Enhancing Predictive Models for Bid Discounts in Brazilian Transmission Auctions: Integrating Technical Variables, Temporal Dynamics, and Strategic Bidder Information

Annual Allowed Revenue (AAR). These discounts reflect a combination of project fundamentals, regulatory parameters, macroeconomic conditions, and bidder strategy. This paper develops predictive models for bid discounts and winning AAR values in Brazilian transmission auctions, using a comprehensive dataset covering all auctions conducted between 1999 and 2024, comprising 58 auctions and 385 awarded lots. The analysis integrates technical, financial, regulatory, macroeconomic, temporal, and strategic information, combining variables available well in advance of the auctions with information disclosed shortly before bidding.

The empirical strategy combines econometric benchmarks with data-driven predictive models, evaluated strictly out of sample using grouped auction-level splits to preserve the institutional structure of the auction process. The results indicate that bid discounts are inherently difficult to predict, reflecting strategic behavior, structural changes in the auction environment, and residual uncertainty. In contrast, winning AAR values are substantially more predictable, as they remain strongly anchored to regulated revenue ceilings and technical cost drivers. These findings highlight the asymmetric predictability of auction outcomes and underscore the relevance of institutional design and regulatory parameters in shaping bidding behavior in regulated infrastructure concessions.