Skip to content
All library documents

Improving GARCH-MIDAS VaR and ES Models with Climate Uncertainty

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

Summary

The document concerns a comparison of GARCH, GJR-GARCH, and GARCH-MIDAS models for FTSE MIB returns, with the Climate Policy Uncertainty Index as the lower-frequency explanatory variable. The researcher reports that conventional GARCH and GJR models perform best in a model confidence set, while GARCH-MIDAS performs worst, and asks whether that outcome is possible.

The response suggests checking whether the model accounts for skewness and fat tails by considering skewed innovations and a Student-t distribution. It also argues that using more simulation or resampling repetitions may improve robustness when calculating VaR and expected shortfall, though at higher computational cost. These are suggestions rather than demonstrated fixes: the answer lacks the data, diagnostics, implementation details, and comparative results needed to confirm that misspecification or repetition count explains the ranking.

Key ideas

  • GARCH-MIDAS can perform worse than standard GARCH or GJR-GARCH in a model comparison.
  • Financial return data may exhibit skewness and fat tails that a normal innovation assumption misses.
  • Skewed innovations and a Student-t distribution are suggested model alternatives.
  • Increasing the repetition count may improve robustness for VaR and expected shortfall calculations, at added computational cost.
  • The proposed changes are hypotheses and are not validated with reported diagnostics or results.

Tags

EquitiesVolatilityRisk managementMachine learning
Full text
# How can I apply the GARCH-MIDAS model to the FTSE MIB using the CPU as an explanatory variable?


# How can I apply the GARCH-MIDAS model to the FTSE MIB using the CPU as an explanatory variable?












I am trying to understand how climate risk impacts the financial market and I am calculating VaR and ES. I am applying the GARCH-MIDAS model to the FTSE MIB, using the Climate Policy Uncertainty Index (CPU) by Konstantinos Gavriilidis as an explanatory variable.

The problem is that within my model confidence set, the best results are given by the GARCH and GJR models, while the GARCH-MIDAS is always the worst. Is this possible?

P.S. I am using functions provided by my professor to calculate the GARCH-MIDAS.

## Answer by Michele Mario Ippolito (score 1)

https://quant.stackexchange.com/a/80724

I would to fix a little stuff in this command: fit_gm_nA <- ugmfit( model = "GM", skew = "NO", distribution = "norm", daily_ret = rendimenti_FTSEMIB_day, mv_m = cpu_mv, K = K, R = 1000, out_of_sample = 639)

- I think that FTSEMIB index have fat tails (but without the values of skewness and kurtosis I don't say that surely) because this is a stylezed fact in financial time series (in particular when we have financial returns). In this case the argument "skew" in your piece of code might be "YES";

- The distribution of your innovation ("distribution") might be "std" for the previous reason. I think that the "shape" parameter that you find in summary statistics of your model might be statistical significant;

- R=1000 is too low. For a more robustness of your model, this argument could be at least 5000. This rise the computation, but could make your model better than the univariate GJR-GARCH model without MIDAS variable when you compute VaR and ES, I think.

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.