BVARs:
the Great Grimpen Mire
Boris Demeshev
Oxana Malakhovskaya
2016-09-22
September Equinox
High predictive power
Underused and confused
BVAR = VAR + Bayesian approach
VAR: \[ \begin{cases} y_t =\Phi_{const}+ \Phi_1 y_{t-1} + \Phi_2 y_{t-2} +\ldots + \Phi_p y_{t-p} + \varepsilon_t \\ \varepsilon_t\sim \mathcal{N}(0,\Sigma) \end{cases} \]
Bayesian approach:
Impose some prior distribution on \(\Sigma\), \(\Phi_{const}\), \(\Phi_1\), \(\Phi_2\), …
Use formula for conditional probability \(p(\theta|data) \sim p(data|\theta) \cdot p(\theta)\) to obtain posterior distribution.
Use posterior distribution for forecasting.
Robert Litterman, 1979, Techniques of forecasting using vector autoregressions
Rao Kadiyala and Sune Karlsson, 1997, Numerical Methods for Estimation and Inference in Baesian VAR-Models
Christopher A. Sims and Tao Zha, 1998, Bayesian methods for dynamic multivariate models
Sune Karlsson, 2012, Forecasting with Bayesian Vector Autoregressions
more than 7 000 hits in scholar.google.com
Note: ARMA gives more than 700 000 hits!
Great Grimpen Mire of prior distributions
Great Grimpen Mire of software
No MCMC in bachelor probability course
Structural: \(Ay_t = B_0+ B_1 y_{t-1} +\ldots + B_p y_{t-p} + u_t\)
Reduced form: \(y_t =\Phi_{const}+ \Phi_1 y_{t-1} + \Phi_2 y_{t-2} +\ldots + \Phi_p y_{t-p} + \varepsilon_t\)
Link: \(B_i = A \Phi_i\).
The parameters \(\Phi_i\) may change over time.
\(y_t =\Phi_{const}+ \Phi_1 y_{t-1} + \Phi_2 y_{t-2} +\ldots + \Phi_p y_{t-p} + \varepsilon_t\)
We should impose prior on \(\Sigma\), \(\Phi_{const}\), \(\Phi_1\), \(\Phi_2\), …
Here \(y_t\) is multivariate: \(m\times 1\).
For \(m=10\) variables and \(p=4\) lags we have more than 400 parameters.
Confusing names for priors
No clear classification of priors
Contradictory notation
Results in:
Coding mistakes
You should struggle a lot to understand
Underuse of BVARs
\[ \begin{cases} y_t =\Phi_{const}+ \Phi_1 y_{t-1} + \Phi_2 y_{t-2} +\ldots + \Phi_p y_{t-p} + \varepsilon_t \\ \varepsilon_t\sim \mathcal{N}(0,\Sigma) \end{cases} \]
We should place prior on:
Covariance matrix \(\Sigma\)
Coefficients \(\Phi_i\)
Prior \(p(\Sigma)\) and prior \(p(\Phi_{const}, \Phi_1, \ldots, \Phi_p)\)
Prior \(p(\Sigma)\) and prior \(p(\Phi_{const}, \Phi_1, \ldots, \Phi_p | \Sigma)\)
\[ \begin{cases} \Sigma \sim \mathcal{IW}(\underline S,\underline \nu) \\ \phi \sim \mathcal{N}(\underline \phi, \underline \Xi) \\ p(\phi, \Sigma) = p(\phi)\cdot p(\Sigma) \end{cases} \]
\[ \begin{cases} \Sigma = const\\ \phi \sim \mathcal{N}(\underline \phi, \underline \Xi) \end{cases} \]
\[ \begin{cases} \Sigma \sim \mathcal{IW}(\underline S,\underline \nu) \\ \phi | \Sigma \sim \mathcal{N}(\underline \phi, \Sigma \otimes \underline \Omega) \end{cases} \]