Supplementary Information for Are Solar Developers Responding to Market Signals? An Empirical Test of Azimuth Optimization in U.S. Utility-Scale Solar

This CSV is the analytical input dataset for "Are Solar Developers Responding to Market Signals? An Empirical Test of Azimuth Optimization in U.S. Utility-Scale Solar.” It contains 6,274 fixed-tilt utility-scale solar arrays installed across the United States between 2005 and 2024, distributed across five ISO/RTO regions (CAISO, ISONE, PJM, MISO, NYISO) and three non-ISO areas. Each row is one array. The 25 columns combine array characteristics from GM-SEUS v1.0 (Stid et al. 2025), including observed azimuth, tilt, capacity, latitude/longitude, installation year; ISO assignment via spatial join with HIFLD/ORNL boundary polygons; revenue-optimal and capacity-factor-optimal azimuths from the node-level photovoltaic simulations of Brown & O'Sullivan (2019); modeled wholesale revenues at observed and optimal orientations; derived azimuth gaps and revenue premiums; and afternoon/midday wholesale price ratios summarized from the LBNL ReWEP tool (Millstein et al. 2025). The companion replication script solar_azimuth_replication_v4_2.py (deposited separately) reproduces every statistic, table, and figure in the paper from this file. Its paper_reconciliation() function validates each computed value against the manuscript line by line. Missing values in the revenue and optimization columns (approximately 46% of rows) correspond to arrays outside the four ISOs covered by Brown & O'Sullivan's simulations (CAISO, PJM, MISO, NYISO) or arrays that did not spatially match a pricing node. This is by design and reflects the paper's analytical scope. Upstream data sources are documented in the replication script's docstring and cited in the paper.