Pre‑2017 1098‑T Mismatches and Prior Civil Audits Can Prove Intent and Support Tax‑Loss Estimates: Fifth Circuit Affirms Broad “Relevant Conduct” in Tax‑Prep Conspiracies

Introduction

In a comprehensive affirmance touching evidence, trial management, and sentencing, the Fifth Circuit upheld the convictions and sentences of three tax return preparers—Adam Earnest (owner of Sunbelt Tax Services), Christopher Randell, and James Klish—arising from a multiyear scheme that claimed millions in false American Opportunity Tax Credits (AOTC, referred to as “Education Credits”) and other fabricated deductions/expenses. The panel’s opinion addresses five disputed areas:

• Admission of prior “other acts” evidence from an earlier employer (American Tax Service) and a Rule 1006 summary chart (Exhibit 605);
• Alleged constructive amendment of the indictment;
• Sufficiency of the evidence on a specific aiding-and-assisting count;
• Sentencing tax-loss estimation (including pre‑2017 AOTC claims lacking Forms 1098‑T and “relevant conduct” spanning two businesses); and
• Denial of a mitigating-role reduction at sentencing for Klish.

The court’s most consequential guidance is twofold. First, it confirms that earlier civil audits and penalties for similar misconduct at a different firm are admissible under Rule 404(b) to show knowledge, intent, and absence of mistake in a later criminal tax-fraud prosecution, with limiting instructions curbing prejudice—even against a co-defendant who did not work at the prior firm. Second, at sentencing, it endorses a broad “relevant conduct” analysis under U.S.S.G. § 1B1.3 for tax schemes that migrate from one business to another, and it approves the use of IRS 1098‑T mismatches (including pre‑2017 filings) as a reasonable proxy for estimating tax loss when considered conservatively and in light of a consistent pattern of fraud.

Summary of the Opinion

After a seven-day trial, a jury convicted Earnest, Randell, and Klish of conspiracy to defraud the United States, and convicted Earnest and Randell of aiding and assisting in the preparation of false tax returns. On appeal, the Fifth Circuit:

• Upheld the admission of extensive Rule 404(b) evidence about prior IRS audits and civil penalties at American Tax Service, finding strong probative value on intent and knowledge and adequate limiting instructions mitigating prejudice, including as to co-defendant Klish;
• Upheld admission of a Rule 1006 summary chart counting thousands of Sunbelt returns that claimed AOTC without a corresponding Form 1098‑T, concluding no expert testimony was required because the chart performed only basic aggregation;
• Rejected claims of constructive amendment, reasoning that careful jury instructions confined the case to Sunbelt conduct and that any “deductions”/“credits” wording variance did not prejudice the defense given the indictment’s table expressly listing “Education Credits”;
• Found sufficient evidence on an aiding-and-assisting count against Earnest where the jury could rationally infer he prepared a return under another preparer’s PTIN; and
• Affirmed a conservative tax-loss estimate exceeding $3.5 million, holding it was reasonable to include pre‑2017 AOTC returns with no 1098‑T on file and to treat closely related conduct at American and Sunbelt as part of the same course of conduct; and
• Affirmed the denial of a mitigating-role reduction to Klish for failure to show he was substantially less culpable than the average participant.

Sentences—100 months (Earnest), 70 months (Randell), and 50 months (Klish)—were affirmed.

