In 1494, the Franciscan friar Luca Pacioli published a bookkeeping treatise with a chapter on what to do when you make a mistake. The instruction is precise. If you posted an amount as a debit when it should have been a credit, you did not scratch it out. You posted a second entry on the correct side, wrote an explanation identifying the error, marked both entries for exclusion from account statements, and then made the proper entry where it belonged. The result, Pacioli wrote, would be "just as if" nothing had been written in the wrong column.
Something had been written, though, and it stayed on the page. Pacioli was codifying practices already in use among Venetian merchants, and the method they had settled on was not the simplest one available. Crossing out the wrong number would have been faster. The convention instead required three entries where one might have done: the mistake, its reversal, and the correction. What the extra work bought was information. A correcting entry tells you that someone noticed an error, roughly when they noticed it, and what they thought the right answer was. An erasure tells you none of that, including the fact that it happened at all.
Five centuries later, personnel at the theatrical production company Livent showed what the convention was guarding against. In the late 1990s, Livent staff removed expenses from the general ledger at quarter-end. The SEC described the items as "literally erased from the company's books." Staff altered invoice dates and account codes, deleted original entries, and reposted fraudulent information so that the changes looked like original transactions. The Commission noted that legitimate adjustments would have gone through journal entries — but journal entries leave visible trails for auditors. Deletion was chosen because it concealed that a change had occurred.
Pacioli's method would have left a record: the original posting, the reversal, the explanation. Livent's method left the impression that nothing had ever been different. Both sets of books were wrong, but only one technique also destroyed the evidence that anything had been fixed. The operation is the same whenever new state replaces old, whether the instrument is a clerk with a knife scraping a figure off parchment or a program overwriting a row. The answer that survives may be correct. The path to it is gone.
After the Arthur Andersen document-destruction scandal, Congress wrote the principle into criminal law. Section 802 of the 2002 Sarbanes-Oxley Act penalizes intentional alteration or destruction of records meant to impede a federal matter. The SEC's implementing rules went further in a way worth pausing on: auditors must retain material that is inconsistent with their own final conclusions — the abandoned reasoning alongside the reasoning that held.
The Venetians were building a checkable system, not a moral one. A balance by itself is an assertion; the sequence of entries behind it is what allows a second party to test the assertion without taking anyone's word. Keeping that sequence costs whoever does the extra posting a few minutes. Not keeping it costs someone else, later, considerably more — and that asymmetry is why the shortcut stays tempting in every system that offers it.