In the past decade, much has been learned about regional climate change from dynamical downscaling done with "pseudo-global-warming" (PGW) techniques. Such techniques generate regional climate change signals by adding mean GCM-simulated climate change signals to boundary conditions derived from reanalysis products. This technique focuses on the impact of warming, thermodynamic effects such as moisture increases, increases in tropospheric stability, and systematic changes in large-scale baroclinicity on regional climate. An alternative to PGW---and one with a longer pedigree---is to downscale GCM data directly to produce a regional solution consistent with the global model's dynamics. This technique would appear to be superior, especially for extreme events, since it includes dynamical changes associated specifically with extreme events in the regional solution. Here we examine whether this is true. We compare simulations in Europe and Western North America done with both PGW and direct downscaling, for the same driving GCMs and historical/future time periods. We find that for anthropogenic changes in temperature extremes, PGW and direct downscaling produce nearly identical results. For changes in hydrologic extremes such extreme precipitation, the PGW technique also performs surprisingly well in both regions, with differences from direct downscaling of less than 15%. Analysis of the driving GCM data reveals the reason PGW performs so well: Even for extremes, much of the anthropogenic signal arises from changes in large-scale thermodynamics, e.g. increases in temperature and moisture. Finally, we provide an overview of those regions where PGW can likewise be used with some confidence to downscale future regional climate change signals, due to small contributions from dynamical effects.