Analysis

A. Precedents Cited and Their Role in the Decision

• United States v. Smith, 804 F.3d 724 (5th Cir. 2015) — Standard of review for evidentiary rulings (abuse of discretion). The panel framed its Rule 404(b)/403 analysis with Smith’s deferential standard, examining whether the district court relied on an erroneous legal view or clearly erroneous evidentiary assessment.
• Fed. R. Evid. 404(b) and 403 — Other-acts propensity bar and permissible purposes (intent, plan, knowledge, absence of mistake). The court treated the American audits/penalties as relevant to knowledge/intent and lack of accident, and balanced prejudice under Rule 403.
• United States v. Fortenberry, 860 F.2d 628 (5th Cir. 1988) — Reversal for unduly prejudicial other-crimes evidence in a dissimilar case. Distinguished because the American evidence involved nearly identical misconduct (unsupported credits) by the same actors, unlike Fortenberry’s disparate, unproven violent/ destructive acts.
• Samia v. United States, 599 U.S. 635 (2023) — Jurors are presumed to follow limiting instructions. The panel relied on Samia to reject “guilt by association” concerns raised by Klish and to find the court’s limiting instructions adequate.
• United States v. Jennings, 724 F.2d 436 (5th Cir. 1984) — Summaries of uncomplicated calculations do not require expert testimony. Supported admission of Exhibit 605 without Rule 702/Daubert expert gloss because it simply counted returns with AOTC but no 1098‑T.
• Fed. R. Evid. 1006 — Summaries of voluminous records admissible to avoid burdening the jury with thousands of individual documents. The court emphasized Rule 1006’s efficiency purpose in admitting the Sunbelt-wide summary.
• United States v. Labarbera, 581 F.2d 107 (5th Cir. 1978) and United States v. Delgado, 672 F.3d 320 (5th Cir. 2012) — Cumulative-error doctrine. Delgado’s principle that non-errors cannot accumulate undercut the cumulative-error claim.
• United States v. Bennett, 874 F.3d 236 (5th Cir. 2017); United States v. Isgar, 739 F.3d 829 (5th Cir. 2014); United States v. Jara-Favela, 686 F.3d 289 (5th Cir. 2012); United States v. McMillan, 600 F.3d 434 (5th Cir. 2010) — Constructive amendment vs. variance; standards of review (including plain error). These authorities framed why the jury instructions avoided a constructive amendment and why any terminology variance was harmless.
• United States v. Bolton, 908 F.3d 75 (5th Cir. 2018); United States v. Sertich, 879 F.3d 558 (5th Cir. 2018); United States v. Grant, 850 F.3d 209 (5th Cir. 2017) — Sufficiency review: highly deferential to the verdict; conviction stands if any rational juror could find the elements beyond a reasonable doubt.
• U.S.S.G. §§ 2T1.1, 2T4.1 and § 1B1.3(a)(2); United States v. Johnson, 841 F.3d 299 (5th Cir. 2016); United States v. Harris, 597 F.3d 242 (5th Cir. 2010); United States v. Ritchey, 117 F.4th 762 (5th Cir. 2024); United States v. Powell, 124 F.3d 655 (5th Cir. 1997) — Loss-calculation standards (reasonable estimates) and broad “relevant conduct” factors (temporal proximity, similarity, regularity). These cases supported the district court’s conservative loss estimate and the inclusion of related American/Sunbelt conduct.
• United States v. Patel, 789 F. App’x 981 (5th Cir. 2019) (unpublished) — Prior Fifth Circuit treatment of AOTC/1098‑T loss estimation affirmed despite imperfections. Not binding (5th Cir. R. 47.5.4), but used comparatively to reinforce the approach.
• United States v. Pena, 91 F.4th 813 (5th Cir. 2024); United States v. Castro, 843 F.3d 608 (5th Cir. 2016); United States v. Thomas, 932 F.2d 1085 (5th Cir. 1991) — Mitigating role reductions: burden to show substantial lesser culpability than the average participant and to be “peripheral” to the offense. These underscore why Klish’s showing fell short.

B. The Court’s Legal Reasoning by Issue

1) Rule 404(b)/403: Admitting Prior Civil Audits and Penalties from a Different Firm

The government introduced extensive evidence of Earnest and Randell’s earlier audits and significant civil penalties at American Tax Service for claiming credits (EIC then) without required documentation. Although those prior incidents involved EIC rather than AOTC, the panel emphasized their near-identity in method: claiming tax benefits without maintaining supporting documentation. That similarity made the evidence probative of specific non-propensity purposes—knowledge of due-diligence expectations, intent to file false claims, and absence of mistake. The court also highlighted:

• Overlap in clients and personnel between American and Sunbelt, making the American evidence probative of mens rea at Sunbelt;
• Detailed limiting instructions cautioning jurors against using American-related evidence to convict on Sunbelt charges;
• Samia’s presumption that jurors follow those instructions; and
• Fortenberry’s inapplicability given the dissimilarity there and the lack of proof in those extraneous incidents, whereas here the American evidence mirrored the charged scheme.

2) Rule 1006 Summaries (Exhibit 605) and No Need for Expert Testimony

The summary chart tallied thousands of Sunbelt returns claiming AOTC without a corresponding Form 1098‑T in IRS data. The court treated the exhibit as classic Rule 1006 evidence: a practical way to present voluminous records without parading thousands of clients through the courtroom. Because the chart performed basic counting—no specialized inferences or complex statistical modeling—no Rule 702/Daubert expert was required (Jennings).

