How Valid Are They An Examination of XBRL Filing Documents with the SEC EDGAR system

1
How Valid Are They?-- An Examination of XBRL
Filing Documents with the SEC EDGAR system

• Kuo-hua Chou
• National Pingtung Institute of Commerce, Taiwan
• Email ckhmike_at_gmail.com

2
XBRL-based Filings

• An important step in XBRL progress.
• At first stage, there are two simpler ways
• A voluntary program e.g., the US SECs Voluntary
  Filing Program (VFP).
  
• A fill-in-the-blank approach e.g., the mandatory
  test filings in template format by Chinese listed
  companies.

3
VFP of US SEC

• The program started running from April, 2005.
• Till Nov. 30, 2006, there are a total of 97 VFP
  filings representing 32 companies.
  
• United Technologies has the most number of
  filings eight.

4
Well-formed and Valid

• According to XML 1.0, an instance document must
  be well-formed, and should be valid.
  
• For an XBRL instance to be valid, it must be
  created in accordance with a DTS.

5
On-line Validation

• The checking schemes in the EDGAR system only
  cover document type, file extension, and presence
  of a particular XBRL tag.
  
• The EDGAR system does not support the on-line
  validation of XBRL instances against their
  respective DTS.

6
Valid and Inconsistent

• XBRL 2.1 only invalidates an instance with syntax
  errors.
  
• XBRL 2.1 does not invalidate an instance with
  only semantic inconsistencies, such as
  calculation errors.
  
• In this paper, an instance will be tagged
  inconsistent when there are calculation errors
  found in it.

7
Validation Tool

• I used Fujitsus Taxonomy Builder Instance
  Creator to validate all VFP filings.
  
• Reason the software is free, reliable and
  multifunctional.

8
Validation Results

• Among the 97 VFP filings, 41 instances are valid,
  the other 56 instances have calculation errors.
  
• Hence, these 56 instances are inconsistent with
  their respective DTS.

9
Inconsistent Rate

• Total number of filings 97.
• Total number of inconsistent instances 56.
• Total inconsistent rate
• 56/9757.73.

10
Examples

• The following companies have at least four
  inconsistent filings
  
• RR Donnelley 4/4.
• EDGAR Online 6/6.
• EMC 5/6.
• Satyam Computer 4/4.
• Xerox 4/4.
• Dow Chemical 4/4.

11
Numbers of Errors

• Number of filings with number of errors
• GTE 60 three.
• LT 60 and GTE 50 three.
• LT 50 and GTE 40 six.
• LT 40 and GTE 30 seven.
• The highest number of errors in one single
  instance is 68, in the second filing of RR
  Donnelley.

12
Reasons for Inconsistencies

• The technical side Misconceptions about
  summation-item relationships.
  
• The accounting side Misconceptions about the
  inherent relationships between accounting
  concepts.
  
• The practical side Intentional omissions on some
  phantom sub-total elements.

13
Case EDGAR Online

• All its six instances are inconsistent.
• The numbers of calculation errors are increasing
  as well 12, 15, 22, 32, 32, 36.
  
• The following five slides explain how to detect
  and correct them.
  
• Context assets of EDGAR Online as of June 30,
  2006, taken out of its fifth filing.

14
The Fifth Filing of EDGAR Online
15
The Keyed-in Values and Calc Values
16
Partial View of Assets Calculation Tree
17
A Partially Corrected Calculation Structure

• The blue stamps represent three missing sub-total
  elements. After filling in values of these three
  elements, the CALC value of Total current assets
  is the same as its keyed-in value.

18
A Fully Corrected Calculation Structure

• EDGAR Online had missed its Other Intangible
  Assets in the XBRL instance. After filling in the
  value and adjusting the values of two
  corresponding sub-total elements, the CALC value
  of Assets is the same as its keyed-in value.

19
Model Companies

• Microsoft All seven filings are valid.
• Adobe Systems All six filings are valid.
• United Technologies Although its first three
  filings are inconsistent, its last five filings
  are valid.
  
• Altria Group, Pepsico,3M and GE each has two or
  three consecutive valid filings.

20
Extension Taxonomies

• Every VFP company builds its extension
  taxonomy. Why?
  
• Companies couldnt find corresponding elements
  for some of their accounts or terms in use. The
  fit issue.
  
• Companies wanted to retain their own labels
  instead of using the standard labels defined in
  base taxonomy. The label issue.
  
• A simpler calculation or presentation structure.

21
The Fit Issue

• Given the diversity of GAAP, the fit between
  taxonomies and financial statements cannot be
  perfect.
  
• One way to measure the fit
• CEU total number of company-defined elements
  used in an instance.
  
• EU total number of elements used in an
  instance.
  
