cover image: BAYESIAN LOCAL PROJECTIONS  Silvia Miranda-Agrippino Giovanni Ricco

20.500.12592/q2whfv

BAYESIAN LOCAL PROJECTIONS Silvia Miranda-Agrippino Giovanni Ricco

19 May 2021

The second type of prior is instead data-based, and follows from the widely held belief the prior, leading to a Gaussian posterior centred at the MLE and with covariance matrix equal to the inverse of the second derivative of the log-likelihood. [...] In large samples the likelihood dominates the prior, leading to a Gaussian posterior centred at the MLE and with covariance matrix equal to the inverse of the second derivative of the log-likelihood. [...] 3.3 Prior Variance Under the two specifications of the prior for the mean of the BLP coefficients, the prior (h) variance is specified in the same way. [...] In setting the parameters of the hyperprior dis- tribution, it is important to observe that at short horizons a VAR (or a RW) is likely to be a good approximation to the DGP, while over medium horizons the bias in the 19 coefficients of the VAR due to model misspecification is compounded and grows due the iteration. [...] This may indicate that some of the characteristics of the responses of the VAR may depend on the dynamic restrictions imposed by the iterative nature of the VAR, rather than being genuine features of the data.

Authors

Sciences Po

Pages
45
Published in
France
Title in English
BAYESIAN LOCAL PROJECTIONS Silvia Miranda-Agrippino Giovanni Ricco [from PDF fonts]

Tables