XBRL-based structured digital financial reporting - Blog: Digital Financial Reporting (using XBRL)

This is not a scientific study, rather it is a prototype and information which will help accountants understand how to create and verify SEC XBRL financial filings. The information also is helpful in understanding the moving pieces of automating the extraction of information from SEC XBRL financial filings. This information is not intended for financial analysis. The following are known limitations of this prototype:

(a) Information provided is for each filing submitted to the SEC during the period indicated. This includes ammended submissions. So, it is a known bug that some submissions are shown multiple times.

(b) The current process only looks for ONE legal entity per filing. However, SEC filings can contain more than one legal entity. As such, the exact amounts which should be shown per filing has not yet been perfected.

An analysis of 100% of the balance sheets of 7,220 SEC XBRL financial filings filed in the months of July, August, and September 2012 show that: (see below for where you can get the raw data)

Balance sheets detected to balance automatically: The balance sheets of 6,967 of the total 7,220 were detected to have assets, were detected to have liabilities and equity, and assets = liabilities and equity (the balance sheets balanced). Of the remaining 253, additional steps were necessary to find total assets and total liabilities and equity. Of that 253, 140 balance sheets were correctly modeled but my algorithms were not correct (i.e. I am not a good enough programmer) but I manually determined that the balance sheets were modeled correctly. The remaining 113 were determined to have one or more errors in the modeling which made discovery impossible. For example, if a filer creates an extension concept for total assets such as SPND:TotalAssets shown below; it is impossible to automatically and reliably detect total assets.
No balance sheet: Only 2 SEC filings did not report balance sheets. Mills Music Trust had cash receipts and disbursements and therefore they had cash. Therefore they could have had a balance sheet. L&L Energy had a statement of changes in equity, therefore they had equity, and therefore they also likely had stuff which would go on a balance sheet.
Balance sheet did not balance: Only 8 SEC filings had balance sheets which did not balance. Not one accountant that I have talked to says there are cases where balance sheets don't balance. In fact, the title "balance" sheet sheds light on that question. So here is an example of one of the 8 balance sheets which did not balance.
Rounding error: Only 6 filers provded balance which had a rounding error. Each of the rounding errors was plus or minus $1. China Fruits Corp is an example of a rounding error.
Reported NIL rather than ZERO: A total of about 122 filers reported a nil value (nil attribute on the fact) rather than zero. About 50 filers reported zero for total assets.
Restatement: 3 filers used a restatement member. Here is one example.
Unaudited/Audited: 22 filers of the total used the audited and/or unaudited member. There were many different approaches to using this. What is a bit odd is that pretty much 100% of all filers are either audited or unaudited. Why would there not be more consistency here? Here is one example.
Obviously inappropriate extension concepts: There were 54 occasions where obviously inappropriate extension concepts were used. For example, this filer created the extension concept "SPND:TotalAssets" (go to the interactive data, go to the balance sheet, go to line item "Total Assets", click and see for yourself). They also extended total liabilities and equity, "SPND:TotalLiabilitiesAndShareholdersEquity". Basically, each of these extension concepts is rather hard to justify.
Incorrect document period end date: There were 7 occasions where an inappropriate document period end date (dei:DocumentPeriodEndDate) was used. For example, this filer used March 31, 2013 whereas the balance sheet date was March 31, 2012. Another way to say this is that the balance sheet date and the document period end date were the same for 7,213 of the total 7,220 SEC filings checked.
Successor/predecessor member: On 14 occasions filers used successor member some times with and some times without a predecessor member. For example, here is one filers use of that member. I don't understand what this is so I have to do some additional digging.
Multiple CIK numbers in one filing: Of all the filings, only 9 had multiple CIK numbers in one filing. Not saying that this is right or wrong, just making an observation. For example, this one filing had 9 CIK numbers used within the filing (the most I have seen).
Multiple "root" entities: About 7 filings had multiple "root" entities. Here is one example. What I mean by root is that generally there is one CIK number and that CIK number has only one legal entity reported. These filings don't have any one legal entity which contains information for all other entities. So, this is a failing in my processing of total assets. I have to take into account EACH legal entity in a filing at least once and one has to be careful to also not double count. Here is an extreme example; this filer has approximately 55 business segments reported on the balance sheet none of which is the root of all other entities. Each has the same legal entity, yet no consolidated numbers are shown, i.e. nothing seems to be reported which contains all the business segments. As such, I wonder if these really are business segments.

Here is the data that I used:

ZIP archive of all filings in Excel spreadsheet: This ZIP file contains an Excel spreadsheet with all the data which allows you to explore these filings for yourself.
HTML page of all filings, by filer: Click to go to an HTML page which lists all 7,220 filings by filer name.
HTML page of all filings, by total assets: Click to go to an HTML page which lists all 7,220 filings by the amount of total assets, highest to lowest.
HTML page of all filings, by generator, then by filer: Click to go to an HTML page which lists all 7,220 filings sorted by "generator" (creation software reported by XBRL Cloud), then by filer name.
HTML page, ERRORS ONLY, by filer: List of ONLY the errors, sorted by filer name.
HTML page, ERRORS ONLY, by generator, by filing: List of ONLY the errors, sorted by generator software (creation software reported by XBRL Cloud), then by filer name.