These appeals form the Tax Court question the propriety of the application here of the net worth method of reconstruction of income and the determination that the Commissioner of Internal Revenue satisfied his burden of proving fraud by the taxpayers.

Taxpayers, owners of two grocery stores in western Pennsylvania,1 petitioned the Tax Court for a redetermination of their federal income tax liabilities after the Commissioner, applying the net worth reconstruction of income method, determined that there had been unreported income, deficiencies in tax, and additions to tax due for fraud. The Tax Court, also using net worth computation, sustained substantial deficiency and fraud penalties for several of the years reviewed by the Commissioner.2

There is no disagreement that taxpayers maintained substantial personal cash reserves.3 Indeed, the net worth method of reconstruction of income was utilized because only a portion of the gross receipts from taxpayers' stores was deposited in bank accounts, there was excessive commingling of taxpayers' personal and business funds, and, at best, only informal bookkeeping methods were employed. Thus, it is the size and source of the cash reserves maintained by taxpayers which underlie this dispute.

This court has heretofore described the net worth income computation formula: the "increase in net worth plus non-deductible disbursements minus non-taxable receipts equals taxable net income." Clark v. Commissioner of Internal Revenue, 253 F.2d 745, 747 (3d Cir. 1958). In Holland v. United States, 348 U.S. 121, 75 S. Ct. 127, 99 L. Ed. 150 (1954), the Supreme Court placed its imprimatur on the use of the net worth method under certain restricted circumstances.4 The compliance with three requirements will legitimate its implementation: first, the Commissioner must reliably establish a net worth for taxpayer as of the beginning of the period under review; second, investigation is required of the taxpayer's explanation for his net worth increases, if such "leads" are "reasonably susceptible of being checked," Holland, supra, 348 U.S. at 135-136, 75 S. Ct. 127;5 finally, the government must suggest a "likely source" of the unreported income.

The Tax Court adopted the Commissioner's finding that the taxpayers' stores constituted the "likely source" of the unreported income. Thus, the validity of the net worth reconstruction truned on a finding that the inventory turnover and gross profit margins were sufficient to generate the quantum of income found to be understated in the returns filed during the years in question. Gross receipts in each store averaged between $500,000 and $1,000,000. The Tax Court found that the gross profit margin of the two stores was fourteen (14%) per cent, and that there was an inventory turnover "generally in excess of 20 to 1." These findings are presumptively correct, and we cannot disturb them unless we find them to be clearly erroneous. Vaira v. Commissioner of Internal Revenue, 444 F.2d 770, 774 (3d Cir. 1971); Farcasanu v. Commissioner of Internal Revenue, 140 U.S.App.D.C. 398, 436 F.2d 146, 148 (1970). We have no difficulty in accepting the Tax Court's determinations. Indeed, taxpayer Lena Runzo testified that the gross profit margin ranged from seventeen (17%) per cent to twenty (20%) per cent, and taxpayers' returns disclosed turnover rates ranging from 18 to 1 up to 56 to 1. Thus, we will not disturb the finding that the grocery stores constituted a "likely source" of the unreported income. See United States v. Hom Ming Dong, 436 F.2d 1237, 1239-1242 (9th Cir. 1971).

In reconstructing taxpayers' annual incomes from 1956 to 1963, the Commissioner found that there had been an overstatement of income in some years during the period. Aware that some asset was not being recognized, both the Commissioner and the Tax Court assumed the asset to be cash income, and plugged the imputed income into their net worth computations. Appellants argue that it would be as consistent to attribute the asset not being recognized to personal cash which the taxpayers urged they had on hand as of January 1, 1956, as it would be to accept the imputations of the Commissioner and the Tax Court. Moreover, appellants contend that the substantial discrepancies between income computed and income reported uncovered in each of these years renders the entire accounting method, as applied to these taxpayers, a complete "mockery."

