A new approach for estimating VAR systems in the mixed-frequency case

Area of Science:

• Econometrics
• Time Series Analysis
• Statistical Modeling

Background:

• Vector Autoregression (VAR) models are crucial for analyzing multivariate time series data.
• Handling mixed-frequency data (e.g., daily and monthly) in VAR systems presents significant estimation challenges.
• Existing methods may lack efficiency or accuracy when dealing with disparate data frequencies.

Purpose of the Study:

• To introduce a novel estimation procedure, MF-IVL, specifically designed for VAR systems with mixed-frequency data.
• To develop a method that leverages the relationship between slow and fast data components for improved estimation.
• To demonstrate the consistency, speed, and accuracy advantages of MF-IVL over traditional methods.

Main Methods:

• The MF-IVL procedure projects slow-moving data components onto contemporaneous and lagged fast-moving components.
• This projection generates instrumental variables essential for the estimation process.
• The method is rigorously analyzed for its statistical properties, including consistency.

Main Results:

• MF-IVL is shown to be generically consistent, providing reliable estimates.
• Simulations indicate that MF-IVL is computationally faster than the extended Yule-Walker procedure.
• The new method demonstrates superior accuracy compared to the extended Yule-Walker approach in mixed-frequency settings.

Conclusions:

• MF-IVL offers a significant advancement in estimating VAR models with mixed-frequency data.
• The procedure's efficiency and accuracy make it a valuable tool for empirical economic research.
• MF-IVL provides a robust alternative for analyzing economic and financial datasets with varying frequencies.