More Revenue, Less Administration? How to Test the Yield and Effects of Tax Reform

A proposed study, not a result or a new business reporting duty. Counterfactual scenarios below ask what would happen under alternative rules; they are not measured outcomes.

I think a simpler, harder-to-manipulate tax base might collect more revenue with less administrative effort. “Might” is doing essential work in that sentence. Neither a wider VAT reverse charge nor a differentiated gross receipts tax can be evaluated by multiplying a convenient headline turnover figure by a preferred rate and calling the result a forecast.

This article proposes a method for testing those ideas. It does not report a completed Slovak microsimulation, a measured reform dividend or an estimated rate schedule. Those outputs require data and analysis not obtained here. Start with a reproducible study design and evidence capable of changing the proposal.

Start with the policy specification

A model cannot rescue an undefined tax. Before collecting data, specify the taxpayers, taxable receipts, territorial allocation, sector rules, exemptions, credits, minimum annual liability and timing of recognition. State whether the proposed tax replaces corporate profit tax completely or leaves defined populations under another regime. Specify the treatment of losses and positions created under the old system.

For reverse charge, specify eligible supplies and recipients, deduction assumptions, invoices crossing the transition date and corrections after that date. Identify the legal route being examined. The Commission describes both specified reverse-charge options and special authorisations under Article 395; neither description establishes that every proposed scope is available.9 10 A model of an unavailable regime can inform negotiations, but it is not an implementation forecast.

The baseline must also be explicit. It should represent the existing rules and already-decided changes over the forecast period, not a deliberately inefficient version of current administration. Include a comparison with narrower improvements. Otherwise the exercise may attribute to radical reform savings that could have been achieved through better forms, fewer duplicate requests or clearer guidance.

Know what the published numbers measure

Slovakia’s Financial Administration publishes a corporate-income-tax data category that includes declared tax, claimed losses, return counts and information on taxpayers with zero or negative tax bases.12 These are useful starting points for understanding available administrative concepts. They are not, by themselves, a firm-level dataset linking receipts, margins, liability, payments and sector characteristics.

The administration’s budget-revenue dataset separately describes cumulative cash receipts, including corporate income tax and VAT.13 That distinction matters. A liability declared for a tax period and cash received in a calendar period are different measures. A replacement model must bridge them rather than compare a theoretical new liability with an unmatched current cash total.

The VAT gap requires similar discipline. The Commission distinguishes revenue not collected under existing rules from the policy effects of reduced rates and exemptions, and notes insolvency and administrative error among the sources of uncollected VAT.4 A receipts-tax forecast cannot simply import a VAT-gap figure as additional corporate revenue. A reverse-charge forecast cannot assume that every component is missing-trader fraud.

Build a protected, auditable analytical dataset

I would ask for a legally authorised analytical file linking relevant tax-return items, payment records and business characteristics through controlled identifiers. Access should be limited, logged and separated from public publication. Researchers need a documented lawful route, not an unrestricted copy of commercially sensitive invoice data. Pseudonymisation does not itself grant access rights. If external transfer of microdata is not authorised, assess calculation by an authorised analytical unit in a controlled environment, with independent review of methodology and legally releasable outputs.

The data dictionary should explain coverage, missing values, corrections and the difference between enterprises, establishments and corporate groups. Related-party transactions and intercompany consolidation require particular care. A total that counts the same economic flow at several stages may be appropriate for one receipts-tax design but inappropriate for another. It cannot be silently changed when the result becomes inconvenient.

Reconcile the analytical sample with published aggregates, explaining the remaining differences. Preserve failing and exiting firms, not only surviving businesses with clean records. Cover different economic conditions. If some categories cannot be observed reliably, show their exclusion and plausible importance rather than filling the gap with invented precision.

Separate mechanical liability from behaviour

The first model should hold observed business behaviour fixed. For each firm, calculate liability under the specified proposal and compare it with the matching baseline. This static result answers a limited question: what would the rules have charged on the same recorded activity? It does not predict what businesses would do after the rules changed.

For a hypothetical distributor, imagine that the new tax falls on receipts without a deduction for purchased goods. Its initial calculation might show a higher liability than its profit tax. The next question is not whether the spreadsheet is correct; it is whether the distributor changes prices, contracts, group structure or activity. Each response affects the eventual revenue and distribution of the burden.

Behaviour should therefore be a separate model layer. Identify the evidence used for each response assumption and test alternatives. Distinguish domestic reclassification from real relocation, changes in reporting from changes in production, and one-off transition behaviour from recurring effects. If credible parameters do not exist, publish scenarios rather than disguising guesses as estimates.

