Test cases for assessing quality of fundamentals feeds

Discussion in 'Data Sets and Feeds' started by Begbie00, Jul 31, 2020.

  1. Begbie00


    Being a sole trader, I don't have access to CompuStat, Refinitiv, etc. I see a lot of inexpensive fundamentals feeds (e.g. Sharadar, Zacks) that I'd like to use for a quality filter on a momentum trading strategy.

    Absent the ability to compare these feeds to industry standard sources like CompuStat, what kinds of tests would you run on them to assess quality?

    Right now my test is to see how they handle big mergers & spinoffs (e.g. movements in the Dow like Alcoa/Arconic/Howmet, HP/HPE, Mondelez/Kraft, Dow/DuPont, UTC/Raytheon). Are there any other "problematic patterns" and test cases that tend to trip up fundamentals data sets?
  2. ZBZB


    You can get refinitiv EIKON rebadged as www.metastock.com Xenith for $99 a month.
  3. jharmon


    For simple companies they're mostly fine. You're going to get a difference of opinion about extraordinary events.

    For complex companies and corporate actions they're all garbage. Yep all of them.

    Spinoffs and mergers are particularly problematic.

    Other points of issue:

    Multinationals that report in a currency different from how they trade - which fx rate do you use? Carefully reconsider that choice for different metrics.

    Change of fiscal year reporting period - yes, some companies just change their boundaries. Not sure how you can compare a 15 month period vs 12, or a 10 versus 12 etc.

    Capital raising - how does an institutional share issue affect per-share fundas? How about when this happens intra-quarter?

    Debt/equity conversion - same as above.

    Bankruptcy emergence - does anything in the past really apply?

    Do I use fundas any more? Nope! After I heard (from an employee) about listed companies using legal mechanisms to defer sales to next quarter to game analyst expecations, I became suspicious. They're all cooking the books.
  4. Begbie00


    All of them meaning even the premium stuff like CompuStat and Refinitiv? Or just the discount ones?

    Thanks for the great set of other test case categories ... definitely want to build them into my test cases.

    Last edited: Jul 31, 2020
  5. Begbie00


    Exactly why I'm using filters first ... scaled total accruals, scaled net operating assets, PROBM/PMAN, probability of financial distress, etc. In theory most are more indicative as earnings management gets worse. TBD on how well it works out-of-sample.
  6. jharmon


    Yep all of them are garbage for complex coprorate actions. Including the premium offerings. I spoke to a rep at one of the companies about a controversial corporate action and he stated along the lines of "yes, we actuallly agree with you, but our competitors did it differently and we decided to match them." That was the Alcoa spinoff from Arconic (former Alcoa). From then I know it was just a "which custmer screams loudsest" contest as to which way the vendor would go. Who knows how they treated the latest spinoff with Howmet and Arconic - I stopped subscribing to fundamentals for these sorts of garbage reasons.

    Low-cost (discount) vendors offering historical fundamentals are particularly trash. You'll spend plenty of time finding problems that they'll never fix despite their rhetoric about so called "quality".

    Not sure any fundamentals work out-of-sample after Covid.

    For me, too much lag and lack of confidence about stated figures (see my previous post in this thread about earnings) for me but that's just my opinion. If you can find an edge that's not invalidated but this corporate postulating then trade it!

    One last thing I forgot to mention is discontinuous reporting. Some companies report all of their fundamentals at the same time each quarter. Others report portions of them over the course of up to 8 weeks. So you might get various earnings items early but balance sheet items a couple of months later. And they might vary it from quarter to quarter too. Urgh - too ugly for me.
  7. Begbie00


    Very much appreciate the context. Thank you. I guess I'll keep downloading all periodic/current reports and build something myself.