Taxes, assertion and lying

Disclaimer: This post does not contain accounting, legal or tax advice, and I am not an accounting, legal or tax professional. Rather, it contains a reflection on ethics and philosophy of language.

Suppose that Sally is doing her Elbonian federal taxes, and she's decided that some of her genuine business expenses would be hard to document, and so she simply does not claim those expenses on her tax forms. Moreover, Sally has lost a portion of her revenue records. She remembers that Frank paid her some amount under $100, but she does not remember what that amount is. Rather than trying to figure out what the amount is, she simply writes down $100, which she knows is excessive. Each of these actions only increases Sally's tax liability, and no one, I think, would say that Sally has done anything dishonest. On the contrary, one might think that she was scrupulously honest in overestimating her liability. One might criticize her on other grounds—for instance if the Elbonian government uses some of the tax revenues for immoral purposes, one might believe that Elbonians have a moral duty to pay the minimum legally required. But even so, to overpay is not a failure in honesty.

However, Elbonian tax forms—like those of many other countries—require one to sign a statement like:

1. To the best of my knowledge and belief, the information given on this form is true, correct, and complete.

Taking (1) literally, it seems that Sally is lying when she signs (1). For she knows that her expense information is not complete and that her revenue information is not correct. But we do not judge Sally to have been dishonest, given the direction of the errors. (I am assuming that the Elbonians do not have some perverse tax system where they tax you on your expenses or lower your tax rate if you have greater income. In those cases, Sally is probably cheating.)

So what is going on? There are actually two problems. One is the problem of the apparently incorrect endorsement made by means of the signature and the other is the problem of the apparently false assertions made by inserting too low a number in an expenses box and too high a number in a revenue box. I think the problem related to the signature is derivative. Even if the Elbonians did not have to sign their tax returns, they would not be permitted to make false statements.

I shall also assume that the Elbonian courts and legislature have not expressly addressed this question. Moreover, I shall assume that Elbonian officials do not make use of the information for statistical purposes that affect policy—if they do, then the incorrect information may be dishonest because it may be important to figuring out economic policies to know the actual revenues and expenses of businesses, and not just over- and under-estimates, respectively. (In this latter way, I assume, Elbonia differs from most countries which, presumably, do use the data statistically.)

There are, I think, three ways to do justice to our intuitions (or at least my intuition) that Sally has not done anything dishonest:

A. The boxes on the form headed "revenues" and "expenses" should be understood, instead, as "amount at least as large as the revenues" and "amount not exceeding the expenses (or: expenses claimed)". This solution implies a significant departure from the ordinary meanings of the language throughout the tax form, since the same point can presumably be made about at least some other parts of the form.

B. We should in general understand questions in terms of salience to the questioner. I do not know how to flesh out the account here. The idea is that the Elbonian tax authorities don't care what your actual expenses and revenues were. They care about the respective contributions to your tax liability from your expenses and revenues, but they only care about it in one direction: they want to have from you numbers that do not underestimate your liabilities. (However, they probably want any such adjustments to be made within a category, to make it easier to verify. Instead of reporting $150 revenue and $50 expenses, one cannot honestly report $105 revenue and $5 expenses.) There should be some general theory of salience that would give this judgment. In effect, this is not very different from (A), but is supposed to be a part of a general theory. Another example of this is the following. Suppose that a policeman is looking for a killer named Mark Kowalski, whom he knows to be hiding in one of the houses in the neighborhood. He comes to your door and asks: "Do you have anybody named 'Mike Kowalski' in your house?" You know this policeman often mixes up the names "Mark" and "Mike", and you know he's looking for Mark Kowalski. Furthermore, you happen to have a guy named Mike Kowalski in your house. If you say "Yes" to the policeman, this will waste the policeman's time, and the real killer may get away. If you say "No" but explain, it will likewise give the real killer time to get away. It seems plausible that an appropriate salience criterion will say that the honest and right thing to say is: "No." I wish I knew how to formulate the criterion.

C. Perhaps the illocutionary act in this case is not simply assertion but something else. Maybe there is an illocutionary acts like assertion with error skewed in interlocutor's favor. In other words, ordinarily the criteria for the moral appropriateness of an assertion have simply to do with truth. But in some contexts, the criteria also specify something about the permissible direction of error. For instance, if I manufacture router bits and I print on the box that this bit becomes unsafe at 12,000 RPM, I have not broken the relevant norm even if the engineering department's best estimate was that the bit only becomes unsafe at 14,000 RPM, which number I lowered to decrease potential legal liabilities and to increase workplace safety. On the other hand, if I changed the number in the other direction even by a smaller amount, and said that the bit becomes unsafe at 14,500 RPM, I would have been dishonest. Maybe the relevant illocutionary act when giving safety information to the user is assertion with error skewed in favor of safety. We could try to modify the content of the assertion in these cases (approaches (A) and (B)), but this might be the best way of handling the cases. Maybe the policeman case in (B) can be handled in some such way: assertion with error skewed in favor of helping law enforcement.

If (C) is the right approach, this means that our moral prohibitions on lying need to generalize to illocutionary acts other than assertion. And once we've done that, maybe we could state some general moral rule that applies to all illocutionary acts, and that generates the prohibition on lying as a special case?