IMPORTANT: found a problem that affects companies with spinoffs

Announcements
1

Dear all,

We found a problem that affects companies that have had spinoffs. It was located thanks to HedgeHog in this this post

In a nutshell it affects market cap, which has ripple effects into other ratios like Price2Sales, Price2Cashflow, and Price2Book. What’s currently in production results in lower market caps (and lower “Price2” ratios) point-in-time. This is caused by spinoff dividend that Compustat generates that compensates for the large drop in price after a spinoff. This spinoff dividend is not real, it represents what happens if you were to sell the spinoff stock immediately.

What it all means is that this could cause an unfair “survivorship-bias” bias. When running a screen that favors low price2sales in the past companies that had a spinoff in the future get better ranks. The “look-ahead” in this is that there’s knowledge in the past that these companies still existed in the future, which usually means they performed better relative to companies that disappeared.

We’re testing the fix and it does not seem to have a major impact in re-runs. A few tests today actually produced better results than what is in production (it’s always better to have sims that understates “real” performance). We’ll do more tests, but this fix has to go live no matter what the consequences.

Below is an example of this problem for UNTD that had a spinoff on 11/1/2013

Very sorry about this.

spinoff-problem2.png
spinoff-problem2.png911×414 28.8 KB

2

For my understanding, and probably others as well:

  • Are just the Price series affected, i.e. Float and Share data are correct?
  • Are formula’s using MktCap such as Enterprise Value (EV) and Enterprise Value per share (EVPS) also affected?
  • Any idea what %, or number, of company’s with fundamental data are affected? I guess the futher back in time the higher…
3

Yes, EV & EVPS are affected as well. Thanks

Note that no matter what, companies after a spinoff will have strange ratios. This is because neither Compustat nor us tries to derive a new set of line-items post spinoff. Doing that can get complex very quickly and in some way only the company itself can do it right.

I think what P123 needs is something to avoid buying companies with a recent spinoff or sell off a position if a spinoff occurred or is about to accur.

4

Marco,

That sounds like a good idea. But probably handle it like mergers. So people can identify new spin-offs and try to find models for them. Just a SPINOFF(30)=1 (with days counter). Unless you think all data around them is junk. Then exclude them.

Best,
Tom

5

Marco, thanks, like the proposed fix. While on this subject, what’s the chance of getting a spin-off filter anytime soon?

6

Correct Marco?

Fixed shortly: Pre-spin-off price series.
To be continued: Post-spin-off statement data issues. But only an issue for the parent company until the next quarterly filing.

Thnx!

P.S. Investing in recently spun off subsidiaries, but not so much the parents, can be rewarding. See paper “Predictability of Long-Term Spinoff Returns”.