Tax Refund Statistics (2026)

The average federal tax refund for the 2026 filing season was $3,275 through the week ending 17 April, up 11.3% from $2,942 at the same point a year earlier.

The IRS issued 90.4 million refunds worth $296.1 billion, against 86.0 million worth $253.1 billion in 2025.

The total refunded rose 17.0% while the count rose only 5.1%, which is the arithmetic behind the higher average: more people got a refund, and the refunds themselves were larger.

Total returns received were 140.2 million, down 0.3%, and returns processed were 138.6 million, up 0.4%.

So the filing population did not grow. The change is entirely in how much came back to the same set of filers.

That pattern points at withholding tables and credit amounts rather than at any change in behaviour by taxpayers, since nobody chose to file differently in the aggregate.

Direct deposit refunds totalled $296.7 billion across 90.8 million refunds, at an average of $3,269.

Both rose sharply on the prior year, by 21.1% in value and 12.0% in count.

As published, the direct deposit count slightly exceeds the total refund count. Both figures carry the IRS's own footnote, and we report them as published rather than adjusting one to fit the other.

137.6 million of the 140.2 million returns received were e-filed, about 98.1%.

The split was 72.8 million through tax professionals and 64.8 million self-prepared.

Self-prepared filings grew faster, up 1.7% against 0.4% for professional filings, continuing a long shift toward software and away from paid preparation for straightforward returns.

Visits to IRS.gov rose 58.4%, from 322.9 million to 511.4 million.

No other line in the table moved by more than 21%.

A jump that size in a season with flat filings usually means people were looking something up rather than filing something new, which is what you would expect if refund amounts or credit rules had changed.

A refund is not a payment from the government. It is the return of money that was withheld from your own pay and should not have been.

A $3,275 refund means roughly $273 a month left your paycheque unnecessarily and sat with the Treasury until you filed.

The Treasury paid no interest on it. At a 4% cash rate, and assuming the overwithholding was spread evenly across the year so each dollar was held for about six months on average, that is roughly $65 of foregone interest on the average refund.

The $65 is real and small, and the behavioural argument on the other side is real and large.

For many households the refund is the only lump sum of the year, and it reliably funds a debt paydown, a repair or a deposit that a slightly larger paycheque would not have.

Forced saving with a bad interest rate still beats intended saving that never happens, which is a defensible position and a very different one from not knowing the tradeoff exists.

The case against a big refund is strongest when the household is simultaneously carrying credit card debt at 20% or more.

There, the overwithheld money is not earning 4% or 0%. It is costing 20%, because the same cash flow would have reduced a balance accruing at that rate all year.

That is the situation where adjusting withholding is worth the paperwork, and it is a much stronger argument than the foregone interest one.

Withholding is set by the Form W-4 you filed with your employer, and it can be changed at any point in the year rather than only when starting a job.

The IRS publishes a Tax Withholding Estimator for exactly this, and the useful time to run it is mid-year, when there is enough of the year left for a change to matter.

Aiming at zero is the wrong target for most people. Aiming at a small refund leaves a margin for error, since the cost of underwithholding is a bill and possibly a penalty, and the cost of a small refund is a few dollars of interest.

The average $3,275 refund invested once and left alone for 30 years at a 7% annual return would be worth roughly $24,900.

Done every year for 30 years, the same $3,275 becomes something over $300,000 on the same assumptions.

Those figures are illustrative arithmetic rather than a forecast, and their real point is the contrast: the refund is treated as found money, and it is the single largest predictable cash event in most households' year.

These are cumulative season-to-date figures compared with the same week of the prior season, not full-year totals.

Returns filed on extension are not in them, and those returns skew toward higher-income and self-employed filers, so late-season data is not a random sample of what is missing.

Weekly comparisons early in a season are particularly unreliable, because the season start date and the position of weekends shift the base.

All filing, refund and direct deposit figures are from the IRS Filing Season Statistics for the week ending 17 April 2026, comparing that point in the 2026 season with the same point in 2025.

The foregone-interest and long-term-growth figures are our own arithmetic on the IRS average refund, clearly labelled as illustrative, and they are the only numbers on this page not taken directly from the IRS table.

Nothing here is tax advice. Withholding decisions depend on individual circumstances and are worth confirming with a tax professional.