Test the motivating cases directly

The study should contain three named test cases, not just an aggregate revenue table. For Slovak produce routed through a Czech distribution centre back to Slovak retail, compare lawful cross-border treatment with a direct domestic chain, controlling for actual logistics. Do not label the route fraudulent. For an automotive exporter, reconcile monthly input VAT, refund claims, settlements and supplier payment dates. A year-end tax receivable is a stock with a particular composition, not evidence of a recurring monthly refund flow.

For the unpaid sole-trader invoice to an accrual-accounting s.r.o., first verify which income-tax expense-deductibility, payment-condition and overdue-liability rules apply to that corporate taxpayer, service and period. The cited personal-income-tax guidance alone does not establish the company’s treatment; only a verified statutory analysis can support an estimate of a timing gap.14 15 Test the author’s illustrative 1.5–5% sector-rate range as a proposal, not as observed American rates or an endorsed Slovak schedule. Include alternatives outside it if the distributional analysis requires them.

High turnover combined with no recorded staff or stock can be a risk signal, not proof of misconduct. Outsourcing or direct delivery could explain that profile. No population count or fraud percentage is inferred from this hypothetical profile. Any future sample needs provenance, coverage and verification against actual transactions.

Measure incidence, not just legal payment

A company writing the cheque does not tell us who ultimately bears its cost. Depending on market conditions, a change may affect selling prices, wages, supplier terms, profits or investment. This is a question for the evaluation design, not an excuse to announce that every new business tax will certainly be passed to consumers.

The study should show results by size, sector, margin, age, export exposure, ownership and supply-chain position where the data allow. Report distributions and vulnerable groups, not only averages. Particular attention should go to low-margin traders, investment-heavy firms, mixed-activity businesses and companies experiencing genuine losses.

A proportionate minimum annual tax should be modelled explicitly rather than hidden inside an average rate. Show whether it is a floor, a creditable payment or an additional charge, who qualifies for relief and how it interacts with the replacement tax. I accept a proportionate minimum annual tax in principle. That does not endorse any particular amount: reliefs, payment capacity and interaction with the main tax determine whether a design meets that principle.

Keep four accounts instead of one headline

The fiscal account should show government receipts by period, transitional costs and continuing administration. The business cash-flow account should show financing requirements, payment timing and refund effects. The compliance account should show work removed and work introduced. The distributional account should show who gains and who loses under each plausible scenario.

These accounts must not be casually added together. Releasing working capital is not the same as creating a recurring annual saving equal to that capital. A payment moved into an earlier month is not necessarily higher long-run revenue. Fewer accounting hours may be a real benefit, but a tax transfer between businesses and the state is not automatically a net gain to society.

Measure actual tasks: returns, receipt classification, corrections, inquiries, software conversion and disputes. For invoice services, compare a public baseline against an improved provider-led design over equal scope and time: total budget and business costs, small-business access and administrative work, practical replaceability, security and operating evidence, usable recourse and competitive space. Include conversion, parallel operation and takeover; test continuity rather than assume ownership determines it. Lost provider revenue is not automatically a social loss, nor a service payment wasted resources. Separate transfers from resource use.

Make uncertainty visible and useful

A credible report should publish a central scenario, adverse scenarios and the assumptions that drive the difference. Test lower compliance gains, more classification disputes, weaker investment, slower software adoption and a difficult economic year. For reverse charge, test displaced fraud and firms without the deduction profile assumed in the simplest liquidity example.

Do not double-count the same improvement under both e-invoicing and a tax-base change. Specify four counterfactual scenarios: the same tax base without the new e-invoicing measure; e-invoicing alone; the tax change alone; and their combination. Keep rate assumptions explicit and vary the economic cycle and enforcement intensity separately. These analytical scenarios do not suspend already-enacted obligations. A static difference is not a causal estimate; choose an empirical comparison group and time window only after establishing data coverage and legal access. Publish code and non-confidential parameter files where possible, with controlled independent replication for protected data. A second team should be able to understand why the result changes when an assumption changes.

A forecast should be allowed to say no

The decision rule should be established before the final numbers arrive. What revenue risk is acceptable? Which groups need protection? How much recurring simplification is required to justify transition costs? What evidence would lead to narrowing, delaying or abandoning the proposal? These are political choices that should be visible, not silently built into a spreadsheet.

I affirmatively favour a public invoicing baseline with optional commercial services and investigation of differentiated receipts taxation and wider reverse charge. Evidence may nevertheless favour improved commercial access or existing taxes. That would defeat a proposed design, not justify erasing the preference before testing it. Publish why the winner performs better against the same criteria, and reject or narrow my proposals where they fail. The objective is an independently checkable decision, not a forecast engineered to confirm my position.