Volatility in exchange rates: Statistical analyses and economic interpretations

Project details

Project manager
Arvid Raknerud
Project term
01.01.2008 - 31.12.2009
Project status
Research field

About the Project

Increased availability of high-frequency data (daily and intra-daily) has led to more widespread interest in continuous time models of financial asset prices (stock prices, exchange rates, etc.) as alternatives to traditional ARCH- and GARCH models in discrete time. Moreover, the financial crisis of 2008 has increased the focus on market risk and volatility. Standard Gaussian models for asset prices, such as the celebrated Black-Scholes-Merton model, do not seem appropriate and the need for implementing more realistic models seems more important than ever.

This project aims to contribute to this research by developing new methods for inference (estimation and prediction) in stochastic volatility models based on non-Gaussian Ornstein-Uhlenbeck (OU) processes with applications to exchange rates. Our approach uses indirect inference methods: First, an auxiliary model based on an approximate Gaussian state space representation of the OU-based model is estimated. Then simulations are made from the underlying OU-model for given parameter values. The parameter value (in the underlying OU-model) which gives the best match between the quasi-likelihood estimate corresponding to the actual data and the quasi-likelihood estimates corresponding to the simulated data, is chosen as the estimate of the true parameter vector.

The project also aims to implement software for indirect inference and quasi-likelihood estimation written in C++ code with user friendly interface in R (which is free software). Some preliminary output from the project can be found on the projects homepage .