Productivity in Procurement Auctions of Pavement Contracts in Mexico
Motivation
A recurring concern in development economics is whether public institutions in low- and middle-income countries are capable of allocating resources efficiently, or whether political incentives and weak accountability lead to systematic misallocation. Infrastructure procurement is a natural setting to study this question: governments spend large sums, contracts are observable in administrative records, and the allocation process is legally defined but often subject to discretion.
In Mexico, as in many developing countries, public construction auctions have a reputation for corruption. The firms that win contracts are not always the ones with the lowest costs or the best track record, they are sometimes the ones with the right political connections. This perception is not unfounded: there is substantial empirical and anecdotal evidence linking procurement discretion to rent extraction and cost overruns in Latin America and elsewhere.
The question motivating this paper is deceptively simple: when the Mexican government bypasses open competition and selects firms directly, is it choosing better firms or favored ones?
A Development Economics Question, Answered with Structural IO Methods
This paper sits at the intersection of development economics and industrial organization. The central question, whether government discretion in procurement serves efficiency or facilitates favoritism, is squarely a development economics concern. It touches on institutional quality, state capacity, corruption, and the efficiency of public spending in a middle-income country.
The method, however, is borrowed from structural industrial organization: a nonparametric auction model (Guerre, Perrigne & Vuong, 2000) extended to allow asymmetric bidders (Flambard & Perrigne, 2006), combined with stochastic frontier analysis. This is a deliberate choice. The question demands a method that can recover firm-level cost distributions from observed behavior without imposing strong distributional assumptions, and the structural IO toolkit provides exactly that. The field of economics does not own questions; methods should follow questions, not the other way around.
The Identification Challenge
A natural approach to answering the question would be to compare costs or productivity across allocation mechanisms directly. But there is a fundamental data limitation: firms that receive contracts through direct allocation are not required to disclose their cost structures. There are no public financial records for these firms at the contract level, and the government’s selection criteria are opaque.
The workaround exploits a key feature of the data: most firms that receive directly allocated contracts also participate in public auctions. In public auctions, bids are observable. The structural auction model allows us to invert observed bids into unobserved costs, using the equilibrium condition that a rational bidder shades their bid above cost by an amount that depends on the competitive environment. By recovering pseudo-costs from public auction bids, we obtain cost estimates for both types of firms, those that only win through public auctions (Type 0) and those that also receive contracts through more discretionary mechanisms (Type 1), without ever observing private cost data directly.
The comparison then becomes: do Type 1 firms, the ones the government selects when it has discretion, have systematically lower costs than Type 0 firms?
GPV Inversion, applied to all public auction bids
\[\hat{c}_i \;=\; b_i \;-\; \frac{1 - \hat{F}(b_i)}{\hat{f}(b_i)}\]
Pseudo-cost \(\hat{c}_i\) recovered for every bidder in both firm types
Key Findings
For projects that include sewage work, firms selected to settings with less competition have lower costs than firms that only participate in public auctions. Economies of scope likely explain this advantage.
For small, simple pavement projects (76% of invitation-only auctions), selected firms have higher costs. These contracts should be put out to public auction.
Governments are 6.2% more likely to bypass public auctions in the year before elections. Firm excess costs rise 13.8% near election time in public auctions.
Study at a Glance
| Public Auction | Invitation (I3P) | Direct Allocation | |
|---|---|---|---|
| Contracts (2011–2018) | 617 | 2,633 | 448 |
| Bids observed | 2,784 | — | — |
| Avg. bidders per auction | 6.9 | ≥ 3 (by law) | 1 |
| Competition level | High | Medium | None |
| Government discretion | Low | Medium | High |