(dissenting as to Point 2).

The difficulties in finding workable applications of the Belo principle are manifested not only in the varying views of the lower courts and the continuing criticisms of text writers, but also in the Supreme Court itself, where it has been constantly distinguished away except for somewhat hesitant support in a single case again emphasizing the sharp division in the Court. Walling v. Halliburton Oil Well Cementing Co., 331 U.S. 17,26,27, 67 S.Ct. 1056. Thus, the Be-lo doctrine has a shaky foundation; while thoroughly accepted is the principle that the actual hourly wage governs, and not the one artificially contrived to perpetuate the pre-statutory wage scale. Walling v. Helmerich & Payne, 323 U.S. 37, 41, 42, 65 S.Ct. 11, 89 L.Ed. 29; Walling v. Youngerman-Reynolds Hardwood Co., 325 U.S. 419, 424, 65 S.Ct. 1242, 89 L.Ed. 1705; 149 Madison Ave. Corp. v. Asselta, 331 U.S. 199, 204, 67 S.Ct. 1178. And the Halliburton case shows that for the Belo case now to govern, there must be a finding that the assumed rate was the actual rate in the light of facts and circumstances making such an inference a rational and reasonable one. It is significant that both cases affirmed — that is, did not feel compelled to reverse — findings made and sustained below. We should not extend the Belo scope.

Here, as the trial court stated in its opinion, what “for all practical purposes” was the same arrangement as the “guaranteed salary plan” adopted in 1940 actually came into existence for a number of employees “after the effective date of the Act, October, 1938,” i. e., to meet its supposed requirements. Moreover, there were changes made in the formula to make it fit more nearly the former wage, and various other anomalies showing that in reality the agreed salary controlled the nominal rate. So the district judge said, referring to those employees who lost by the change to the assumed hourly rate: “It, therefore, becomes evident that as to such employees the regular hourly rate, and consequently the overtime rate were illegally computed and im*678properly applied.” And he adds this apt quotation: “ ‘It was derived not from the actual hours and wages but from ingenious mathematical manipulations, with the sole purpose being to perpetuate the pre-statutory wage scale.’ Walling v. Helmerich & Payne, Inc., 323 U.S. 37, at page 41, 65 S. Ct. 11, 89 L.Ed. 29.”

This conclusion seems to me inescapable on the evidence; the only error the judge made, in my judgment, was in believing that this artificiality of rate was cured as to those who either received at least as much as before or came on the pay roll later. As is well settled, the employer’s honest belief that he is complying with the law is not a defense. Cases cited supra.

The suggestion in the opinion, made the foundation for reversal here, that the fluctuations in wages but for the guarantee would have been greater than in the Belo case cannot be accepted as either accurate or controlling. There was no finding on this issue below and almost no discussion in the briefs here except for the Administrator’s argument by way of conclusion that the record “suggests a regularity of normal working hours readily susceptible to adjustment to the statutory policy.” The table relied on in the opinion is made up from an exhibit introduced by plaintiff for another purpose. It was never claimed to be, and obviously is not, a fair sample for the use here made. Indeed, it represents only about 10 per cent of the time worked on the guarantee. And defendant’s own statement compiled from the actual evidence demonstrates that the time worked less than the statutory maximum was very substantially under 12%, instead of the 20% stated in the table.1 We can take it as clear on this record that the variation was rather less than half (not more than) that asserted for the Belo case.2 Moreover, we can easily overestimate the force of delusively simple numbers.3 At most this would be but one element out of a complex total supporting the inference of “actuality,” rather than “artificiality.” Fact finding at best is exceedingly difficult; neither it nor the Belo principle can be confined within the bounds of a mathematical chart of wage deviations from an assumed norm. In re Fried, 2 Cir., 161 F.2d 453, 462-464; Frank, A Plea for Lawyer Schools, 56 Yale L. J. 1303, 1307, 1308, 1327. Actually the facts here destroy the base for the Belo doctrine.

On Petition and Cross-Petition for Rehearing.

PER CURIAM.

Petitions denied.

This is an action for an injunction to prevent defendant from violating the Fair Labor Standards Act, 29 U.S.CA.. § 201 et seq., with respect to 59 of defendant’s employees out of a total which varies from 2,200 to 3,000. As the district judge found, defendant, which manufactures, sells and distributes textile products, operates nine mills; it “produces a fluctuating percentage of its capacity, depending upon factors which seem to be common to the textile industry.”

1. As to 14 of the 59 employees in question, defendant asserted in the district court, that they were not within the coverage of the Act because they serve in a “bona fide executive * * * capacity” within the meaning of § 13(a) (1) and the Regulations *672made, pursuant thereto, by the Administrator.1 *The district judge held that but one of these persons was within the exemption. Defendant appeals from that part of the judgment against it which relates to 5 of the other 13 employees.

The district court found as a fact that defendant had not affirmatively proved that, of these five, two, who are mill foremen, “spent not to exceed 20 percent of the number of hours worked in the workweek by the non-exempt employees under their direction in doing work of the same nature as that performed by non-exempt employees.” As there is ample evidence to support this finding, it is not “clearly erroneous.” Accordingly, defendant, which had the burden of proof, did not show that these employees came under clause (f) of the pertinent Regulations. As to the other three employees, operating engineers in defendant’s power plant, the district court found that they “were operating engineers or turbine operators, whose primary duty was the operation and maintenance of important, valuable and complicated machinery. Supervision of firemen and other employees was secondary to their main duty. Their primary duty does not consist in the management of a department of the establishment in which they are employed.”

“Sec. 541.1 Executive. The term ‘employee employed in a bona fide executive * * * capacity’ in Section 13(a) (1) of the Act shall mean any employee,

“(a) whose primary duty consists of the management of the establishment in which he is employed or of a customarily recognized department or subdivision thereof and

“(b) who customarily and regularly directs the work of other employees therein, and

“(c) who has the authority to hire or fire other employees or whose suggestions and recommendations as to the hiring or firing and as to the advancement and promotion or any other change of status of other employees will be given particular weight, and

“(d) who customarily and regularly exercises discretionary powers, and

“(e) who is compensated for his services on a salary basis at not less than $30 per week (exclusive of board, lodging, or other facilities), and

On the evidence, this finding must be sustained. Consequently, these men did not come under clause (a) of the pertinent Regulations. As to these five employees, the judgment is affirmed.

2. Forty-five other employees are paid guaranteed salaries under a so-called guaranteed-salary plan. This plan, which came into existence in October or November, 1940, is expressed in letters signed by these persons, of which the following is typical: “This will confirm the previous verbal understanding between us relative to my wages. I am receiving a guaranteed weekly salary of $59 per week for a scheduled work week of 45 hours. My hourly rate is $1.242 per hour for the first 40 hours and time and a half or $1.863 per hour for each hour worked in excess of 40 hours per week. If I work less than my scheduled 45 hours in any work week, I receive my guaranteed weekly salary of $59. If I work over my scheduled 45 hours in any work week, I receive $1.863 per hour in addition to my guaranteed weekly, salary of $59. In addition to the above base and overtime pay, I receive any production bonus to which I may be entitled.” 2 Plaintiff contends that payments under this plan do not comply with the overtime requirements of § 7(a) of the Act.3 ****The district court held that *673the plan did not meet the statutory requirements in so far as it applied to any of the 45 employees whose average hourly earnings, before they were “placed on the plan,” were greater than the hourly rates under the plan, but that the plan was valid as to the remainder of the 45, whether they had previously worked for the defendant or were first hired after the plan became effective. The court entered judgment based on these conclusions. Defendant appeals from that portion of the judgment holding the plan partly invalid, and plaintiff cross-appeals from that portion holding the plan partly valid.

*672“(f) whose hours of work of the same nature as that performed by non-exempt employees do not exceed 20 per cent of the number of hours worked in the work week by the non-exempt employees under his direction: Provided, That this paragraph shall not apply in the case of an employee who is in sole charge of an independent establishment or a physically separated branch establishment.”

*673Our decision must turn on the applicability of Walling v. A. H. Belo Corp., 316 U.S. 624, 62 S.Ct. 1223, 86 L.Ed. 1716, in which the Court, because of a guarantee, significantly refused to apply the doctrine, and the “regular rate” formula, announced the same day in Overnight Motor Transp. Co. v. Missel, 316 U.S. 572, 62 S.Ct. 1216, 86 L.Ed. 1682. Before the recent decisions in Walling v. Halliburton Oil Well Cementing Co., 331 U.S. 17, 67 S.Ct. 1056, and 149 Madison Ave. Corp. v. Asselta, 331 U.S. 199, 67 S.Ct. 1178, several circuit courts, including this court, had intimated that, in the light of decisions subsequent to Belo— i. e., Walling v. Helmerich & Payne, 323 U.S. 37, Walling v. Youngerman-Reynolds Hardwood Co., 325 U.S. 419, 65 S.Ct. 1242, 89 L.Ed. 1705; Walling v. Harnischfeger Corp., 325 U.S. 427, 65 S.Ct. 1246, 89 L.Ed. 1711, 65 S.Ct. 11, 89 L.Ed. 29; — the Supreme Court had narrowed the Belo doctrine almost to the vanishing point.4 But the opinion in Halliburton, and comments on Belo and Halliburton in Asselta, showed that this view was mistaken. As an intermediate appellate court, occupying a satellite position with a restricted orbit, it is our function to interpret not the statute directly but the Supreme Court’s interpretation of the statute.5 We must, therefore, now take the Belo doctrine as an established gloss on the Act, one which constitutes an exception to the usual rule — as to the actual “regular rate” — announced in the Missel, Helmerich & Payne, Youngerman-Reynolds and Harnischfeger cases. In other words, a contract rate which, in line with those cases, would otherwise be deemed an artificial regular rate, is not so when the contract provides for a guarantee a la Belo, So here, several facts which undoubtedly would have invalidated the contracts will not do so, if, because of the guarantee, the contracts come within the Belo exception.6

We incline to believe that that exception has these limits indicated in the Belo opinion (316 U.S. 635, 62 S.Ct. 1229, 86 L.Ed. 1716) :7 The mere fact of a guaranteed wage does not suffice; in addition, there must be, as there was in Belo and Halliburton, a condition of irregularity or instability of work, so that the guaranty yields the employees a stability of employment and income otherwise absent. The Su*674preme Court has not explicitly stated the needed quantum of irregularity, and we must ascertain it with the aid of such guides as the Court has implicitly provided. Since the Court has unequivocally said that Belo was correctly decided, we believe that, if the irregularity in any case is at least as great as in Belo, a fair guarantee will bring that case within the exception.

Whether this case is thus exceptional is a question of fact. In Belo, there was no finding by the district court of the fact of instability, nor did the Supreme Court state any of the details of the evidence pertinent thereto. We have examined the Belo record for such evidence and have set it forth in the Appendix to this opinion; that evidence consists solely of a sampling of time sheets. In the instant case, the trial judge did not discuss this question, nor did he make any finding which bears on it.8 All the evidence here of a kind similar to that in the Belo record we have also set forth in our Appendix. As that evidence is entirely documentary, no issue of witness’ credibility arises; therefore, we can pass on the facts as well as could the trial judge,9 and need not remand for a finding by him. In comparing Belo and this case, we have taken as a yardstick the number of weeks during which the employees, in the respective cases, worked less hours than the statutory minimum (i. e., 44 in Belo and 40 here). On the basis of that comparison, we conclude that the irregularity here was somewhat greater than in Belo. Consequently, and because we consider the guarantee here fair in the circumstances, we hold that the contracts here meet the statutory requirement as construed by the Supreme Court.10 We may add that, in the record in this case, there is an exhibit showing that employees, under the plan, in 12 percent of the total weeks worked less than the scheduled (i. e., contract) hours;11 this, in and of itself, indicates considerable instability; but, as no such comparable total figures appear in the Belo record, we are unable to make a comparison in that respect between the two cases.

Plaintiff makes much of a sentence in the Asselta opinion (331 U.S. at pages 209, 210, 67 S.Ct. at page 1184), in which the Court said that the Belo-Halliburton exception does not apply unless there is “provision for a guaranteed weekly wage with a stipulation of an hourly rate which under the circumstances presented could properly be regarded as the actual regular rate of pay.” We do not agree with plaintiff that this means that the hourly rate must be identical with the actual regular rate in accord with Missel, Asselta, Helmerich & Payne, and Youngerman-Reynolds. We think it means that, where there concur (a) enough irregularity and (b) a fair guarantee, then the contract rate is to be “regarded as” the equivalent of the actual rate, i. e., as a substitute therefor. Plaintiff also seeks to differentiate the instant case from Belo and Halliburton on the ground that the agreements there provided for overtime compensation at “not less than one-and-one-half times” the contractual hourly rates, while here that phrase is not in the contracts. We see nothing but a verbal difference, resulting in no practical, distinguishing consequences. In the case at bar, as in those two other cases, the employee is paid at the fixed overtime rate of 150% of the contract rate.

We do not agree with the district court that the plan is invalid as to those employees who had previously received hourly earnings greater than the hourly rates they received under the plan. In both Belo and Halliburton, the guarantee plans resulted in some such reductions.

*675To the extent that it held the plan partly invalid, the judgment is reversed; otherwise it is affirmed.

Appendix

1. Instability of employment in Belo Case and in the case at bar.

Belo casei

Total Weeks Worked

Weeks worked 44 hours or over

Weeks worked under 44 hours

Anderson 33 26 7

Bagwell 30 28 2

Bates 29 27 2

Batts 30 30 0

Brewer 28 24 4

Burleson 24 23 1

Cornkle 24 15 9

Crume 29 17 12

Eastus 44 37 7

Fair 34 34 0

Greeley 30 28 2

Lewis 29 19 10

Lockart 26 20 6

Lovell 31 IS 13

Martin 28 11 17

Mills 30 28 2

Moffet 30 28 2

Monroe 30 29 1

Montigue 28 25 3

Morrow 28 18 10

Newman 26 23 3

Prasifka 32 32 0

Simmons 18 18 0

Snodgrass 34 33 1

Vickers 30 16 14

West 30 22 8

Wilson 25 16 9

Walker 30 28 2

— — .—

Total 820 673 147

Percentage 100% 82% CO 3*

Utica Knitting Mills:

Total Weeks Weeks Weeks worked worked Worked 40 hours under or over 40 hours

Ahrens 25 20 5

Angelhow 30 23 7

Atkinson 20 16 4

Blackowiak 15 .14 1

Total Weeks Worked

Weeks worked 40 hours or over

Weeks worked under 40 horns

Bronk 31 26 5

Cipriano 21 19 2

Clark 22 19 3

Cunningham 24 22 2

Davies 20 19 1

Fanelli Z1 19 2

Giglio 16 15 1

Haight 19 18 1

Haven , 22 21 1

Heckert 20 20 0

Helliwell 19 17 2

Hess •29 23 6

Hitchcock 24 21 3

Jerzak 24 20 4

C. Julian 9 6 3

M. Julian 19 14 5

Keys 19 14 5

Laney 13 13 0

Lannuti 5 4 1

Leuthauser 10 5 5

McCann 33 24 . 9

.

Mabbett 33 24 9

Madrid 17 13 4

Monescalchi 22 21 1

Moorehead 15 14 1

Munson 10 10 0

Murdock 22 19 3

Noell 23 15 8

Payne 19 12 7

Petrie 79 67 12

Roberts 10 5 5

Salmon 11 10 • 1

Sanger 11 6 5

Scannel 38 27 . 11

Shelton . 23 11 12

Simpson 3 2 1

Sinsabaugh 8 3 5

Slater 24 14 10

Smith 11 8 3

Starr 19 17 2

Tison 14 7 7

Treannie 21 13 8

Turczen 4 4 0

Wallace 16 15 1

Wehnau 31 27 4

Williams • 8 2 6

Wirth 17 16 1

Total 1019 814 205

Percentage 100% 80% 20%

*676These tables have been prepared from the time sheets of the employees shown in the records of the two cases; while the time sheets of all employees for all the weeks worked are not in the record, the sampling is broad enough in each case to make possible a reasonably reliable estimate of the instability of employment.

The variation is noted below 44 hours in the Belo case, since at the time that case arose, the statutory work week was 44 hours. The variation is noted below 40 hours in the case at bar, since the statutory work week is now 40 hours.

2. Tabulation showing percentage ,of total weeks that employees on plan in the instant case have worked more or less thorn "scheduled” hours; there are no comparable figures in the Belo record.

