central_bank_communication
Replication materials for the 'The effect of central bank communication on sovereign bond yields: The case of Hungary' article
This repository contains the neccesary source codes to reproduce the analysis in the article. The R source code is in the scripts folder, the needed data is in the data/input and data/output folders. For a detailed description of each script see below.
01_pr_sentiment_analysis.R
This script replicates the sentiment scores using the press release text data and the dictionary we created. The text analysis is done in R with the quanteda package. The sentiment scores for the Newey-West OLS robustness check using the L-M and AGH dictionaries are also computed in this script.
02_mc_cohesion_index.R
This script replicates the monetary council cohesion index calculation. It uses as input the excel sheet available at the webpage of the Central Bank of Hungary (MNB): https://www.mnb.hu/en/monetary-policy/the-monetary-council/voting-records-of-the-monetary-council-members (last accessed: 2020/05/06).
The scripts contains the step by step instructions for the index reproductions. The input file used for this script is data/input/mnb-mt-voting.xlsx.
03_yield_scores.DO
The do file creates the scores.xlsx from the data/input/yields.xlsx file, which contains the PCA scores for the short, medium and long term bond yields.
04_yield_scores_transform.R
The R script transforms the PCA scores from the Stata output by rescaling and log transforming the variables.
05_final_dataset_prep.R
The output of this script is the final dataset which is used for creating the figures and tables in the article. NB: Missing data is coded as '.' (for Stata compatibility).
This script assembles the parts created above and merges them into the macro controls. The macro controls are in the data/input/data_macro.csv file.
Variables in the dataset
- date: Year-month string
- i: Central bank base rate
- USD: USD/HUF exchange rate
- EUR: EUR/HUF exchange rate
- u: Unemployment rate
- m3: M3 money stock
- core_i_yoy: Core inflation, year on year change
- pmi: Purchasing Managers' Index
- log_total: logarithm of the rescaled combined yield factor scores (all time horizon)
- log_in_year: logarithm of the rescaled r3m r6m yield factor scores
- log_out_year: logarithm of the rescaled r1y r3y r5y r10y r15y yield factor scores
- log_intra_year: same as log_in_year
- log_1to3year: logarithm of the rescaled r1y r3y yield factor scores
- log_long_term: logarithm of the rescaled r5y r10y r15y factor scores
- total_resc: rescaled combined yield factor scores (all time horizon)
- in_year_resc: rescaled r3m r6m yield factor scores
- out_year_resc: rescaled r1y r3y r5y r10y r15y yield factor scores
- short_term_resc: same as in_year_resc
- medium_term_resc: rescaled r1y r3y yield factor scores
- long_term_resc: rescaled r5y r10y r15y factor scores
- r3m: yields of bonds with maturity rates of 3 months
- r6m: yields of bonds with maturity rates of 6 months
- r1y: yields of bonds with maturity rates of 1 years
- r3y: yields of bonds with maturity rates of 3 years
- r5y: yields of bonds with maturity rates of 5 years
- r10y: yields of bonds with maturity rates of 10 years
- r15y: yields of bonds with maturity rates of 15 years
- mt_cohesion: The cohesion index of the monetary council
- date_text: The release date of the press release
- nethawkish: the Hawkish sentiment score computed with our dictionary on the press release corpus
- lm_baseline: sentiment scores with dictionary baseline using the Loughran-McDonald dictionary (net positive) on the the press release corpus
- net_hawk_agh: the Hawkish sentiment score computed with the AGH dictionary
- monthly_mlf: ECB interest rates
- fed_i: US FED interest rates
- effective_fed_i: Effective US FED interest rates
- ecb_bs: ECB balance sheet
- fed_bs: FED balance sheet
- Simor: Central bank governor dummy. 1 for months when Andras Simor was the governor
- Matolcsy: Central bank governor dummy. 1 for months when Gyorgy Matolcsy was the governor
06_figures.R
This script recreates the Tables 1-2, Figures 1-3; 7-9.
07_ardl_analysis
This is the source code for the ARDL analysis and the corresponding test statistics. It also produces the QQ plots (Figs 4-6) in the appendix and the contents of Tables 3, 4 and 5.
08_robustness_analysis.R
The code for the Newey-West OLS robustness check in Appendix C, and produces the results in Table 9.
09_sentiment_examples.R
The code to replicate the sentiment calculation highlights in Table 10 in Appendix D.
System information
R Session Info
> sessionInfo()
R version 4.0.0 (2020-04-24)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 18363)
Matrix products: default
locale:
[1] LC_COLLATE=English_United States.1252
[2] LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C
[5] LC_TIME=English_United States.1252
system code page: 1250
attached base packages:
[1] stats graphics grDevices utils datasets methods base
loaded via a namespace (and not attached):
[1] compiler_4.0.0
Stata version info
Stata/IC 16.0 for Windows (64-bit x86-64)