MEMORANDUM RE: LOSS CALCULATION FOR SENTENCING
This Memorandum discusses the issue raised by Defendants Dana Siciliano and June Kodiak regarding loss calculation for their sentencing, and elaborates on this Court’s ruling on the issue at the February 27, 2009 hearing.
I. Background
On July 18, 2006, Dana Siciliano and June Kodiak, along with eight other individuals, were indicted for their involvement in a scheme led by co-defendant Mahn Huu Doan. See (Doc. 1 — Indictment). The following facts summarize the indictment, to which all defendants have entered guilty pleas. Over a period of several years, Doan purchased or borrowed the identities of many individuals in the Asian — American community and used those identities to purchase properties *625throughout Philadelphia. In conducting his scheme, Doan would first buy the homes from the existing owners with one of his recruited identities in the “A” transaction.1 After the A purchase, Doan would then sell the properties to another acquired identity (in effect, to himself) in the “B” transaction at an artificially inflated price. Doan enlisted the help of an appraiser for the B transaction; the appraiser attested to the inflated values of the homes by claiming that they had been renovated or were in the process of being renovated. Using this sham appraisal, Doan would obtain a mortgage at the artificially inflated value of the home.
For the financing of the transactions, Doan utilized Encore Mortgage company, where both Dana Siciliano and June Kodiak were employed. In addition to conventional ones, many of the mortgages that Doan obtained for the B transaction were FHA-insured loans through HUD. From September 2001 until February 2003, Sici-liano and Kodiak knowingly accepted fraudulent documentation in approving these FHA-insured and conventional mortgages. At one point in the scheme, several of the mortgages that were still partially unpaid were refinanced at an even higher value; the proceeds of these new mortgages were used to maintain payments on other outstanding mortgages. At another point in the scheme, Doan was incarcerated for another crime and was no longer able to maintain the mortgage payments on the properties. Siciliano’s boss, Vincent Sirolli, with the knowledge that foreclosure on the properties might arouse the suspicion of law enforcement, directed Sici-liano to keep track of all the outstanding mortgages in the scheme and maintain the payments.
As to the two Defendants at issue here, Siciliano pled guilty on September 12, 2006 to conspiracy, wire fraud, making false statements to HUD, and identity fraud. (Doc. 78, 81). Kodiak also pled guilty on September 26, 2006 to the same four counts. (Doc. 87, 88).
In anticipation of all defendants’ sentencing hearings, on September 23, 2008, the Government filed the “Government’s Sentencing Memorandum Concerning Loss Calculation for All Defendants.” (Doc. 205). In that Memorandum, the Government discussed the method it used to estimate the loss caused by the scheme for purposes of § 2B1.1 of the Sentencing Guidelines.2 The Government later filed a revised Memorandum and spreadsheet on October 6, 2008, which included additional paid off loans in its loss estimation. (Doc. 212). On January 12, 2009, the Government filed individualized Sentencing Memoranda for Dana Siciliano (Doc. 237) and June Kodiak (Doc. 238); those Memoranda apply the loss calculation methodology to both Defendants in particular.
Siciliano filed her own Sentencing Memorandum in response on January 16, 2009. (Doc. 242). Siciliano took issue with the methodology offered by the Government to calculate loss under the Guidelines. To rebut the Government’s methodology, she offered an expert report concerning the amount of loss.
On January 21, 2009, the Court held a sentencing hearing for Siciliano (Doc. 268), where she presented the expert testimony of Eugene Pasymowski, MAI, a real estate appraiser, who offered Siciliano’s alternative method of estimating losses, described below. During the sentencing for Kodiak on January 22, 2009 (Doc. 263), Kodiak *626agreed to adopt the expert testimony and arguments offered by Siciliano subject to agreement by Siciliano’s counsel, which was forthcoming. Both sentencing hearings were adjourned to allow the Government to advocate its position on the case law for loss calculation under the Sentencing Guidelines in similar cases. (Doc. 289 — Govt’s Letter of Jan. 28). Siciliano responded (Doc. 290 — Siciliano Letter of Feb. 11), and the Government provided its reply (Doc. 291 — Govt’s Letter of Feb. 12).
The sentencing for both Siciliano and Kodiak resumed on February 27, 2009. This Court ruled on the relevant loss calculation for the sentence and then issued a sentence for both Defendants. This Memorandum elaborates on the rationale underlying that ruling, and it is limited to the record established at the hearings for these two Defendants.
