Tropical cyclones (TCs) and their economic cost risk under climate change are significant concerns globally. Previous studies on TC damage functions and risk assessment are mostly performed based on modeling TC-level damage and thus obtaining the annual average loss for a country or region. The scalability of these damage functions at finer scales has been less systematically explored. In this study, we examine how the model structure, estimated parameters, and model performance of TC damage functions vary with spatial scale. The comparisons are illustrated by fitting two types of damage functions based on reported damage data at the county, province, and TC scales. We find that the newly proposed precipitation-calibrated sigmoidal damage function significantly outperforms the wind-calibrated sigmoidal damage function at three scales of county, province and TC event. Another type of power-law damage function that integrates hazard, exposure, and vulnerability complements the typical sigmoidal damage function because it yields a better fit when estimating direct economic loss above the province scale. Our work provides an empirical assessment of the role of spatial scale and damage function in TC economic impact evaluation and demonstrates the importance of spatially scale-specific policy-making in TC risk management and climate adaptation strategies.