Shared Bidding Algorithms and Competition: Evidence from Electricity Markets
This paper provides empirical evidence that shared autobidding software synchronizes bids across batteries owned by different operators, especially after the ERI scarcity-disclosure reform . Bids from the same provider co-move more strongly in the same region due to provider-driven forecasts and implementation, and co-movement intensifies with provider concentration . A dynamic Bellman analysis estimates the marginal value of stored energy and, via counterfactual reclearing, tests whether observed bidding reflects owner-level incentives or internalisation of same-provider profits; results reject owner-level only deviations for near-margin shares roughly 30-70% and support internalisation across the path . The conduct test finds evidence of internalisation when the focal provider's near-margin MW share is about 30-70%, with lambda around 0.75-1, while outside this range the test loses power . The identified conduct implies a consumer-cost of about \$5.5 million per year, concentrated in the two major autobidder providers (Tesla Autobidder and Fluence Mosaic) . FCAS deviations show no conduct signal, suggesting the observed effects are primarily in energy bidding rather than FCAS services . Overall, the findings contribute to the policy debate on algorithmic collusion and highlight that concentration in the upstream provider layer can shape market outcomes .
