Better earnings estimates?

Report a problem or error
1

Any chance to get some form of ‘analyst accuracy weighted earnings estimates’? Either give us access to the ‘raw analyst level data’ and let us build models for them somehow…or maybe a partnership with someone like these ‘crowdsource’ estimates sites like estimize (their community covers 900 or so stocks at least, and has API feeds and claims 69% more accurate):
http://estimize.com/estimates

Looks like this one might even be free. Maybe some tools on this side could be worked out for current higher level members / developers?

2

What you describe could be done in three ways

We build it from scratch but we would need detailed estimate data and we only license consensus. It’s a big expensive project.

Any data added to P123 needs to be historical point-in-time. I doubt estimize has that. Also, external data has additional hurdles, like linking it to the right company which is very hard in the past.

I believe Reuters has this type of data when they bough a company (forgot the name), but they would ask for a lot of money (to also make up for P123 switching to S&P!)

Thanks

3

Marco,

I would be interested in this but also anything that broadened or improved our estimate data.

I ran a screen of Zacks earnings estimates revisions today (free screener: not Zacks Rank) and ran the same one on P123. There were 12 stocks that passed both screens. This was out of 44 in the Zacks screen and 50 in the P123 screen. I believe this is just due to different analysts being followed.

I think there are a number of ways the additional information could improve ports. One could look for more consensus and probably improve the return for each stock or could increase the number of stocks in the port without affecting returns: more analyzed stocks.

This may be too costly and it works well as is.

Edit: is it just different analysts or a delay in updating the consensus estimates for CompuStat or Zacks?

4

It’s a lot more. To some extent, it’s different analysts, but the bigger difference is in the methods of collection and analysis, particularly in deciding which estimates should be included in consensus. For example, how long can an estimate be outstanding before it’s considered stale and is removed from consensus? What’s the difference between an estimate that is truly stale or just unchanged? How do vendors go about tracking analyst movement among firms? How do vendors go about ascertaining that estimates are truly apples to apples, assuming they do so at all? Etc.