Main Results¶

This page presents the core empirical findings: OLS price regressions, CADE validation, IV estimates, network-split analysis, and Bajari--Ye tests.

OLS Price Regressions (H1)¶

FL presence is associated with significantly higher negotiated prices across all specifications.

Specification	Coefficient	SE	Effect (%)	$N$
(1) Item + Year FE	0.087*	(0.024)	+9.1%	1,654,447
(2) Item + Year + PBU FE	0.064*	(0.022)	+6.6%	1,654,447
(3) Pregao only (all FE)	0.089*	(0.025)	+9.3%	1,334,729
(4) Convite only (all FE)	0.037	(0.022)	+3.8%	319,718

H1 supported

FL-present tenders exhibit 4--9% higher prices, robust across specifications. The preferred specification (column 2, all FE) yields a 6.6% price premium.

Non-FL (Genuine) Firms¶

Specification	Coefficient	SE	Effect (%)
(2) All FE	0.143*	(0.010)	+15.4%

FL-present tenders have more, not fewer, genuine competitors---ruling out competitive displacement and consistent with cover bidders being added on top of genuine competition.

Coefficient summary — **Figure 3.** Coefficient on FL presence across outcomes, specifications, and estimation methods. OLS estimates in gray; IV estimates in red.

CADE External Validation¶

We cross-match FL firms against 65 CADE procurement cartel convictions (47 firms matched to BEC).

Metric	Value
FL firms co-participating with CADE convicts	193 / 2,735 (7.1%)
Expected rate (permutation, 1,000 draws)	2.0%
Ratio	3.5x
Permutation $p$-value	$< 0.001$

FL detects beyond known cartels

Excluding all CADE-involved markets, the FL coefficient is 0.062 (vs. 0.064 baseline)---virtually identical. The FL screen captures price anomalies beyond the cases already prosecuted by CADE.

Instrumental Variable Estimates (H2)¶

First Stage¶

The leave-one-out instrument is strongly relevant: $F = 396$ (preferred specification).

2SLS Estimates (Preferred: Panel A)¶

DV	OLS	2SLS	First-stage $F$
Log price	0.064	0.194*	396
Log firms	0.167	0.501*	396
Log bids	0.169	0.614*	396
Log non-FL firms	0.143	0.404*	396

IV Magnitude Interpretation¶

The implied signal-to-total-variance ratio:

\[\hat{\lambda} = \frac{\hat{\beta}_{\text{OLS}}}{\hat{\beta}_{\text{IV}}} = \frac{0.064}{0.194} \approx 0.33\]

Given that the FL dummy is a deliberately coarse screen defined by a distributional threshold, a signal-to-noise ratio of one-third is plausible. The OLS estimate is treated as a conservative lower bound and the IV estimate as an upper bound.

LATE interpretation

The IV estimate captures the Local Average Treatment Effect for complier tenders---those whose FL participation is shifted by supply-side availability. The LATE may exceed the ATE if compliers are disproportionately affected by cover bidding.

Network-Split Results¶

FL firms are classified into concentrated-market (high winner HHI) and competitive-market (low winner HHI) subgroups based on co-bidding networks.

Group	$N$ FL firms	Coefficient	SE	Effect (%)
All FL	2,735	0.064*	(0.022)	+6.6%
Concentrated-market FL	1,356	0.032	(0.024)	+3.3%
Competitive-market FL	1,379	0.126*	(0.031)	+13.4%

Price effect concentrates in competitive markets

The FL--price effect is four times larger in competitive markets than in concentrated markets. Cover bidding is redundant where dominant firms already control pricing---but most valuable where genuine competitive threat exists.

Network split — **Figure 5.** FL price coefficient by market concentration level. The price effect concentrates among FL firms operating in competitive markets (low winner HHI).

Bajari--Ye Tests¶

Three Claims¶

Claim 1 --- Exchangeability rejected. The KS test yields $D = 0.15$ ($p < 0.001$): FL bid residuals are drawn from a different distribution than non-FL residuals.

Claim 2 --- Conditional independence rejected. The mean pairwise product of FL residuals is 5.16 (bootstrap $p < 0.001$): when one FL bid is R$1 above its predicted value, the paired FL bid is on average R$5.16 above its prediction.

Claim 3 --- Placebo passes. Randomly assigned fake groups among non-FL bidders yield a pairwise product of 5.80, consistent with common tender-level shocks rather than coordination. Adding tender FE absorbs these shocks, dropping both products substantially while preserving the FL vs. non-FL contrast.

Bid-level evidence of coordination

The Bajari--Ye diagnostics indicate that FL bids are neither drawn from the same distribution as non-FL bids nor independently determined within tenders. This pattern is consistent with cover bidding.

Mechanisms¶

Three mechanism tests jointly narrow the space of alternative explanations:

Mechanism	Test	Result	Interpretation
M1: Selection	Non-FL firm count	+15.4% more genuine competitors	Rules out competitive crowding-out
M2: Reference price	Winning-bid-to-ref-price ratio	-4.1% closer to reference	Consistent with coordination anchor
M3: Reverse causality	Lagged price on FL entry	Elasticity $\approx$ 0.004	Two orders of magnitude too small

Regime densities — **Figure 6.** Bid spread densities: simulated Regime 1 (complementary), simulated Regime 2 (coordinated), and empirical FL distribution. The empirical pattern matches Regime 1.