The panel rejected the defense suggestion that pre‑2017 returns should be excluded due to then-laxer 1098‑T requirements, noting the chart’s probative value to show a pattern. The lack of a 1098‑T match, while not conclusive in isolation, was one piece in a body of evidence demonstrating systematic claiming of unsupported credits.

3) Constructive Amendment vs. Variance

The defendants argued the trial “morphed” into a prosecution for an uncharged conspiracy at American. The court disagreed. Meticulous jury instructions pinned the charges to Sunbelt conduct starting in 2015 and explicitly told jurors not to convict for any American-related acts. This foreclosed a constructive amendment (McMillan; Bennett).

A second claim said counts alleging “deductions” shifted to proof of “credits.” The indictment’s own table specified “Education Credits,” and any drafting inconsistency between “deductions” and “credits” did not prejudice the defense under variance analysis (Bennett).

4) Sufficiency of the Evidence (Aiding and Assisting Count)

A taxpayer (Stamps) testified that Earnest prepared one of her returns, though the 2016 return bore a different preparer’s PTIN. Trial evidence showed Earnest routinely filed returns under others’ PTINs. Viewing the record in the light most favorable to the verdict, the panel held a rational juror could find beyond a reasonable doubt that Earnest prepared the 2016 return, satisfying the aiding-and-assisting elements (Sertich).

5) Tax-Loss Estimation and “Relevant Conduct”

The government’s total loss estimate ($10,078,767) would have yielded a higher Guidelines base offense level, but the district court took a conservative approach and placed the loss between $3.5 million and $9.5 million. The Fifth Circuit affirmed both the estimation framework and the inclusion of two contested components:

• Pre‑2017 AOTC returns lacking 1098‑T: Even acknowledging imperfect proxies, the court approved relying on IRS 1098‑T mismatches as part of a reasonable estimate where the evidence showed a consistent pattern of unsupported credits. The panel compared this approach to its (nonprecedential) decision in Patel, concluding the estimate was not clearly erroneous—especially as the district court discounted uncertainty and excluded Schedule A/C losses.
• Conduct at a different firm (American) as “relevant conduct” for Sunbelt: Applying § 1B1.3 and Powell’s factors (temporal proximity, similarity, regularity), the court affirmed the inclusion of earlier American returns when sizing loss exposure for the Sunbelt conspiracy. The district court found “the exact same scheme” carried forward—often with the same taxpayers and preparers. Importantly, the panel also stressed that Sunbelt’s returns alone (>4,500 affected returns; nearly $5 million in AOTC without 1098‑T) would have kept the loss above $3.5 million, making any dispute over American returns non-dispositive in this case.

6) Mitigating-Role Reduction (U.S.S.G. § 3B1.2)

Klish prepared 547 returns with unsupported credits and earned the same commission as others. He did not establish a baseline for the “average participant,” nor did he show he was “peripheral” to the offense (Thomas). The district court’s factual findings—that he understood the scheme, made decisions on fraudulent returns, and was neither more nor less involved in planning than peers—were unrebutted. Denial of the reduction was affirmed.

C. Impact and Forward-Looking Implications

1) Evidence and Trial Practice

• Expect more robust use of civil audit/penalty records as admissible Rule 404(b) evidence to prove knowledge and intent. Where prior and charged schemes mirror each other, defendants will have a steep hill limiting such evidence.
• Limiting instructions remain powerful. Samia’s presumption that jurors follow instructions gives district courts confidence to allow probative other-acts evidence even in multi-defendant trials.
• Rule 1006 summaries built from IRS data (e.g., 1098‑T mismatch lists) can be admitted without experts if they simply compile counts or totals. Defense challenges should focus on input integrity or complexity, not on the mere use of aggregated summaries.
• Cumulative-error arguments will fail absent actual errors; appellate courts will not aggregate non-errors.

