normal_diffuse_favar_twostep

Two-Step FAVAR with Normal-Diffuse Prior (NormalDiffuseFAVARTwostep)

Overview

The Two-Step FAVAR with Normal-Diffuse Prior combines factor-augmented VAR analysis with Normal-Diffuse priors in a two-step estimation framework. This estimator first extracts factors via principal components, then applies Normal-Diffuse priors to the factor-augmented VAR, providing a balance between informative priors and computational efficiency while handling large datasets.

Methodology

The Normal-Diffuse FAVAR Two-Step specification features:

• Two-step estimation: factor extraction followed by VAR estimation
• Principal component factor extraction in first step
• Normal priors on VAR coefficients with diffuse priors on covariance matrix
• Semi-conjugate structure allowing analytical treatment of coefficients
• Efficient posterior computation through conditional conjugacy

This approach combines the information advantages of factor extraction with flexible prior specification on the covariance structure.

Parameters

Prior Configuration

• Sigma: Method of calculating priors on covariance matrix
  • Values: "ar" (autoregressive) or "eye" (identity-based)
  • Default: "ar"
  • Controls how the prior covariance matrix is constructed

Minnesota-Style Prior Parameters

• Autoregression: Prior on first-order autoregression

  • Range: [0, 1]
  • Default: 0.8
  • Controls the belief in random walk behavior for own lags

• Lambda1: Overall tightness of priors

  • Range: (0, ∞)
  • Default: 0.1
  • Smaller values impose tighter priors (more shrinkage)

• Lambda2: Variable weighting

  • Range: (0, ∞)
  • Default: 0.5
  • Controls relative weights between own-variable and cross-variable lags

• Lambda3: Lag decay

  • Range: (0, ∞)
  • Default: 1
  • Controls how quickly prior importance decays with lag length

• Lambda4: Exogenous variable tightness

  • Range: (0, ∞)
  • Default: 100
  • Controls prior tightness on exogenous variable coefficients

• Lambda5: Block exogeneity shrinkage

  • Range: (0, ∞)
  • Default: 0.001
  • Controls shrinkage when block exogeneity is imposed

Advanced Options

• Exogenous: Priors on exogenous variables flag

  • Default: false
  • Enables specialized priors for exogenous variables

• BlockExogenous: Block exogeneity flag

  • Default: false
  • Imposes block exogeneity restrictions

MCMC Settings

• It: Total number of iterations

  • Range: [1, ∞)
  • Default: 20000
  • Total MCMC iterations for sampling covariance matrix

• Burnin: Number of burn-in draws

  • Range: [0, ∞)
  • Default: 5000
  • Required burn-in for covariance matrix sampling

• StabilityThreshold: Threshold for maximum eigenvalue magnitude

  • Range: (0, ∞)
  • Default: null (no stability check)
  • Ensures VAR stability by rejecting draws with eigenvalues above threshold

• MaxNumUnstableAttempts: Maximum number of unstable sampling attempts

  • Range: [1, ∞)
  • Default: 1000
  • Maximum attempts to find stable draw before stopping

Usage Guidelines

When to Use

• Flexible covariance modeling: When diffuse priors on covariance are preferred
• Large datasets: When dealing with many observable series
• Moderate computational requirements: When full conjugacy is not essential
• Robust inference: When avoiding strong assumptions on covariance structure

When Not to Use

• Fast computation required: When conjugate alternatives are preferable
• Strong covariance priors desired: When informative priors on covariance are important
• Very large systems: When computational burden becomes excessive
• Simple applications: When simpler priors are sufficient

Practical Considerations

• Requires MCMC for covariance matrix but analytical updates for coefficients
• More flexible than fully conjugate approaches
• Moderate computational requirements between conjugate and fully non-conjugate
• Good balance of prior flexibility and computational efficiency

Model Specification

Two-Step Procedure

Step 1: Factor Extraction

F_t = Λ X_t + e_t

Factors extracted via principal components from observable dataset X_t.

Step 2: FAVAR with Normal-Diffuse Prior

[Y_t]   [Y_{t-1}]
[F_t] = A [F_{t-1}] + ε_t

Prior Specifications

A | Σ ~ N(M, Σ ⊗ Ω)
Σ ~ Diffuse (improper prior)

where M, Ω follow Minnesota-type restrictions, Σ has diffuse prior.

