We noticed a problem with Yield for many ADRs, perhaps over 100 of them. The Indicated Annual Dividend (IAD) goes missing for about a month or so in the data . It might coincide with earnings announcement. We think it’s a data problem and we have reported it. Depending on what they say we might have to work around it or wait for the fix.
For example Toyota’s (TM:USA) and Sony (SONY:USA) Yield factor keeps going to 0 and back historically. See images below.
Sorry for not noticing this sooner. For now it seems to only affect ADRs.
I heard back from data provider. They set the Indicated Annual Dividend to NULL (0 for our purposes) if 210 days pass since last announcement for semi-annual dividends. But if you look at the dividend data the ex-dates (which are usually very close to the announce date) are consistent for Toyota, every 6 months or so.
In other words, being a Japanese company, the announcement is somehow missed . Then , around the divided pay date, the exdate is being backfilled. But by then it’s too late since the IAD went to 0 for a month or two.
I doubt they will change their methodology so we will have to do something ourselves for ADRs and override them.
The workaround is now live. We simply delay an additional 60 days setting the IAD to 0 – FOR ADRs ONLY. This is to account for the data provider seemingly missing the announcement, and expiring the IAD after time has passed since the last pay date. We think they are backfilling subsequent dividends, but they do not agree it’s an issue with their process, and they will not do anything.
Our delay cleans up a lot of the jittery yields with Honda and Toyota, but it’s not a fool-proof fix. Figuring out the future Yield or IAD is quite involved and requires accuracy in processing press releases from the companies. That’s well beyond our capacity. So it’s safer to do this workaround only for ADRs where it’s conceivable that the press release is not being picked up.
Is this fix also effecting ADRs that did not have this problem? As noted in another thread this change had a significant negative impact on one my strategies that utilizes Yield as its primary ranking factor. If I eliminate ADRs from my universe I get my performance back, however this implies to me that this fix might have destroyed a valid signal as opposed to just exposing overfitting in my model.
