Papers from this Research Program¶
This page lists the paper currently in submission preparation and a ranked set of follow-up paper ideas that can be built using the same underlying data (BEC-SP procurement panel 2009–2019, CADE adjudication records, LANCES bid-level export, plus optional linkages to RAIS, IBGE municipal panels, and reference-price data). Each entry includes three target-journal options with my probability estimates of revise-and-resubmit (R&R) at each.
Probabilities are honest reviewer-style assessments; "Confirmed" is not an option for any single-jurisdiction paper from this data lake without cross-data replication. All probabilities assume a clean manuscript at submission.
Reading the R&R probability columns
R&R = invited to revise and resubmit (not Accept). Acceptance probability is roughly half of R&R for the listed top journals. "Strong fit" = my reviewer-side reading of the journal's recent publication mix; "Marginal fit" = stretch.
1. Current paper (in submission preparation)¶
Cheap Signals, Costly Proof: Award-Layer Evidence Triage in Cartel Enforcement¶
Authors: Darcio Genicolo-Martins & Paulo Furquim de Azevedo (Insper)
Version: v20 (May 2026, JLEO submission-clean) manuscript PDF · online appendix
Distinctive contributions:
- Triage as evidence allocation — reframes cartel screening as a resource-allocation problem under costly observability rather than a stand-alone classification problem. Award-layer triage gates costly bid-layer forensics.
- Loser-side cobidder concentration — among always-loser firms, the FL14 ranking achieves AUC 0.924 against adjudication-anchored cobidders; AUC ≈ 0.49 against direct CADE defendants (the predicted null defining the loser-side scope).
- Sequential architecture — Sequential award → bid gatekeeping recovers 131 of 193 cobidders (recall 0.679) while opening bid microdata for only 2,000 of 11,676 firms — an 83% footprint reduction at an 8% recall cost relative to full-observability joint scoring; sequential recall is markedly more robust than joint under temporal holdout.
- Selection + mechanism decomposition of price evidence — the FL-price coefficient sign-reverses between broad sample (+0.064) and overlap-cell ATT (−0.097). The reversal decomposes into a positive selection component (cartels concentrate in high-price cells, ΔQ5−Q1 = +5.58 log-points among non-treated items) and a negative mechanism component (cover-bidders inflate bidder counts by 66% within cell, depressing the winner price relative to reference). The sign-reversal is the empirical signature of cover-bidding theory.
Target journals:
| Tier | Journal | R&R prob. | Fit rationale |
|---|---|---|---|
| 1 | JLEO (J. of Law, Economics, and Organization) | 65–70% | Institutional economics core; cost-of-evidence framing is JLEO-resonant; loser-side scope discipline + mechanism decomposition give two distinct contributions. Single-jurisdiction cap is the binding constraint. |
| 2 | JLE (J. of Law and Economics) | 55–60% | Adjacent home journal. JLE referees tend to want stronger causal identification; the within-cell mechanism (Test 2) is associative not causal, so JLE referees with strong IV preferences may downgrade. |
| 3 | AEJ: Applied Economics | 40–50% | Strong general-purpose journal. Would value the selection + mechanism decomposition methodologically, but the institutional framing is less natural; AEJ:A referees prefer cleaner identification or larger external validity. |
Current submission gating constraints:
- Single-jurisdiction limitation — every audit lives in BEC × CADE.
Cross-jurisdiction replication (ComprasNet federal) would push R&R
to 75–80% at JLEO. See
COMPRASNET_PATH_TO_CONFIRMED.md. - CADE selection bias — cobidder labels reflect CADE's adjudication choices; not within-data testable. Same fix path.
2. Eight follow-up paper ideas using the same data¶
Ideas 2.1–2.5 are ordered by overlap with the current paper, from highest to none. Ideas 2.6–2.8 (added 2026-05-25) round out the set with enforcement-evaluation, reference-price, and structural-damages angles. Several are extensions of the cover-bidding framework; the rest are deliberately disjoint, taking the BEC + CADE + RAIS data lake into different research programs. Where a number of adjudicated cartels caps precision, the entry says so up front — these are reviewer-honest assessments, not pitch decks.
Idea 2.1 — Adaptive Cartels: Strategic Bid-Rigging Response to the 2018 Decreto¶
Overlap with current paper: High — uses FL14 framework directly, just pivots from static detection to dynamic adaptation.
