Publishing the
Yardstick
Benchmark transparency, data-matching notice, and the design of voluntary compliance instruments in developing economies
Publishing the criteria you select on lets every taxpayer self-correct before enforcement begins. But publication alone is approximately neutral — the dividend appears only when the yardstick is paired with verification notice and disclosure economics.
The argument
Audit coverage of the SME segment rarely exceeds a few per cent in any year, and the deterrence effect decays outside the audited population. Improving an instrument that touches two per cent of taxpayers does not move the other ninety-eight.
A complementary instrument works through information. Australia has published its yardsticks for over a decade — industry benchmark bands, advance notice of data-matching programs, annual compliance focus areas — converting selection criteria into signals that reach every taxpayer at near-zero marginal cost. This paper asks what the instrument consists of, where it fails, and how to build it.
Download the paper PDFHeadline result
Publishing bands alone is approximately neutral — index 98.0 against an opacity baseline of 100. Error correction among honest firms is offset by a licensing effect: strategic firms, now certain of the criterion, report precisely to the visible band edge.
Bands + advance data-matching notice + a penalty differential favouring pre-notification disclosure cuts the modelled gap by roughly one quarter.
before enforcement contact
Transparency is a system, not a document.
Three regimes
200,000 firms; honest (72%), responsive over-claimers (23%), brazen (5%). The administration audits the top two per cent by deviation above the band.
| Regime | Gap index | Strike rate | Honest audited | Bunching | Disclosed |
|---|---|---|---|---|---|
| A — OpacityNO PUBLICATION | 100.0 | 76.3% | 18.4% | 3.8% | — |
| B — Bands publishedALONE | 98.0 | 54.7% | 34.5% | 5.6% | — |
| C — Bands + notice + disclosureFULL ARCHITECTURE | 77.3 | 50.4% | 36.7% | 4.7% | 35.9% |
Table 1. N = 200,000, fixed seed. Gap index relative to regime A = 100. Illustrative parameters — comparative statics, not level prediction.
Publication alone is neutral
Correction among honest firms is cancelled by evasion to the band edge. Bunching rises 3.8% → 5.6%. Robust across threefold variation in perception noise.
Selection quality falls
Culpable behaviour relocates inside the band; residual outliers are increasingly honest firms with unusual ratios. An argument for evidence-chained selection.
The system delivers
Notice plus disclosure differential cuts the gap by 23%, with over a third of over-claims disclosed pre-contact — revenue at correspondence cost, not audit cost.
Design principles
Publish the system, not the number. Bands with methodology, at least one live matching program on a label the bands touch, and a penalty position making early disclosure visibly cheaper than being found.
Confidentiality thresholds. Suppress bands where the population risks identification — and publish the suppression rules. Start with fifteen to twenty-five well-covered industries; a flawed first release discredits the instrument for years.
The informality boundary. Benchmarks reach the lodging population, not the informal sector. A boundary, not a defect — claims beyond it should not be made.
Channels and language. The signal travels through intermediaries or not at all: agents, associations, banks and software providers, in the languages taxpayers use.
Explainable selection. Every case traceable to a published band, notified program or announced focus area — addressing the strike-rate deterioration directly.
Legal instruments. A lawful publication vehicle for the notice regime, plus the data-acquisition powers and penalty provisions to sustain a credible disclosure differential.
Measurement and delivery
Staged publication supplies its own identification strategy: covered industries against similar uncovered ones, difference-in-differences on lodged-ratio distributions. Baselines must be measured before first publication — behavioural effects are unattributable retrofitted. Bunching estimators at the band edge double as the gaming tripwire.
Delivery fits existing structures — PFTAC, ATAF, CIAT and SGATAR for convening and peer review; IMF technical assistance and the TADAT cycle for capacity and measurement anchoring. A Papua New Guinea pilot is developed as a reference implementation in a companion concept note.
The design task is not whether to publish, but to refuse to publish less than the system.
Selected references: Allingham & Sandmo (1972) · Kleven et al. (2011) · Pomeranz (2015) · Naritomi (2019) · Slemrod (2019) · Luttmer & Singhal (2014) · Saez (2010) · Chetty et al. (2011) · Okunogbe & Santoro (2023) · TADAT Field Guide · ATO small business benchmarks.
Maitras Working Paper MWP-2026-02. Simulation code accompanies this paper (N = 200,000; fixed seed 42). Figures are simulation-based illustrations, not empirical estimates. Correspondence: Smart Resources Qld / Maitras.ai, Port Moresby. © 2026 Maitras.ai.