general_tv_favar

General Time-Varying FAVAR (GeneralTVFAVAR)

Overview

The General Time-Varying FAVAR combines factor-augmented VAR analysis with general time-varying parameter modeling in a two-step framework. This estimator first extracts factors via principal components, then applies flexible time-varying coefficient priors to model parameter evolution in the factor-augmented VAR, allowing coefficients to evolve freely over time following random walk or mean-reverting processes.

Methodology

The General Time-Varying FAVAR specification features:

• Two-step estimation: factor extraction followed by time-varying VAR estimation
• Principal component factor extraction in first step
• General time-varying coefficients with flexible evolution patterns
• Random walk or mean-reverting parameter dynamics
• Minnesota-style priors combined with time-varying parameter structure

This approach captures both common factor dynamics and flexible parameter evolution without distributional constraints.

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

Time-Varying Parameters

• TVPVolPrior: Prior variance for time-varying parameter innovations

  • Range: (0, ∞)
  • Default: 10000
  • Controls prior belief about parameter variation over time

• TVPVolPriorDims: Prior for parameter evolution variance

  • Range: (0, ∞)
  • Default: 2
  • Controls tightness of priors on parameter change rates

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

• Burnin: Number of burn-in draws

  • Range: [0, ∞)
  • Default: 5000
  • Required burn-in for time-varying parameter sampling

• It: Total number of iterations

  • Range: [1, ∞)
  • Default: 20000
  • Total MCMC iterations including burn-in

• 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 parameter evolution: When coefficient changes may follow various patterns
• Unrestricted structural change: When no specific constraints on parameter evolution
• Exploratory analysis: When investigating general time-variation patterns
• Maximum flexibility desired: When avoiding distributional constraints on parameters

When Not to Use

• Bounded parameters preferred: When economically meaningful parameter bounds are important
• Small sample sizes: When insufficient data for unrestricted time-variation estimation
• Interpretability crucial: When simpler, more interpretable time-variation is needed
• Computational constraints: When computational efficiency is paramount

Practical Considerations

• Most flexible time-varying FAVAR approach
• May lead to parameter paths that are difficult to interpret economically
• Requires careful monitoring of parameter evolution reasonableness
• Benefits from informative priors on parameter evolution variance

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: General Time-Varying FAVAR

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

General Time-Varying Structure

A_{i,j,t} = A_{i,j,t-1} + ω_{i,j,t}
ω_{i,j,t} ~ N(0, Q_{i,j})

or with mean reversion:

A_{i,j,t} = ρ_{i,j} A_{i,j,t-1} + (1-ρ_{i,j}) μ_{i,j} + ω_{i,j,t}

Parameter Evolution Flexibility

• Accommodates random walk parameter evolution
• Allows for mean-reverting parameter dynamics
• Can handle different evolution patterns for different coefficients
• No distributional constraints on parameter values

Computational Characteristics

Performance

• Speed: Moderate to slow due to unrestricted parameter path sampling
• Memory: Higher requirements for storing flexible parameter paths over time
• Convergence: Variable convergence depending on parameter evolution patterns
• Scalability: Scales moderately with system size and sample length

MCMC Structure

• Requires sampling of unrestricted time-varying parameter paths
• May need specialized sampling schemes for efficiency
• Parameter paths may exhibit poor mixing without proper tuning
• Benefits from adaptive MCMC approaches

Factor-Augmented Analysis

Flexible Factor Loading Evolution

• Factor loadings can evolve in any pattern over time
• No restrictions on factor importance changes
• Can capture complex factor transmission evolution
• Maximum flexibility in modeling factor-variable relationships

Economic Interpretation Challenges

• Parameter paths may be difficult to interpret economically
• Evolution patterns may not align with known economic theory
• Requires post-estimation analysis for economic interpretation
• May reveal unexpected time-variation patterns

