The evolution of the monetary transmission channels in Turkey: evidence from a TVP-VAR model

ABSTRACT

This article investigates the evolution of the monetary transmission mechanisms in Turkey for the period from January 1986 to December 2016. To this aim, the impacts of monetary variables on the prices and economic activity are investigated with a time-varying vector autoregressive model based on. The evidences from the time-varying responses indicate that the adoption of inflation targeting policy has markedly affected the functioning of transmission channels. The results also suggest that local and global financial crises may magnify the impact of monetary policy shocks on the overall economy.

KEYWORDS:

• Monetary policy
• transmission channels
• TVP- VAR model
• Turkey

JEL CLASSIFICATION:

• C11
• C32
• E52

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See Angeloni, Kashyap, and Mojon (2003) and Mahadeva and Sinclair (2002), for a detailed review of the literature on the monetary transmission channels.

² The variables are seasonally adjusted using Census X-13 method.

³ The variables are found integrated of order one, I(1), according to ADF and Phillips and Perron unit root tests. Lee and Strazicich (2003) unit root test, allowing for two endogenous structural breaks in the series is applied. This test also rejects the null hypothesis of a unit root with two breaks for all variables. The results are available from the corresponding author.

⁴ Since the random walk model is not stationary, we imposed stability constraint on the evolution of the time-varying parameters based on Cogley and Sargent (2005).

⁵ As outlined by Primiceri (2005), this assumption simplifies the inference and increases the efficiency of the estimation algorithm.

⁶ In order to determine the number of lags in the VAR, we estimate the model from one to four lags and select the appropriate lag with the lowest Akaike Information Criterion, the minimum value is obtained when the model is estimated with two lags.

⁷ We use the same following priors in Nakajima (2011) in the Bayesian estimation of the TVP-VAR model ${\sum }_{\beta }\sim IW\left(25,0.01I\right)$,${\left({\sum }_{\alpha }\right)}_i^{-2}\sim G\left(5,0.02\right)$,${\left({\sum }_{\sigma }\right)}_i^{-2}\sim G\left(5,0.02\right)$,${\left({\sum }_{\alpha }\right)}_i^{-2}\dots {\left({\sum }_{\sigma }\right)}_i^{-2}$, are $i^{th}$diagonal elements of the ${\sum }_{\alpha }\dots {\sum }_{\sigma }$ matrices, respectively. IW and G represent the respective inverse Wishart and Gamma distributions. In the determination of the initial values of the time-varying parameters we use the flat priors as follows: ${\mu }_{{\beta }_0}={\mu }_{{\alpha }_0}={\mu }_{{\sigma }_0}=0$ and ${\sum }_{{\beta }_0}={\sum }_{{\alpha }_0}={\sum }_{{\sigma }_0}$. Further details about the estimation of the TVP-VAR model based on MCMC algorithm can be found in Nakajima et al. (2011).

⁸ The further diagnostics for the time-varying parameters reported in Figure A1 including the sample autocorrelation function, the sample paths and the posterior densities for selected parameters indicate that the MCMC estimation leads to stable and uncorrelated samples.

⁹ The responses obtained from corresponding linear VAR model covering different subsamples selected according to the crisis periods, namely 1994:5, 2001:3 and 2008:12, are also plotted in Figures A2, A3 and A4 in the appendix. Figure A2The variation in the sign and in the magnitude of the linear responses across the different estimation periods corroborates the time-varying impact of monetary policy variables on the prices and economic activity.