Adaptive grids for the estimation of dynamic models

Abstract This paper develops a method to flexibly adapt interpolation grids of value function approximations in the estimation dynamic models using either NFXP (Rust, Econometrica: Journal Econometric Society, 55, 999–1033, 1987) or MPEC (Su & Judd, Econometric 80, 2213–2230, 2012). Since requires grid structure for approximation be hard-coded into constraints, one cannot apply iterative node insertion refinement; NFXP, adaption by (iteratively) inserting new nodes will generally lead discontinuous likelihood functions. Therefore, we show how continuously moving nodes, technique referred as r -adaption. We demonstrate obtain optimal based on balanced error principle, and implement this approach including additional constraints maximization problem. The is applied two models: (i) bus engine replacement model 1987), modified feature continuous mileage state, (ii) content consumption original data from world’s leading user-generated networks domain music.