Though taxpayers seek shelter in Holland, their arguments provide them no safe passage.6 Whether it be as consistent to conclude that the source of the unrecognized asset is that suggested by the taxpayers, we emphasize that the respective theories cannot at this juncture be weighed in apposition: "[o]ur review of such findings is narrowly limited, for the findings of the Tax Court are presumptively correct." Vaira, supra, 444 F.2d at 774. And though the findings by the Commissioner and the Tax Court may, when based on mere inference, generate difficult problems relating to burden of proof in criminal matters, Holland, supra, 348 U.S. at 126, 75 S. Ct. 127, the Court there expressly recognized that no such impediments inhere in civil proceedings for the recovery of deficiencies.7 Indeed, upon a careful review of the record, we find no reason to reject the Tax Court's refusal completely to credit taxpayers' accounts of their personal cash reserves as of the beginning of the period. See Marcello v. Commissioner of Internal Revenue, 380 F.2d 509, 511 (5th Cir. 1967); 2 Rabkin & Johnson, Federal Income Gift and Estate Taxation, Sec. 12.03A, p. 1231a. In any event, the Tax Court correctly recognized that any error in computation for those years challenged by appellants could only inure to taxpayers' benefit; had the Commissioner or the Tax Court failed to acknowledge the asset not being recognized, the net worth increase for any subsequent year would increase the taxpayers' liabilities in each of the years immediately following those in which the income was found to be overstated.

We must reject as well appellants' contention that the magnitude of the discrepancies between income computed and income reported in the years in which income was found to be overstated is so substantial that the computations for the entire period are rendered defective. As the government suggests, where income for a period of successive years is subjected to retrospective examination, use of the net worth method of computation may indeed produce unlikely variations in attributed income as between individual years. But as the Fifth Circuit noted in Conway v. United States, 278 F.2d 710, 712 (1st Cir. 1960), only the weight of the testimony is affected thereby:

Risk of anachronous allocation is inherent in any case where a taxpayer fails to maintain full and accurate records. He cannot be heard to complain if the record shows sufficient basis for the findings, in spite of possible criticism of the pattern of the result. * * * In these circumstances we would not hold plainly wrong a rule of thumb allocating the income to the year in which it first came to light.

We hold that use of the net worth method was fully warranted by the circumstances of this case. Holland, supra, 348 U.S. at 125, 75 S. Ct. 127; see United States v. Hom Ming Dong, supra, 436 F.2d at 1240. Its implementation here adequately comported with the procedural safeguards required by Holland. Hence, the deficiency findings of the Tax Court are not clearly erroneous.

Although the Commissioner's findings usually are presumptively correct, fraud must be proved to the Tax Court by clear and convincing evidence. Valetti v. Commissioner of Internal Revenue, 260 F.2d 185, 188 (3rd Cir. 1958); Goldberg v. Commissioner of Internal Revenue, 239 F.2d 316 (5th Cir. 1956). Nonetheless, "because the question of fraud is an issue of fact, the Tax Court's findings settle the matter unless they are clearly erroneous." Estate of Upshaw v. Commissioner of Internal Revenue, 416 F.2d 737, 741 (7th Cir. 1969).8 As the court in Upshaw there noted:

Some of the factors the Tax Court may take into consideration in determining whether the taxpayer was guilty of fraud with intent to evade taxes are: persistent understatement of income over a period of years; failure to keep adequate records and books; misstatement of facts to a revenue agent and concealing ownership of property. All of these factors are present here.9

Here, too, there is a series of findings suggesting fraud: repeated and substantial understatement of income; an abject failure to keep adequate books and records; misstatements to revenue officers regarding cash reserves at the beginning of the investigated period; excessive dealings in suspiciously large amounts of cash; and a basic lack of credibility in taxpayers' account of the alleged cash hoard. Taxpayers argue that all the indicia of fraud in this case establish no more than "averments" or "suspicion" of deliberate understatement of income with the intent to defraud the government. But there is more than ample evidence to satisfy the rigorous standards heretofore rehearsed. And as the Fifth Circuit noted in Webb v. Commissioner of Internal Revenue, 394 F.2d 366, 380 (5th Cir. 1968):

Tax evaders seldom leave tracks and therefore circumstances can be convincing. Though an isolated erroneous tax figure cannot be escalated or pyramided into fraud, a confraternity of similar errors can take on more sinister tax aspects.

The decisions of the Tax Court will be affirmed.