With seven recognized methods of proof. One is direct and generally works from identified transactions and records. The other six use circumstantial evidence to reconstruct income from assets, expenditures, deposits, cash activity, business ratios, or operational data.
The methods are set out in Internal Revenue Manual 9.5.9 and constrained by Supreme Court case law that is often left out of shorter summaries. Understanding which method is being used against you is the first step in defending it, because each one has its own required proofs and its own weaknesses.
There are seven methods, and the IRS prefers the first one because it is the hardest for a taxpayer to rebut.
| Method | What it works from | Typical use | IRM authority |
|---|---|---|---|
| Specific item, also called the direct method | Identified transactions in your own records and third-party records | Understated income, overstated expenses, false credits | IRM 9.5.9.4 |
| Net worth | Change in your assets and liabilities across a period, plus living expenses | Missing or inadequate records, unexplained wealth | IRM 9.5.9.5 |
| Expenditures | What you spent, compared with reported income and net worth | Cash income spent on living rather than saved | IRM 9.5.9.6 |
| Cash method | Total cash receipts for the period, compared with cash spending and deposits | Cash-intensive businesses where deposits alone don’t capture total receipts | IRM 9.5.9.8 |
| Bank deposits | Total deposits, plus cash spending, less non-income items | Businesses that deposit receipts | IRM 9.5.9.7 |
| Percentage markup | Industry ratios applied to your purchases or inventory | Retail, limited use only | IRM 9.5.9.9 |
| Unit and volume | Units handled multiplied by price or profit per unit | Pizzerias, laundromats, funeral homes | IRM 9.5.9.10 |
The manual explains the preference plainly at IRM 9.5.9.4: the specific item method “is the easiest to understand, included in a prosecution report, and present at trial,” and “is the hardest for the subject to rebut.”
The government identifies specific transactions you omitted or deductions you were not entitled to, and proves each one.
IRM 9.5.9.2.1 lists three categories suited to it: understatement of income, overstatement of expenses, and fraudulent claims for credits or exemptions. IRM 9.5.9.4 adds that the government “must show that the specific omissions of income were made willfully.”
Expect the agents to interview the people around the transaction. Your bookkeeper, your employees, your customers, and your accountant are all fair game. 26 U.S.C. § 7525 gives accountant communications a limited privilege in noncriminal tax matters, but it does not extend to criminal cases or IRS-CI interviews, which is why sensitive discussions are often better structured through counsel.
The government establishes what you were worth at the start of a period, what you were worth at the end, adds what you spent, and treats the excess over reported income as unreported income.
The Supreme Court described it in Holland v. United States, 348 U.S. 121 (1954). The government “attempts to establish an ‘opening net worth’,” then “proves increases in the taxpayer’s net worth for each succeeding year,” and “the taxpayer’s nondeductible expenditures, including living expenses, are added to these increases.” Where the result substantially exceeds reported income, the government claims the excess is unreported.
The Court was not enthusiastic about it. In the same paragraph, it warned that the method “is so fraught with danger for the innocent that the courts must closely scrutinize its use.”
Four, and each one is a place to attack the government’s case.
| Safeguard | What Holland requires |
|---|---|
| A reliable opening net worth | The government must establish opening net worth “with reasonable certainty,” because the result “depends entirely upon the inclusion in this sum of all assets on hand at the outset” |
| Investigation of your leads | Where you furnish leads “reasonably susceptible of being checked, which, if true, would establish the taxpayer’s innocence,” and the government does not investigate them, “the trial judge may consider them as true and the Government’s case insufficient to go to the jury” |
| A likely taxable source | The government must show “a likely source, from which the jury could reasonably find that the net worth increases sprang.” It is not required to negate every possible non-taxable source where you furnish no leads |
| A specific jury instruction | Charges “should be especially clear, including, in addition to the formal instructions, a summary of the nature of the net worth method, the assumptions on which it rests, and the inferences available both for and against the accused” |
One point of divergence worth knowing. The IRS itself, in IRM 9.5.9.5.1.1, describes Holland as setting out three requirements rather than four, omitting the jury-instruction point from the list. The instruction requirement is in the opinion regardless, and it matters at trial.
Holland also limited the willfulness inference. Where the books “appear correct on their face, an inference of willfulness from net worth increases alone might be unjustified, especially where the circumstances surrounding the deficiency are as consistent with innocent mistake as with willful violation.”
