Discussion

What This Paper Adds

Public procurement in developing countries sits at the intersection of two literatures that rarely speak to each other. The industrial organization literature has developed powerful tools for recovering unobserved costs from strategic bidding behavior. The development economics literature has long documented how political distortions shape public spending. This paper builds a bridge between them.

Structural IO / Auction Theory
  • GPV inversion: bids → costs
  • First-price sealed-bid auctions
  • Asymmetric bidder types
  • Nonparametric identification
This paper
Development Economics / Governance
  • Political distortions in spending
  • Corruption and connected firms
  • Efficiency of public institutions
  • Procurement reform design

The methodological contribution goes beyond this paper. Development economists studying public auctions often rely on reduced-form comparisons of prices or winning bids. Structural models allow recovery of the latent object that actually matters — the firm’s true cost — which cannot be read off from bids alone because of strategic markup. The approach taken here is transferable to other developing-country settings where governments allocate public works contracts and where some firms have privileged access to non-competitive mechanisms.


When Does Discretion Make Sense?

The core empirical finding is not a simple verdict on government discretion. The answer depends on the type of project and the political context.

Simple projects
~62% of public contracts
Complex projects
~38% of public contracts
Are selected firms more cost-efficient?
No — Type 0 firms have lower costs
Yes — Type 1 firms have lower costs
What drives allocation?
Political cycle + party alignment Elections ↑ discretion by 6.2 pp
Project characteristics Sewage work ↓ discretion by 9.3 pp
Policy implication
Open to public competition — discretion is misused here
Discretion may be justified — but should require technical documentation

The disconnect is the key insight: for the majority of contracts, the allocation decision is driven by political factors rather than cost considerations. The law’s flexibility is being used where it hurts most.


Limitations and Caveats

NoteUnobserved quality

The paper studies production costs, not quality. A firm that builds cheaply but builds poorly is not truly efficient. This concern is real but deliberately minimized by the choice of contracts: small, standardized pavement projects where the scope for quality variation is limited. Road surface composition and thickness are contractually specified and easier to verify than, say, the structural integrity of a bridge. The assumption is not that quality is irrelevant, but that it is less of a first-order concern in this setting than it would be for larger infrastructure.

For studies of larger or more heterogeneous projects, a Common Value framework may be more appropriate — one where project quality and complexity affect all bidders’ costs rather than just private information.

NoteUnobserved heterogeneity at scale

The GPV inversion assumes that cost variation across bids is driven by private information observable to each firm but not to the econometrician. For larger, more complex contracts, project-level unobservables — soil conditions, local subcontractor availability, logistical factors — may be correlated across bidders and with the bids themselves. This introduces a common value component that the current model does not account for. Extending this approach to larger contracts would require incorporating methods for handling unobserved auction-level heterogeneity, such as the framework developed by Krasnokutskaya (2011).

NoteIdentification relies on public auction behavior

Cost estimates for Type 1 firms are recovered from their bids in public auctions, where they compete against Type 0 firms. If Type 1 firms anticipate receiving I3P or direct contracts regardless of their public auction performance, they may bid less aggressively — inflating the estimated strategic markup and, with it, the recovered cost. The direction of this bias, if present, would make Type 1 firms appear more expensive than they truly are in simple projects, and potentially less expensive than they truly are in complex ones.

NoteThe implementation gap

For these methods to be genuinely useful, they need to be applicable within public institutions — not just academic research environments. Estimating structural auction models and stochastic frontiers requires econometric expertise, access to bid-level data, and software infrastructure that most procurement agencies do not currently have. The declining cost of writing and running code, aided by recent advances in AI-assisted programming, may reduce this barrier over time. Development of accessible software tools for evaluating contract allocation efficiency — analogous to what has happened in other areas of applied econometrics — would be a valuable next step.


Open Questions

These methods open two directions for future work that go beyond the scope of this paper.

Regional diagnostics

Where are inefficiencies largest — and why?

The firm-level efficiency estimates can be aggregated to the municipality or state level to map where procurement costs are furthest from the frontier. This is not an accusation of corruption — many gaps reflect administrative capacity, thin contractor markets, or bureaucratic delays rather than political manipulation. But it identifies where reform efforts should be concentrated and where the returns to institutional improvement are highest.

A call to the profession

Structural tools for development questions

Public procurement is one of the largest fiscal activities of developing-country governments, yet it is understudied with structural methods relative to other settings. The approach here — using equilibrium bidding theory to recover costs, then asking whether those costs are efficient — is applicable wherever bid-level data from competitive auctions exists alongside a discretionary allocation mechanism. The invitation is for development economists to bring these tools to bear on governance questions in their own contexts.