Journal of Risk

Regularization effect on model calibration

Mesias Alfeus, Xin-Jiang He and Song-Ping Zhu

  • This paper considers the regularization effect on model calibration
  • It compares two methods to calibrate two popular models that are widely used for stochastic volatility modeling, i.e., the SABR and Heston models, with the time series of options written on Nasdaq 100 index
  • It performs empirical studies to examines the regularization effect on the out-of-sample pricing accuracy.
  • Overall results agree with the calibration literature that adding an extra penalty to an objective function for calibration raises in-sample pricing errors to a higher level, while on a long-time horizon, parameters obtained from a regularized calibration yield better out-of-sample performance
  •  The paper concludes that regularized calibration is only to be recommended when considering out-of-sample pricing for a long-time horizon.

As is well known, the centerpiece of model calibration is regularization, which plays an important role in transforming an ill-posed calibration problem into a stable and well-formulated one. This realm of research has not been explored empirically in much detail in the literature. The goal of this paper is to understand and give an answer to a question concerning pricing accuracy using the parameters resulting from a correctly posed calibration problem in comparison with those inferred from a relaxed calibration. Our empirical findings indicate that regularized calibration is only to be recommended when considering out-of-sample pricing for a long time horizon.

Sorry, our subscription options are not loading right now

Please try again later. Get in touch with our customer services team if this issue persists.

New to View our subscription options

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here