IBES is most followed for estimates. Bloomberg doing decent job. CapitalIQ doing ok. Don’t know others.
Shaun,
Thank you so much! I actually rejected 60% of the recommendations today because Yahoo showed negative revisions for those recommended stocks today. So this seems significant although not necessarily related to Marco’s original post.
I do not know if I will stick with Yahoo: I got the impression that Thompson Financial Network (Yahoo’s data source for estimates) may have a relationship with I/B/E/S but I will look into this further: or would appreciate further comments from anyone.
Specifically, my impression is based on the following from Wikipedia: " In the mid-1980s I/B/E/S began to expand its dataset to include companies trading in international markets. Lynch, Jones was sold to Citigroup in 1986. Barra bought I/B/E/S in 1993, selling it to Primark two years later. Thomson Financial purchased Primark in 2000; the database is currently owned by Thomson Reuters." Thompson Reuters’ web page would seem to confirm this as I/B/E/S Estimates data is listed as one of their products.
It still seems like buying because I think a stock has a positive revision when I/B/E/S says it is has a negative revisions may not be that smart.
BTW, I would not have put Zacks on that list either. At a minimum, fewer analysts, it seems. But then if I new what I was doing I wouldn’t be asking the question.
US and Asian companies seem to agree with you. Companies that were rated by analysts were surveyed (in a survey I found online). The US companies thought IBES (and their analysts) did the best at rating their companies. Asian companies liked Bloomberg. The results were similar overall for IBES and Bloomberg.
Thanks again!!!
Regards,
Jim
Marco,
Can’t you solve this issue by generalizing the existing functions?
For example, you could support positive offsets in EPSEst for future estimates, and use different values for the ‘type’ parameter (or introduce a third parameter) to let the user indicate to which source the estimate should be aligned (compustat or capitaliq).
CurQEPSMean etc are then just a shorthand for EPSEst with some set of default parameters, just like other (fundamental) factors. You can change the defaults, but then make it clear to users how to change their formulas to the correct EPSEst formula that is equivalent to the current definitions. This would satisfy users who don’t want their backtests to be changed, power users who want as many factors as they can get their hands on, and it doesn’t clutter the list of factors for beginners.
Thanks,
Peter
Marco,
Is there an easy way to determine whether an estimate is pointing to the correct year/quarter?
Is the data in the panels always for the correct quarter/year or can it occasionally be incorrect for the same reason? It seems like ShowVar and WeeksIntoQ or QtrComplete will not work to determine which quarter is being reported for fundamental data as it can be at least a few days off, I think. I’m not aware of another source that I have access to for CapitalIQ’s Earnings Estimates.
I’m not against doing some occasional research on my own. I’m just trying to determine the easiest way to do it.
It seems like it might be easy to look at next quarter or next year in the panels–if I knew when to do it. Or knowing that panels refers to the correct quarter/year (if it does) would work: I could just confirm that the estimates are in line with what I would expect for my port.
Thank you in advance for your help on this and for keeping us informed.
Regards,
Jim
Jim - I found that when LatestActualPeriodDate > PeriodDateQ, current qtr EPS estimates are actually in NextQEPSMean. This appears to affect only a very small% of database.
-debbie
Debbie,
Thank you so much!
I will check that out.
Yes. If I did the screen right only 120 stocks in the fundamental universe today.
Very helpful!!
Best regards,
Jim
We will add specific functions that will let you know if something is stale. Also functions that will tell you exactly what period CurrQ , CurrY , etc , correspond to.
Marco,
Thank you!!!
Very much appreciated.
Regards,
Jim
Thanks Marco!
Marco:
This seems like a good solution.
Bill
Marco, would you have an ETA for when this work will be completed?
Thank you!
Its coming along. I think by end of next week we should be able to rebuild the estimates and have ways to determine if fundamentals are “stale” (price reacting to data not yet updated in compustat) . I tested rebuilding a date in the past and about 300 companies , mostly small ones , had a “current q or y” difference. About what I expected.
It wasn’t easy. Compustat has a useful date for when a period was publicly announced (even if compustat didn’t process it yet), but it doesn’t work for the present. Compustat is mainly geared for backtesting, to write academic papers. They don’t even have future splits. Heck, they didn’t even collect open prices until 2004.
We are planning to buy a dataset from CapIq called “future events” that would solve the issues. But I think we can resolve it by looking at other data points for now.
Thanks for the update Marco!
Marco,
You said: [quote]
Heck, they didn’t even collect open prices until 2004.
[/quote]
A few years back this was discussed on the forum after the conversion from Reuters data to Compustat data, where Compustat was using the prior close price for the next open price. P123 introduced a function for calculating a pseudo open price for data prior to 2004 (I forgot the function). At the time you said that P123 had the actual open prices from (I think) CapIq, and in the future would replace the open function with the actual open prices.
I don’t remember a forum discussion in which the pseudo open prices were ever replaced with the actual open prices. Was this ever done, and if so when, and if not, are there still plans to do so?
Denny:
My memory is the function was (Hi + Low)/2. As far as I know we are still using the pseudo-open. Actual open prices were deemed far less important than R2G's and even the "Popular with Our Users" feature which I think should be taken down.
Best,
Bill