Emergency? DE

Evasion computation: the trial court must do the sums — estimation, target-actual comparison and the input VAT question

Bundesgerichtshof (German Federal Court of Justice), order of 5 February 2026 — 1 StR 510/25, ECLI:DE:BGH:2026:050226B1STR510.25.0 · Full text (PDF, German)

The case

The defendant ran a restaurant as sole shareholder and managing director of a GmbH. According to the findings of the Aurich Regional Court, she and her husband used the software of two till systems and a manipulation app to delete substantial parts of the daily takings on a total of 1,248 days, pocketing them covertly (para 2). The tax adviser — acting in good faith and not privy to the plan — filed annual VAT, corporation tax and trade tax returns for 2015 to 2017, each on the same day, declaring turnover that was too low; the hidden profit distributions of €211,200, €422,392 and €410,551 also went undeclared in the defendant’s income tax returns; and output transactions were likewise concealed in the VAT advance returns for January 2018 to February 2019 — total evasion, on the Regional Court’s calculations: €1,003,643.76 (para 2). The concealed turnover for June 2015 to February 2017 was estimated at €41,198.50 gross per month, based on an undeleted till file (the “L-Bon table”) (para 3); the saved income tax of €259,860, computed at the 25% flat rate for capital income, was confiscated (para 4). The Regional Court imposed an aggregate custodial sentence of three years six months. The First Criminal Senate set aside the convictions for tax evasion (27 completed and two attempted counts), the aggregate sentence and the confiscation order, and recharacterised the till manipulations as falsification of technical records; confiscation of the till systems as instruments of crime stood (operative part; paras 5, 16).

The court’s reasoning

The Senate states the standard which it would repeat verbatim a few weeks later in 1 StR 557/25: “To determine the scale of the tax evasion, the tax arising by operation of law on truthful declarations (the ‘target tax’) must be compared with the tax actually assessed too low as a result of the untruthful declarations (the ‘actual tax’); the difference between these two results yields the amount evaded” (para 7; unofficial translation).

Measured against that, the judgment holds up in no tax category. For corporation and trade tax 2015 to 2017, the contents of returns and assessments are missing; the monthly estimate of €41,198.50 is “not arithmetically traceable”; the annual deficits found contradict the assessment of the evidence; above all, the Regional Court failed to deduct from the profit taxes the VAT evaded — and hence to be additionally assessed — in the same taxable period (para 8, citing BFH, judgment of 8 December 2021 — I R 24/19, para 48). For VAT, the Senate requires findings on the essential content of the returns, including the input VAT amounts claimed: only then is it apparent whether the result was a net liability or an input VAT credit — in the credit case, the offence is completed only upon the tax office’s consent (section 168 sentence 2 of the Fiscal Code) (para 9). The Regional Court had confined itself to “additional turnover”; a coherent computation of the target tax was missing, and given the substantial annual deficits declared, the Senate could not rule out an input VAT surplus in individual periods (para 10). For income tax, the till withdrawals were correctly treated as hidden profit distributions (section 20(1) no. 1 sentence 2 of the Income Tax Act); but here too a traceable target-actual presentation was lacking, and the estimation error infected the quantum of the withdrawals (para 11). Whether the flat computation with 15% corporation tax and 25% flat-rate income tax at least produced non-detrimental minimum amounts, the Senate left open — it would not be so if losses had to be taken into account (para 12).

Two further course-settings follow. The offences committed through the good-faith tax adviser were committed by the defendant as an indirect perpetrator (section 25(1) alternative 2 of the Criminal Code); the company returns filed on the same day therefore stand in unity of offence, not as separate offences (para 13, citing BGH, order of 10 December 2025 — 1 StR 387/25, para 11). And for the retrial, the Senate recalls the nemo tenetur limit: the criminal sanction attaching to filing obligations is suspended for the taxable periods that are the subject of the criminal proceedings notified to the taxpayer (para 18). The confiscation of €259,860 fell with the income tax counts on which it rested (para 15); the conviction for falsification of technical records (section 268(1) no. 1 variant 2, (2) of the Criminal Code) in 1,248 counts and the confiscation of the till systems (section 74(1) of the Criminal Code) stood (para 16).