• CEU/EU the higher the rate, the lower the fit.

22
How Poor is the Fit?

• EDGAR Online the average rate of CEU/EU for its
  six filings is 29.38.
  
• Microsoft the average rate of CEU/EU for its
  seven filings is 54.51.
  
• Altria Group the average rate of CEU/EU for its
  two filings is 71.13.

23
Why is the Low Fit?

• It needs further work to analyze what constitute
  the low fit. Does it come from low fit in the
  main financial statements (if so, which one has
  the lowest fit), or from low fit in the
  footnotes, MDA, MR, etc.?

24
Case Microsoft

• Some elements are used in two or more scenarios.
• It seems that the low fit in Microsoft may come
  from footnote, mda, etc.

25
The SECs Reaction

• The SEC has awarded 5.5 million to XBRL US Inc.
  to complete the writing of XBRL taxonomies.
  
• Hopefully, this can drastically raise the fit.

26
The Label Issue

• There arent standard chart of accounts
  prescribed by regulatory agencies.
  
• In practice, companies may use different labels
  to represent the same concept.
  
• In the label linkbase of base taxonomy, there is
  a standard label defined for every concept.
  
• Companies may prefer using their own labels to
  standard labels.

27
The Label Issue

• Through the use of priority and use attributes of
  labelArc, company-defined labels can get higher
  priority and replace the standard labels.
  
• Seven companies have used the above scheme in
  their label linkbases Microsoft, PepsiCo, GE,
  Ford Motor Credit, Ford Motor Co., and ENGlobal.

28
The Impact of XBRL

• Among the 32 companies, there are only 7
  companies using the priority scheme in the label
  linkbases.
  
• Q Has the implementation of XBRL narrowed the
  range of choices in account labels?
  
• A It needs more work to analyze the impact.

29
Succinctness

• For succinctness sake, some companies build
  extension taxonomies by importing the upstream
  usfr-pte or usfr-fste taxonomies, and define
  their own calculation or presentation
  structures.
• In this way, companies can avoid a lot of phantom
  sub-total elements.

30
Against the Rule?

• According to the SECs VFP rule, filers must use
  one of the standard US GAAP Version 2.1 based
  taxonomies developed by the XBRL consortium as
  part of the basis for creating their instance
  document.
• The way of succinctness seems to violate the
  rule.

31
Case Ford Motor Co.

• Ford Motor Co. builds its extension taxonomy by
  importing usfr-pte taxonomy.
  
• The calculation structure by Ford Motor Co. is
  much simpler than the structure defined in the
  usfr-ptr and us-gaap-ci taxonomies.
  
• e.g., the calc structure of balance sheet

32
Ford Motors Calculation Structure of Balance
Sheet
33
Taxonomy Versioning

• For the time being, few companies maintain a
  relatively stable taxonomy.
  
• Most VFP companies file a different taxonomy
  every time they do the filing.
  
• If the contents are the same, the file names
  should be the same. On the other end, if the
  contents have changed, the file names should be
  changed.

34
Versioning Problem

• Some VFP companies, e.g., Microsoft, EMC, EDGAR
  Online, and Adobe Systems, have used the same
  file names for schema and linkbase files of two
  consecutive filings even though the contents have
  changed.
• This will cause problems in validation process.

35
Case Microsoft

• Among its seven filings, the file names of the
  main schema files in its last six extension
  taxonomies are exactly the same
  msft-20050228.xsd, and linkbase files have the
  same phenomenon, too.
• Except for the seventh filing, the contents of
  the schema and linkbase files had changed every
  time they were filed with the EDGAR system.

36
My View

• In most software, the validation process is
  triggered by pushing the Validate button. If
  errors occur after validation process, users tend
  to feel that the instance is invalid.
• Before an agreement is reached and rules are
  promulgated in certain specification (e.g.,
  FRIS), its better to correct all the calculation
  errors before filing the instances.

37
Suggestions

• A full-fledged on-line validation function should
  be developed and installed in the EDGAR system in
  the near future.
  
• Before that happens, the validation task may be
  performed by staff members of the EDGAR system.

38
Suggestions

• The SEC may consider adding a column showing the
  status of validity on the web page that lists all
  the XBRL filings chronologically.

39
Suggestions

• Since there are already four companies that have
  used same names for consecutive files that are
  not exactly the same, there should be some
  warnings on the individual web page that lists
  files for a single filing that has the above
  problem.

40
Suggestions

• SEC may consider requiring companies to provide
  the delta contents, which are differences
  between two consecutive taxonomies.
  
• This information is very important both for
  outsiders to perform validation task, and in the
  future, for auditors to perform XBRL attestation
  services.

41
The End