Mind Map

Interactive overview of the paper's logical structure. Click and drag to explore.

mindmap
  root((Frequent Losers<br/>as Cover Bidders))
    **Data**
      BEC Platform
        Sao Paulo state
        2009–2019
      4.5M tender-items
      41K unique firms
      Phases 2 and 3
        Convite sealed bid
        Pregao e-auction
    **FL Definition**
      Always-losers
        win rate = 0%
        16,843 firms
      IQR threshold
        median + 1.5 x IQR
        threshold ~ 14 tenders
      2,735 FL firms
        Treatment: tender has >= 1 FL
    **H1: Screening Marker**
      OLS price premium
        4–9% higher prices
        4 DVs x 4 specs
      CADE validation
        3.5x overrepresentation
        3 of 5 convicted firms
      Sensitivity
        Oster delta = 3.6
        Sensemakr RV = 17.5%
    **H2: Cover Bidding**
      Instrumental Variables
        IGYR shift-share IV
        First stage F = 396
        21% markup LATE
      Bajari-Ye tests
        Exchangeability rejected
        Independence rejected
        Tender FE: coeff 0.38
      Network analysis
        Competitive > concentrated
        FL amplify bid rigging
    **Mechanisms**
      M1 Selection
        +15% genuine firms
      M2 Calibration
        -4% bids closer to ref
      M3 Reverse causality
        Elasticity ~ 0
    **Robustness**
      Threshold sensitivity
        IQR multiplier 1.0–2.5
      Matching
        CEM and IPW
      Cross-fitting
        DML welfare bounds
      DiD
        Sun and Abraham
        Underpowered MDE 7.8%
      Welfare bounds
        2.4% of spending
    **Policy Implications**
      Flag FL participation
      Triage suspicious tenders
      Investigate patterns
      Monitor over time