Paul Warren recently revisited the popular discussion of negative values in XBRL . Many experts consider negative values the biggest quality/comparability issue in XBRL. If it’s that important, I want to chime in! This post is hopefully the first of a few on the subject. This post will focus on my experience in creating XBRL and the rationale I apply to the problem of “negative values”.
I came up with an memory-aid to remember what “negative” means in different contexts. In the workflow for XBRL creation, these contexts are like different “layers” or “stages” of creating XBRL. In each of these 3 stages, “negative” has a slightly different meaning:
- Presentation – Printed Financials – HTML
- Data – Instance Document – Raw XBRL
- Presentation – Interactive Financials – Rendered XBRL
For most preparers, XBRL is derived from the HTML. Printed FinancialsThat’s why we start with the Printed Financials.*
Printed Financials: What does “Negative” mean?
Paul Warren creates a fictitious financial statement to illustrate his point. He chooses the three most essential line-items from the income statement:
It took me 10 tries on the SEC’s most-recently-filed list to find a filer who follows Paul’s convention; Ralph Lauren Corporation states their “Costs” equivalent (Cost of Goods Sold) as negative values. For example, a a cost of goods sold of 766.7 million for the three months ended 9/29/2012:
Why does Ralph print Costs as negative values? They are printed as negatives to visually reinforce the relationship between the last three line-items in the screenshot; Cost of goods sold is subtracted from Revenues to arrive at Gross Profit.
In other words, in the printed financials negative value can mean subtraction. I was surprised to find most of the filings I browsed do not use this “subtraction” convention. None of Microsoft, Apple, or Google, use this “subtraction” convention. See screenshots of Microsoft and Apple, below; all Cost values are stated as positive amounts.
Paul Warren, and Ralph Lauren choose to print costs as negative values, and Microsoft and Apple choose to print costs as positive values. Paul and Ralph’s financials are arguably easier to interpret (because the negative values reinforce the relationship between Revenues, Costs, andProfits). This “subtraction” convention makes the relationships of line-items more explicit, but it does not change the core meaning of the values themselves. I can understand that Ralph Lauren had 766.7 million in costs and Apple had 87,845 million in costs. The “subtraction” convention allows the printed financial presentation to “relate” that; but if I already knew this relationship, the negatives are extraneous.
So if negatives for subtraction are extraneous, why do some filers use them, and other filers ignore them? It seems the choice to use “subtraction” convention is arbitrary. In this BNSF example** the subtraction convention is used on the “Interest income” line-item, but not on the “Income tax expense” line-item. In other words, judging from the first subtotal (arriving at Income before income taxes), positive numbers are added, and negative numbers are subtracted. So I expect the same for the total (Net income). Yet for the total… positive numbers are actually subtracted. Is this internally inconsistent?
So “subtraction” makes the relationships among line-items in printed financials easier to understand. While it’s a convenient reminder of the relationships; it’s extraneous if the relationship is already understood by the reader. It’s arbitrary in that some filers use it — others don’t — and some use it inconsistently. It’s irrelevant to the understanding of the values themselves. Given all this, I think it’s fair to say that the “confusion” around negative values is not an XBRL issue, but an issue in general. Indeed, as we’ll see later, XBRL improves on this “muddling of meaning”. The XBRL instance document itself does not use extraneous and arbitrary negatives to convey “subtraction” relationships.
Let’s analyze a second “meaning” of a negative values in the “printed” layer. In this example from Vanguard’s most recent income statement, negative values are seen side-by-side with positive values.
Here Vanguard prints “Income (loss) from operations” as a negative value (149,878) in one period, and a positive value, including 516,532, in three other periods. This example is different from the “subtraction” examples above; in the “subtraction” example, a single line-item was all positive or all negative. In this example, a single line-item contains both positive and negative values. What does it mean?
The label explains its meaning — the label is like a “legend“. The legend tells us that a positive value can be interpreted as “Income” and a negative value can be interpreted as “loss“. This makes our interpretation of the values consistent with our spoken language; we would say Vanguard had a loss of 149,878 million (It would sound awkward to say they had “negative income” even thought it is technically correct) In other words, this line-item could have two-meanings.
Paul Warren explains the importance of “two-meanings” using a question-answer analogy. If some asked you about Vanguard’s Income for the most recent period, you would respond “Vanguard’s did not have income; in fact they had a loss of 149,878 million”. Paul says, “the sign should be negative” because, “in your answer you need to correct the question”
This “two-meaning” is not true for something like Costs. While someone could interpret a “negative Income” as “a loss”, the idea of “negative Cost of Goods Sold” is nonsense; there is no way to correctly interpret that — there is no way to correct that question!*** While “Income” has “two-meanings” — it is just as likely to be a “loss” — “Cost” has only one meaning. That’s how we distinguish the two uses of “negative values”.
