Abstract: This study envisions a future of operating shared autonomous vehicles (SAVs) in 374 small and medium-sized urban areas (population < 375,000) in the United States. Specifically, the study has three main objectives: 1) generating large-scale synthetic trip data for 374 urban areas, 2) employing an agent-based modeling framework to simulate SAV fleet operations in these areas, and 3) developing models to explain variations in SAV operational performance across study areas. The primary goals are to compare SAV performance across urban areas and identify the factors influencing operational performance. The study generated over 30 million synthetic commuting trips for 374 urban areas based on open-access socioeconomic data. The agent-based modeling results were analyzed to evaluate SAV fleet performance across different areas. The results revealed substantial spatial variations in SAV fleet performance across areas. Through statistical modeling, this study quantified the correlates of SAV fleet performance, which are significantly related to regional road network and travel demand. A larger, denser, and more connected network was associated with greater fleet efficiency.