Time-Varying Parameter VAR (TVP-VAR)
Also known as: Time-Varying Parameter Vector Autoregression, TVP-SVAR, Stochastic Coefficient VAR, Zamana Göre Değişen Parametreli VAR
TVP-VAR is a Bayesian multivariate time-series model in which both the VAR coefficients and the shock covariance matrix are allowed to evolve continuously over time as random walks. Introduced by Primiceri (2005) to study U.S. monetary policy transmission, the model captures structural changes and regime shifts without requiring ex-ante knowledge of when breaks occurred, making it indispensable for macroeconomics, finance, and any setting where economic relationships are suspected to be unstable across time.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use TVP-VAR when theory or historical evidence suggests that the relationships among macroeconomic or financial variables shift over the sample period — for example, across monetary policy regimes, financial crises, or structural reforms. The model requires a moderately long time series (typically T ≥ 80 quarters) to identify time variation reliably. Key assumptions include random-walk parameter evolution and log-normally distributed stochastic volatility. It is less appropriate for short samples, high-dimensional systems without shrinkage priors, or purely cross-sectional questions. Alternatives include threshold VAR for discrete regime changes, factor-augmented VAR for large variable sets, or time-varying DFM for high-dimensional panels.
Strengths & limitations
- Simultaneously captures time variation in both coefficients and shock variances without pre-specifying break dates
- Grounded in a coherent Bayesian framework that naturally quantifies parameter uncertainty via posterior credible intervals
- Produces time-varying impulse response functions that reveal how policy or shock transmission evolves across regimes
- Widely adopted benchmark in empirical macroeconomics, ensuring replicability and comparability with a large literature
- Computationally intensive MCMC estimation becomes prohibitive as the number of variables or lags grows (curse of dimensionality)
- Requires long time series — sparse data make it difficult to distinguish genuine time variation from estimation noise
- Random-walk assumption for parameters may be too restrictive when structural changes are abrupt rather than gradual
- Posterior results can be sensitive to prior choices, particularly on innovation variances of the state equations
Frequently asked
How does TVP-VAR differ from a standard VAR with structural breaks?
A VAR with structural breaks assumes coefficients are piecewise constant and requires the researcher to specify — or statistically detect — exact break dates. TVP-VAR instead lets coefficients evolve continuously as random walks, so no breakpoints need to be fixed in advance. This flexibility comes at the cost of higher computational demand and a need for longer samples to identify smooth variation reliably.
Is stochastic volatility mandatory in a TVP-VAR?
No, but it is strongly recommended. Without stochastic volatility the model assumes constant error variances, which is implausible for macroeconomic data spanning different volatility regimes such as the Great Moderation or the Global Financial Crisis. Omitting it can cause the time-varying coefficients to absorb heteroskedasticity, biasing estimates of genuine parameter change. Primiceri (2005) includes stochastic volatility by default for this reason.
What software can estimate TVP-VAR?
Matlab code accompanying Primiceri (2005) is widely circulated and forms the basis of many extensions. R packages such as bvartools and BGVAR implement variants with convenient wrappers. Python users can leverage PyMC or Stan for custom specifications. Many central bank research departments maintain internal Fortran or Julia implementations optimised for speed when running large-scale TVP-VAR analyses.
Sources
- Primiceri, G. E. (2005). Time varying structural vector autoregressions and monetary policy. Review of Economic Studies, 72(3), 821–852. DOI: 10.1111/j.1467-937X.2005.00353.x ↗
How to cite this page
ScholarGate. (2026, June 2). Time-Varying Parameter VAR (TVP-VAR). ScholarGate. https://scholargate.app/en/econometrics/tvp-var
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
Compare side by side →