Comments on: Lipper: Active vs. Passive, Round 3,462

By: GerardVanDam

GerardVanDam — Tue, 10 Jul 2012 08:20:52 +0000

Thank you for the offer to send me your report. I’ll definitely contact Joel.

Some information for other readers. This is probably no news to you:

Carhart (1997) has split ‘survivorship bias’ into ‘survivor bias’ and ‘look-ahead bias’.

‘Survivor bias’ = Dead portfolios are excluded from the sample.

‘Look-ahead bias’ = methodology requires funds to exist for a specified period of time (in this blog: 1, 3 or 10 years).

In using rolling 1 year returns (to grasp ‘survivorship bias’) the effect of ‘look-ahead bias’ was reduced. ‘Survivor bias’ was already covered in using the closed/dead fund database.

By: ed.moisson

ed.moisson — Fri, 06 Jul 2012 16:06:12 +0000

The analysis I carried out did indeed include funds that have since closed.

I used rolling 1 year returns so as to get an insight into returns for funds that did not have 3 and/or 10 year returns, and to see how the findings vary over different periods, i.e. together accounting for as many funds over as many different periods as possible.

For any given 1 year period, my approach ensures the data is robust. As for comparing 1992 with 2011 one year returns, then certainly I could have run the analysis looking only at the 10% of actively managed equity funds in Europe have a track record that stretches back to 1992 (and I may do this in future). However, I wanted to get as much of an ‘industry-wide’ view as possible.

Please email Joel your email address and I’d be happy to send you the report that fleshes out the analysis a little more.

By: GerardVanDam

GerardVanDam — Fri, 06 Jul 2012 13:15:38 +0000

Could you please elaborate on your methodology? Especially on the issues of ‘survivor(ship) bias’, ‘back-filling bias’ and ‘(mis)classification / factor bias’ concerning the Lipper database.

To my knowledge Lipper has a ‘dead’ fund database. Why not use it to prevent ‘survivor(ship) bias’?

You’re trying to grasp ‘survivorship bias’ by taking funds’ rolling returns (for every year from 1992 to the end of 2011) and looking at the percentage out-performers over 1-year periods. The numbers you mention for the first year (59.1%) and for the last year (26.7%) are completely in line with a database that suffers from ‘survivorship bias’. A part of the underperformers from 1992 will have disappeared from the database (especially the consistent underperformers), so with the ‘survivorship biased’ database it looks as if a larger percentage will have outperformed than in reality has been the case.

Very curious to hear your thoughts on these issues.