AN-009: SME-winner share and extensions¶
Intuition (plain-language)
Letting cheaper non-SMEs bid means SMEs win less often — the composition cost of open competition. It shows up equally in direct- and indirect-administration buyers, so the channel is institutionally uniform rather than driven by one type of buyer behaving differently.
Reduced-form motivation layer
The numbers below are from the v1–v4 reduced-form DiDiR pipeline
(scripts/02_analysis.R + companions), which the v8 manuscript
carries as motivation in §1 but does not headline. The canonical
v8 result is the structural counterfactual decomposition — see
AN-010 (decomposition) and
AN-011 (welfare arithmetic).
Question¶
Under open competition (pre-period in switched group 65), what fraction of winning firms are SMEs? Does the open-vs-SME-only difference in SME winner share vary across PBU types?
Design¶
- Sample: completed items in BEC parquet cache, 18-month window.
- Variation: DiDiR with SME-winner indicator as outcome;
heterogeneity cross-cut by
pbu_class(direct administration vs indirect administration). - Specification: linear probability model with item FE; item-clustered SE; PBU-type interaction in the second specification.
- Outcomes: probability winner is SME.
Results¶
The probability that the winning firm is an SME decreases under open
tenders, directly supporting the competition mechanism: open tenders
attract larger non-SME firms that outbid SMEs on price (see
docs/extensions.md and
output/tables/tab_sme_winner.tex).
Interacting the treatment indicator with a direct-administration dummy
reveals broadly similar effects across buyer types
(output/tables/tab_heterog_pbu.tex),
suggesting the competition channel operates consistently across the
institutional landscape.
Output: output/tables/tab_sme_winner.tex,
output/tables/tab_heterog_pbu.tex.
Interpretation¶
The SME-winner share decline is the direct counterpart of the Gelbach composition channel in AN-008. The symmetry across PBU types says the composition shift itself is not driven by institutional-buyer heterogeneity — it tracks the type-conditional cost gradient that the structural decomposition in AN-010 builds on.
Reading: the composition shift is real and pervasive; it is not a buyer-specific artifact. This strengthens the interpretation that non-SMEs are systematically lower-cost on the price-forming margin, not just lower-cost for sophisticated buyers.
Confidence: yellow. The direction is robust; the magnitude is informative but not as load-bearing as the structural decomposition. The PBU-type symmetry is a useful descriptive cross-cut but is not by itself an identification argument.
Follow-ups¶
- Composition-margin Lee bounds (extending the AN-005 trimming logic to the SME-winner indicator) would tighten the composition reading against differential completion across PBU types.
- Cross-cut by non-SME firm size (RAIS-employment quartile of winners) would test whether the composition shift is concentrated on large non-SMEs or on marginal non-SMEs barely above the SME threshold.
- Within-PBU repeated-purchase patterns: do specific PBUs shift their SME-winner share more than others within the broad PBU-type classification? This would identify procurement-officer-level heterogeneity.