Traditional studies typically evaluate the economic feasibility of individual projects either analyzing a specific maintenance plan or assessing the economic feasibility of a new road expansion in the Amazon. While these project-level evaluations are vital, my approach here shifts the focus to a broader network-level dilemma: how a social planner should allocate a single, fixed annual budget between these two competing demands. In this context, the core question is not whether roads matter, but how that fixed budget should be split between keeping the existing network functional and expanding it into new frontiers.

In the technical report, “Welfare-Based Road Investment Choices in the Brazilian Amazon,” I developed a dynamic welfare model analyzing 1,242 federal corridors across the region. Under the model‘s specifications, a planner chooses one core number each year—the maintenance share of the budget. The model then simulates 12 years into the future, weighing mobility gains against direct fiscal costs and environmental damages from carbon emissions and non-carbon ecosystem services.

Under the main calibration, social welfare is maximized at an interior solution of 95% maintenance and 5% expansion. Both pure expansion and pure maintenance deliver worse overall welfare outcomes. But why does maintenance dominate so heavily? Three key mechanisms drive this result:

1. Diminishing Returns: The economic gains from road quality and access level off over time (i.e., each extra unit of quality or access adds less than the last).
2. Quality–Access Complementarity: New expansion projects are only as good as the network supporting them; connectivity gains fundamentally depend on roads being properly maintained.
3. Absorptive Spending Caps: Corridors face hard institutional and physical capacity ceilings. Pouring money beyond what a road corridor can actually absorb results in fiscal waste.

The policy message here is practical rather than absolute. Brazil should treat maintenance-first budgeting as the default institutional posture for federal highways in the Legal Amazon. At the same time, it should preserve a small, highly targeted expansion margin for corridors where access gains are justified and environmental risks are manageable.

It is worth mentioning that this model isn’t designed to replace project-level environmental licensing (EIA/RIMA) or to cherry-pick individual roads, but it does offer a vital network-level benchmark.

Currently, the Brazilian National Department of Transport Infrastructure (DNIT) spends about 74% of its national construction-and-maintenance budget on upkeep. For the Legal Amazon federal highway budget, the analysis suggests tilting slightly further toward maintenance (closer to the model’s 95% optimum) without sacrificing welfare.