Central question: When the 2018 Decreto 9.412 raised the procurement cap from R$80K to R$176K (and CADE detection methodologies had matured), did cartels adapt their cover-bidding strategies to remain undetected? Did some cartels exit the market, others change tactics?
Empirical strategy: Difference-in-differences around 2018, using items that crossed the cap as treated and items always above/below as controls. Outcomes: cobidder participation patterns, bid-distribution moments, FL turnover. Triangulate with CADE adjudication dates to identify cartels likely aware of detection risk.
Data reuse: All from current paper + AN-020 (DiD around decreto) expanded. Approx 6-8 weeks of work; mostly script extension.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | RAND (Journal of Economics) | 30–35% | Structural IO; cartel adaptation under enforcement risk fits RAND's IO core; small N of adapted cartels is the cap. |
| 2 | AEJ: Applied Economics | 35–40% | Clean DiD design at a real policy threshold; AEJ:A appreciates well-identified shock-response work. |
| 3 | IJIO (Int'l J. of Industrial Organization) | 45–55% | Empirical IO with strong fit for cartel-dynamics literature; IJIO is a more reachable target. |
Why this idea exists: The 2018 Decreto is an underexploited natural experiment. Current paper uses it as a robustness check (DiD null); this paper would use it as the identifying variation for a dynamic question.
Idea 2.2 — The Architecture of Cover Bidding: Cross-Sector Heterogeneity¶
Overlap with current paper: Medium — same theoretical framework (cover bidding), same data (bid-level), but pivots to industrial heterogeneity rather than enforcement triage.
Central question: Does the cover-bidding signature (bid-level gap-to-winner, bidder-count inflation) appear uniformly across procurement sectors, or is it sector-specific? Tests whether cover bidding is a generic cartel technology or specialized by industry structure.
Empirical strategy: Use bid_level_full_v14 to characterize sector-by-sector bid-distribution moments (CV, skewness, gap-to- winner). Sectors: healthcare (group 65, 37), construction, IT services, transportation, paper-and-office. Test whether the selection + mechanism decomposition from AN-039/AN-040 holds within each sector. Cross-sector comparison anchored on CADE adjudications.
Data reuse: bid_level_full_v14 (1.5GB, already loaded for AN-040), plus sector classification from item codes. ~8-12 weeks of work because new sectoral analyses required.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | IJIO (Int'l J. of Industrial Organization) | 50–55% | Strong fit for sectoral empirical IO; bid-distribution literature appears here. |
| 2 | RAND (Journal of Economics) | 35–40% | RAND values structural IO and cross-sector comparison; competes with the harder structural papers there. |
| 3 | J. of Industrial Economics | 40–50% | Good secondary fit; smaller journal but solid for sectoral heterogeneity. |
Differentiator from current paper: focus on industrial structure, not enforcement architecture. The current paper's sequential gatekeeping contribution is absent here; the cover-bidding signature decomposition is the lead.
Idea 2.3 — The Geography of Cartel Networks in Brazilian Procurement¶
Overlap with current paper: Low — same cartel substrate but entirely different methodology (spatial econometrics + network analysis) and question (geography, not detection).
Central question: How do procurement cartels organize spatially? Are they regional clusters, sectoral networks, or politically structured (around state capitals, governing-party alignment)? Maps the cartel-adjacency network across municipalities and PBUs.
Empirical strategy: Build a firm × buyer network from BEC, weight by tender shared with CADE-adjudicated cartels. Spatial autocorrelation analysis on the cartel-adjacency density. Combine with IBGE municipal data (population, political alignment, fiscal capacity) to test whether cartel concentration follows institutional weakness. Could also use TSE political data.
Data reuse: BEC firm × buyer location (PBU coordinates), CADE adjudications, IBGE municipal panel (already in monorepo). ~10-12 weeks of work; requires new GIS / spatial-econ pipeline.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | J. of Economic Geography | 50–60% | Perfect fit for spatial structure of economic activity; smaller journal but high-impact within geography. |
| 2 | Regional Science and Urban Economics | 45–55% | Strong methodological fit for urban-procurement work. |
| 3 | AEJ: Applied Economics | 35–40% | Possible if the spatial story has welfare implications + a clean shock; harder sell otherwise. |
Differentiator: descriptive spatial-network paper. The detection/triage contribution of the current paper is absent; the question is "where" not "how to detect".
