Rounding in accounting is the practice of replacing precise figures with simpler ones to make numbers readable and comparable, and it follows different rules depending on whether you are filing a tax return, running payroll, preparing financial statements, or ringing up sales tax. Two methods do most of the work: round half up, which is what most people learned in school, and banker’s rounding, which computers use by default. The rules governing which one applies, and when the rounded number has to tie back to the unrounded one, are more specific than most preparers realize.
The Two Rounding Methods You’ll Actually Encounter
Round half up, sometimes called arithmetic rounding, is the familiar one. If the fractional part is 0.5 or greater, round up; if it is less, round down. To the nearest dollar, $100.49 becomes $100 and $100.50 becomes $101. This is the default convention for tax returns and general-purpose accounting.
Banker’s rounding, also called round half to even, treats the exact midpoint differently. When a number lands precisely at 0.5, it rounds to the nearest even digit. So $101.50 rounds to $102, and $102.50 also rounds to $102. Over thousands of transactions, always rounding 0.5 upward creates a slight upward bias in totals. Banker’s rounding removes that drift because roughly half the midpoint values round up and half round down.
This matters because banker’s rounding is the default in the IEEE 754 floating-point standard used by virtually all modern computers. Most accounting software, database engines, and programming languages use round half to even internally unless overridden. When a financial system produces a total that differs by a penny from what you calculated by hand, that is usually why: you rounded 0.5 up, and the system rounded to even.
How to Round on a Federal Tax Return
Rounding on federal tax returns is optional. The IRS lets you round cents to whole dollars, but if you round, you have to round every amount on the return the same way.
The authority is 26 U.S.C. ยง 6102, which allows taxpayers to drop fractional dollars or round up when the fraction is fifty cents or more. The statute also makes clear that rounding applies to the final amount on each line, not to the intermediate calculations behind it.1Office of the Law Revision Counsel. 26 U.S. Code 6102 – Computations on Returns or Other Documents
The Form 1040 instructions put it in plain terms: drop amounts under 50 cents and increase amounts from 50 to 99 cents to the next dollar. So $1.39 becomes $1 and $2.50 becomes $3. When two or more amounts have to be added to figure a single line, include the cents in the addition and round only the total.2Internal Revenue Service. Publication 17 (2025), Your Federal Income Tax That last point is where people slip. Round each component first, add them, and the sum can drift by a dollar or more from the right figure.
You can also skip rounding and report exact dollar-and-cent amounts. Form 1040 and Form 1040-SR do not have a cents column, so you would need to write in the decimal point.2Internal Revenue Service. Publication 17 (2025), Your Federal Income Tax
Payroll and the Fractions-of-Cents Problem
Payroll creates a rounding issue that individual returns do not. Applying Social Security and Medicare rates to each paycheck almost always produces fractional cents, and the employer has to withhold a whole-cent amount. A fraction disappears on every check. Across hundreds of employees and a full quarter, those slivers add up.
Form 941, the quarterly employment tax return, handles this on Line 7, designated for the “current quarter’s adjustment for fractions of cents.” The adjustment reconciles the difference between the employee share of Social Security and Medicare actually withheld from paychecks and the amounts you get by applying statutory rates to total taxable wages for the quarter. It can be positive or negative.3Internal Revenue Service. Instructions for Form 941 (03/2026)
The adjustment is usually small, often just a few dollars even for large employers. Skipping it, though, leaves the return out of balance and can trigger IRS notices. Payroll software makes the adjustment automatically for most employers. Anyone running payroll manually or trying to reconcile a software discrepancy needs to know the line exists and what it is for.
Rounding on Financial Statements
Rounding in published financial statements is driven by materiality. An item is material if omitting or misstating it could change a reasonable person’s economic decision. A company with $4 billion in revenue does not need to show the last dollar. That is why large public companies typically present figures in thousands or millions.
