SMU SOE Seminar Series (May 7, 2025): Estimating State Space Models: Simple Corrections for Finite Sample Bias

TOPIC:

ABSTRACT

This paper introduces new methods to adjust for fi…nite sample biases in potentially nonlinear and non-Gaussian state space models without repeatedly optimizing the log-likelihood function. Stability of the bias adjusted estimates is enforced using either a data-driven shrinkage procedure or the indirect inference approach to bias adjustment. The methods are tested on i) a factor model, ii) a Gaussian affine term structure model, iii) a multivariate stochastic volatility model, and iv) a quadratic term structure model. The simulation results show that the methods give lower bias and lower root mean squared errors when compared to the (quasi) maximum likelihood estimator.

Click here to view the CV.