It looks at what you spent rather than what you accumulated, which captures income that was consumed rather than saved.
Judge Goodrich gave the classic description in United States v. Caserta, 199 F.2d 905, 907 (3d Cir. 1952): the analysis “starts with an appraisal of the taxpayer’s net worth situation at the beginning of a period,” and if “his expenditures have exceeded the amount he has returned as income and his net worth at the end of the period is the same as it was at the beginning,” then the return “shows less income than he has in fact received.”
The Supreme Court approved the theory in United States v. Johnson, 319 U.S. 503, 517 (1943), where private expenditures “exceeded his available declared resources.” The civil-side analog is the source-and-application-of-funds method in IRM 4.10.4.5.3.1, which treats any excess of applications over sources as an understatement.
A citation caution, because this one is misreported constantly. The expenditures-method discussion attributed to Taglianetti comes from the First Circuit opinion, 398 F.2d 558, 565 (1st Cir. 1968), not from its later trip to the Supreme Court. That Supreme Court disposition, 394 U.S. 316 (1969), is a one-paragraph per curiam about electronic surveillance disclosure and says nothing about methods of proof.
The government totals your deposits, adds cash you spent without depositing, subtracts identifiable non-income items, and treats the remainder as income.
The premise, from IRM 9.5.9.7, is that unexplained deposits to a bank account may represent unreported income. Eliminate the identifiable non-income sources, and what is left equals corrected gross income.
The classic case is Gleckman v. United States, 80 F.2d 394 (8th Cir. 1935), an Eighth Circuit decision, and the case IRM 4.10.4.5.4.1 itself credits as the origin of the bank deposits method’s four-part test.
Four things have to be shown, per IRM 4.10.4.5.4.1:
IRM 9.5.9.8 treats the cash method as a separate approach and reserves it for last resort: the manual limits it to investigations where “the subject maintains little or no books and records” and the evidence will not support any of the other six methods, and states the method “should only be used as a last resort when other methods of proof are not feasible.” The underlying theory tracks the expenditures method rather than bank deposits: both show an understatement by comparing spending against reported income, where “the source of the funds used to make the expenditures is unexplained.” On the civil side, IRM 4.10.4.5.4 folds this into a single Bank Deposits and Cash Expenditures Method, which is why some summaries list six methods rather than seven; the criminal investigation manual treats them as two.
Both estimate income from operational data rather than financial records, and both are constrained.
Percentage markup applies industry ratios to your purchases or inventory to derive sales and profit. IRM 9.5.9.2.2 restricts it sharply: the method “should only be used as a primary method of proof, on a limited basis, and not used to corroborate other methods.”
Unit and volume multiplies the number of units you handled by the price or profit per unit. IRM 4.10.4.5.6 applies it when records of units sold are reliable and can be verified through third-party sources like shipping records or customer invoices.
On the civil side, the Supreme Court blessed an aggregate-estimate approach in United States v. Fior D’Italia, Inc., 536 U.S. 238 (2002). Where the taxpayer’s markup is known, IRM 4.10.4.5.5 requires the IRS to use it rather than an industry average, which is often the strongest objection available.
It depends entirely on whether the case is civil or criminal, and the two standards are far apart.
The government bears the burden and must meet it with clear and convincing evidence.
Two authorities work together. 26 U.S.C. § 7454(a) places the burden on the Secretary: “In any proceeding involving the issue whether the petitioner has been guilty of fraud with intent to evade tax, the burden of proof in respect of such issue shall be upon the Secretary.” Tax Court Rule 142(b) supplies the quantum: the burden “is to be carried by clear and convincing evidence.”
Cite both. The statute alone does not state the standard, and the rule alone does not state the statutory basis.
Once the government clears that bar as to any part of the underpayment, § 6663(b) treats the entire underpayment as attributable to fraud, except for any portion the taxpayer establishes, by a preponderance of the evidence, is not. The penalty itself is 75% of the portion attributable to fraud, and under § 6663(c) it does not automatically reach a spouse on a joint return.
Beyond a reasonable doubt, on every element, including willfulness.
Holland drew the contrast directly: “Unlike civil actions for the recovery of deficiencies, where the determinations of the Commissioner have prima facie validity, the prosecution must always prove the criminal charge beyond a reasonable doubt.” The constitutional anchor is In re Winship, 397 U.S. 358, 364 (1970).