II. Framework of the Sentencing Guidelines
The primary issue disputed by the parties concerns the amount of loss incurred by the victim entities, either HUD or the conventional mortgagee, which were entitled to the balance due on the mortgages on which Defendants defaulted. Although agreeing that the applicable section of the Sentencing Guidelines is § 2B1.1 and that the base offensive level is 6, the parties disagree as to the enhancement that Defendants Siciliano and Kodiak deserve under the Sentencing Guidelines for the amount of loss caused by their conduct, which factors into the applicable offense level.
There are two types of loss that a district court can use for the loss calculation in § 2Bl.l(b)(l): actual loss or intended loss. “Subject to the exclusion in subdivision (D), loss is the greater of actual loss or intended loss.” USSG § 2B1.1, app. n. 2(A) (Nov. 1, 2002). In this case, both parties appear to agree that this Court should calculate the actual loss, rather than the intended loss.
Actual loss is defined as “the reasonably foreseeable pecuniary harm that resulted from the offense.” USSG § 2B1.1, app. n. 2(A)(i). Pecuniary harm is determined to be “reasonably foreseeable” when “the defendant knew or, under the circumstances, reasonably should have known, [that it] was a potential result of the offense.” USSG § 2B1.1, app. n. 2(A)(iv). A loss estimate by a district court should be based on several factors:
(i) The fair market value of the property unlawfully taken or destroyed; or, if the fair market value is impracticable to determine or inadequately measures the harm, the cost to the victim of replacing that property.
(ii) The cost of repairs to damaged property.
(iii) The approximate number of victims multiplied by the average loss to each victim.
(iv) More general factors, such as the scope and duration of the offense and revenues generated by similar operations.
USSG § 2B1.1, app. n. 2(C). “In a case involving collateral pledged or otherwise provided by the defendant,” the loss is then reduced by “the amount the victim has recovered at the time of sentencing from disposition of the collateral, or if the collateral has not been disposed of by that time, the fair market value of the collateral at the time of sentencing.” USSG § 2B1.1, app. n. 2(E)(ii).
The Guidelines are clear that the loss calculation does not need to be perfectly accurate, though it can not be based on pure speculation. “The court need only make a reasonable estimate of the loss. The sentencing judge is in a unique position to assess the evidence and estimate the loss based upon that evidence. For *627this reason, the court’s loss determination is entitled to appropriate deference.” USSG § 2B1.1, app. n. 2(C) (citing 18 U.S.C. § 3742(e) and (f)) (emphasis added). The Third Circuit’s standard of review for a loss calculation also provides for deference to the district court’s estimate. If appealed, “[t]he appropriate standard of review of a district court’s decision regarding the interpretation of the Sentencing Guidelines, including what constitutes ‘loss,’ is plenary. Factual findings, however, are simply reviewed for clear error.” United States v. Napier, 273 F.3d 276, 278 (3d Cir.2001) (internal citations omitted). In determining whether to apply credit to a loss calculation in these cases, a District Court’s decision to accept a valuation method as more or less reliable is “essentially a fact-based determination,” and the Third Circuit will “review for clear error.” Id.
Finally, as with other sentencing enhancements, there is a burden-shifting framework for loss calculation during sentencing. “The burden of proving the amount of loss in this case rests with the Government.” United States v. Sweitzer, 2005 WL 3806056, at *5 (M.D.Pa. Apr. 22, 2005) (Rambo, J.) (citing United States v. Evans, 155 F.3d 245, 253 (3d Cir.1998); United States v. McDowell, 888 F.2d 285, 291 (3d Cir.1989)).
[T]he ‘burden is ... on the government to prove ... the facts in support of a sentence enhancement; the defendant does not have to prove the negative to avoid the enhanced sentence.’ If, however, the Government establishes a pri-ma facie case, the burden of production shifts to the defendant. At this point, the defendant is ‘required to come forward with evidence tending to cast doubt on the government’s evidence.’ The ultimate burden of persuasion, though, remains with the Government. Sweitzer, at *5 (quoting Evans, 155 F.3d at 253).
