By applying power law scaling, we propose an extremely parsimonious modeling framework to capture the interest rate term structure movements across all frequencies. We estimate a model with merely five parameters on the U.S. dollar LIBOR from one to 12 months and swap rates from two to 30 years. Due to the extreme parsimony, the five model parameters are estimated with strong statistical significance. Meanwhile, by capturing movements of all frequencies, the model prices all interest rate term structures to near perfection, with the mean absolute pricing error averaging around half a basis point. The model also generates much better out-of-sample forecasting performance on the short-term interest rates than either the random walk assumption or an autoregressive specification. Further specification analysis shows that the power law scaling assumption matches well with data.