Replication Codes

This repository contains the codes used to replicate the results in Marques and Trucíos (2025).

Empirical Application

• Empirical_Application.R computes one-step-ahead conditional variances using GARCH, SV, GAS, and MSGARCH models, each with both standardized Normal and Student-t innovation distributions.
• Tables_App.R generates the tables and performs the Model Confidence Set procedure for the empirical application results.

Five-minute realized variances are freely available from the CaPiRe database. Daily returns were obtained from Economatica.

Monte Carlo Simulation

• MonteCarlo_GARCH-GAS-SV-MS.R runs the one-step-ahead forecasting experiment. To use the code, modify the parameters accordingly, or execute it in batch mode using the following command:
  R CMD BATCH "--args n=2500 type=BR outliers=FALSE" MonteCarlo_GARCH-GAS-SV-MS.R MonteCarlo_2500_BR_FALSE.txt &
  (You can change BR to US, FALSE to TRUE, or adjust the sample size as desired.)
• Tables_MC.R reproduces the results shown in Tables 3 to 6 of the paper.
• Model_Confidence_Set_MC.R performs the Model Confidence Set procedure for the simulation study.

Auxiliary Functions

• DGPs.R defines the data-generating processes used in the simulations.
• Utils_GARCH-GAS-SV.R contains additional functions for model estimation and forecasting.
• Descriptive_Statistics displays the descriptive statistics in Table 7.

References

Marques, F. and Trucíos C. (2025). "GARCH, GAS, SV, and MSGARCH models: Do we really need all of them for forecasting daily volatility?". Submitted