Three approaches to hindcasting low-frequency wind-driven shelf currents, bottom pressure, and adjusted sea level are compared at two locations along the west coast of the United States. The ''purely empirical'' approach is a multiple input linear regression in which alongshelf wind stress records at various time lags are fit to the observations. The ''empirical FOWE'' (first-order wave equation) approach assumes the dynamics of a single wind-forced, frictionally damped, coastal-trapped wave and adjusts the choice of free-wave phase speed and frictional decay time to produce the best hindcast of the oceanographic observable. The ''dynamical FOWE'' approach uses the full theory of wind-forced, long coastal-trapped waves to hindcast observables with no adjustment of the free-wave parameters to improve the fit. The results are similatr for both hindcast locations. The purely empirical approach produces the best fit (in a least squares sense) but offers little dynamical insight. At both locations using more than two wind records (one local and one ~500 km to the south) does not significantly improve the hindcast. The other two approaches perform nearly as well as the purely empirical approach. However, the empirical FOWE approach can lead to dynamical inconsistencies because the optimal phase speed may be different for different variables or for the same variable observed at different cross-shelf locations. Further, local forcing along with remote forcing can bias the results, leading to lower phase speed and friction time estimates than are appropriate for the remotely forced signal. The dynamical FOWE approach offers the most physical insight and produces results which are consistent between variables and at different cross-shelf locations. Further, it is the only approach of the three which can be used as a true forecasting tool. Its performance is, however, also degraded by strong local forcing, suggesting that more dynamical sophistication is required. ¿ American Geophysical Union 1988