III. Parties’ Contentions
A. Government
In providing its initial loss calculation, the Government divides the approximately 180 properties in the scheme into two categories: (1) properties where foreclosure has been completed and the properties resold; and (2) properties where foreclosure has not been completed or the properties have not yet been resold. As to Category 1, there are around 60 properties that have gone through foreclosure and where resale has occurred. The Government contends that the losses to be used in the sentencing calculation for these Category 1 properties are equal to the amount of the principal that remains on the mortgage, plus the costs spent as a result of the foreclosure, less the amount that the property was actually resold for. The calculation is: Loss = Mortgage Balance [plus] Foreclosure Costs [minus] Property Resale Amount. The Government has provided loss statements from HUD for each of these homes in the first category. Using this method, the total loss for Category 1 properties, as applied to Siciliano and Kodiak, is $1,858,429.18. See (Doc. 237-Govt’s Siciliano Sentencing Memorandum, at 5); (Doc. 238 — Govt’s Kodiak Sentencing Memorandum, at 5).
The more complicated situation arises from the Category 2 properties, those that have not yet been resold. Of the 180 properties total, there are about 120 Category 2 properties. In attempting to estimate the losses for these Category 2 properties, the Government uses the losses of the Category 1 properties. It argues that the “vast majority” of the properties in the scheme were in similar areas of Philadelphia, were similarly priced, and were similarly valued. Given these similarities, the *628Government first calculated the average percentage loss for all of the Category 1 properties that have been resold. The calculation for the average percentage loss is: Percentage Loss = Average Dollar Loss for Category 1 [divided by] Average Mortgage Balance for Category 1. After calculating the Category 1 average percentage loss, 38%,3 the Government argues that an identical 38% loss should be applied to the remaining balance of each of the Category 2 mortgages. Using this methodology, the estimated loss is $2,089,997 for the FHA-insured Category 2 properties, and $1,198,748 for the commercial mortgage Category 2 properties. Together, the Government’s estimated loss for the Category 2 properties is $3,288,745. When combined with the actual loss of the Category 1 properties, the Government estimates the total loss applicable to Siciliano and Kodiak to be about $5,147,174. See (Doc. 237 — Govt’s Siciliano Sentencing Memorandum, at 5); (Doc. 238 — Govt’s Kodiak Sentencing Memorandum, at 5).
Finally, the Government argues that its methodology for loss calculation benefits Defendants Siciliano and Kodiak in two ways. First, the calculation includes those B transaction mortgages that were later fully paid off by the defendants, even where the mortgage was paid off with another fraudulent sale. Including these mortgages would mean several of the mortgages in the calculation are assigned a $0, or 0%, loss; these paid-off mortgages reduce the average loss for the properties. Next, the Government also notes that it did not include the debenture interest that HUD pays the servicer on the insurance claim. It does, however, argue that the Sentencing Guidelines would permit the inclusion of such interest.
B. Siciliano and Kodiak
Siciliano and Kodiak, on the other hand, challenge the Government’s methodology for both the Category 1 and Category 2 properties. Siciliano’s expert testified at the January 21, 2009 sentencing hearing about a regression analysis that he conducted to predict the home values for a sample of the scheme properties, all located in South Philadelphia. As to the Category 1 properties, Siciliano and Kodiak use the predicted values from Mr. Pasymow-ski’s regression analysis to show that the actual sale prices that the Government uses are much lower than their predicted market values, which they argue would prevail in a conventional, arms-length transaction.4
As to the Category 2 properties, Sicili-ano and Kodiak argue that home prices and changes in home prices in Philadelphia are not the same across neighborhoods, particularly South Philadelphia, which, they argue, has seen an unusually strong increase in prices during the past several years; they contend that it would be inappropriate to use such a rough average loss for properties in certain neighborhoods, such as South Philadelphia. In support of their argument, Siciliano and Kodiak again use Mr. Pasymowski’s regression analysis *629and the predicted values of 43 of the Category 1 and 2 properties in South Philadelphia. In formulating the regression analysis, Mr. Pasymowski first inspected all of the 48 properties by observing their exterior condition. Next, by using the sale prices of similar properties in the same neighborhoods, then considering the variation on certain characteristics of the properties in the regression analysis, Mr. Pasy-mowski predicted the value of those 43 properties if they were sold in an arms-length transaction. Mr. Pasymowski’s analysis uses the variables of market conditions/time, gross living area, building condition, number of blocks away from Washington Avenue, and number of blocks away from Front Street.
