Spectral analysis is a useful tool for a limited set of economic hypotheses--those for which frequency-domain interpretation is plausible, behavior across frequencies is not consistent, and the primary frequencies do not largely determine the amount of variance. In addition to these theoretical restrictions, there are practical limitations: the large amount of data required, the limited number of variables that can reasonably be included, and the lack of summary statistics (e.g., regression coefficients that summarize the entire spectrum). The advantages are that the time series is broken down into individual frequencies that can be studied independently or together and that symmetric consideration can be given to both lags and leads between series. The real obstacle to spectral analysis is not technical but in interpretation of the estimates. 26 pp.