Posted on December 22, 2011 by PerTrac Research Team

It is perhaps time to admit that eliminating the instant-history bias is arguably as much an art as it is a science

Instant-history bias occurs when funds decide to incubate (i.e., develop a track record) for some time before openly reporting their returns to industry databases. When funds do decide to report, they report their complete history since inception, thus resulting in an ‘instant-history.’ The bias comes into existence because only funds that have a solid track record will decide to report results to databases, whereas funds that have undergone a poor performance run are likely to keep their returns silent and not report to industry databases until they have put together some offsetting good performance periods. Hence, the ability to choose when an incubation period ends and when to disclose historical returns leads to an upward performance bias for younger funds.
 
We do, however, have methods to choose from that will help mitigate the effect of instant-history bias. The straight forward procedure would be to simply remove funds with less than 12 months of returns data.  A second method would be to determine the median incubation period for funds within a universe and then remove that number of monthly periods from each fund.  A third way would be to eliminate all performance figures prior to a fund’s entry date into a database. Lastly, an approach can be fashioned out of a method for determining the impact of the instant-history bias. This involves comparing a fund’s first year monthly average excess returns relative to the S&P 500 Index to its five year monthly average excess returns.  Presumably, one could define the difference between the one year and five year periods’ returns as the upward performance bias and then simply subtract this figure from all of the fund’s annual returns.  

While all these methods are used in research they each have shortcomings. By eliminating funds with less than 12  months of data we may be removing funds without any backfill. Finding the average median incubation period for funds is entirely dependent on data vendors; they would have to provide both inception date and data start (reporting) date for each fund. As a result of the non-standardized way in which data vendors collect their information, not all data providers provide both of these fields (this is an example of a “spurious bias”).

One may argue that the only true way to eliminate the instant-history bias would be to eliminate all of a fund’s performance figures prior to its entry date into a third party database. However, this would once again introduce a dilemma. Since not all database vendors provide both inception and data start dates, by choosing to use only the ones which provide both dates, we would be eliminating the instant-history bias but at the same time reintroducing a selection bias (i.e., some funds may choose not to report to these databases so this universe would not present a holistic view of the industry).

Finally, by examining a fund’s first year excess returns relative to its five year excess returns, we would be excluding hundreds of funds that have been in business for less than five years. We would have effectively truncated the total universe of funds to study.

So our dilemma as researchers is to choose one of the above methodologies or to develop a new approach. We are planning to incorporate the first approach into our future studies by eliminating those reported funds with less than 12 months of data. We know it is not perfect but it can be applied across all third party data bases. We are very interested in your comments regarding instant history bias with the use of third party reported data. Let us know your thoughts on this matter.