Where the decision sits in the case-law

The order consolidates the Senate’s settled presentation case-law (inter alia BGH, order of 8 May 2019 — 1 StR 242/18, para 8; order of 19 February 2025 — 1 StR 482/24, paras 6 et seq.) and became, within weeks, the cited authority of the parallel order 1 StR 557/25 (at para 7). Its VAT profile is sharper than the sister case’s: the liability-or-credit switch of section 168 of the Fiscal Code decides not merely the presentation of the computation but the line between a completed and an attempted offence — a point regularly overlooked in till-manipulation cases featuring high declared losses. Noteworthy, too, is the duty to credit the VAT to be additionally assessed against the profit-tax evasion: the trial court must think the tax categories together instead of adding up evasion figures. Finally, the correction of the concurrence assessment joins the recent line on indirect perpetration through good-faith advisers — with immediate consequences for individual sentences and the aggregate sentence. Here as well: no change of course on substance — the till manipulation as such remained punishable and the tills remained confiscated; the order disciplines the arithmetic, not criminal liability.

The literature

The defence literature anticipated the fault lines of this order. Spilker demands that criminal courts must not adopt tax estimates unexamined: required are full judicial conviction on an objectively verifiable basis, a minimum scale of guilt, and in dubio pro reo also as to quantum (Spilker, UR 2026, 121 [127 f.]) — it was precisely on the arithmetical traceability of the estimate that the Regional Court failed here (paras 8, 10). On the relationship of output and input VAT, the same author stresses that input VAT with a direct economic link reduces the scale of the evasion and that what was obtained must be determined autonomously under sections 73 et seq. of the Criminal Code (Spilker, UR 2025, 521 [524]). Dorn warns against transferring the set-off ban schematically to VAT settings, lest economically distorted confiscation amounts result (Dorn, Einziehung im Steuerstrafverfahren, 2025, pp. 118–140; on determining what was obtained, pp. 92–116). Quedenfeld/Füllsack urge coordination with the parallel security instruments (section 324 of the Fiscal Code, section 111e of the Code of Criminal Procedure) (Verteidigung in Steuerstrafsachen, 6th ed. 2025, Part 4 B.IV.3 marginal no. 1115, p. 519) — setting aside the confiscation simultaneously removes the basis of ongoing attachments at their previous level.

Three levels — kept strictly apart

Official headnote: No official headnote has been published. Operative propositions: the scale of the evasion follows from the comparison of target and actual tax (para 7); for VAT, the trial court must establish the essential content of the returns including input VAT and make apparent whether the result was a net liability or a credit — in the credit case, completion depends on the tax office’s consent (section 168 sentence 2 of the Fiscal Code) (para 9).

Administrative practice: No express administrative guidance exists (as at 19 August 2026).

Our conclusion for the defence: Expressly our own assessment: this order is the working manual for the defence in every till and estimation case. Where the indictment builds on tax-audit estimates, every stage must be attacked: the arithmetical derivation of the estimate, its consistency with the court’s own findings, the coherent target-tax computation per tax category and period, the deduction of the VAT to be additionally assessed from the profit taxes, and the input VAT side of the VAT returns. The liability-or-credit question can shift counts from completion to attempt; the concurrence correction reduces the number of individual sentences. No criminal-law promise: the manipulation itself remained punishable as falsification of technical records, and the tills remained confiscated — the order shifts the battle to quantum, completion and asset access.

Defence levers

Defence assessment: POSITIVE · Keywords: evasion computation; estimation; input VAT surplus; section 168 sentence 2 AO; confiscation; unity of offence

FAQ

Why was the confiscation of €259,860 set aside?

It rested on the income tax saved in the 2015 to 2017 counts — and it was precisely that computation which failed review (paras 11–12, 15). Without a viable evasion computation there is no quantifiable proceed of crime; confiscation will be decided afresh at the retrial.

Is the defendant now in the clear?

No. The conviction for falsification of technical records in 1,248 counts and the confiscation of the till systems stand (para 16); the tax evasion charges will be retried before a different economic-crime chamber. The order corrects the computation and the concurrences — it contains no prediction of the outcome.

Source box

Your next step

Till inspection, search or an indictment built on estimates? We take the computation apart — target-actual, input VAT side, confiscation volume — before it becomes final. → Confidential first assessment

Put till management, filing processes and documentation on an audit-proof footing. → VAT CMS Quick Scan

Book your free 15-minute assessment →Or pick a slot now (Mon·Wed·Fri 10–12)
◈ Explore this structure live — infinitely deep