Idea 2.4 — Information Frictions in Public Procurement Auctions¶
Overlap with current paper: None — same data, completely different research program (auction theory + market design rather than cartel detection).
Central question: Why don't more firms compete in Brazilian procurement auctions? Tests whether the limited participation reflects information frictions (firms don't know about distant tenders, can't evaluate reference prices, lack capacity for distant logistics), or strategic abstention. Could the procurement system increase competition by changing information disclosure?
Empirical strategy: Auction-level analysis. Outcome: number of bidders. Predictors: distance from firm to buyer, reference-price information availability, modality (Convite vs Pregão), tender value, month-of-year cyclicality. Mechanism tests using policy variation (e.g., reference-price disclosure changes, modality threshold boundaries).
Data reuse: BEC + bid_level + Firms_final (firm characteristics). Distance computed from PBU and firm CEP. ~10 weeks of work.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | J. of Public Economics | 30–40% | Standard home for public-procurement market design; would need strong identification (the modality threshold variation could provide it). |
| 2 | AEJ: Applied Economics | 35–45% | Well-fit for applied empirical work on auction outcomes; would value the policy variation. |
| 3 | Int'l Tax and Public Finance | 40–50% | Smaller but appropriate for procurement-design papers; less competitive. |
Differentiator from current paper: no cartel framing, no adjudication anchor, no FL ranking. This paper is about auction participation and market design — a different intellectual project entirely.
Idea 2.5 — Procurement and Municipal Economic Development: Welfare Effects of Concentrated Supplier Markets¶
Overlap with current paper: None — different theoretical framework (development economics + fiscal federalism), different outcomes (municipal employment, fiscal capacity).
Central question: How does the structure of municipal procurement (supplier concentration, average winning prices, supplier diversity) shape local economic development outcomes? Specifically: does more concentrated procurement (fewer winners getting more business) benefit municipalities through lower prices and faster delivery, or harm them through reduced supplier capacity and fewer local jobs?
Empirical strategy: Municipality panel. Outcomes: formal-sector employment growth, fiscal balance, supplier diversity (HHI of who gets contracts). Predictors: procurement HHI, average winning prices, modality mix. Identification: 2018 Decreto cap shift OR cross- municipality variation in procurement architecture.
Data reuse: BEC + RAIS firm-level employment + IBGE municipal fiscal data + DATASUS health outcomes (optional). Heavy data assembly; ~12-16 weeks of work, larger lift than other ideas.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | World Development | 45–55% | Perfect fit for Brazilian municipal context; less competitive than top journals; values practical-policy work. |
| 2 | J. of Development Economics | 35–45% | Top development journal; would want strong identification + theory; cleaner story needed. |
| 3 | J. of Public Economics | 25–35% | Stretch fit; JPubE referees prefer US/OECD contexts and tighter identification. |
Differentiator: completely outside the cartel-detection framework. A development paper using the same data lake for a fundamentally different question.
Idea 2.6 — Adjudication and Deterrence: What Happens to Cartel Firms After CADE?¶
Overlap with current paper: Low–Medium — reuses the CADE adjudication anchor and the FL ranking, but turns the question from detection to post-detection firm behavior.
Central question: Once CADE adjudicates a procurement cartel, what happens to the firms? Do convicted firms and their cover bidders exit BEC, persist unchanged, or reorganize (new CNPJ roots, recombined bidding partners) to keep operating below the detection radar? Is enforcement deterring conduct or merely displacing it?
Empirical strategy: Event-study around each cartel's CADE adjudication date (observed). For adjudicated firms and their cobidders, track post-event participation intensity, win rates, FL-stratum membership, and co-bidding-partner turnover. A "phoenix" test links pre-event CNPJ roots to post-event entrants sharing addresses, partners (via RAIS), or bidding fingerprints. Untreated always-losers with comparable pre-event participation form the control.
Original contributions to the literature:
- First firm-level recidivism/displacement measurement for procurement cartels in an emerging-market e-procurement setting — the cartel-duration and recidivism literature (Levenstein & Suslow 2006) is built largely on international price-fixing cartels; procurement-specific, firm-level post-adjudication dynamics are thin.
