A Dynamic Nelson-Siegel Model for Treasury Bond Yields with Irregular Mixed-Frequency Macroeconomic Information

Description

This replication package provides all data files, Python code, and documentation required to reproduce every empirical result reported in the manuscript. The package is organized into seven functional directories covering both the China and U.S. Treasury bond markets: (1) raw data sourced from CCDC, WIND, and FRED, including weekly Treasury bond yields and macroeconomic variables (FX, CPI, M2, IAV); (2) model scripts implementing five specifications — DNS, DNS-TVL, DNS-TVL-GARCH, DNS-IMF-TVL, and DNS-IMF-TVL-GARCH — estimated via the Unscented Kalman Filter (UKF); (3) table- and figure-generating scripts producing Tables 1–11, Figures 1–6, and Appendix Tables B1–B3, C1, D1, E1–E4; (4) Bootstrap standard error computation; (5) in-sample fit evaluation (RMSE, adjusted R²) and out-of-sample forecasting comparison (DM test, CW statistic, R²_oos); and (6) a comprehensive README file and replication documentation with step-by-step instructions. All scripts include extensive English comments and clear header labels indicating the corresponding output. The code requires Python 3.11+ with dependencies specified in the README. Pre-saved outputs are included for computationally intensive steps.

Steps to reproduce

See README_Replication_Package.pdf and Replication_Documentation.pdf in zip files.

Institutions

• Xi’an Jiaotong-Liverpool University

  Jiangsu, Suzhou

Categories

Licence