We use a nonlinear, nonparametric method to forecast unemployment rates. This method is an extension of the nearest-neighbor method but uses a higher-dimensional simplex approach. We compare these forecasts with several linear and nonlinear parametric methods based on the work of Montgomery et al. (1998) and Carruth et al. (1998). Our main result is that, due to the nonlinearity in the data-generating process, the nonparametric method outperforms many other well-known models, even when these models use more information. This result holds for forecasts based on quarterly and on monthly data.

A new information-based approach for estimating systems of many equations with nonnegativity constraints is presented. This approach, called generalized maximum entropy (GME), is more practical and efficient than traditional maximum-likelihood methods. The GME method is used to estimate an almost ideal demand system for five types of meat using cross-sectional data from Mexico, where most households did not buy at least one type of meat during the survey week. The system of demands is shown to vary across demographic groups.