After computing the predicted values, Mr. Pasymowski then calculates the gain/ loss for these 43 properties by taking the resale price in Doan’s B transaction, then subtracting the predicted value. Using these numbers, he predicts that there is not a loss on the 43 properties, but a gain of $955,627.
IV. Case Law on Property Valuation for Loss Calculation in Sentencing
A. United States v. Sharma, 190 F.3d 220 (3d Cir.1999)
To this Court’s knowledge, only three Third Circuit cases have considered the issue at hand. The earliest case was United States v. Sharma, 190 F.3d 220 (3d Cir.1999). In Sharma, a father and his three adult sons were convicted for providing “material and false statements that misrepresented their financial resources to two banks in order to obtain loans and lines of credit.” Id. at 223. In particular, the defendants “grossly misrepresented the value and profitability of their assets through submission of false financial statements.” Id. The jury found the defendants guilty of conspiracy and bank fraud. Id. The primary issue on appeal related to the sentencing of the defendants and the calculation of loss under the Sentencing Guidelines; the court considered whether “the district court erred by not giving [the defendants] credit for, and thereby not reducing the calculation loss by, the value of assets pledged to [the bank that it] recovered or can be expected to recover.” Id. at 228.
Judge Rosenn, writing for the court, first noted that the court had “plenary review over the district court’s refusal to give the defendants the claimed credits, but [reviewed] the court’s factual findings for clear error.” Id. at 229 (internal citations omitted). Considering the district court’s treatment of the value of a parcel of land that the bank recovered after suing the defendants, the court noted that the bank “only credited the defendants $40,000 toward the loan’s principal, and the [district] court likewise credited the defendants for $40,000.” Id. However, an appraisal of the land by the bank’s appraiser estimated that the land was worth $80,000. Id.
The court held that the district court’s decision to only provide a $40,000 credit was made in error, finding that it was not supported by the evidence. “Because [the bank] ultimately obtained the parcel of land by suit after the defendants’ default and gave the defendants a partial credit for its value, the defendants are entitled to a credit for the full value of the land less SBI’s expenses in the litigation to acquire title.” Id. However, the court also determined that the error was harmless since the additional credit would not have affected the total offense level of the defendants under the Sentencing Guidelines. The court did not mention whether a resale value of the property existed, only the appraisal value and the value of the bank’s credit to the defendants.
*630B. United States v. Napier, 273 F.3d 276 (3d Cir.2001)
Shortly after Sharma was decided, the Third Circuit considered a similar issue in United States v. Napier, 273 F.3d 276 (3d Cir.2001). In Napier, the defendant pleaded guilty to four counts of bank fraud; he admitted to sending false documentation to a bank in order to get a loan. Id. at 277-78. After a few payments, Napier defaulted, and the bank foreclosed on the home. Id. The property was later purchased at the sheriffs sale by the bank for $305,000 and resold a month later for $307,500; Napier originally borrowed $384,000 with an appraisal value of $480,000. Id. At the initial sentencing, the District Court based the loss on the sheriffs sale price of $305,000 and calculated the total offense level at sixteen. Id. at 278. After the sentencing, Napier appealed. Id. The Third Circuit remanded with an instruction to the district court judge to re-sentence the defendant in light of the decision in Sharma. Id. After Napier was re-sentenced, he again appealed. Id. He argued that “the District Court erred in interpreting this court’s mandate on remand by failing to reduce the loss calculation by the appraised value of the property pledged to secure the loan.” Id. at 277. On remand, the district court had based its calculation of loss on the amount that the bank recovered on its subsequent sale of the property after the sheriffs sale, $307,500. Id. at 278. It rejected the defendant’s proposed value of the appraisal made in preparation of Napier’s mortgage, finding that the resale price was a “more reliable indication of the property’s fair market value.” Id.
