![IconGems-Print](images/IconGems-Print.png)
463 – SPEED: Methodological Advances in Time Series: BandE Speed Session, Part 1
Efficient Prediction Under Model Instabilities
Tae-Hwy Lee
University of California, Riverside
Shahnaz Parsaeian
University of California, Riverside
Aman Ullah
University of California, Riverside
This paper aims to improve the prediction under model instabilities. Model instability is defined as a permanent change in the parameters of the model. We introduce a combined estimator of the post-break data and full-sample data and show that this combined estimator has a lower MSFE compared to the post-break estimator, which is a standard solution under model instabilities. The combination weight lies between zero and one. Monte Carlo experiment demonstrates our theoretical findings.