Role of stylized features in constructing better estimators

ABSTRACT

This article discusses the role played by stylized features of financial time series in constructing better estimators for the model parameters. We study in detail one such estimator for the transition probabilities of a simple regime switching model. The estimator is based on the squared autocovariances of the time series, which has been discussed in several empirical studies of economic and financial time series. The effectiveness of this estimator in improving the estimation accuracy is investigated, using both finite sample and asymptotic computations. We also report simulation results to confirm our findings and to extend our conclusions over a bigger region of the parameter space.

KEYWORDS:

• Estimation
• Markov switching
• method of moments
• regime switching
• stylized feature
• volatility clustering.

MATHEMATICS SUBJECT CLASSIFICATION:

• 62M09
• 62M10
• 62C15
• 62F12
• 62P05

Acknowledgments

The author thanks John Kolassa, Shubhasis Dey, Kausik Gangopadhyay, and the anonymous referees for many valuable comments and suggestions, which greatly improved the presentation.