B0354
Title: Nonparametric estimation of non-anticipative optimization strategies
Authors: Bart Claassen - University of Groningen (Netherlands) [presenting]
Diego Ronchetti - Audencia Business School (France)
Abstract: An empirical estimation method is introduced for a class of intertemporal stochastic optimization models for the replication of a benchmark function through a non-anticipative strategy subject to constraints. Examples are replications of asset payoffs and reference utility levels for price-taker investors. The method exploits the entire structural information, and it allows for consistent estimation of the values of the structural parameters that cap the expected magnitude of the replication error at reference levels. It is a kernel-based local GMM approach that minimizes an average local quadratic distance of non-linear functionals of the probability density functions of the state variables from chosen reference levels. The properties of the method in a Markovian setting are described. It is illustrated in the replication of a function of an unspanned stochastic volatility of asset returns in a financial market. It is shown how to estimate the minimal initial endowments and costs for trade execution that bound at chosen levels the expected magnitude of the replication error for price-taker investors with different levels of risk aversion.