This appeal presents an unsettled question: is the government a “victim” within the purview of the Victim and Witness Protection Act, 18 U.S.C. §§ 3363-3364 (VWPA or the Act), and, thus, entitled to restitution, when it provokes the commission of a crime that, by design, directly results in depletion of public coffers? We answer this question in the negative, concluding that, in such circumstances, the sovereign is not entitled to restitution under the Act. At the same time, we resolve a more pedestrian sentencing issue which, although much bruited by appellant, has little substance.
I. BACKGROUND
Defendant-appellant Leroy Gibbens is a shoemaker who did not stick to his last. Instead, Gibbens developed a sideline as a broker of second-hand food stamps. In April 1992, the United States Department of Agriculture (USDA) mounted an investigation into food stamp trafficking in Lewiston, Maine. The targets of the investigation included appellant and his son, Zachary J. Gibbens.
In due course, an undercover agent approached Gibbens the younger and his confederate, Joseph R. Beaulieu III, offering to sell food stamps at roughly twenty-five cents on the dollar. 1 The junior Gibbens, who had *30 followed in his father’s footsteps in more ways than one, consummated a few small transactions with the agent, reselling the bootleg food stamps in saloons and other local haunts for thirty or forty cents on the dollar. He also told his father of the agent’s overtures, and, at his father’s urging, put the two men in contact with each other.
Appellant, having recently repaired to Florida, dealt with the agent by telephone, wire, or mail, or by using his son as an internuncio. In a half-dozen transactions during the spring and summer of 1992, appellant bought stamps that had an aggregate face value of $12,895, paying the agent approximately one-fourth of that amount, and resold them at a profit. In their communications throughout this period, appellant continually importuned his vendor to furnish more stamps at more frequent intervals. He also boasted about a putative partner, albeit vaguely. Then, suddenly, to appellant’s apparent dismay, the stream of sales stopped in July of 1992.
Toward the end of that year, the agent renewed contact. Appellant bought two more batches of food stamps at deep discounts. The redemption value of the stamps acquired during this period totalled $8,100. The second of these transactions marked the initial face-to-face meeting between appellant and the agent.
The government subsequently dropped the other shoe: all three cobblers were arrested and a federal grand jury handed up a fourteen-count indictment. Appellant pleaded guilty to one count of conspiracy to acquire and use food stamps in an unauthorized manner, 18 U.S.C. § 371, and six counts alleging unlawful possession of food stamps in violation of 7 U.S.C. § 2024(b). The government agreed to dismiss the only other counts in which appellant was featured.
The district court sentenced appellant on October 22,1993. In constructing the guideline sentencing range (GSR), the court started at offense level six. See U.S.S.G. § 2Fl.l(a). It then factored in a four-level upward adjustment for amount of loss, see U.S.S.G. § 2Fl.l(b)(l)(E) (specifying increment for fraud cases involving losses ranging from $20,000 to $39,999.99), a two-level enhancement for more-than-minimal planning, see U.S.S.G. § 2Fl.l(b)(2)(A), and a two-level credit for acceptance of responsibility, see U.S.S.G. § 3E1.1. These computations yielded an adjusted offense level of ten. For a defendant with a negligible record of prior criminality (Criminal History Category I), this adjusted offense level produced a GSR of six-to-twelve months in prison.
The court imposed a six-month inearcera-tive sentence, to be followed by three years of supervised release. The court eschewed any fine, but ordered appellant to pay $15,-230 to the government as restitution. The court computed the amount of restitution by aggregating the face value of the food stamps handled by appellant (i.e., the sums owed by the USDA to the retailers who ultimately presented those stamps for redemption) and then subtracting the monies appellant paid to acquire the stamps on the black market.
Appellant now challenges his sentence. He showcases several assignments of error. The first two entries are merely alternative formulations of a claim that the USDA engaged in impermissible sentencing factor manipulation — a claim which we find lacking in merit. The other items relate, in one way or another, to the order for restitution. Because we conclude that the government does not qualify for statutory restitution on the facts of this case, we need not address the remaining challenges to the restitution order.
II. SENTENCING FACTOR MANIPULATION
The doctrine of sentencing factor manipulation is a kissing cousin of the doctrine of entrapment.
See United States v. Connell,
Though phrased in various ways, appellant’s theory boils down to an assertion that the USDA revived the investigation, after soft-pedaling it for four months, not with a view toward bringing the conspirators to *31 heel, but for the sole purpose of boosting appellant’s offense level (and, hence, ensuring a prison sentence). In support of this theory, appellant notes that the GSR rose once the amount of loss exceeded $20,000, see U.S.S.G. § 2Fl.l(b)(l)(E); that the last transaction, which exposed him to this increase by bringing the amount of loss over the $20,000 mark, was superfluous, as the government had him dead to rights four months earlier; and that, as soon as the government reached the $20,000 plateau, it halted the sting.
