Skip to content

The deep-market null is not universal: a residual within-firm gap in the earlier period

🟡 The deep-market within-firm null is not uniform across the urgent panel, but the heterogeneity has two dimensions that behave differently. Splitting the within firm-buyer-item comparison, the administrative urgent coefficient is essentially zero where trade is deep — above-median quantity −0.005, SUS-formulary items −0.001 — and turns positive in below-median-quantity (+0.066***), earlier-period (+0.117***), and non-formulary (+0.101) cells (AN-003, AN-004). Disambiguating those two axes is what this finding is about.

Economic intuition

Two candidate "thin-market leverage" stories look alike in raw splits but mean different things, and only one survives. The quantity axis is scale — small orders cost more because they forgo bulk discounts, not because the same firm charges more (the within-triple quantity elasticity does the work). The period axis is a genuine within-firm gap, but it runs the "wrong" way — the administrative channel is the dearer one, and it fades as the market matures — so it reads as early-market maturation, not a court order handing suppliers power over the sanctioned buyer. Reading the sign honestly is what separates a real non-null from a convenient one.

The quantity axis is scale, not same-firm pricing. A quartile decomposition (61_h4_quantity_quartiles.R) shows the within-triple coefficient varies with order size only through bulk discounts: the within firm-buyer-item log-quantity coefficient is −0.259 (SE 0.074), and once quantity is held fixed there is no systematic same-firm price gradient across quartiles. The order-size heterogeneity is the scale channel (Lost scale), not a markup.

The period axis is a genuine within-firm gap — but read its sign honestly. The earlier-period coefficient survives a within-triple quantity control (62_h4_period_axis.R): it moves from +0.117 (SE 0.037) to +0.168 (SE 0.062, p = 0.007), while the late period stays null. So a real within-firm price gap persists in the earlier period. But the surviving gap is positive — the administrative channel is dearer within firm-buyer-item, not the litigated one — and it fades as the urgent market thickens (a continuous administrative×year interaction declines from −0.025 to −0.019 with the quantity control). That is the opposite direction from "a court order lets the supplier squeeze the sanctioned buyer," so the gap is reported as a bounded non-null, not asserted as litigated-buyer leverage.

What to conclude. The no-broad-same-firm-markup result in No broad same-firm markup is a deep-market statement, not a universal one: a residual within-firm gap persists in the earlier period. But the two natural "thin-market leverage" candidates do not both survive scrutiny — the quantity dimension is scale, and the period gap is administrative-dearer and time-declining, consistent with early-market maturation rather than supplier power over the litigated buyer.

Caveat. These are conditional contrasts within the selected administrative urgent channel — the closest feasible urgent-procurement comparison — not causal estimates of urgency in thin markets. The splits are observational cuts, and the "earlier period" differ on more dimensions than timing alone. The reading is 🟡 because it is a single-source own-project estimate in São Paulo BEC.

Sources.

  • Own analysis: AN-003 (within-firm heterogeneity by quantity, timing, formulary status), AN-004 (market-depth splits).
  • Disambiguation: 61_h4_quantity_quartiles.R (quantity axis = scale), 62_h4_period_axis.R (period axis survives scale control; sign and time-decay).
  • Cross-refs: H:thin-market-supplier-leverage; H:no-broad-same-firm-markup.
  • Validation: backing scripts 48_mechanism_evidence.R, 50_v9_outputs.py.