While the “subtraction” meaning of negative is a matter of preference and does not affect meaning, the “two-meaning” use of negative is vital to interpreting the values properly. In this case, if Vanguard printed four positive values, they would completely change the meaning of these statements; a reader would wrongly understand that Vanguard had income of 149,878 million. This is why the XBRL instance document must use the “two-meaning” meaning of negative.****
There is a third, indirect way to interpret negative values. The “two-meaning” teaches us that it is necessary to print a negative value if you want readers to understand that you mean “loss”. However, you might forego the corresponding “label/legend” used to interpret the “two-meaning”; Vanguard’s latest annual Income Statement (below) uses negative values to convey “two-meanings” like their Income Statement (above). However, unlike the Income Statement above, with the label “Income (loss) from operations“, the label below is not clear– it says only “Income from operations.”
While the explicit label/legend “Income (loss) from operations” is arguably easier to understand, it does not change the meaning of the values. An experienced financial reader would infer that a negative income represents a loss, even if the label only tells her “Income from operations”. The “label/legend” is another convention which is indirectly related to negative values, but it seems somewhat extraneous if the “two-meanings” are already understood by the reader. And the use of “label/legend” [at least for Vanguard] seems arbitrary; they use an explicit “two-sided” label/legend in one case, and an implicit “one-sided” label/legend in another case.*****
In summary, we looked at the first of three “layers” or “stages” where negative values have meaning. This is the “Printed Financial Presentation” layer. In this layer, we analyzed three****** different meanings:
- “Subtraction” (optional clarification of the calculation relationship)
- “Two-meaning” (necessary to distinguish two-meanings)
- “Label/legend” (optional clarification of two-meanings)
Throughout this post, I hinted at the way each meaning is used in the next layer, the “Raw Data /XBRL Instance Document” layer. The “two-meaning” is the only one which is necessary for the XBRL Instance Document. In the next post, I’ll clarify what I mean.
*The SEC requires Printed Financials to be submitted in HTML filetype. But filetype is not relevant; this could be a PDF or a Word Doc or a Google Doc.
**BNSF may be using the subtraction convention for the Net Interest calculation because 1) This calculation is not as well-understood as the calculation of Profit and/or 2) It is more complex; there are more than two summands; unlike Revenue-Cost=Profit, or Income-Taxes=Income After Taxes
***This idea of “nonsense” is subjective; could we conceive of negative Costs ? Maybe if some reversal exceeded all actual costs. Or negative Revenues? Perhaps. We’ll talk about this idea of the “natural” or “expected” state of a line-item later– as Paul notes, it’s hard to have this discussion without implying some “statistical occurrence”.
****I insist XBRL must use negatives because I’m ignoring the possibility of creating two separate concepts [i.e. two separate line-items]; one line-item to represent “Income”, and one line-item to represent “Loss”. I’m ignoring this possibility because it diverges further from the printed financial, which use just one line-item to represent “both” (aka “two-meaning”). It isn’t just a matter of using fewer concepts; it is a matter of the best representation (most consistent with the way we think about financials, and the way we prepare financials.)
*****If the filer has only income for all financial periods, like this BNSF 10-Q, the filer only needs to communicate one of the two-meanings; all values represent “Operating income“. If the filer has only losses for all financial periods, like this Hoku 10-Q, the filer only needs to communicate the other one of the two-meanings; all values represent “Net loss”. This scenario justifies using a “one-sided” label. In some extreme cases (Nate to find a filing), filers may even forego the negative signs for a series of only losses; they let the one-sided “label” of “Losses” tell the story. But that doesn’t change the possibility of the other-meaning “Income” — that doesn’t change that “income” has two-meanings.
******In some cases, a negative could mean both “two-meanings” and “subtraction”. In other cases, a positive or negative value may not mean what you think. This Cheniere Energy example is mystifying unless you know how Noncontrolling Interest income relates to Income ; they use neither words “Income” nor “Loss”; simply “Noncontrolling-interest”. This is minimalist in the extreme. This makes it especially hard to understand that the positive values are actually Losses; they are printed as positives because they are being deducted. Watson does a better job; they explicitly describes these positive numbers as losses; but they mean for positive numbers to be added. Berkshire Hathaway describes these positive numbers as income; and they mean for the positive numbers to be subtracted. I discuss this extra-confusing “doubling-up” of “two-meanings” and “subtraction” in my series of posts about upside-down calculations. I’m thoroughly convinced that confusion around negative values is rooted in the confusion of different kinds of meaning and nonstandard uses of negatives, from HTML printed financials; I hope I can convince you that XBRL does better to disambiguate these meanings.