To be sure, the sequence of events is suggestive — but there is another side to the story. By the USDA’s account, the press of other agency business necessitated a temporary suspension of the investigation following a sale on July 20, 1992. The hiatus ended four months later because the agency’s workload had eased and the government needed proof, beyond a reasonable doubt, of appellant’s conspiratorial intent. 2 Moreover, the USDA was hoping, based on appellant’s allusions to a supposed business partner, to land a bigger fish.
The prosecution also suggests that appellant’s predisposition to deal in food stamps on a long-term basis, as evidenced by his incessant demands for more stamps at more frequent intervals, undermines his claim that he was blindsided by unfairly manipulative conduct. Although the district court made an express, fully warranted finding that appellant remained ready, willing, and eager to continue dealing bootleg food stamps indefinitely and on an escalating scale, the government’s point is nonetheless of modest relevance. When an accusation of sentencing factor manipulation surfaces, the judicial gaze should, in the usual case, focus primarily — though not necessarily exelusively — on the government’s conduct and motives.
3
See Brewster,
Undercover operations comprise a valuable, and generally lawful, weapon in the government’s armamentarium.
See Connell,
We can plot no bright line to separate the government’s ordinary conduct in a conventional sting operation from extraordinary misconduct of a sort that might constitute sentencing factor manipulation. We believe the subject must be approached on a case-by-case basis, albeit with due regard for the potential dangers of sentencing factor manipulation,
see Connell,
In an effort to hoist this burden, appellant intimates that the present situation is inherently susceptible to manipulation and, therefore, gives rise to a conclusive presumption of official misconduct. We do not agree. The inquiry must proceed as a stereotypical exercise in factfinding, linked to an allocation of the burden of proof but uncluttered by artificial presumptions.
Putting matters in this perspective reveals the fundamental weakness in appellant’s position. The government’s explanation of the sequence of events, apparently credited by the district court, is at least as plausible as the adverse inference that appellant would have us draw. We have held, time and again, that when a sentencing court is confronted with two reasonable views of the record, and chooses to credit one such view rather than the other, its choice cannot be termed clearly erroneous.
See, e.g., United States v. Ruiz,
III. THE GOVERNMENT AS VICTIM
In his most portentous assignment of error, appellant posits that, on the facts of this case, the USDA is not a “victim” within the meaning of the restitutionary provisions of the Victim and Witness Protection Act, 18 U.S.C. §§ 3663-3664. This proposition presents a pure question of statutory interpretation and, as such, invites
de novo
review.
See, e.g., Gifford,
A. Conceptualizing the Problem.
This ease falls into a grey area that separates two established legal principles. On one hand, although once problematic,
see infra
p. 35, it is by now settled that a government entity (local, state, or federal) may be a “victim” for purposes of the VWPA (and may be awarded restitution) when it has passively suffered harm resulting directly from the defendant’s criminal conduct, as from fraud or embezzlement,
See, e.g., Ratliff v. United States,
On the other hand, the federal courts are consentient to the effect that the government is not a “victim” for purposes of VWPA (and may not be awarded restitution) to the extent that it incurs costs in the clandestine provocation of a crime that, if carried to fruition under ordinary circumstances, would not directly harm the government.
4
See, e.g.,
*33
Gall v. United States,
What makes this case difficult is that it falls somewhere between the two ends of the spectrum. While we deal with a crime provoked by an undercover investigation, the crime was designed to inflict harm on the government. If consummated under circumstances net involving official participation, the crime would have resulted in direct loss to the government in exactly the manner that the government here experienced loss. Nonetheless, the government instigated the particular incidents for. which it now claims the right to restitution — indeed, had there been no official participation, the claimed losses would not have eventuated. This means that here, unlike in either of the more familiar prototypes, the difference between the face value of the food stamps and the amount appellant paid for them was both a calculated consequence of the defendant’s crime and a calculated cost of the government’s investigation. As a result of the hybrid nature of the loss, each side argues that this difficult situation more closely resembles the prototype that favors its position — and neither argument can easily be debunked.
B. Statutory Interpretation.
We envision the task of resolving this conundrum as an exercise in statutory construction. Our role, of course, is as interpreters of the words chosen by Congress, not as policymakers or enlargers of congressional intent. This role requires that we start with the statutory text.
