In June 1957, Joseph Cirillo and his wife, Martha, jointly filed their first income tax returns for the years 1945 to 1954, inclusive. They acted after having learned that revenue agents were investigating the husband’s tax liability. Throughout the years in question, the husband had received both a regular salary as a municipal employee, from which appropriate sums had' been withheld on account of income taxes, and modest fees earned in the part-time practice of law. The wife had received no income. The Commissioner disallowed certain deductions claimed in the delinquent returns and imposed penalties. The controversy ultimately reached the Tax Court,, which sustained the deficiencies determined by the Commissioner and held that part of each annual deficiency had resulted from fraud with intent to evade taxes. The Tax Court held both spouses liable for the amounts of the deficiencies not covered by salary withholdings and imposed an additional 50% fraud penalty for *481each year upon the husband alone.1 T.C. Memo 1961-192. The case is here on cross petitions wherein the spouses urge that the fraud penalty was both unjustified and erroneously computed, while the Commissioner claims that fraud penalties should have been imposed upon both husband and wife.
Before the Tax Court, Joseph Cirillo, whom we shall call the taxpayer, testified that he had failed to file timely returns, not because of any intent to defraud, but because he believed, on the basis of a summary mental calculation made at the end of each year, that the with-holdings from his salary alone were enough to cover all income taxes payable •on both his salary and his small net income from private law practice. Cf. First Trust & Sav. Bank v. United States, 8th Cir. 1953, 206 F.2d 97. However, he concedes that he kept no systematic records of receipts and expenditures in the course of his practice and made no detailed computation of his tax liability at the end of each taxable year. In the absence of such records, he urges that the good faith and reasonableness of his asserted belief that his taxes were covered by salary withholdings are substantiated by the subsequent detailed computations which appear in the delinquent returns filed in 1957. In summary, the tax liabilities, withholdings and consequent obligations and overpayments shown on those returns are as follows:
Although the Tax Court disallowed about half of the deductions claimed in these returns, the taxpayer relies on the court’s finding that “there is no contention or showing that the delinquent returns involved here are fraudulent”.
*482The taxpayer does not challenge the existence of deficiencies or their amounts as found by the Tax Court. Therefore, we have to consider only whether and to what extent fraud penalties were justified.
The evidence in this case strongly indicates that the taxpayer’s failure to file returns was a willful neglect of a statutory duty. As a lawyer, he must have been aware that one who earns several thousand dollars a year is obligated to file income tax returns. Each year he was reminded of his duty to file a return by the standard W-2 Form which he received from his employer. Moreover, he has been convicted of the misdemeanor of willful failure to file income tax returns for the years 1953 and 1954. United States v. Cirillo, 3d Cir., 1957, 251 F.2d 638, cert. denied, 1958, 356 U.S. 949, 78 S.Ct. 914, 2 L.Ed.2d 843.2
But willful failure to file a timely return, which may create both criminal liability and an additional civil liability, does not in itself and without more establish liability for a fraud penalty, though it may be relevant in that connection. Jones v. Commissioner, 5th Cir., 1958, 259 F.2d 300. This is true because a fraud penalty can be imposed only where proof of a deficiency is supplemented by proof that the “deficiency is due to fraud with intent to evade tax”. Int.Rev.Code of 1939, § 293(b), ch. 2, 53 Stat. 88; cf. Int.Rev.Code of 1954, § 6653(b). First Trust & Sav. Bank v. United States, supra. The critical question is whether the circumstances attending a particular failure to file warrant an inference of intention to evade taxes. Moreover, the evidence must be evaluated in the light of the settled rule that fraud can be established only by clear and convincing proof or, as we have put it, “by something impressively more than a slight preponderance of evidence”. Valetti v. Commissioner, 3d Cir., 1958, 260 F.2d 185, 188. Accord, Goldberg v. Commissioner, 5th Cir., 1956, 239 F.2d 316. To justify a fraud penalty the circumstances surrounding the failure to file returns must strongly and unequivocally indicate an intention to avoid the payment of taxes. Powell v. Granquist, 9th Cir., 1958, 252 F.2d 56; cf. Bender v. Commissioner, 7th Cir., 1958, 256 F.2d 771.
Taxpayer contends that his willful failure to file was not and could not have been the result of a scheme to cheat the government out of taxes which he was obligated to pay, because he did not believe that he owed any taxes. We are aware that the Tax Court disbelieved this testimony and that it is not for this appellate court to assess taxpayer’s credibility as a witness. However, the Commissioner’s heavy burden of proof on the issue of fraud cannot be satisfied by mere disbelief of taxpayer’s testimony. The record must contain some convincing affirmative indication of the required specific intent.
Taxpayer’s delinquent returns lend considerable credibility to his testimony concerning the years from 1945 through 1951. Taking into account the expenses which he believed he was entitled to deduct plus the credit for income taxes withheld, taxpayer calculated that he had overpaid his taxes for each of *483those years by amounts in the order of fifty to one hundred seventy-five dollars. These delinquent returns were not fraudulent, and there is no showing or argument that any of the deductions claimed but disallowed were fictitious. Only because taxpayer was unable to substantiate them with the kind of records which the Commissioner properly demanded3 was it determined that he owed additional taxes for each of these years except 1948. And for that year it is agreed that he owes nothing. All of this corroborates taxpayer’s testimony that at the end of each of these years he believed that he owed the government no taxes beyond what had already been withheld from his salary.
