Nonparametric estimation of time varying parameters under shape restrictions

In recent years, a lot of econometric literature has been devoted to estimating time varying coefficients in regression models. Here, a new method based on smoothers is proposed, which is able to introduce shape restrictions over the coefficients. The statistical properties of the estimator are obtained for very general situations, including locally stationary regressors. In particular, the procedure provides consistent results for time varying autoregressive models. The practical problem of implementation is also addressed. A data-driven method for selecting the control parameters is provided, together with an algorithm that reduces the computational cost. A simulation study and an application to real data supports the theoretical results. (C) 2004 Elsevier B.V. All rights reserved.