Adjustments Incorporating Filing Date

[ijyliu]

may want to use most recent filing date item as the financial data for the fixed quarter date

[ijyliu]

For roughly 8% of the data, filing date from the tabular financial statement variables is outside of the fixed quarters.

This is concerning

What was our threshold for earnings call date

Potential action only drop some or all of these cases really

see https://github.com/current12/Stat-222-Project/blob/main/Code/Exploratory%20Data%20Analysis/All%20Data/All%20Data%20-%20Analyze%20Filing%20Date.ipynb

[ijyliu]

Other option detach earnings call and financial data

Make data be fixed quarter date and most recent call and most recent statement, not necessarily associated

Do we trust the filing dates more or the filing date year and quarter more

Probably best to just check a few filings the dates and the year and quarters and see which is right

[ijyliu]

Both year + quarter and filing date seem accurate from looking at a few items

technically it's possible for a company to have a call without releasing financial results, a point in favor of switching to most recent filing date rather than attaching calls and financial statements

[ijyliu]

Summary

I recently re-discovered variable filing date from the tabular financial statement data, which theoretically represents the day the financial data was released via SEC Edgar.

Note that technically there are 3 different filing dates for the 3 kinds of financial statements (cash flow, income, balance, sheet). However, for a given company, year, and quarter, these dates differ less than 1% of the time. For this post I looked at the version of filing date in all_data, which I think is from whichever of the 3 statements was joined last.

Potential Actions

Restructure all data to be credit rating on fixed quarter date, most recent earnings call, and most recent tabular financial statement data (rather than data attached to the call by year and quarter). In other words, join financial data on the condition that filing date is the latest one for a given fixed quarter date.

In the current setup joining financial variables on by attaching them to earnings calls, roughly 4.5% of the time, the filing date is after the fixed quarter date, an inconsistency. Another 2.5% of the time, the filing date is more than 92 days before the fixed quarter date. 54% of the time, the filing date is more than 2 days away from the earnings call date (though only ~3% of the time is it more than 45 days away). It's theoretically possible for a company to have an earnings call without releasing accompanying financial statements for that quarter, though this is supposedly uncommon. A fair amount of the time it does seem to happen that financial results are released without a call.

It's not super clear which of year + quarter or filing date are more accurate. Browsing the EDGAR website, both seemed good in the (admittedly few) cases I looked at. The current method puts more faith in year + quarter and the assumption earnings calls and financial statements are meant to go together. Changing things up would put more faith in the quality of the filing date variable.

An alternative potential action is that of dropping all cases where filing date is not in the fixed quarter.

Code

https://github.com/current12/Stat-222-Project/blob/main/Code/Exploratory%20Data%20Analysis/All%20Data/All%20Data%20-%20Analyze%20Filing%20Date.ipynb

[ijyliu]

Proposed: drop observation if filing date more than 45 days after earnings call date

[ijyliu]

@seanzhou1207 opinion on this? https://github.com/current12/Stat-222-Project/issues/53#issuecomment-2000469232

[ijyliu]

Note: signing up for Libor's office hours to discuss this, since he probably knows a lot about if earnings calls and financial statements should align - for join structure

[ijyliu]

note: dropping cases where "fillingDate" does not agree between sources

[ijyliu]

implemented drop of items with filling date outside of fixed quarter date and more than 45 days after call date