After recognizing the standard of review, the Third Circuit, on the second appeal, considered the defendant’s argument that the district court erred in using market value, rather than the appraisal value, of the property for calculating the fraud loss. Judge Sloviter, writing for the court, first stated that its instruction to remand the case in light of Sharma was not an instruction to use appraisal value, rather than market value. Id. at 279. Considering the Sharma opinion, the court noted that the district court judge who sentenced Napier “found the $307,500 valuation to be most reliable as it ‘represents the result of an arm’s length transaction between sophisticated parties.’ It reasoned: ‘Such valuation of property, if available, should be used in preference to appraisals, which are, by definition, merely estimates of what property will cost in an arm’s length transaction.’ ” Id. (internal citations omitted). Napier argued that the subsequent sale was not an arm’s length transaction and was not the most reliable method of valuation. The court, in response, only provided a brief holding: “Admittedly, there does not seem to be much evidence on the record as to value, but the government proved to the District Court’s satisfaction that the $307,500 value was more reliable than the appraised value determined more than five years prior to the discovery of the fraud. We have no basis to hold that the District Court’s decision was clearly erroneous.” Id. at 280. The court therefore affirmed the sentence imposed by the district court judge. Id. at 281.
The parties in this case dispute the significance of Napier. The Government contends that Napier instructs district court judges to use the valuation of property in arm’s length transactions, rather than the estimated value of an appraisal. Siciliano and Kodiak, on the other hand, focus on the language of the district court in Napier during the re-sentencing. In the opinion concerning the re-sentencing, Judge James Kelly stated that, in Sharma, “the Third Circuit concluded that an appraisal value of property should be used in preference to the purchase price of that property *631paid by a creditor at a forced sale.” United States v. Napier, 2001 WL 33569, at *2 (E.D.Pa.2001) (Kelly, J.). However, concluding that Sharma does not require the use of an appraisal in every instance, Judge Kelly used the resale of the property after the sheriffs sale to determine the loss in that case. Id. Siciliano and Kodiak contend that the Third Circuit’s decision to affirm Judge Kelly’s sentencing shows that an appraisal is a preferred valuation over a forced sale. While Judge Kelly certainly did make this statement, this Court is unable to find any passage in the Sharma opinion that comes to such a holding. The subsequent Anderson decision by the Third Circuit, although not precedential, also indicates that Judge Kelly’s preference for appraisal over a forced sale is not the current law on loss calculation in the Third Circuit. Finally, the Government notes in its February 12, 2009 letter that the resale values used for the Category 1 properties are not sheriffs sales, but were sales from HUD to a third party on the open market. See (Doc. 291 — Govt’s Letter of Feb. 12, at 1).
What is clear to this Court is that the Napier opinion suggests that the valuation derived from the sale of a property, such as the values used by the Government for the Category 1 properties, are preferred to appraisal valuations.
C. United States v. Anderson, 216 Fed.Appx. 258 (3d Cir.2007)
In United States v. Anderson, 216 Fed. Appx. 258 (3d Cir.2007), the Third Circuit provided the most recent, albeit non-prece-dential, opinion on the issue. The defendant, Delroy Anderson, pleaded guilty to conspiracy and wire fraud. Id. at 259. Anderson had engaged in a scheme with several others whereby the conspirators falsified credit documents and inflated property appraisals for mortgage applications, then defaulted on the loans. Id. At Anderson’s sentencing, the district court found that Anderson was responsible for a loss of $2,599,409 for his involvement in the scheme. Id. at 260. The district court’s loss calculation was initially developed by an FBI agent; the agent’s calculations were created by “startling] with the principal of each loan; subtract[ing] any amount [the bank] was able to recover through foreclosure sales or refinancing; subtracting] any origination fees earned by [the bank]; addfing] any costs associated with foreclosure; and addfing] any loss of bargained-for interest.” Id. at 262. The defendant argued that “the District Court should not have relied on the amounts [the bank] recovered from foreclosure sales because [the bank], in its haste to minimize its losses, sold the relevant properties for less than market value.” Id.
In affirming the district court’s loss calculation, the Third Circuit noted that Anderson did not provide any evidence of the actual market value of the properties to rebut the Government’s proposed methodology. Id. The court also stated that “it defies logic to suggest that [the bank] would do anything less than its utmost to recoup its losses, especially when it had no other means of doing so.” Id. With no evidence that the actual market values of the properties were greater than the amount recovered, the court held that the district court’s loss calculation was not clear error. Id.