Willfulness has its own definition. In Cheek v. United States, 498 U.S. 192 (1991), the Court held the standard is “the voluntary, intentional violation of a known legal duty,” and that a good-faith misunderstanding of the law negates it “whether or not the claimed belief or misunderstanding is objectively reasonable.” Our page on what prosecutors must prove in tax evasion cases covers the elements.
Signs the IRS treats as worth developing, organized into six categories. The manual calls them indicators rather than badges.
| Category | Representative indicators from IRM 25.1.2.3 |
|---|---|
| Income | Omitting entire sources of income; inability to explain substantial net worth increases over a period of years; personal expenditures exceeding reported resources; concealing domestic or foreign accounts, brokerage accounts, or digital assets; failing to deposit receipts into a business account contrary to established practice; cashing income checks at check-cashing services |
| Expenses or deductions | Fictitious or substantially overstated deductions; personal expenditures claimed as business expenses; dependency exemptions for nonexistent or self-supporting persons; false documents supporting refundable credits; disguising trust fund loans as expenses |
| Books and records | Multiple sets of books or none; concealment or refusal to produce records; false or back-dated entries; irregularly numbered or altered invoices; amounts on the return not agreeing with the books |
| Allocations of income | Distributing profits to fictitious partners; shifting income or deductions to a related taxpayer where rate differences matter |
| Conduct of the taxpayer | False statements about material facts; attempts to hinder the examination; failing to disclose relevant facts to the return preparer; destruction of records after the examination began; transferring assets to conceal them; a pattern of consistent underreporting over several years |
| Methods of concealment | Placing asset ownership in other names; transfers made in anticipation of assessment; retaining possession or use of a transferred asset; secret bank accounts; deposits into nominee accounts. |
The qualifier matters as much as the list. IRM 25.1.2 states that an indicator “serves as a sign or symptom” and that “no one indicator of fraud is determinative that fraud is indeed present.” An indicator is not an affirmative act, and the manual is clear elsewhere that fraud cannot be established without affirmative acts.
The list now includes digital asset indicators, added in 2024. Cryptocurrency transactions have their own subsection.
Our page on negligence versus tax fraud covers where careless mistakes end, and fraud begins.
Because the Code says it does, in one sentence and with no time limit.
26 U.S.C. § 6501(c)(1): “In the case of a false or fraudulent return with the intent to evade tax, the tax may be assessed, or a proceeding in court for collection of such tax may be begun without assessment, at any time.”
Two neighboring provisions do the same thing. § 6501(c)(2) covers a willful attempt to defeat or evade non-income taxes, and § 6501(c)(3) covers a failure to file.
This is why a fraud allegation on a single old year is more serious than the dollar amount suggests. It reopens the year permanently, and it is what makes the 75% penalty under § 6663 assessable long after an ordinary deficiency would be time-barred.
By attacking the assumptions the method rests on rather than arguing about the total.
Each method has a soft point. A net worth case turns on the opening figure, so cash on hand at the start of the period, loans, gifts, and inheritances are the battleground, and Holland puts a real burden on the government to run down the leads you give it. A bank deposits case turns on whether transfers between your own accounts, loan proceeds, and redeposits were correctly identified as non-income. A percentage markup case turns on whether your actual margin was available and ignored.
Two practical points matter here. Furnishing leads early, specifically and in writing, matters because their evidentiary value under Holland depends on the government having had a fair chance to check them. Taking the willfulness question seriously as a separate issue matters too, because a deficiency the government can compute is not the same as fraud it can prove.
Our page on the penalties for tax fraud covers what is at stake, and whether you can go to jail covers the criminal exposure.
Indirect-method cases are built from explanations. Every account of where the money came from becomes a lead the government tests, and an inconsistent one becomes evidence.
Ayar Law handles civil fraud examinations and criminal tax defense across Michigan from offices in Farmington Hills and Grand Rapids, and represents taxpayers facing IRS fraud allegations nationwide. Call (248) 262-3400 to request a confidential case review, or contact us through our website.
Last updated August 2026.
This page is general information about federal tax procedure and criminal tax law. It is not legal advice, and reading it does not create an attorney-client relationship. Case law and IRS internal guidance change, and how any method of proof applies depends entirely on the facts. Prior results do not predict future outcomes. If you are under examination and fraud is raised, speak with a licensed tax attorney before providing further information.
