Tuesday, February 03, 2009

All your data are belong to us

What if the data from which you derived your conclusions changes over time?

I recall reading the previous paper discussing this phenomenae, and am happy to report the authors have drafted a follow-up paper that sheds some light on what appears to be a wide-spread practice of retroactive rating changes from sell-side analysts.

The conclusions are still disturbing, but going forward this little game appears to be over.

We document widespread ex post changes to the historical contents of the I/B/E/S analyst stock recommendations database. Across a sequence of seven nearly annual downloads of the entire recommendations database, obtained between 2000 and 2007, we find that between 1.6% and 21.7% of matched observations are different from one download to the next. When we use a cleaned-up version of the 2007 tape as a point of comparison, we find that between 10% and 30% of all observations on the earlier tapes are now recorded differently on the 2007 tape.


No comments: