dissenting: This case is, pure and simple, an example of a district judge’s unwillingness to aeeept the strictures imposed by the Sentencing Guidelines. Although most frequently district judges are frustrated by the necessity to impose unreasonably harsh and unjust sentences as the result of the inflexible and mechanistic sentencing system that now binds the judiciary, in this ease the district judge surprisingly was motivated by a contrary concern. He thought that a sentence *827of less than six years was too little for the securities fraud involved. Accordingly, he decided to disregard what the Guidelines clearly contemplate and created his own formula, one that he thought would justify the sentence he decided to impose. To make matters worse, he then erred by misapplying the newly created formula. When the district judge added four levels because the amount of loss substantially exceeded $5,000,000, he did not realize that the guideline offense level with which he started already included one of the levels he thought necessary to add. Therefore, he erroneously added one level more than even his own formula would permit.1
While it is clear that the district court had the authority to depart upward because of the amount of loss involved, unlike the majority, I believe it equally clear that the extent of the departure is contrary to law. In reviewing the extent of departure, we consider whether the district court’s action comports with the “structure, standards, and policies of the Act and Guidelines.” United States v. Hicks, 997 F.2d 594, 599 (9th Cir.1993); United States v. Lira-Barraza, 941 F.2d 745, 751 (9th Cir.1991). We also examine whether the district court has explained “in detail” the reasons underlying its decision as to the extent of departure. Hicks, 997 F.2d at 599.
In this case, the district judge stated that he was “[djeparting upward by one level for each $5,000,000 increase in loss.” The first problem is that, even if we apply the district court’s stated formula, it would yield an upward departure of three levels and not four. For losses totalling $2,000,001-$5,000,000, the 1987 Guidelines prescribe a 10-level increase; for losses over $5,000,000, the Guidelines prescribe an 11-level increase. U.S.S.G. § 2F1.1(b)(1) (1987). We must give recognition to the increases already incorporated into the Guidelines, and the district court’s formula, properly applied, would do so. It establishes increments in the amount of $5,000,000 starting from the sum of $5,000,001. Under the loss table contained in section 2F1.1, $5,000,000 is the ceiling for the category that receives a 10-level increase. The 11-level category commences at $5,000,-001. Under the district court’s formula of $5,000,000 increments, the ceiling for that 11-level category would be $10,000,000. What the district court failed to recognize is that no departure is required for losses in that category; the Guidelines already provide an 11-level enhancement.2
*828The district court’s formula, if properly applied, would result in an upward departure when the amount of loss exceeds $10,000,000. If one level is added for each additional $5,000,000, a loss over $10,000,000 would yield a one-level departure (resulting in a total increase of 12 levels due to the amount of loss); a loss over $15,000,000 would yield a two-level departure (for an increase of 13 levels); and a loss over $20,000,000 would yield a three-level departure (for an increase of 14 levels).
In this case, the district court erred in departing by four levels, which were added to the 11-level increase already prescribed by the applicable guideline section for losses over $5,000,000. Apparently, the court simply failed to realize that a one-level increase for at least the first substantial increment of losses over the $5,000,000 threshold is already incorporated into the 11-level increase prescribed by section 2F1.1 and that a departure could not be given for a loss of $5,000,-001, the starting point for the 11-level category. See U.S.S.G. § 2F1.1 (n.10) (1987). Any departure must commence at the ceiling for that category. Because of his error, the district court both increased the offense level by one (as prescribed by the loss table) and departed upward by one level for the first increment of losses that exceeded the $5,000,000 threshold (i.e., the $5,000,001-$10,-000,000 range). In short, the district court double-counted. Although it was apparently under the impression that it was imposing a one-level increase at $5,000,000 intervals, the court actually imposed a two-level increase for the first additional $5,000,000. The actual departure is, for this reason, inconsistent with the formula the district court stated it was following, as well as with the Guidelines.
An even more serious problem, however, is that the district judge offered no satisfactory explanation of why a four-level departure is reasonable in view of the structure, standards, or policies of the 1987 Guidelines. Instead he stated that: “Departing upward one level for each $5,000,000 increase in loss reasonably addresses the gravity of the harm not otherwise accounted for in the applicable version of the guidelines.” Although the majority relies on this statement in upholding the departure, I fail to see how the judge’s summary proclamation that his chosen formula is correct explains anything. Aside from saying that he’s right because he’s right, the district judge has provided no clue as to why he believed $5,000,000 steps to be reasonable. Why not, for instance, increase the offense level at increments of $5,000 or, for that matter, $500,000,000? The district court offers no reasoned explanation. It just declares that its chosen formula is correct.3
*829Most important of all, even if the district judge had attempted to provide an explanation of his reasons, it is clear that we could not affirm. The formula announced by the district court and approved by the majority is directly contrary to the approach adopted by the Sentencing Commission. The loss table contained in section 2F1.1 follows a general pattern of increasing the offense level by one as the amount of loss doubles. If the structure of the 1987 loss table were followed, one would expect an additional one-level increase for losses over $10,000,000 and a two-level increase for losses over $20,000,000. Given that the amount of loss in this case was found to be “no less than $20,000,000,” a departure of two levels — not four levels — would be appropriate here.4 That is not, of course, to say that a departure is lawful only if it follows precisely the pattern established in the Guidelines. Rather, the Guidelines’ approach provides a framework and guide by which to measure the departure adopted. Reasoned variations from the Guidelines’ general pattern at particular levels might well be consistent with the approach we have endorsed. See Lira-Barraza, 941 F.2d at 751. A totally contrary formula is not.
The structure of section 2Fl.l’s loss table reflects the Commission’s recognition that, as the total amount of the loss increases, a specific incremental loss becomes less significant for purposes of determining the appropriate sentence, and the additional punishment imposed is of lesser magnitude. Thus, a $500,000 increase from a base loss of $20,-000 (to $520,000) yields a five-level increase. On the other hand, a $500,000 increase from a base loss of $520,000 (to $1,200,000) yields only a one-level increase. And the next $500,000 increase from $1,200,000 (to $1,700,-000) yields no increase at all. See U.S.S.G. § 2Fl.l(b)(l) (1987). Under the Guidelines’ approach, step-ups in the offense level are based not on fixed amounts of increased losses, but on the percentage by which the additional loss increases the base loss — as a general rule, a level is added for each 100 percent increase in loss. The district court offers no reasonable basis for reversing the Commission’s methodology, ignoring the fixed percentage of total loss approach, and increasing the punishment by one level for increases in amounts that constitute a constantly declining percentage of the total loss.
There may, in some cases, be good reasons for not using the Commission’s exponential approach for losses that substantially exceed $5,000,000. It was clearly impermissible, however, to adopt a far more punitive— though no less mechanical — approach to losses over this amount without any reasoned justification for doing so. Additional losses at the higher levels would appear to be less serious than, or equally serious to, losses in the same amount at a lower level, rather than more serious. At the very least, the district court was required to provide some rational explanation for its decision to adopt a formula applicable to additional losses that is so contrary to and so much harsher than that set forth in section 2F1.1. If the district court had given some reasons for its conclusion that a linear approach was warranted, we could at least evaluate those reasons. Here, however, the district court has provided no explanation whatsoever for its total reversal of the Commission’s approach.
I believe it highly unlikely that any reasonable explanation for a four-level upward departure is possible in this case. At any rate, the district court, after being given two chances, has failed to come up with one — it has simply informed us that its chosen sentence is correct. Under these circumstances, the rule of lenity requires that “we infer the rationale most favorable to the [defendants-]appellants, and construe the guidelines accordingly.” United States v. Martinez, 946 F.2d 100, 102 (9th Cir.1991). Following this rule, I would conclude that the district court should not have departed up*830ward by more than two levels. I would, accordingly, perform the duties vested in us under the misguided sentencing procedure that now controls both our actions and those of district judges, review the district court’s actions in the manner required by law, and vacate and remand the sentence with instructions that the district court depart upward, if at all, by no more than two levels.