Comparison with Related Approaches

vs. Beta Time-Varying FAVAR

• Bounds: Unbounded vs. bounded parameter evolution
• Interpretation: More flexible but potentially less interpretable
• Stability: May be less stable due to lack of bounds
• Applications: Better when parameter bounds are not theoretically motivated

vs. Constant Parameter FAVAR

• Flexibility: Maximum flexibility vs. no time-variation
• Complexity: Much more complex estimation and interpretation
• Data requirements: Requires much more data for reliable estimation
• Applications: Better when strong evidence of time-variation exists

vs. Regime-Switching FAVAR

• Evolution pattern: Continuous vs. discrete parameter changes
• Identification: May be harder to identify specific change points
• Interpretation: Continuous evolution vs. clear regime interpretation
• Complexity: Similar computational complexity

Implementation Strategy

Factor Extraction Stage

• Standardize data and handle missing observations
• Select number of factors using information criteria
• Validate factor stability assumptions
• Assess factor interpretability for time-varying context

General Time-Varying Estimation

• Initialize parameter paths using constant parameter estimates or prior information
• Use efficient sampling schemes for unrestricted parameter evolution
• Monitor convergence and mixing of parameter paths
• Implement adaptive MCMC where necessary

Parameter Evolution Monitoring

• Continuously monitor parameter path reasonableness
• Implement checks for explosive parameter behavior
• Use shrinkage priors to prevent unrealistic parameter evolution
• Consider robustness to different prior specifications

Model Validation

Parameter Evolution Analysis

• Examine parameter paths for economic interpretability
• Test for significant time-variation using formal tests
• Compare parameter evolution patterns across different specifications
• Assess parameter evolution predictability and patterns

Time-Varying Model Performance

• Compare in-sample fit with simpler alternatives
• Assess out-of-sample forecasting performance
• Evaluate parameter evolution forecasting accuracy
• Test robustness to different evolution variance priors

Factor Analysis with Time-Variation

• Examine time-varying factor loadings and importance
• Assess factor interpretation consistency over time
• Compare factor contribution evolution patterns
• Validate factor extraction robustness to parameter time-variation

Applications

Unrestricted Structural Change Analysis

• Investigation of parameter evolution without prior assumptions
• Identification of unknown structural change patterns
• Exploratory analysis of time-variation in economic relationships
• Discovery of previously unknown parameter evolution patterns

Flexible Policy Analysis

• Monetary or fiscal policy rule evolution analysis
• Changing policy transmission mechanism identification
• Policy effectiveness evolution over time
• Regime change detection without prior regime specification

Financial Market Evolution Studies

• Unrestricted analysis of changing financial relationships
• Market efficiency evolution without structural assumptions
• Risk factor importance evolution analysis
• Financial integration pattern identification

Macroeconomic Relationship Evolution

• Investigation of evolving macroeconomic relationships
• Business cycle propagation mechanism changes
• International spillover evolution analysis
• Productivity and growth relationship changes

Advantages and Limitations

Key Advantages

• Maximum flexibility in parameter evolution modeling
• No restrictive assumptions on parameter dynamics
• Can reveal unexpected time-variation patterns
• Suitable for exploratory time-variation analysis

Main Limitations

• May produce economically uninterpretable parameter paths
• Requires large datasets for reliable estimation
• Computationally intensive with potential convergence issues
• May overfit to sample-specific patterns

Economic Interpretation Guidelines

Parameter Path Analysis

• Look for economically meaningful parameter evolution patterns
• Identify periods of rapid vs. gradual parameter change
• Connect parameter evolution to known economic events
• Assess parameter path stability and predictability

Structural Change Identification

• Use parameter paths to identify potential structural break dates
• Analyze parameter evolution correlation with policy changes
• Examine parameter evolution during crisis vs. normal periods
• Test parameter change significance and persistence

Factor Evolution Patterns

• Interpret time-varying factor importance in economic terms
• Connect factor loading changes to changing economic structures
• Assess factor evolution correlation with business cycles
• Examine factor interaction evolution patterns

References

• Bernanke, B.S., J. Boivin, and P. Eliasz (2005). "Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) Approach"
• Primiceri, G.E. (2005). "Time Varying Structural Vector Autoregressions and Monetary Policy"
• Koop, G. and D. Korobilis (2010). "Bayesian Multivariate Time Series Methods for Empirical Macroeconomics"
• Eisenstat, E., J.C.C. Chan, and R.W. Strachan (2016). "Stochastic Model Specification Search for Time-Varying Parameter VARs"