Wks. on

Wks. worked less than scheduled hrs. Earnings controlled by guarantee

Wks. worked more than scheduled hrs. Earnings controlled by hourly rates

Wks. worked exactly scheduled hours

Employees Guarantee Number Percent Number Percent Number Percent

E. Angelhow 193 49 25.4% 142 73.6% 2 1%

W. Wallace 137 14 10.2% 94 68.6% 29 21.2%

G. Slater 193 33 17.1% 121 62.7% 39 20.2% '

A. Williams 226 62 27.4% 122 54. % 42 18.6%

J. Hess 226 13 5.8% 181 80. % 32 14.1%

C. Wehnau 226 22 9.7% 148 65.5% 56 24.8%

L. Bronk 226 25 11.1% 147 65. % 54 23.9%

F. Ahrens 226 21 9.3% 143 63.3% 62 27.4%

S. Keyes 226 15 6.6% 197 87.2% 14 6.2%

E. Davis 226 8 3.5% 201 88.9% 17 7.6%

A. Monescalchi 226 13 5.7% 192 85. % 21-9.2%

E. Heckert 199 8 4. % 137 68.8% 54 27.2%

J. Scannel 226 24 10.6% 98 43.4% 104 46. %

C. Munson 226 6 2.6% 213 94.2% 7 3.2%

E. Haven 185 32 17.3% 116 62.7% 37 20. %

J. Hitchcock 226 24 10.6% 155 68.6% 47 20.8%

M. Cunningham 226 11 4.9% 163 72.1% 52 23. %

E. Clarke 226 29 12.9% 131 58. % 66 29.2%

R. Murdock 226 31 13.7% 146 64.6% 49 21.7%

M. Haight 226 8 3.5% 201 88.9% 17 7.6%

A. Wirty 226 12 5.3% 18 8. % 196 86.7%

F. Giglio 226 10 ' 4.4% 82 36.3% 134 59.3%

S. Fanelli 226 5 2.2% 145 64.2% 76 33.6%

F. Moorehead 226 25 11.1% 127 56.2% 74 32.7%

E. Madrid 175 13 7.4% 1 00.6% 161 92. %

J. Blackowiak 226 2 0.9% 195 86.3% 29 12.8% '

T. Petrie 226 173 76.5% 43 19. % 10 4.5%

C. Smith 226 8 3.5% 196 86.7% 22 9.8%

F. McCann 226 36 15.9% 150 66.4% 40 17.7%

J. Jerzak 226 21 9.3% 151 66.8% 104 23.9%

R. Cipriano 173 19 11. % 61 35.3% 103 53.7%

W. Shelton 226 44 19.5% 141 62.4% 41 18.1%

L. Starr 226 41 18.1% 158 70. % 27 11.9%

0. Atkinson 171 33 19.3% 126 73.7% 12 7. %

A. Laney 108 51 47.4% 50 46.5% 7 6.1%

J. Tison 226 32 14.2% 179 79.2% 15 6.6%

*677Wks. on

Wks. worked less than scheduled hrs. Earnings controlled by guarantee

Wks. worked more than scheduled hrs. Earnings controlled by hourly rates

Wks. worked exactly scheduled hours

Employees Guarantee Number Percent Number Percent Number Percent

W. Sanger 226 21 9.3% 181 80.1% 24 10.6%

J. Leuthauser 226 21 9.3% 86 38.1% 119 52.6%

M. Julian 226 23 10.2% 82 36.3% 121 53.5%

J. Roberts 226 13 5.8% 187 82.7% 36 11.5%

T. Sinsabaugh 184 16 8.6% 69 37.5% 99 53.9%

C. Julian 198 18 9.1% 166 83.8% 14 7.1%

J. Treannie 162 14 8.6% 137 84.6% 11 6.8%

E. Race 19 1 5.3% 18 94.7% 0 0

J. Toczdlowski 23 2 8.7% 21 91.3% 0 0

J. T. Kane 148 8 5.4% 140 94.6% 0 0

E. Helliwell 202 13 6.4% 0 0% 189 93.6%

D. Braun 27 0 0% 27 100. % 0 0

J. Lanutti 137 4 2.9% 128 93.4% 5 3.7%

W. Schmutzler 79 9 11.4% 1 1.3% 69 87.3%

J. Payne 20 2 10. % 18 90. % 0 0

V. E. Payne 93 21 22.5% 67 72. % 5 5.4%

G. Salmon 30 7 23.3% 23 76.7% 0 0

Totals 9862 1166 6222 2474

Aggregate percentages (100%) 12% 63% 25%

(dissenting).

The Administrator, on petition for rehearing, now shows that his exhibit from which were deduced the computations made the sole basis for reversal here was a selected one “to include every week in which the guarantee was paid for 40 hours or less in order to illustrate the situations in which •the payment of the guarantee could not be attributed to ‘overtime.’ ” That the exhibit was thus compiled to show only the departures from the standard, leaving the large bulk o'f the workweeks quite unconsidered, 'necessarily means that the computations are even more inaccurate and misleading than *6791 pointed out in my original dissent. This the defendant substantially concedes in its reply, arguing that the decision should rest upon more general grounds. The Administrator goes on to state: “The complete records would disclose that approximately 2 per cent, rather than 211 *per cent, of the workweeks were of less than 40 hours.” Hence, unlike the Bclo situation, the departures from the norm were quite negligible. We should recall that this issue was not tried or considered below,2 and only came into the case via the opinion, seemingly in response to the Administrator’s suggestion, quoted in my dissent, of “a regularity of normal working hours” readily adjustable to the statutory policy. The Administrator’s present request for a return of the case to demonstrate these facts yet more conclusively seems therefore a very modest one indeed. Appellate fact finding is dubious at best; it becomes dangerous when it is indulged in as a surprise to the parties; it takes on elements of the fantastic when it is persisted in against a showing of quite contrary facts.