Interest in estimating the potential socioeconomic costs of climate change has led to increasing use of dynamical downscaling: nested modeling in which regional climate models (RCMs) are driven with general circulation model (GCM) output to produce fine-spatial-scale climate projections for impacts assessments. We evaluate here whether this computationally intensive approach significantly alters projections of food supply. Our results suggest that it does not. We simulate U.S. maize yields with the widely-used DSSAT crop model, driven by two GCMs, each in turn downscaled by two RCMs. While RCMs correct some GCM biases related to fine-scale geographic features, errors in yield are dominated by broad-scale (100s of kms) GCM systematic errors that RCMs cannot compensate for. Once a simple statistical bias correction is applied, GCM- and RCM-driven U.S. maize yields are essentially indistinguishable. These results support previous suggestions that the benefits for impacts assessments of dynamically downscaling raw GCM output may not be sufficient to justify its computational demands.