1. Text. The VWPA states that restitution may be awarded only to a “victim of the offense.” 18 U.S.C. § 3663(b)(1). A “victim of an offense” is defined as “any person directly harmed by the defendant’s criminal conduct in the course of the scheme, conspiracy, or pattern.” Id. § 3663(a)(2). In the idiom of the Act, the question we decide today is whether the government is a “victim” in the sense that it is “harmed by the defendant’s criminal conduct” when it experiences loss that is the direct, foreseeable consequence both of the criminal’s conduct and of the government’s own machinations. Conceived in this way, the question is one of first impression. 6
We believe that the key phrase, “harmed by,” as it appears in the VWPA, is ambiguous. Under one reading of the phrase, the statute is satisfied when, as now, an entity experiences a loss directly and foreseeably caused in whole or in part by the criminal’s conduct. But this reading represents one choice out of several. For example, it is also entirely possible that the word “harm” denotes “aggregate harm” — a construction which, if adopted, would require the phrase to be read with a view toward some type of cost-benefit analysis. In that event, the very *34 fact that the government knowingly incurred the loss could be taken as signifying that, in its estimation, the game was worth the candle. Put another way, the fact could, evidence the government’s belief that the overall gain — incapacitating the targets of the investigation and deterring others from embarking on similar schemes — outweighed the out-of-pocket loss.
A second, more intriguing possibility is that “harmed by” connotes passivity. In ordinary usage, “harm” is suffered at-the hands of another, while “loss” may be merely experienced or sustained. It defies common usage to envision an entity that planned and provoked a crime as a victim in the same sense that a passive sufferer of harm is a victim, notwithstanding that the entity may have experienced loss. Courts cannot ignore legislative decisions to use one particular word instead of another.
See, e.g., United States ex rel. Springfield Term. Ry. Co. v. Quinn,
A statute is ambiguous if it reasonably can be read in more than one way.
See United States v. O’Neil,
2. Legislative History. The VWPA was first enacted in 1982 in an effort to afford greater protection to victims and witnesses, and to enhance their stature in the criminal justice system. See S.Rep. No. 532, 97th Cong., 2d Sess. 30, reprinted in 1982 U.S.C.C.A.N. 2515-16. The object of the restitution provisions in particular was to help “restore the victim to his or her prior state of well-being.” Id. at 2536. Although the word “victim” was not precisely defined in either the original Act or its accompanying commentary, it is pellucid that, in the eyes of the enacting Congress, the prototypical victim was a private individual. The preamble to the Senate Report laments that the victim is all too often the “‘forgotten person’” in the legal process. Id. at 2516. With regard to the restitution provisions, the only specific example of a victim describes an elderly lady who, after being mugged, had to forgo surgery because the prosecutors did not seek restitution in a sufficient amount. See id. at 2536-37.
Absent a clearly marked trail leading in some other direction, courts should presume that words used in a statute are to be given their ordinary meaning.
See United States v. Dawlett,
To' be sure, Congress amended the VWPA in 1990, adding a statutory definition of “victim” as one “directly harmed by the defendant’s criminal conduct.” 18 U.S.C. § 3663(a)(2). However, we resist the conclusion that, by specially defining “victim” on this occasion, Congress meant to stray far enough from the common meaning of the word to eliminate the element of passivity. Rather, the legislative history attests that highly idiosyncratic concerns motivated Congress’s action.
The amendment first surfaced in the House and Senate versions of the proposed “Banking Law Enforcement Act” under the caption, “Enhancement of Ability to Order Restitution in Certain Fraud Cases.” See 136 Cong.Rec. H 5996 (daily ed., July 31, 1990); 136 Cong.Rec. S 18322 (daily ed., Nov. 2, 1990). It was then incorporated into, and passed as part of, the sprawling Crime Control Act of 1990, P.L. 101-647,104 Stat. 4789. In that incarnation, the definition comprised *35 one of nine disparate provisions grouped in a single title under the appellation “Banking Law Enforcement.” In floor remarks, the sponsor of that title explained that its “purpose” was “to enhance the enforcement powers of the Department of Justice and the federal financial institution regulatory agencies with respect to unlawful activities affecting federally insured financial institutions.” 136 Cong.Rec. E 3684 (daily ed., Nov. 2, 1990) (remarks of Rep. Sehumer).