In the face of this evidence, the Commissioner offered little to sustain his burden of proving fraud for the years 1945 to 1951. Apart from evidence which, as already pointed out, builds a very strong ease of willful failure to file without demonstrating the intent which accompanied that failure, the only evidence offered by the Commissioner was the taxpayer’s failure to keep systematic and detailed records of income and expenditures of his law practice. The significance of this omission is minimized by the fact that from 1945 through 1951, taxpayer’s practice was very small, never yielding receipts in excess of $2200 in any year and, in several years, yielding less than $1000. Indeed, government agents conceded that, even without such records, they were able to determine the amount of taxpayer’s earnings from bank deposit statements retained by him, and that there was no indication that taxpayer had received additional income not reflected in these statements. Thus, the only needed information which was lacking because of the taxpayer’s failure to keep books was a record of the expenditures which he made in the regular course of his law practice. In these circumstances, any inference that taxpayer’s unbusinesslike procedure was intended to conceal tax liability is too weak to achieve the clear and convincing character which proof of fraud must exhibit.
The years 1952, 1953 and 1954 present a different picture. In 1952, taxpayer’s gross income from his law practice increased very substantially to $4,169.70, four times as much as it had been in any year before 1951 and twice as much as it had been in 1951, while his deductible expenses did not increase proportionately. Moreover, withholdings did not increase substantially. As a consequence, the tax withheld from taxpayer’s salary in 1952 was more than $500 short of covering his total tax liability. Even his own calculations for 1952, made several years later, failed to disclose an overpayment such as those he consistently claimed for the years before 1952. And for 1953 and 1954, taxpayer’s delinquent returns admitted substantial deficiencies. Thus, the record indicates that at the end of 1952 and thereafter, even a rough calculation, honestly made, would at least have shown the taxpayer that there was need for a more careful analysis of income and expenditures to substantiate or dissipate his hope that no additional taxes were due. His disingenuous avoidance of accurate knowledge when the need for such knowledge must have been apparent was in itself a substantial indication of fraudulent intent.
We conclude, therefore, that the evidence of fraud was sufficiently clear and convincing to justify the imposition of fraud penalties for the years 1952, 1953 and 1954. Not so, however, for the years from 1945 through 1951.
Our next consideration is the method employed by the Tax Court in computing fraud penalties. Section 293(b) of the 1939 Code provides that “if any part of any deficiency is due to fraud”, a fraud penalty shall be imposed at the rate of “50 per centum of the total amount of the deficiency”. In this case the Tax Court added 50% to the taxpayer’s total liability for each year as a fraud penalty, without having subtracted either *484the amount withheld from salary or the amount of tax reported on the delinquent return.
In section 271(a) as amended, 58 Stat. 245 (1944), “deficiency” is defined as “the amount by which the tax imposed by this chapter exceeds * * * the amount shown as the tax by the taxpayer upon his return, if a return was made * * Section 271(b) (1) explicitly provides that “[t]he tax imposed by this chapter and the tax shown on the return shall both be determined -* * •» without regard to the credit under section 35 [for amounts withheld from wages] * * This language makes it clear that neither the existence nor the amount of the “deficiency” of a taxpayer who has failed to file a return is affected in any way by the existence of a withholding credit partially or fully offsetting his tax liability. Moreover, in the circumstances of this case the “total amount of the deficiency”, upon which the fraud penalty is based, must be the entire amount of the “tax imposed” for the year unless, under the section 271 definition of deficiency, it is permissible to deduct from that total tax liability the amount of tax shown on the delinquent return filed several years after the due date. We think this is not permissible.
1952 is the first year as to which we have concluded that fraud was proved. When the deadline for filing a 1952 tax return passed, a “deficiency” in the amount of “the tax imposed” came into existence. Because no return had then been filed, there simply was no “amount shown as the tax by the taxpayer upon his return” to be deducted from the total tax liability in computing the “deficiency”. And since the failure to file was fraudulent the taxpayer’s liability at that time included a fraud penalty measured by that deficiency. Thus, taxpayer has to take the position that although, from 1953 until his delinquent filing in 1957, he was liable for a 1952 fraud penalty measured by his total 1952 tax liability, he could terminate that liability for fraud by filing a delinquent return admitting his full tax liability. It is not surprising that the Tax Court has consistently refused to give any such effect to a delinquent filing, reasoning that the “return” contemplated by section 271 is a timely return. Charles F. Bennett, 1958, 30 T.C. 114; Maitland A. Wilson, 1946, 7 T.C. 392; see George M. Still, Inc., 1953, 19 T.C. 1072, aff’d mem., 2d Cir., 1955, 218 F.2d 639 (collection of fraudulent return by amended return does not affect computation of fraud penalty) ; cf. Simon v. Commissioner, 8th Cir., 1957, 248 F.2d 869 (carryback deduction does not affect computation of fraud penalty). We think this conclusion is consistent with the language and the scheme of the 1939 Code. It validates the Tax Court’s computation of 1952 and 1953 fraud penalties. The parties recognize that section 6653(c) (1) of the 1954 Code, which is applicable to the 1954 fraud penalty, has now explicitly provided that only a timely return is to be considered in determining the existence and amount of the underpayment which is the measure of the fraud penalty.