2) Substantive Tax-Fraud Sentencing

• Loss Estimation: Courts may use IRS information return mismatches (such as missing 1098‑Ts) as a reasonable proxy for fraudulent claims when buttressed by evidence of a pattern and applied conservatively. Defense teams should be prepared with affirmative evidence showing legitimacy of specific categories of credits if they wish to narrow loss estimates (Johnson places the burden to rebut on defendants).
• Relevant Conduct Across Entities: When a scheme continues from one employer to another—especially with overlapping clients and preparers—earlier conduct may be swept into the loss calculus via § 1B1.3. Corporate form changes will not necessarily cabin relevant conduct.
• Drafting Indictments and Jury Instructions: Careful drafting (including specific tables listing credit types) and thorough limiting instructions can blunt constructive-amendment arguments—particularly when proof tracks the indictment’s particulars.

3) Defense Strategy Considerations

• Contemporaneous Objections: Constructive-amendment challenges are stronger with timely objections; otherwise, review shifts to plain error (Isgar).
• Tax-Loss Rebuttal: Provide concrete counter-evidence (e.g., actual 1098‑Ts; school billing statements; bursar records) to show that challenged credits were legitimate, especially in pre‑2017 filings. Without such proof, courts may deem government estimates reasonable.
• Role Reduction: Demonstrate with facts how the defendant’s culpability compares to the “average participant” and why the defendant was truly “peripheral.” Mere non-leadership status is insufficient (Thomas).

Complex Concepts Simplified

• Rule 404(b) “Other Acts” Evidence: The government generally can’t prove “you did it before, so you did it again.” But it can use prior conduct to prove intent, knowledge, plan, or absence of mistake—especially when the prior conduct is similar to the charged conduct.
• Rule 403 Balancing: Even relevant evidence can be excluded if its unfair prejudice substantially outweighs its probative value. Limiting instructions help reduce unfair prejudice.
• Rule 1006 Summaries: When evidence involves thousands of documents (like tax returns), the rules allow a summary chart to help the jury. The underlying documents must be admissible and available to the other side for inspection.
• Constructive Amendment vs. Variance: A constructive amendment occurs when the trial allows conviction for a different crime than the one charged—automatic reversal. A variance is a difference between charge and proof that does not change the offense—reversal only if it prejudices the defense.
• Sufficiency of the Evidence: On appeal, the verdict stands if any rational juror could find guilt beyond a reasonable doubt when the evidence is viewed in the prosecution’s favor.
• Relevant Conduct (U.S.S.G. § 1B1.3): Sentencing considers not just the convicted counts but all related acts forming the same course of conduct or common scheme—assessed by time, similarity, and regularity.
• Tax Loss (U.S.S.G. § 2T1.1): Courts need not compute exact loss; they may reasonably estimate based on available facts, using conservative assumptions where data are imperfect.
• Mitigating Role (U.S.S.G. § 3B1.2): A defendant must prove they are substantially less culpable than the average participant—being “less involved” than leaders is not enough unless the defendant was essentially peripheral.
• Form 1098‑T and AOTC: Universities issue 1098‑T to report tuition/related expenses; the IRS uses it to verify AOTC. While compliance and due-diligence rules evolved over time, the absence of a 1098‑T in IRS data can be probative of an unsupported claim when viewed with other evidence.

Conclusion

United States v. Randell confirms a pragmatic, evidence-centered approach to prosecuting and sentencing tax-preparer fraud within the Fifth Circuit. The decision has three principal contributions:

• It solidifies that prior civil audits and penalties for closely similar misconduct are admissible under Rule 404(b) to show knowledge, intent, and absence of mistake in later criminal cases, provided the court deploys careful limiting instructions;
• It endorses the use of 1098‑T mismatch data— including pre‑2017 returns— as part of a reasonable, conservative tax-loss estimate when the record shows a consistent pattern of unsupported education credits; and
• It interprets “relevant conduct” broadly for tax-fraud schemes that migrate across business entities, focusing on temporal proximity, similarity, and regularity to capture the full scope of the scheme at sentencing.

The opinion’s practical message is clear: where the government can demonstrate continuity of method and actors, courts will admit prior civil tax enforcement history to establish intent, will accept summary charts to present sprawling tax-return data, and will allow conservative, pattern-based estimates of tax loss—particularly in the face of a defendant’s failure to produce countervailing records. For defense counsel, the case underscores the necessity of timely objections, targeted evidentiary rebuttal, and a fact-driven showing for any role reduction. Randell thus stands as a significant clarifier in Fifth Circuit tax-fraud jurisprudence on both trial evidence and sentencing methodology.