Discretion at the Margin: Threshold Manipulation and Electoral Bans in Colombian Public Procurement

Public procurement laws trade rules against discretion with value thresholds: below them, buyers use faster and less competitive procedures. Colombia offers an unusually clean setting to measure how much buyers manipulate estimated values to stay below the line and what it costs. Thresholds are entity-specific (five budget bands set by law) and move every January with the minimum wage, while the round numbers that usually confound bunching estimates stay fixed. Using the universe of Colombian procurement records (SECOP I and II, 2011-2026; 8,569,218 SECOP II processes) and each entity-year's own threshold, we find sharp bunching of estimated values just below both legal cutoffs: 49,631 excess processes below the small-purchase threshold (86% of the mass observed in the last five percent below it) and 3,157 below the tender threshold. The spike tracks the threshold, not the round numbers: it moves with the minimum wage, appears only at each entity's own band, and is an order of magnitude smaller for procedures that carry no value cap. Territorial governments, run by elected mayors and governors, bunch more than appointed national agencies at the small-purchase threshold, even after reweighting for budget band, object and year. Processes bunched below the tender threshold attract half as many bidders as those just above it, obtain smaller discounts, and are more often enlarged after award. A second source of variation, the four-month ban on direct contracting before presidential elections, shows that buyers respond to rules on the timing margin: direct contracting falls by 95% during the ban and rises by 209% in the two weeks before it, in each of four electoral cycles, with a partial shift toward below-threshold procedures. Chile's fixed UTM thresholds display the same pattern, and a cutoff abolished in 2024 shows no bunching afterwards.

Source: arXiv:2610.28965v1 - http://arxiv.org/abs/2610.28965v1 PDF: https://arxiv.org/pdf/2610.28965v1 Original Link: http://arxiv.org/abs/2610.28965v1