Finally, the Commissioner asks us to reverse the Tax Court’s holding that Martha Cirillo, unlike her husband, is not liable for fraud penalties. The Commissioner argues that the wife’s action in joining her husband in filing non-fraudulent delinquent returns for the years in question was sufficient to make her too liable for the fraud penalties.
Earlier cases have decided that a wife who is a party to a fraudulent joint return may be held liable for the fraud penalties assessed on account of her husband’s fraud in preparing that return, notwithstanding that she herself had no income, did not entertain any fraudulent intent, and, indeed, did not know that the return was fraudulent. Estate of Ginsberg v. Commissioner, 5 th Cir., 1959, 271 F.2d 511; Furnish v. Commissioner, 9th Cir., 1958, 262 F.2d 727, 731-734; Kann v. Commissioner, 3d Cir., 1953, 210 F.2d 247, cert. denied 1954, 347 U.S. 967, 74 S.Ct. 778, 98 L.Ed. 1109; Boyett v. Commissioner, 5th Cir. 1953, 204 F.2d 205; Howell v. Commis*485sioner, 6th Cir., 1949, 175 F.2d 240; Meyer J. Safra, 1958, 30 T.C. 1026, 1037. The feature which distinguishes the present case from those earlier cases and makes it one of first impression is that the joint returns here involved were themselves not fraudulent; rather fraud occurred and penalties attached earlier when the husband, with fraudulent intent, failed to file timely returns.
Whether this distinction warrants a different result depends upon the interpretation given to section 51(b) (1) of the 1939 Code, ch. 2, 53 Stat. 27, as amended, 62 Stat. 115 (1948) (now Int. Rev.Code of 1954, § 6013(d) (3)), which provides that where a joint return is filed “liability with respect to the tax” shall be joint and several. It is reasonable to view “the tax” for which the return-signing wife is liable as including both the amount stated in the joint return and any deficiency assessments on account of the incorrectness, inadequacy or bad faith of that filing. If a wife, however innocently, joins in a fraudulent return, any additional assessment for fraud is a tax obligation created by that filing and measured by the difference between the tax that should have been reported and the amount the spouses jointly reported. This rationale supports the cited cases involving fraudulent joint returns.
But in the present case the fraud penalty is neither imposed for nor measured by any deficiency based upon the joint return. Indeed, we have already pointed out in another connection that the deficiency upon which the present fraud penalty is based arose when the husband, with intent to evade taxes, failed to file a timely return. That deficiency did not subject the wife to any liability. Joseph A. Mundy, 1955, 14 CCH Tax Ct.Mem. 1067. In these circumstances, the fraud penalty is in no way dependent upon the joint return. For this reason it may reasonably be regarded as distinct from the tax liability which section 51(b) (1) imposes as a consequence of signing a joint return. Of course, this analysis leaves a wife who elects to sign a joint return liable for the tax shown on the return and any deficiencies which may be determined in connection with that filing. Neither the language of the statute nor any consideration of equity or tax policy provides a persuasive reason for imposing a broader liability. Accordingly, we agree with the Tax Court that Mrs. Cirillo did not become liable for fraud penalties.
So much of the Tax Court’s decision as imposed fraud penalties for the years 1945 to 1951, inclusive, will be reversed. Otherwise, the decision will be affirmed.
. The Tax Court’s computation was as follows:
* For each year the additional fraud penalty was 50% of the amount shown in. this column.
. Taxpayer complains that the admission into evidence of the record of his conviction of willful failure to file returns was error, and that, in any event, the court should have allowed him to show the limited scope of the evidence on which that conviction was based. He is mistaken. His conviction was admissible as evidence of the fact that his failure to file returns for the two years in question was willful, Richard F. Smith, 1958, 31 T.C. 1, 9; see Stagecrafters’ Club, Inc. v. District of Columbia Division of American Legion, D.D.C.1953, 111 F.Supp. 127, aff’d mem., 94 U.S.App.D.C. 74, 211 F. 2d 811; cf. Masters v. Commissioner, 3d Cir., 1957, 243 F.2d 335, 338-339 (admissibility of conviction on nolo contendere plea), if not as a conclusive determination of that fact, compare Meyer J. Safra, 1958, 30 T.C. 1026, 1034-35, nonacq., 1962 Int.Rev.Bull. No. 43, at 8, with Emich Motors Corp. v. General Motors Corp., 1951, 340 U.S. 558, 568-569, 71 S.Ct. 408, 95 L.Ed. 534 (dictum).
. Of course, tlie burden of proof on this issue was on the taxpayer. Burka v. Commissioner, 4th Cir., 1950, 179 F.2d 483.