. The district judge computed Vargas’ total offense level under the 1987 Guidelines as follows:
6 Base offense level (§ 2Fl.l(a))
+ 11 Amount of loss over $5 million (§ 2Fl.l(b)l))
More than minimal planning (§ 2F1.1(b)(2)) (N +
Role in the offense (§ 3B 1.1(c)) (N +
Victim-related adjustment (§ 3A1.1) (N +
Acceptance of responsibility (§ 3E1.1) CN I
Upward departure for loss over $20 million *T +
Total offense level 25
As explained below, one of the four levels by which the court upwardly departed was already included in the 11-level increase for the amount of loss.
. Starting with a base level of six, the 1987 loss table provides for the following increases in the offense level as the amount of loss increases:
Loss Increase in Level
(A) $2,000 or less no increase
(B) $2,001-$5,000 add 1
(C) $5,001-$10,000 add 2
(D) $10,001-$20,000 add 3
(E) $20,001-$50,000 add 4
(F) $50,001-$100,000 add 5
(G) $100,001-$200,000 add 6
(H) $200,001-$500,000 add 7
(I) $500,001-$ 1,000,000 add 8
(J) $1,000,001-$2,000,000 add 9
(K) $2,000,001-$5,000,000 add 10
(L) over $5,000,000 add 11
U.S.S.G. § 2F1.1(b)(1) (1987).
The district court's stated formula would yield the following additional increases:
$5,000,001-$10,000,000 11 (no departure)
$10,000,001-$15,000,000 12 (1 level departure)
$ 15,000,001-$20,000,000 13 (2 level departure)
$20,000,001-$25,000,000 14 (3 level departure)
$25,000,001-$30,000,000 15 (4 level departure)
The formula actually applied by the district court, however, yields an additional level for each category:
$5,000,001-$10,000,000 12 (1 level departure)
$10,000,001~$15,000,000 13 (2 level departure)
$15,000,001-$20,000,000 14 (3 level departure)
$20,000,001-$25,000,000 15 (4 level departure)
$25,000,001-$30,000,000 16 (5 level departure)
. Although the district court mentioned three other purported justifications for its chosen departure, it nowhere provides a coherent explanation of why the departure is reasonable in view of the structure, standards, or policies of the 1987 Guidelines. First, the district court relied upon the subsequently enacted version of the Guidelines. As the majority recognizes, this rationale is foreclosed not only by our precedent, see United States v. Martinez, 946 F.2d 100, 102 (9th Cir.1991), but by the explicit instructions in our earlier memorandum disposition.
The district court's remaining justifications are variations on a common and conclusory theme. The court states that a 35% increase in the offense level (from 11 to 15) is warranted, given that the amount of loss increased by 200%. That explains nothing. Why not increase the offense level by 20%, 50%, 100%, or even 200%? Why 35%? As with its statement that a one-level increase for each additional $5,000,000 is appropriate, the court fails to provide a coherent explanation, grounded in the structure, standards, or policies of the Guidelines, of why a departure yielding a 35% increase in the offense level is reasonable.
The court’s final justification, while ostensibly relying on the 1987 loss table, quite clearly conflicts with the prescribed pattern of increases. The court states that a four-level increase is reasonable in view of the $15,000,000 increase in the loss amount, because a four-level decrease corresponds to a $4,500,000 decrease in the loss amount. Aside from all the obvious problems with this non-comparison, an examination of the fraud table shows that a $4,500,000 change in the amount of loss does not uniformly or generally correspond to a four-level increase or decrease; to the contrary, it all depends on the point from which one measures. A $4,500,000 decrease from a $5,000,000 loss would correspond to a three-level decrease in sentence, while a $4,500,000 decrease in loss from a $4,600,000 starting point would lead to a decrease of five levels. Therefore, the relevance of the district court's observation escapes me. However, given the fact that under his basic approach the district judge should have departed by three and not four levels, the non-comparison is particularly unenlightening. What, one might ask, do these figures show as to the validity of a three-level in*829crease? Probably as little as they reveal with respect to a four-level increase.
. Following the pattern established by the 1987 loss table would yield the following departures:
$10,000,001-$20,000,000 add 12 (1 level departure)
$20,000,001-450,000,000 add 13 (2 level departure)
$50,000,001-$100,000,000 add 14 (3 level departure)
As set forth in note 2, supra, the district court’s stated formula yields far steeper increases.