V. Analysis of the Third Circuit Opinions and this Court’s Methodology
While the deferential standard of review may blur some of the conclusions that can be drawn from the holdings, these three opinions outline an apparent order of preference in calculating the value of real estate for purposes of loss calculation. First, the Third Circuit’s decision in Shar-ma indicates that the use of an appraisal *632value for a property is an appropriate means of calculating loss. The subsequent decision in Napier, however, suggests that the resale price of a property is more accurate than an appraisal price. Finally, Anderson indicates a presumption that the actual realized resale price of a foreclosure property is the value to be used, that the bank has every incentive to maximize the price of the property at this sale, and that defendants can not overcome this presumption without providing some evidence to show that these resale prices were not genuine arms-length values.
Considering the evidence and holdings of these three decisions, the Court calculated the loss calculations for Defendants Siciliano and Kodiak as follows.
A. Category 1 Properties
First, the Court decided to distinguish, for purposes of the loss calculation, between Category 1 properties and Category 2 properties. As for those 60 or so properties in Category 1 that have gone through foreclosure and been resold, the Government’s suggested methodology-using the actual prices from the resale — was used, rather than the appraisal prices offered by Siciliano and Kodiak’s expert. Thus, the loss for Category 1 is $1,858,429.18.
B. Category 2 Properties
As to the Category 2 properties, Mr. Pasymowski’s expert report calls into question the accuracy of the Government’s rough averaging.5 Mr. Pasymowski’s opinions are certainly subject to substantial criticism on several grounds. First, Mr. Pasymowski does not include any predicted values for incidental fees and expenses that will result from the foreclosure and *633resale of the properties in computing his predicted gain or loss.6
Second, in creating his formula to determine gain or loss, Mr. Pasymowski simply took the price used for the B sale (rather than the amount remaining on the mortgage) and subtracted his predicted value of the 43 properties to estimate the gain or loss: Loss = B Transaction Resale Price [minus] Predicted Value.7 Using Mr. Pa-symowski’s formula would actually work against Siciliano and Kodiak by increasing the loss attributed to the Defendants.8
Third, Defendant Siciliano initially remarked in her Sentencing Memorandum, in arguing that different Philadelphia neighborhoods have different resale characteristics, that South Philadelphia has been an unusually strong market for the resale of property over the past several years, compared to the other neighborhoods where the remaining properties are located. However, Mr. Pasymowski’s 43 properties are all located in South Philadelphia, and he did not provide any predicted values for the other Category 2 properties.
Fourth, the Government also argues that these properties, because they are foreclosure properties, will be less desirable for potential buyers, thereby reducing the resale price, compared to non-foreclosure properties since the potential buyers will likely suspect that the properties remained vacant for some time and were part of a fraud.
Finally, there is a dispute between the parties as to whether Mr. Pasymowski properly assumed that the properties, whose values he predicted, were in average condition. However, he personally inspected the exterior of the 43 properties prior to performing the regression analysis.
Despite these concerns with Mr. Pasy-mowski’s report and conclusions, the regression analysis that he performed to predict the home values was more sophisticated than the Government’s across the board 38% average loss. As the Third Circuit’s decision in Anderson suggests, a defendant may rebut the factual basis of the Government’s proposed sentence enhancement and ultimately prevent the Government from satisfying its burden of persuasion on the issue. Anderson, 216 Fed.Appx. at 262. While the Government’s calculation is not necessarily wrong, Mr. Pasymowski’s report and testimony, specifically the regression analysis that he performed, gave the Court some doubt about the Government’s average of a 38% loss for all properties.
This Court’s doubt on the calculation is significant. For only the Category 1 properties, at a loss amount of $1,858,429.18, the Sentencing Guidelines *634call for a 16 point offense level enhancement. If the Court were to fully adopt the Government’s loss calculation methodology for the Category 2 properties, the Sentencing Guidelines would instead call for an 18-point enhancement to Siciliano and Kodiak’s offense level for the total loss on all properties. See USSG § 2Bl.l(b)(l) (providing that a court should apply a 16 point offense level enhancement if the loss is more than $1,000,000, or should apply an 18 point offense level enhancement if the loss is more than $2,500,000).
C. Conclusion
Applying the principles underlying the rule of lenity,9 because the presentation by Siciliano and Kodiak gave the Court some doubt about the accuracy of the Government’s methodology for the Category 2 properties, this Court only utilized the methodology and loss total proposed by the Government for the Category 1 properties; no additional loss was included for the Category 2 properties, and the total loss was calculated at $1,858,429.18.10