artificially inflated. It would be like running a 4 by 100-metre relay and dropping the time of the two slowest runners and just multiplying the times of the two fastest runners by two to get your 4 by 100-metre leg times. Not cool.
Another performance enhancing “substance” has been less publicly discussed but perhaps just as important, if not more, and that is known as backfill bias. It’s almost the complete opposite of survivorship bias. Whereas survivorship bias eliminates the losers, backfill bias add winners retroactively.
This happens because databases of active management returns can require funds to volunteer the return information to them. But if a new fund doesn’t submit performance data to start and rather waits to see if they have good returns first, they can then retroactively submit the past performance data upon being included in the database at their discretion. You’ll see this more in the hedge fund space versus the mutual fund space around the world as hedge funds are subject to much less regulatory oversight.
So just how big are these biases?
In a lecture by David Swensen the chief investment officer of Yale University’s endowment fund, he cites a few studies with shocking numbers. If you look at domestic U.S. equity fund returns for the year 2000 – looking at data available in the year 2000 – they were -3.1 per cent. But, if you look at the same fund category returns for the year 2000, with data available in the year 2005, the average return had magically increased to +1.2 per cent.
Why? You had five years to cut out losers and add some winners.
He notes that Burton Malkiel, author of the seminal finance book A Random Walk Down Wall Street, looked at hedge funds for an eight-year period and estimated the survivorship bias to be 4.4 per cent and the backfill bias to be 7.3 per cent per year. Also cited was Roger Ibbotson, co-author of the annually updated Stocks, Bonds, Bills and Inflation market return reference guide for investment professionals, who looked at a 10-year period and saw survivorship bias of 2.9 per cent and backfill bias of 4.6 per cent per year.
You can watch Mr. Swensen’s lecture for free as part of Yale University’s open learning initiative.
When looking at historical group returns, not only do you have to check to see if the data corrects for survivorship bias, but it needs to address backfill bias too. Just like we can’t drop our worst runners on a relay team, we also can’t recruit faster runners from another team after the race has already begun.