We think that this history, coupled with the division of opinion that originally existed in the courts on whether a government entity could ever be a “victim” under the Act, makes it highly probable that the newly emergent definition was intended to accomplish two things. Broadly, the amendment was meant to clarify that, in appropriate cases, a government entity, say, FSLIC or FDIC, could be regarded as a “victim” under the Act. More narrowly, the amendment was designed to clarify the government’s entitlement to restitution for losses suffered qua insurer as a consequence of savings-and-loan fraud, that is to say, as a passive sufferer of the harm caused by , such fraud. Although special definitions sometimes are taken wholly to supplant common usage, see 2A Sutherland Statutory Construction § 47.28 (5th ed. 1992), this special definition is not of that ilk; it strengthens, rather than dissipates, the force of our point anent common usage. In other words, notwithstanding the 1990 amendment, the presumption in favor of ordinary meaning continues to apply in this ease. And the ordinary meaning of the word “victim” poses an obvious problem for the government’s view of the VWPA universe.
To sum up, nothing in the legislative history of either the organic Act or its amendments indicates that losses incurred in government sting operations should be subject to recoupment under the VWPA. Conversely, there is some basis in the legislative history of the VWPA for believing that the enacting and amending Congresses both viewed the word “victim” in a more restrictive manner than the government urges here. We do not mean to suggest that the benefits of the VWPA should be confined to widows and orphans; but we are constrained to note that, as the status of victimhood is expanded beyond passive sufferers of harm, we move further and further away from the concerns that drove Congress to pass the statute.
C. The Rule of Lenity.
We recognize that the Act’s language and legislative history, though suggestive, do not speak unequivocally to the question at hand. In light of this uncertainty, we have examined more recondite sources. We confess, however, that our quest has proven unrewarding; by and large, the government’s claim resists analogy. We have considered analogies from the doctrines and case law of civil restitution, criminal restitution through probationary conditions, tort law, and a variety of other sources. 7 None offer compelling guidance.
When all else fails to bring sufficient lucidity to the meaning of a penal statute, the rule of lenity casts the decisive vote. That rule, which mandates the resolution of ambiguities in a criminal statute favorably to the defendant,
see, e.g., United States v. Bass,
This is not only the proper time to invoke the rule of lenity, but also the proper place; after all, the rule of lenity played the decisive role on the one occasion that the Court ventured to interpret the VWPA.
See Hughey v. United States,
We retrace the Court’s steps here. On the principle of lenity, we resolve lingering doubts as to the statute’s meaning in favor of the defendant. We hold as follows: a government agency that has lost money as a consequence of a crime that it actively provoked in the course of carrying out an investigation may not recoup that money through a restitution order imposed under the YWPA.
We add an eschatocol of sorts. As courts reaching similar conclusions have observed,
see, e.g., Salcedo-Lopez,
IV. CONCLUSION
We need go no further. We direct the district court to modify the defendant’s sentence by deleting the award of restitution; and, with that modification, we affirm the judgment below.
Affirmed as modified.
Notes
. Zachary Gibbens and Joseph Beaulieu were employed by appellant at his shoe repair shop in Lewiston.
. On this scenario, the final- transaction assumed particular importance because previous deals had been conducted from afar, and, without a face-to-face encounter, the government might be hard pressed to verify appellant's identify in court. Cf., e.g., B. Franklin, Poor Richard’s Almanac (1758) (warning that "for want of a nail the shoe is lost”).
. To be sure, a defendant’s predisposition, or the lack thereof, may have evidentiary significance in an assessment of the government's motives and conduct. Moreover, one can imagine different species of sentencing factor manipulation, in some of which predisposition may be of greater relevance.
See, e.g., Connell,
. Courts interpreting analogous state statutes have divided on this type of question. For example, some courts hold that, when a government agency disburses money in a drug sting, it is not a "victim" entitled to restitution.
See, e.g., State v. Newman,
. The relevant provision of the Act states that restitution may be ordered "in the case of an offense resulting in damage to or loss or destruction of property of a victim of the offense.” 18 U.S.C. § 3663(b)(1),
. We realize that
United States v. Dougherty,
. The interested reader may wish to consult various works that afford broad-gauged historical perspectives on the subject. See, e.g., Stephen Schafer, Compensation and Restitution to Victims of Crime (2d ed. 1970); Richard E. Laster, Criminal Restitution: A Survey of its Past History and an Analysis of its Present Usefulness, 5 U.Rich. L.Rev. 71 (1970).
. While Hughey's precise holding, denying restitution for losses resulting from offenses other than the offense of conviction, has been super *36 seded partially by the 1990 amendment to 18 U.S.C. § 3663(a)(3), this development does not throw the slightest doubt on Hughey's herme-neutical approach. We, therefore, regard Hu-ghey as impeccable authority for the purpose at hand.
. Courts are divided on whether drug buy money may be recovered in restitution as a condition of supervised release.
Compare Daddato,