- A behavioral test of deterrence vs displacement — separates Becker-style deterrence (conduct stops) from displacement (conduct migrates to new shells), a distinction the leniency/enforcement literature (Miller 2009) rarely draws with bid-level micro-data.
- The "phoenix firm" channel, operationalized — a reproducible linkage of pre- and post-adjudication identities through bidding fingerprints + RAIS partner/address overlap, usable as an enforcement-evaluation instrument.
Data reuse: Current-paper CADE anchor + FL ranking + AN-020 event scaffolding, plus RAIS partner/address linkage for the phoenix test. ~8–10 weeks; the main lift is the identity-linkage pipeline.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | JLE (J. of Law and Economics) | 30–35% | Enforcement effectiveness is core JLE; the small number of adjudicated SP cartels is the binding constraint, and JLE referees will press on power. |
| 2 | IJIO (Int'l J. of Industrial Organization) | 40–45% | Empirical-IO home for cartel-dynamics and enforcement-response work; more forgiving on N if the displacement mechanism is clean. |
| 3 | Int'l Review of Law and Economics | 50–55% | Strong topical fit and more reachable; values applied enforcement evaluation over heroic identification. |
Differentiator: the current paper detects; this paper asks whether detection changes behavior. No triage-architecture or cost-of-evidence content — the contribution is enforcement evaluation. Binding constraint, stated honestly: the count of distinct adjudicated SP cartels is small, so the event-study is power-limited; cross-state CADE records would be the natural fix.
Idea 2.7 — Reference Prices as a Collusive Instrument¶
Overlap with current paper: Medium — same cover-bidding substrate, but isolates the reference-price (preço de referência) channel that the current paper deliberately sets aside.
Central question: BEC tenders are anchored by an administratively set reference price. Do winners and cover bidders exploit the reference-price process — clustering winning bids just below the anchor, or inflating the anchor itself — so that the reference price becomes a collusive coordination device rather than a competitive ceiling?
Empirical strategy: Characterize the distribution of winning-bid-to-reference-price ratios and test for bunching just below the anchor (McCrary / Cattaneo density tests) in FL-present vs FL-absent tenders. Where reference-price revisions are observed across re-tendered items, test for systematic upward anchor drift in cartel-adjacent environments. Modality split (Convite vs Pregão) as the institutional contrast.
Original contributions to the literature:
- Reference prices as a coordination device, not just a ceiling — the procurement-design literature on awarding and screening rules (Decarolis 2014) treats reference/reserve prices as mechanism parameters; their use as a collusive anchor is largely untested empirically.
- Anchoring (Tversky & Kahneman 1974) imported into procurement collusion — bridges behavioral economics and bid-rigging detection by treating the administrative anchor as the object cartels manipulate, not just a passive ceiling.
- A bunching-based screen at the reference-price margin — a new, microdata-light screen complementary to the loser-side FL screen of the current paper, deployable wherever a reference price is recorded.
Data reuse: BEC + bid_level + the reference-price field (where populated — see caveat). ~8–10 weeks.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | J. of Public Economics | 25–35% | Procurement market-design home; the bunching identification must be airtight and JPubE referees will demand it. |
| 2 | AEJ: Applied Economics | 30–40% | Likes clean bunching designs at administrative thresholds; the behavioral-anchor framing is a plus. |
| 3 | J. of Industrial Economics / IJIO | 40–50% | Reliable IO home for a reference-price-manipulation screen. |
Data-availability caveat: these R&R figures assume the reference-price field is populated for a usable share of tenders. If coverage is thin, the idea collapses to a robustness note inside Idea 2.2 — confirm coverage before committing any time.
Differentiator: the current paper reports the price coefficient as descriptive scope and refuses a damages reading; this paper makes the price-setting process itself the object of study.
Idea 2.8 — What Did the Cartel Cost? A Structural Overcharge Estimate¶
Overlap with current paper: Medium — uses the same cover-bidding framework and bid-level data to do precisely what the current paper refuses to do: estimate damages.
Central question: The current paper deliberately reports no overcharge. This paper takes on the structural counterfactual directly: absent the cover bidders, what would the winning price have been, and what is the implied overcharge in CADE-adjudicated procurement cartels?
