I understand why we might want to exclude preliminary data for sims.
But I do not fully understand any possible advantages to excluding preliminary data for PORTs.
Obviously there cannot be a PIT problem for ports.
So there is the problem that the information could get revised later (still not a PIT problem for a port).
In addition, I believe Marco has mentioned that there may be problems with NAs. I do not understand how using preliminary data creates NAs.
Could someone please expand on how this may create more NAs. Also please explain any other possible advantages to excluding preliminary data in ports.
I went ahead and funded a port today (using preliminary data) so the is not just a theoretical concern. I am wondering whether I should continue to use preliminary data next week or whether I might want to change this.
In addition to any details, any recommendations or personal preferences would be appreciated.
There are two minor disadvantages to Use Prelims for live strategies.
You’ll probably get a few more NAs in certain functions. It might also negatively impact the value of BeneishMScore and PiotFScore. The reason Use Prelims will create more NAs is as follows. Companies often announce only a small amount of data in their preliminary announcements. They’ll often announce only the income statement and not the balance sheet or cash flow statement. For almost all line items, if the preliminary announcement is NA we automatically fall back to the previous final statements for our figures. But there are some line items where that doesn’t happen, for a variety of complicated reasons. This has always been the case–it’s not new behavior. One way to explore this is to create a screen with CompleteStmt = 0. List a number of important line items using ShowVar. Then try Use Prelims and Exclude Prelims and compare the number of NAs you get with the former.
Because many companies do not issue cash flow and balance sheet statements in preliminary announcements, once the final statement has been released, your ranking position will change if it’s dependent on those items. So Exclude Prelims will mean lower turnover. For example, let’s say the accrual ratio–(NetInc - OperCashFl)/AstTot–plays a significant part in your screen or ranking system. A company announces its net income and then four weeks later files a statement that includes its operating cash flow and total assets. If you use prelims, your accrual ratio for that four-week interval will be very different from what it was before the announcement and what it will be after the final statement.
If neither of those disadvantages bother you too much, I would recommend Use Prelims for live strategies.
Yuvol - I haven’t really been following this issue (prelim data). I wasn’t aware that Factset supported preliminary data. In any case, how does one turn on or off prelim data? I want it off always but I would like to know where the control is.
Each screen (settings tab), simulation (general tab), and ranking system (performance tab, rank tab) has “PIT Method - Prelim” and a Use or Exclude option. We are working to get this functionality in place for the fundamental charts too.
Preliminary data can be overwritten once the final data is released. In addition, we do not have any effective dates for FactSet data prior to 2020. So the simple answer to your question is that "true PIT" is not possible.
Let's take the following scenario as an example. A company announces preliminary earnings on October 3, 2018. It releases official earnings six days later, on October 9, 2018. FactSet processes the preliminary earnings two days after release, on October 5. It processes the final earnings three days after their release, on October 12. If you use Prelims, you will get the final earnings as of October 4, one day after the initial release. If you don't use Prelims, you will get the final earnings as of October 10, one day after the final release, but you won't be able to see the preliminary earnings that were actually available from FactSet on October 5. (This is just made-up stuff by way of example; I did not check to see whether these dates are weekdays.)
Just to be safe, when backtesting I exclude all companies that released anything within the last five to seven days. But that may be extreme.
TL;DR: It would be nice to be able to lag the 10-k data and the analyst earnings estimate data effectively while keeping the pricing data current. SOME PREVIOUSLY DISCUSSED METHODS LIKE REBALANCING SIMS ON TUESDAY OR WEDNESDAY AFFECT THE PRICING DATA USED IN THE RANKING SYSTEM, INCLUDING VALUE RATIOS.
I just used weeksintoq > 0 in the buy rules and excluded PIT data. Inspired by this post. Does that ensure a PIT sim for the most part? You did something different or in addition to this, perhaps?
Hmmm…analysts' earnings estimates are separate and not affected by weeksintoq , but P123 has a lag still? Maybe we can ensure PIT-ness in a sim, with confidence? That would be good to know and be sure about for me, I think.
This can be addressed and any concerns put to rest? Can we create a 5-7 day lag in earnings release data at least, and what is the best way to do this?
And what lag do we use for analysts' earnings estimates at P123 now if that information is available to members?
I believe the data at P123 is good enough to test sims and have ports perform well out-of-sample but it has taken a couple of years of good results in out-of-sample ports to confirm this for me. There is an easier way? What is the best way to do this for future reference?
Here is another post where Yuval points out problems created by Factset (not P123) data delays. I am just interested in workable solutions to known Factset data issues where possible,: Monday morning overkill
What kinds of data are in scope for what's considered Prelim, as far as the P123 PIT Method == Prelim setting is concerned? Is it just financials from 10Q/10K reports? Or does it include analyst data too? What about actuals? Would be helpful to have a better sense of what's in and what's out.
The PIT method setting affects only data from financial reports and announcements. It does not affect analyst data.
As the tooltip says,
Companies may release preliminary financial data before their complete earnings report. By selecting "Exclude" for the "PIT Method - Prelims," you ensure that only the final earnings data is considered, omitting any preliminary figures. Choosing "Include" can enhance your trading system's responsiveness to shifts, though it may compromise consistency if only a few factors are updated.
What happens if the financial 10Q filing is filed before the news?
I had a case last week that i want to understand. In my live ports the company INBP had no TotalLiabilities from Thursday until after the complete data run on Saturday. The live port wanted to sell the stock on Friday, but when i run a sim now with daily rebalancing this doesn't happen. In both live and sim PIT Method - Prelim is "Use".
Is it possible that the "news data" (with incomplete data) has overwritten the filing data (with full data), because the filing was a day earlier than the news or what is a possible explanation here?
A simulation, even with daily rebalancing, will run according to the weekend rankings. Simulation rankings are not updated daily. So whatever you saw on Thursday will not be reflected in the simulation. Also, a huge number of companies announce their data right after filing it with the SEC, so what you're suggesting (incomplete data overwriting full data) seems highly unlikely. You can spot check for incomplete data by looking at the quarterly values for various items.