Full posterior MCMC chains for "The Effect of New Housing Supply in Structural Models: A Forecasting Performance Evaluation"
Year: 2026
Type: dataset
Abstract: Complete Markov chain Monte Carlo output for the paper "The Effect of New Housing Supply in Structural Models: A Forecasting Performance Evaluation" (International Journal of Forecasting). 50 posterior chains of 500,000 to 1,000,000 draws each, plus 14 posterior-mode files, for a two-country-free medium-scale DSGE model estimated on real-time US data at six Great-Financial-Crisis forecast origins (2007-Q4 to 2009-Q1) and four COVID-era origins (2020-Q1 to Q4). This dataset is the companion to the code repository, which contains everything else needed to reproduce the paper: https://github.com/stefangirstmair/HousingSupply_Replication_IJF The repository already ships thinned copies of these chains (5,000 draws each, 84 MB), which reproduce the forecast layer bit-identically and suffice for every table and figure except Table 11. These full chains are needed for the MCMC convergence diagnostics of Table 11, whose inefficiency factors and effective sample sizes are properties of the full serially correlated chain and are meaningless after thinning, and for reproducing Tables 9 and 10 to the last digit. To use: unpack the archive at the root of the cloned repository. This creates estimation/results/, which the pipeline detects automatically. Chain lengths are not uniform: the extended model on the HP-filtered treatment at the 2007-Q4 to 2008-Q4 origins was run for 1,000,000 draws, all other chains for 500,000. Software: MATLAB R2023a. Input data are real-time vintages from ALFRED (Federal Reserve Bank of St. Louis).
Author: Girstmair, Stefan
Language: English
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204557798