Empirical strategy: Structural first-price estimation (Guerre–Perrigne–Vuong nonparametric recovery of latent valuations) on competitive (FL-absent, non-adjudicated) tenders to identify the competitive bidding function; counterfactually price the adjudicated cartel tenders against it. Cover-bidder bids are treated as non-informative (Asker 2010) and excluded from the competitive-fit step. Bound the estimate with the selection + mechanism decomposition (AN-039 / AN-040) so the structural number inherits the paper's honesty about cartels selecting into high-price cells.
Original contributions to the literature:
- First structural overcharge estimate for Brazilian public- procurement cartels — the structural collusion-damages literature rests on a handful of settings (Asker 2010, stamp dealers; Porter & Zona 1993 and Bajari & Ye 2003, US highway/milk procurement); an emerging-market e-procurement estimate is new.
- A selection-corrected structural counterfactual — embeds the current paper's selection/mechanism decomposition into the structural step, netting the damages number of the "cartels choose high-price cells" selection that contaminates naive overcharge regressions.
- Cover-bid exclusion as an identification device — turns the loser-side screen into a structural-estimation input (cleaning the competitive-fit sample), not merely a detector — a reusable move for any structural collusion study with a credible cover-bidder flag.
Data reuse: bid_level_full_v14 + CADE anchor + AN-039/040 decomposition. ~12–16 weeks; the structural pipeline is the heavy lift.
Target journals:
| Tier | Journal | R&R prob. | Rationale |
|---|---|---|---|
| 1 | RAND (Journal of Economics) | 25–35% | Structural-IO home; high value if the counterfactual is credible, but RAND referees are demanding and the cartel N is small. |
| 2 | IJIO (Int'l J. of Industrial Organization) | 40–50% | Best risk-adjusted target for an applied structural overcharge paper. |
| 3 | JLE (J. of Law and Economics) | 35–45% | Damages estimation is squarely law-and-economics; referees will want exactly the selection correction the current paper pioneers. |
Differentiator: this is the damages paper the current one refuses to be. Where the current paper stops at "scope, not damages," this one takes the structural identification head-on — and is candid that the small number of adjudicated cartels bounds how precise any overcharge estimate can be.
Summary: portfolio view¶
| Idea | Overlap | Top journal | Top R&R |
|---|---|---|---|
| Current paper | — | JLEO | 65–70% |
| 2.1 Adaptive Cartels | High | IJIO | 45–55% |
| 2.2 Cover-Bidding Architecture | Medium | IJIO | 50–55% |
| 2.3 Geography of Cartels | Low | J. Economic Geography | 50–60% |
| 2.4 Information Frictions | None | AEJ: Applied | 35–45% |
| 2.5 Municipal Development | None | World Development | 45–55% |
| 2.6 Adjudication & Deterrence | Low–Med | Int'l Rev. Law & Econ | 50–55% |
| 2.7 Reference-Price Instrument | Medium | J. Ind. Econ / IJIO | 40–50% |
| 2.8 Structural Overcharge | Medium | IJIO | 40–50% |
The portfolio mixes:
- Highest R&R targets: J. of Economic Geography (Idea 2.3, smaller but high-fit), then IJIO (Idea 2.2), then JLEO (current paper).
- Most ambitious targets: AEJ:Applied (could fit current paper as Tier 3, or Idea 2.4 as Tier 2 if identification cleaned up).
- Cleanest reach: World Development (Idea 2.5) and J. of Economic Geography (Idea 2.3) both ~50% with good positioning.
- Disjoint papers (2.4 and 2.5) avoid intellectual cannibalization with the current paper at the cost of needing more data work.
- The three added ideas (2.6–2.8) are the natural intellectual continuations once the current paper is in review: 2.6 evaluates whether enforcement works, 2.7 attacks the reference-price channel, and 2.8 is the structural damages paper the current one deliberately declines to be. All three are power-capped by the small number of adjudicated cartels — the same single-jurisdiction constraint that binds the current paper, and the same cross-jurisdiction fix.
Sequencing recommendation: complete current paper R&R cycle first; then pursue Idea 2.1 (Adaptive Cartels) as the natural extension if R1 reviewers like the FL framework, or pivot to Idea 2.3 (Geography) for a clean break that uses similar data without competing intellectually with the current paper.
Last updated: 2026-05-25