I recently stumbled over an interesting paper by Ljungqvist, Malloy and Marston (LMM). It seems the old soviet-style air-brushing of pictures is still alive and kicking on Wall Street. In short bad forecasts (bad in terms of being far away from what really happened) have been deleted from an analyst’s performance history, while good ones are being kept.
What LMM did was looking at a database (called I/B/E/S run by Thompson Financial) where many Wall Street analysts post their buy and sell recommendations. Some post their recommendation anonymously, while some include their names. Now, this database can be used retrospectively to check how well each analyst is performing. Did her forecasts hit the target? Was her buy-recommendation a good one? By searching through all the research that one particular analyst has issued, one could get an opinion of how well she is doing.
LLM checked all analyst reports covering the period 1993-2002. They look at the entries in the database in 2002, and again in 2004 (for the same analyst reports covering the period 1993-2002, so they would expect to find exactly the same information bar corrections of factual errors). They find that for some reports the author disappears from the database after first having been included. That means that for some reports that had an author in 2002, the name of the author had disappeared when they searched again in 2004 (looking at the same analyst report). Interestingly they find that reports were the analyst was furthest away from the consensus, and particularly bad reports were more likely to become anonymized in this fashion. Apparently neither the database provider, the investment banks or the analyst themselves have any idea how this could happen.
It will be interesting to see what reactions they will get from Thompson and the investment banks.