UNITED STATES of America, Plaintiff-Appellee, v. Isaac AHANMISI, Defendant-Appellant, 324 F. App'x 258

Affirmed in part; vacated and remanded in part by unpublished opinion. Judge SHEDD wrote the opinion, in which Judge NIEMEYER and Judge SCHROEDER joined. Unpublished opinions are not binding precedent in this circuit. SHEDD, Circuit Judge: Isaac Ahanmisi was indicted on 23 counts of aiding and assisting in the preparation of false tax returns in violation of 26 U.S.C. § 7206(2) and two counts of witness tampering in violation of 18 U.S.C. § 1512(b)(3). Ahanmisi was ultimately convicted of 19 counts of aiding and assisting in the preparation of false tax returns. He now appeals his conviction and sentence.

For the following reasons, we affirm Ahanmisi’s conviction, vacate his sentence, and remand to the district court for a new sentencing hearing. Ahanmisi owned and operated a company that prepared tax returns. At sentencing, the evidence established that Ahanmi-si electronically filed 1,729 tax returns during the relevant time period. The government represented to the court that the IRS analyzed the 1,729 returns with a computer program that flagged returns if they met two criteria: (

a) the person’s adjusted gross income was at least $15,000 and (

b) at least 35 percent of the adjusted gross income consisted of

Schedule A deductions. J.A. 555-56. Using this program, the IRS selected 619 returns for further examination, ultimately auditing 512 of those returns. Of that group, 164 of the audited tax payers were contacted and ultimately agreed to pay an additional IRS assessment. The government calculated the average tax loss of those 164 returns, along with the 29 returns of the witnesses who testified at trial, and multiplied that average loss by the 1,729 returns electronically filed by Ahanmisi to estimate the aggregate tax loss for sentencing purposes.

The government estimated the aggregate tax loss to be between $2,500,000 and $7,000,000 dollars. The district court accepted the government’s estimate of Ahanmisi’s aggregate tax loss. As a result, it determined Ahan-misi’s base offense level to be 24. U.S.S.G. § 2T4.1. The court also imposed a two-level enhancement for obstruction of justice pursuant to U.S.S.G. § 3C1.1, yielding an offense level of 28 and a Guidelines range of 78-97 months. Ahanmisi received a 78-month sentence, followed by one year of supervised release on each count, and a special assessment in the amount of $1,900. This appeal followed.

Ahanmisi contends that the district court erred in accepting the government’s estimate of his aggregate tax loss. Specifically, he contends that it was unreasonable for the government to calculate the average tax loss based on a nonrandom sample of returns. The Sentencing Guidelines provide that the court “will simply make a reasonable estimate [of the tax loss] based on the available facts.” Federal Sentencing Guidelines Manual, § 2T1.1 Application Note 1. The government must establish loss by a preponderance of the evidence. United States v. Harris, 882 F.2d 902, 907 (4th Cir.1989).

It is undisputed that the government used a non-random sample to calculate the average tax loss per return. Because the sample was not randomly selected, we cannot conclude that the average tax loss figure of the selected returns was representative of the entire universe of 1,729 returns filed by Ahanmisi. Further, the criteria used by the government may have skewed the average tax loss toward a larger average loss per return than a random sample would have produced.

This non-representative and potentially skewed average tax loss figure does not provide a “reasonable estimate based on the available facts.” Guidelines, § 2T1.1 Application Note 1. Consequently, we find that the government did not meet its burden of establishing a reasonable estimate of the aggregate tax loss. We therefore vacate Ahanmisi’s sentence and remand for resentencing. AFFIRMED IN PART; VACATED AND REMANDED IN PART. . We have examined the issues raised by Ahanmisi concerning his conviction and find them to be without merit. . At oral argument, it was clear that this meant the return's

Schedule A deductions had to amount to at least 35% of adjusted gross income. .The government’s lawyer proffered at trial and reiterated on appeal that the initial sample pool of 619 returns was not chosen randomly. . The sample’s non-randomness distinguishes this case from cases like United States v. Maye, 205 F.3d 1335 (4th Cir.2000) (unpublished), where we upheld a tax loss estimate that was extrapolated from a random selection of tax returns.