Most of the prior analysis which I have done related to the ability to discover the identity of a reported fact from the perspective of the concept characteristic of that reported fact. For example, this web page summarizes information I was able to extract from the 30 companies which make up the DOW, the Fortune 100, and the S&P 500. If you look for "cells" which are orange, I could not find that piece of information.
So for example, if you look on the DOW list, you see that I cannot find only one value: Net cash flow for General Electric. This is because GE created an extension concept for that commonly reported fact. If you look on the Fortune 100 list you see 3 information points which I could not find. If you look on the S&P 500 list you see two things. First you see about 35 missing information points, that is about 1.4% as compared to .6% of the total number of information points for the DOW and Fortune 100.
But if you look at row 232 which is Host Hotels you see numerous missing information points. Row 270 which is Legg Mason has a similar issue. What is going on here which is different?
So, here is the deal. Of 7,160 reporting entities which I analyzed (I trimmed my set of 7,199 down slightly due to duplicate CIK number issues and multiple CIK issues which don't relate to XBRL); I could not idenfity ANY INFORMATION for 54 reporting entities. It also means that I could effectively discover information for 7,106 of the SEC XBRL financial filings, or 99.2% of all filings.
Why? Well, the graphic below summarizes the reasons and this PDF provides all the details.
Basically, 99.2% of SEC filers model their information and you can discover the "root" reporting entity or what is called the "economic entity" or "accounting entity". That accounting entity might be broken out in the report as different legal entities.
And so the vast majority of SEC filers allow for the discovery of the root accounting entity and may provide additional details for different legal entities, variable interest entities, or other reporting entity or accounting entity breakdowns.
There are some very specific things that these reporting entities do which make it challenging to identify the root of the reporting entity/accounting entity. Generally, these are inconsistencies with the ways other filers report information (the remaining 7,106 where you CAN discover the information). Here are a few examples (go through the entire PDF above if you are interested in the details and want to become a digital financial reporting master craftsman):
- Extension for consolidated entity: Legg Mason does something different than most filers. Click on the link to go to their balance sheet and look at the value of the Legal Entity [Axis]. Rather than using the Entity [Domain] to represent the consolidated entity, they created an extension concept, lm:ConsolidatedLeggMasonGroupMember. But that WOULD have worked, however they also made another mistake, they provided the Entity [Domain] report element which would have represented the consolidated group in their modeling, but they did not USE that report element. The dynamics of both these things make the root reporting entity unidentifiable explicitly like most other filers.
- Successor/predecessor: There are two ways SEC reporting entities articulate successor/predecessor related information: using the Scenario [Axis] (18 filers use this approach, here is one) and using the Legal Entity [Axis] (4 filers use that approach, here is one). There could be others, I did not check this. Having two approaches is unnecessarily inconsistent. These 22 filers stood out because again, each of them also modeled some sort of root entity but reported no facts for that entity which they modeled. This is similar to Legg Mason above, leading to not being able to detect the root entity using automated approaches.
- Equity component [Axis] used on balance sheet: I have no idea why but 11 SEC filers saw fit to model the balance sheet using the "Equity component [Axis] which is both rather strange and leads to not being able to discover the root accounting entity. Here is one example.
There are other funky things that SEC filers do, again look through the PDF. A great way to learn how to do things correctly is by learning from the mistakes of others.
One thing that I find particularly interesting is that if you consider the fundamental accounting relations which I had mentioned in a prior blog post and consider these reporting entity/accounting entity breakdown approaches; that frames the primary financial statements nicely into a neat little "package". Of the 7,160 SEC filings in my set of 2012 10Ks, there are only about 210 which have specifically identifiable errors/issues which, if fixed, would provide a solid set of financial information. The list of errors/issues is finite. These are all very fixable. In fact, I believe that all of these errors are detectable using automated verification processes. This set of reporting entity/accounting entity errors is likewise finite and fixable.
While it might be challenging to get the SEC to implement the inbound verification necessary to make these issues go away (and achieve a solid information set) at least for the primary financial statements; maybe the market would do it. Maybe XBRL Cloud would. It just builds on the framework which they already have in their EDGAR Dashboard.
When the FDIC implemented their XBRL-based system, on day one the FDIC said 18,000 mathematical errors simply disappeared from call reports. (Read page 26 in XBRL for Dummies) Granted, the SEC XBRL financial filings are more complex that FDIC call reports. However, it seems like we could be getting close to a similar benchmark for these SEC filings. Granted, it is only an incremental step but is seems both achievable and useful if achieved. I would go as far as predicting that primary financial statement information (balance sheet, income statement, cash flow statement) reuse could be very real within one to two years even if the SEC does nothing.