For quite some time I have wanted to update the way we have made the Comment Letter available. Because of this I have delayed updating the collection. I received an email from a researcher last week who needed our comment letters to be updated and since I have been struggling with another project (and she needed these updated) I turned my focus to the Comment Letters.

I had two primary goals with this project. First, I discovered sometime ago that the SEC sometimes included a field in the header associated with the Comment Letters that indicated (pointed to) the filing that was the subject of the letter <PUBLIC-REFERENCE-ACC>. I wanted to add these into the metadata that we include with the filings. The second thing I wanted to do was to systematically replace all of the pdf comment letters with a txt version since the txt version is friendlier to your use of Python to do finer work with these filings. Sometimes the SEC includes a TEXTEXTRACT version of the comment letter and my intention was to use that version if available. Otherwise we will run the pdf form of the letters through one of our tools and use that rather than the pdf version.

I will have to plead incompetence for not knowing earlier that the <PUBLIC-REFERENCE-ACC> was available and we were not incorporating it into the platform for your use. When we develop features we take what we think is a reasonable sample of the population and based on what we discover as we are working with that sample we make the decisions about how to move forward. You have read my comments in the past where I acknowledge – I don’t know what I don’t know.

Now that I know of this tag I set to work this weekend to begin incorporating it into the platform. Guess what, the SEC stopped using this tag before January 2021. From what I can tell, the last comment letter that included this field was uploaded on 11/12/2020. Frankly, the frequency of this field being included in the comment letters dropped off significantly after September 2020.

I am building a database of all of the metadata for these filings and so when the code first finished running on Saturday I was sure I had made some Burch error (this is an error that only I can make). After I spot checked a few of the later header files and did not see the field in these files I decided to check all of the header files for filings made since 2020 by year to see if maybe (I was hopeful) there was a new field. So I ran code to identify each unique field in every header file by year. Guess what, I did not make a Burch error, there were no filings with either the <PUBLIC-REFERENCE-ACC> field or a new field that provided this info. There were 29 unique fields across all of the filings made in 2020 and only 28 unique fields made in all of the filings in 2021 and 2022. Serious bummer.

{'acceptance-datetime', 'filing-values', 'private-to-public', 'type', 'conformed-name', 'business-address', 'state', 'street1', 'cik', 'fiscal-year-end', 'assigned-sic', 'filed-for', 'form-type', 'date-changed', 'sec-header', 'phone', 'public-document-count', 'city', 'zip', 'former-company', 'filing-date', 'company-data', 'mail-address', 'accession-number', 'state-of-incorporation', 'former-conformed-name', 'street2', 'irs-number'}

I did some searches on the SEC website and could not find any indication about this. The latest version of the PDS Technical Specification document still contains reference to this field. This is interesting because the latest version is 12/14/2020, which is after the field stopped showing up.

It is really interesting, that was a very useful feature and it is gone. I have not found anyone who is happy with the redesign of the EDGAR website/interface. The project I have been stuck on is trying to improve metadata collection from 10-K filings and I have finally had to make a decision that we can’t capture everything we want to capture with the confidence that is needed to distribute through the platform. I will post more on that later. I am just curious though who at the SEC makes these decisions and why?

Enough ranting. Let me address what has been accomplished. First, there is a new index UPLOAD Y2016-Y2025. The label is a little bit misleading. Today it only has filings from 2020 – the end of 2022. The balance will be moved from the UPLOAD Y2016-Y2020 index within the next week. That process is running now but I wanted to have the filings that our client needed available now.

Second, there is a new database – COMMENT_LETTER_DETAILS:

This contains a number of fields that I think might be useful. The SEC’s decision to stop providing the tag is going to make this more painful to complete than I hoped. However, for all the 2020 filings with the tag present we added a link to the EDGAR filing that was the subject of the Comment Letter. We also added the type of filing that was the subject. We will do this for every filing we can back to 2005 (which is when Comment Letters first became available).

When you do a search remember that the RDATE/CDATE have a different meaning with Comment Letters. The SEC does their research and sends a Comment Letter to a registrant. The day the letter is send is considered the ‘Filed Date’ for purposes of securities law and on EDGAR you will find the Comment Letters filed date to match the letter date. However, the Comment Letters are not visible on EDGAR for review until sometime later. The window has shortened considerably since the early days however it can still extend several months. The RDATE value we use is the date the header indicates the filing status changed from PRIVATE-TO-PUBLIC. The CDATE is the SEC EDGAR reported filing date. Thus, if you expect a market reaction, presumably that would come around the CDATE.