Skip to content

Manuscript

Contribution

This paper makes three main contributions:

  1. Cost estimation of SME set-asides: Provides credible causal estimates of the procurement costs imposed by restricting public tenders to SMEs, using a quasi-experimental variation in Sao Paulo, Brazil.

  2. Heterogeneity analysis: Demonstrates that the costs of restricting competition are not uniform---they are approximately 40% larger for high-value items, offering guidance for differentiated policy design.

  3. Competition mechanism: Documents the full causal chain from restricted tenders to reduced competition (fewer firms and bids) to higher prices, with supporting evidence from winner composition and procurement efficiency analyses.

Institutional Background

Since 2005, all public buyer units (PBUs) in the state of Sao Paulo have been required to purchase common goods and services through Bolsa Eletronica de Compras (BEC), an electronic procurement platform. In 2014, the federal SME law made it mandatory to execute exclusive public tenders for SMEs for items valued at R$80,000 or less.

However, between 2014 and 2018, the state of Sao Paulo exempted group 65 (medical, dental, and hospital supplies) from this requirement, based on a joint agreement that health items were strategic goods warranting open competition. In March 2018, a legal opinion from PGE-SP reversed this interpretation, subjecting group 65 to the same SME-favoring rules as all other product groups.

Groups Before March 2018 After March 2018
Group 65 (switched) Opt-out costs = 0 Opt-out costs > 0
Others (always treated) Opt-out costs > 0 Opt-out costs > 0

Data and Sample

Feature Detail
Source BEC administrative records (SEFAZ/SP)
Coverage All standardized goods procurement in Sao Paulo state
Period January 2016 -- December 2019
PBUs 1,344 public buyer units
Items 82,569 distinct items across 76 groups
Transactions 832,984 successful transactions
Treatment group Group 65 (medical/hospital supplies, ~27% of purchases)
Treatment date March 2018

Empirical Strategy

The identification strategy exploits the timing of the policy change (March 2018) that affected only group 65, using a difference-in-differences in reverse (DiDiR) design. DiDiR identifies pre-switch-period effects by comparing the switched group (group 65) with the always-treated group (all other groups).

The main specification is:

\[y_{pigt} = \eta_i + \gamma\,\text{Pre}_t + \beta\,(g65_{pgt} \times \text{Pre}_t) + x\,\delta + \varepsilon_{pigt}\]

where:

  • \(y\) is the outcome (log price, log firms, log bids, or distance)
  • \(\eta_i\) are item fixed effects
  • \(\text{Pre}_t = 1\) if month < March 2018
  • \(g65_{pgt} = 1\) if item belongs to group 65
  • \(x\) includes log quantity and tender type (sealed bid/auction)
  • \(\varepsilon_{pigt}\) is clustered at the item level

The coefficient \(\beta\) captures the pre-switch-period effect of the SME policy on group 65 outcomes.

Time windows: 6 months (Sep 2017--Aug 2018), 12 months (Mar 2017--Feb 2019), and 18 months (Sep 2016--Aug 2019).

Software and Estimation

Component Detail
Language R 4.5
Fixed effects fixest::feols() with lean estimation
Data handling data.table + arrow (parquet cache)
Clustering Item-level (main), group-level (event study)
Figures ggplot2 with cairo_pdf, grayscale theme
Threads 16 (fixest and data.table)