The SEC permits rounded figures in filings like the 10-K and 10-Q. Standard practice is to disclose the convention near the heading of each table, usually with a parenthetical like “(in thousands)” or “(in millions, except per share data).” Consistency is important. If the income statement is rounded to the nearest thousand, the balance sheet and cash flow statement should follow the same convention. Mixing precision across statements confuses readers and raises audit questions.
Rounding at that scale can hide meaningful changes. A $400,000 revenue swing at a company reporting in millions will not register in the rounded figures. That is acceptable because the amount is immaterial at that scale, but whether a given amount is material depends on context, not just size.
When Rounding Becomes a Legal Problem
SEC Staff Accounting Bulletin No. 99 rejected the idea that materiality can be reduced to a percentage threshold. The bulletin states that “exclusive reliance on this or any percentage or numerical threshold has no basis in the accounting literature or the law.” A numerically small misstatement can still be material if it masks a change in earnings trends, hides a miss on analyst expectations, turns a loss into income, affects loan covenant compliance, or increases management compensation.4U.S. Securities and Exchange Commission. Staff Accounting Bulletin No. 99 – Materiality
This is where rounding stops being bookkeeping and becomes legal exposure. Earnings per share is reported to the nearest cent, and a fraction of a cent can decide whether a company rounds up to a figure that meets or beats market expectations. Academic research has identified statistical patterns suggesting some companies time adjustments so EPS lands just above the midpoint needed to round up to the next penny. The SEC’s Division of Enforcement has investigated whether issuers improperly rounded up EPS in quarterly reports, sending inquiries about accounting adjustments that could have inflated reported earnings.
The takeaway for anyone in a financial close: rounding needs to follow a documented policy applied the same way every period. If a rounding call happens to produce a more favorable number, the rationale has to hold up on its own, not be reverse-engineered from the result.
Making the Numbers Tie
Round a column of numbers, then round the total separately, and the parts often will not equal the whole. Three line items might round to $10, $10, and $10, totaling $30, while the unrounded values sum to $30.97, which rounds to $31. It is a mathematical fact of rounding, not an error, but the statements still have to balance.
The usual fix is a dedicated general ledger account, often called Rounding Adjustment or Rounding Difference. When rounded components do not sum to the required rounded total, the small difference posts there. It is then typically allocated to the largest line item or the least sensitive component so the reported figures tie. On the balance sheet, this keeps assets equal to liabilities plus equity down to the displayed unit.
The account also creates an audit trail. An auditor reviewing it should see only small, routine entries. A large or unusual balance is a red flag, because someone may be burying a real discrepancy in what looks like a mechanical rounding difference.
Spreadsheet and Software Traps
Spreadsheet formatting is the most common source of rounding confusion in practice. Format a cell as currency with two decimals, and the sheet displays a rounded number but stores the full unrounded value. Every formula referencing that cell uses the full-precision figure, so visible totals may not match what the displayed components appear to add up to. Wrapping the formula in ROUND makes the stored value match the display.
Data moving between systems is the other trap. One system might use banker’s rounding, another might use round half up, and a third might truncate toward zero. The differences stay invisible until someone reconciles the outputs and finds small, persistent gaps that are painful to trace. Documenting which method each system uses, and building a reconciliation step into the close, keeps it from becoming a quarterly problem.
Sales Tax Rounding
Businesses collecting sales tax hit rounding on every transaction. An 8.25% rate on a $4.99 purchase produces $0.411675, which has to become a whole-cent amount on the receipt. The general rule across jurisdictions is to round the tax to the nearest cent using standard mathematical rounding: third decimal place of 5 or higher rounds up, otherwise round down. Some states publish bracket tables that specify the exact tax for price ranges, effectively pre-computing the rounding.
The rounding applies to the tax calculation, not to the final total. If a cash transaction cannot be settled to the exact penny, the total may be rounded to the nearest nickel for change-making, but the sales tax reported to the state still reflects the cent-level calculation. On periodic sales tax returns, small rounding variances across many transactions are expected, and state revenue departments generally treat them as routine.