Animated Overview¶
A narrated visual walkthrough of the paper's key ideas and findings.
This animation reflects an earlier version of the paper
The narrated video predates the current "Cheap Signals, Costly Proof" framing and reports superseded numbers (e.g. an AUC of 0.94 against CADE, and a "screen-works" reading). For the current figures see the Paper, Results, and Findings pages. Under the current reproducible, non-circular label (651 adjudication-anchored cobidders), the raw award-layer score reaches only ROC 0.761 / PR-AUC 0.143, and once procurement opportunity is held fixed the within-stratum AUC is ≈ chance (0.471) — there is no robust residual signal net of opportunity. The contribution is a deflationary decomposition framework supporting incumbent-firm triage with retrospective validation, not a working cartel detector.
Note (2026-06-06): this animation was produced under the single-platform framing; the two-platform extension — the federal ComprasNet audit reported in §5, where the deflation replicates under partially overlapping legal anchors — postdates it and is not reflected here. The animation also predates the v24–v25 reframe: the two positive modeled objects — led (v25) by the over-crediting bias characterization (a size-bias object, signs only, with \(\mathrm{CV}(T)\) as a leading-order sufficient statistic / diagnostic and the magnitude read from a synthetic surface anchored at one empirical point per platform, not an estimated curve), plus the enforcer stopping rule stated modestly as a standard cost–benefit (MB = MC) tangency — postdate it as well, as does the new three-document package (paper + online appendix + online supplement). See the Changelog (v23, v24, and v25 entries).
Controls
- Click Play to start the narrated animation (English, via browser text-to-speech)
- Use Prev / Pause / Next buttons or arrow keys to navigate
- Best experienced in Chrome for highest-quality speech synthesis
Scenes covered:
| # | Topic | Key number |
|---|---|---|
| 1 | Title | |
| 2 | The problem: cartels in public procurement | |
| 3 | The detection gap: limited data | |
| 4 | The strange pattern: firms that always lose | |
| 5 | Cover bidders: faking competition | |
| 6 | The screening rule | 41K → 16.8K → 2,735 firms |
| 7 | Evidence: higher prices | +3.6--7.7% |
| 8 | Evidence: validated against CADE | AUC = 0.94 |
| 9 | Evidence: strategic pattern | Pregão +9.3% vs. Convite +3.8% |
| 10 | Three-stage enforcement pathway | Screen → Triage → Investigate |
| 11 | Closing statement |