Posterior Structure

Semi-conjugate structure allows:

A | Σ, Data ~ N(M*, Σ ⊗ Ω*)
Σ | A, Data ~ IW(S*, ν*)

Gibbs Sampling

1. Draw coefficients A given Σ and data (analytical)
2. Draw covariance Σ given A and data (Inverse-Wishart)

Computational Characteristics

Performance

• Speed: Moderate speed due to semi-conjugate structure
• Memory: Moderate requirements for factor-augmented system
• Convergence: Good convergence properties with shorter chains
• Scalability: Scales reasonably with number of factors and observables

Semi-Conjugate Benefits

• Analytical updates for VAR coefficients
• MCMC only needed for covariance matrix
• Faster than fully non-conjugate approaches
• More flexible than fully conjugate methods

Factor-Augmented Analysis

Factor Interpretation

• Extracted factors represent common components in large dataset
• Factor loadings reveal which variables are driven by common factors
• Factor dynamics capture aggregate economic fluctuations
• Factor shocks can be interpreted as aggregate disturbances

Covariance Modeling

• Diffuse priors allow data to determine covariance structure
• More flexible modeling of residual correlations
• Better accommodation of heteroskedasticity patterns
• Robust to covariance misspecification

Comparison with Related FAVAR Approaches

vs. Normal-Wishart FAVAR Two-Step

• Covariance priors: Diffuse vs. informative Inverse-Wishart priors
• Flexibility: More flexible covariance modeling
• Computation: Slightly slower due to covariance sampling
• Robustness: More robust to covariance prior misspecification

vs. Flat FAVAR Two-Step

• Coefficient priors: Informative Minnesota vs. flat priors
• Shrinkage: Better performance through coefficient shrinkage
• Forecasting: Typically superior forecasting performance
• Interpretation: Minnesota priors aid coefficient interpretation

vs. One-Step Normal-Diffuse FAVAR

• Estimation: Two-step vs. simultaneous factor-VAR estimation
• Computational cost: Much faster than one-step approach
• Factor uncertainty: Ignores factor uncertainty but practically efficient
• Implementation: Easier to implement and debug

Implementation Strategy

Factor Extraction Stage

• Standardize data and handle missing observations
• Select number of factors using information criteria
• Validate factor stability across sample periods
• Assess factor interpretability and economic meaning

VAR Estimation Stage

• Apply Normal-Diffuse priors to factor-augmented system
• Use Gibbs sampling for covariance matrix
• Exploit analytical updates for VAR coefficients
• Monitor convergence of covariance parameters

Model Validation

Factor Assessment

• Examine factor loadings for economic interpretation
• Test factor stability using subsample analysis
• Compare factor forecasting performance
• Assess factor contribution to overall model fit

FAVAR Performance

• Compare forecasting accuracy with standard VAR
• Evaluate impulse response functions and their stability
• Assess structural interpretation through factor identification
• Cross-validate factor selection and prior specification

Posterior Analysis

• Examine coefficient posterior concentration
• Assess covariance matrix posterior properties
• Validate VAR system stability
• Check sensitivity to hyperparameter choices

Applications

Macroeconomic Analysis

• Business cycle analysis with many economic indicators
• Monetary policy transmission through multiple channels
• International spillover analysis using global factors
• Real-time nowcasting with factor updates

Financial Applications

• Portfolio risk analysis with factor decomposition
• Stress testing using macroeconomic factors
• Asset pricing with latent factor extraction
• Systemic risk assessment through factor dynamics

Policy Research

• Central bank forecasting with large information sets
• Fiscal policy impact assessment
• Structural change analysis through factor evolution
• International policy coordination analysis

Advantages and Limitations

Advantages

• Balances prior informativeness with covariance flexibility
• Computationally more efficient than fully non-conjugate approaches
• Robust to covariance prior misspecification
• Good forecasting performance through coefficient shrinkage

Limitations

• Still requires MCMC for full posterior inference
• More computationally intensive than fully conjugate alternatives
• Two-step approach ignores factor uncertainty
• May be sensitive to factor extraction methodology

References

• Bernanke, B.S., J. Boivin, and P. Eliasz (2005). "Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) Approach"
• Doan, T., R. Litterman, and C. Sims (1984). "Forecasting and Conditional Projection Using Realistic Prior Distributions"
• Kadiyala, K.R. and S. Karlsson (1997). "Numerical Methods for Estimation and Inference in Bayesian VAR-models"
• Stock, J.H. and M.W. Watson (2002). "Forecasting Using Principal Components From a Large Number of Predictors"