Data and algorithms for the project of sterilized intervention effectiveness evaluation on Russian data 2014-2017
Description
This project presents tests for sterilized intervention effectiveness evaluation for small open commodity exporting economy. The time series used in the estimation includes 618 daily observations from 6 November 2014 to 20 April 2017. The oil price is the price of the OPEC basket in US dollars per barrel. The forex rate is the average spot rate (with delivery TODAY) on the Moscow Interbank Currency Exchange (MICEX) measured in Russian rubles per US dollar. The calculation of the sterilized intervention shock is based on the results of repo auctions which were related to average daily export of oil, gaz and oil product taken from Russian Balance of Payments. The data are stored in the data_daily_till_20_04_17.mat file which is in the Dynare_Matlab.zip archive. They are also contained in JMulti project (VECM_Mendeley.zip archive). In both sources the data include information about repo auctions Bank of Russia (BoR) conducted in 2014-2017. Repo maturity: 7 days (series dz7, dzpos7, dzneg7 in Matlab; series Z7, Zpos7, Zneg7 in JMulti) 28 days (series dz28, dzpos28, dzneg28 in Matlab; series Z28, Zpos28, Zneg28 in JMulti) 1 year (series dz1y, dzpos1y, dzneg1y in Matlab; series Z1y, Zpos1y, Zneg1y in JMulti) In all estimation procedures I use aggregated series: (series dz, dzpos, dzneg in Matlab; series Z, Zpos, Zneg in JMulti). "pos" means positive interventions (more USD distributed by BoR) "neg" means negative interventions (less USD distributed by BoR) The project includes 1. Estimation of theoretical general equilibrium model constructed for exchange rate determination in small open commodity exporting economy. It is performed in Dynare/Matlab software. 2. Estimation of vector error correction model. It is performed in JMulti software. Both models are used for evaluation of sterilized intervention effectiveness. It means that stable relationship between exchange rate and interventions is explored. To get more information about the model used and about results I report see my working paper: https://ideas.repec.org/p/hig/wpaper/170-ec-2017.html
Files
Steps to reproduce
To reproduce results of Dynare/Matlab estimation unpack archive Dynare_Matlab.zip in some folder and run script Runfile_Baseline_int2_v2.m This script runs Baseline_2int_v2.mod file. Note that Dynare software should be properly installed (see http://www.dynare.org/). In current version of Baseline_2int_v2.mod file: 1. Modes are not calculated but supplied from the mode file=Mode_lag8_MCMCMH. To recalculate modes you should set mode_compute=5 and remove mode file=Mode_lag8_MCMCMH in estimation command 2. MCMCMH algorithm is not running. Data are taken from mh_flie. To repeat calculation you may set mh_replic=100000 and remove load_mh_file in estimation command. It take some time. 3. Simulated IRF are not calculated. To repeat calculation of simulated IRF you may uncomment deterministic simulation block ( remove /* in lines 119 and 157) Results of model estimation for different lags could be found in archive IRFs. Unpack this archive and run script Graphs_IRFs.m in Matlab. You will see different impulse-response functions for different lags in the equation 5 (line 31 of Baseline_2int_v2.mod file) in comparison with IRFs in VECM. To reproduce results of VECM estimation unpack VECM_Mendeley archive and run JMulti (see http://www.jmulti.de/ ). Then open VECM_Mendeley project and repeat standard steps for VECM estimation in JMulti. Speps needed: 1. Choose "Two stage procedure" and Specify "Estimate coint relations by S2S with possible restrictions on beta" 2. Specify restrictions on beta. (2 restrictions: betas, corresponding to Zneg and Zpos should be zeros) 3. Run estimation 4. Estimate SVEC model ("Estimate with boot Std. Err.") 5. Bootstrap confidence intervals with Hall and Efron algorithms for 50 periods 6. Display Impulse Responses 7. If you wand to retreive IRF series you may use Control/Symbol Control button in JMulti