term: preferred habitat with endogenous price of risk

Julia code for solving and simulating a term structure model with preferred habitat demand and an endogenous price of risk, and generating the paper’s tables/figures.

Repository layout

src/: core model code (solver, simulations, impulse responses, outputs).
scripts/: entrypoint scripts and (optional) Slurm wrappers.
data/: empirical inputs (CSV files used for figures/QE paths, etc.).
output/: generated results, tables, and figures (default location).

Requirements

Julia: this project is developed against Julia 1.9.x (see cluster scripts for the exact module used).
Packages: managed via Project.toml (a Manifest.toml may be present locally, but is not required).

Setup

From the repository root:

julia --project=. -e 'using Pkg; Pkg.instantiate()'

Running locally

The main entrypoint is scripts/run.jl (it calls into the term module in src/term.jl).

Solve + simulate (default)

julia --project=. scripts/run.jl

Run a single calibration (or a list/range)

The first positional argument is a calibration index (1-based).

# single
julia --project=. scripts/run.jl 7

# comma-separated list
julia --project=. scripts/run.jl 1,3,7

# inclusive range
julia --project=. scripts/run.jl 3:9

Output only (tables/figures)

julia --project=. scripts/run.jl --mode=output

Configure output locations

scripts/run.jl supports both CLI flags and environment variables:

Results folder (JLD2 files):
- --results_dir PATH or TERM_RESULTS_DIR=PATH
- default: output/results
Output folder (tables/figures):
- --out_dir PATH or TERM_OUT_DIR=PATH
- default: output
Mode:
- --mode solve|output|both or TERM_MODE=...

Examples:

julia --project=. scripts/run.jl --results_dir output/results --out_dir output --mode both

TERM_RESULTS_DIR=output/results TERM_OUT_DIR=output TERM_MODE=both julia --project=. scripts/run.jl

Reproducibility / runtime knobs

The driver script calls term.configure_runtime!() to make batch behavior explicit:

Threads: set via Julia -t N / JULIA_NUM_THREADS.
BLAS threads: TERM_BLAS_THREADS=1 (default is 1).
Headless plotting: TERM_HEADLESS=true (default is true).

Run metadata (Julia version, threads, git commit when available, TOML SHA1s, etc.) is saved alongside solution_*.jld2 / results_*.jld2.

Data inputs

Most empirical inputs are CSV files in data/. The code resolves paths robustly via data_path(...) (independent of the working directory).

In particular, QE path inputs are loaded from:

data/Figure20_1.csv … data/Figure20_8.csv

If required data files are missing, ImportData() errors with a list of missing filenames.

Outputs

By default, outputs are written under:

Results: output/results/ (JLD2)
Tables: output/tables/
Figures: output/figures/

Citation

If you use this code, please cite it (see CITATION.cff).

Notes on HPC / Slurm

This repo includes Slurm wrapper scripts under scripts/ for the authors’ cluster setup. If you are running on a different cluster, you will typically need to adjust module/depot paths, resource requests, and result/output directories.

Term

Install / Use

README