Skip to content
All library documents

Using GARCH-X Covariates to Forecast Exchange-Rate Volatility

Article Quant Q&A · Author: Luigi87
Read the original

Summary

The note considers how to forecast currency exchange-rate volatility and compares broad approaches, including simple purchasing-power-parity ideas, factor models, univariate GARCH, and multivariate dynamic conditional correlation models. The author is seeking a model that is practical to implement without being overly simple, and asks for alternatives and references.

The response suggests GARCH-X: a GARCH model that captures volatility persistence while adding external covariates to the mean, volatility, or both. Candidate explanatory variables include deviations from interest rate parity, purchasing power parity and its changes, and productivity differences. This offers a way to combine a time-series volatility process with economic information. The document provides a modeling suggestion rather than empirical comparisons, validation results, or a specification for selecting predictors, so forecasting performance would need to be tested for the currency pair, data, and forecast horizon of interest.

Key ideas

  • GARCH models can represent persistence in exchange-rate volatility.
  • GARCH-X augments that framework with external predictors in the mean or variance equation.
  • Potential covariates include parity deviations, purchasing power parity changes, and productivity differences.
  • The suggested approach is not accompanied by comparative evidence or a defined forecasting specification.

Tags

ForexVolatilityStatisticsMachine learning
Full text
# how to model the volatility of the currency exchange rate


# how to model the volatility of the currency exchange rate












I want to estimate/predict the volatility of the currency exchange rate. I have checked in literature a few models from very simple PPP to econometric factor model forecasting, to GARCH (for univariate), to DCC (dynamic conditional correlation for multivariate). Are there more suitable models? if yes may you mention them and indicate some references? Currently my choice would be to go for a factor model of the exchange rate because it seems to me a not too simple and not too complicated thing to implement. However I would like some advices. Thanks

## Answer by kurtosis (score 1)

https://quant.stackexchange.com/a/55879

Probably your best approach is to use something like a GARCH-X model. That gives you a GARCH setup to capture the persistence of volatility; however, you also bring in other exogenous-ish covariates to better model the mean and/or volatility. That would allow you to bring in (for example) the size of interest rate parity violations, PPP, change in PPP, productivity differences, and more.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.