Credit-Market Sentiment and the Business Cycle

Credit-Market Sentiment and the Business Cycle#

Last updated: Aug 20, 2026, 2:05:03 AM

Project Notes

Table of Contents#

Notebooks 📖

Pipeline Charts 📈

Pipeline Dataframes 📊

Pipeline Specs#

Pipeline Name

Credit-Market Sentiment and the Business Cycle

Pipeline ID

P01

Maintainer

Fernando Raffo, Bangjie Xu

Contributors

Fernando Raffo, Bangjie Xu

Repository

fernando-raffo/p03_lopez_salido_stein_zakrajsek_2017

Pipeline Web Page

Pipeline Web Page

Date of Last Code Update

2026-08-19 22:04:54

OS Compatibility

Windows, Linux, MacOS

Linked Dataframes

P01:fred_macroeconomic_variables
P01:shiller_market_variables
P01:greenwood_hanson_hys
P01:mergent_fisd_bond_data
P01:fred_processed_monthly_series
P01:fred_processed_annual_series
P01:shiller_processed_annual_series
P01:greenwood_hanson_hys_processed_series

About this project#

This project replicates key results from:

Lopez-Salido, D., Stein, J. C., and Zakrajsek, E. (2017), “Credit-Market Sentiment and the Business Cycle.” The Quarterly Journal of Economics, 132(3): 1373-1426. https://doi.org/10.1093/qje/qjx014

The paper shows that elevated credit-market sentiment in year t - 2 (proxied by narrow, aggressively priced credit spreads and a high share of junk-bond issuance) forecasts a subsequent widening of credit spreads and a decline in economic activity in years t and t + 1, using U.S. data from 1929 to 2015. This repo pulls the underlying data (FRED, Greenwood-Hanson credit-spread and issuance data, and Shiller’s stock-market data), reconstructs the paper’s credit-market sentiment measure via a two-step forecasting regression, and reproduces select tables and figures from the paper, including:

  • Figure 1: the Baa-Treasury credit spread over time

  • Figure 2: credit-market sentiment and economic growth

  • Table 1: forecasting economic growth with credit spreads and stock prices

  • Table 2: the two-step regression results linking financial-market sentiment to economic growth

Extension. The original paper builds its credit-market sentiment proxy from the spread between Moody’s seasoned Baa-rated (lowest investment-grade) corporate bond yields and the 10-year Treasury yield. This project extends the replication by rebuilding the same figures and regressions using the spread on Moody’s Aaa-rated (highest-grade) corporate bonds in place of Baa. The Aaa-based outputs are produced alongside the original Baa-based ones throughout the pipeline.

Case study. A second extension applies the paper’s Table II first-step sentiment regression - estimated and held fixed on 1929-2015 data - out of sample to the 2020-2022 COVID cycle, comparing its predicted change in the Baa-Treasury credit spread against the realized change.

Data Sources#

Source

Description

FRED

Moody’s Aaa/Baa seasoned corporate bond yields, 10-year Treasury yield, 3-month T-bill rate, CPI, population, real GDP, and the NBER recession indicator

Robert Shiller’s Data Website

Monthly S&P Composite price, dividends, earnings, CPI, the 10-year Treasury rate, and the cyclically adjusted price-earnings ratio (CAPE / P/E10)

Greenwood & Hanson (2013), via Harvard Business School

Published historical annual high-yield share of nonfinancial corporate bond issuance, 1926-2008

Mergent FISD via WRDS

Bond-level issuance and rating data used to reconstruct the high-yield share from the early 1980s onward, spliced onto the published Greenwood-Hanson series

Quick Start#

0. Software & Access Prerequisites#

  1. Conda Package Manager (e.g. via Anaconda)

  2. Python 3.12 or above

  3. MacTeX or TeX Live

  4. WRDS Subscription

1. Create & Activate Virtual Environment#

You can create a conda environment and all dependencies direcly using the command below if you have the conda package manager:

conda env create -f environment.yml
conda activate p03_lopez_salido_stein_zakrajsek_2017_env

Alternatively, we also include a requirements.txt file to create an environment with alternative package mangers or a simple Python virtual environment:

conda create -n p03_lopez_salido_stein_zakrajsek_2017_env python=3.12
conda activate p03_lopez_salido_stein_zakrajsek_2017_env
pip install -r requirements.txt

python -m venv .venv
source .venv/bin/activate 
pip install -r requirements.txt

2. Configure WRDS Credentials#

Copy .env.example into a new file called .env in the project:

cp .env.example .env

Then edit .env with your WRDS credentials. It should look like:

WRDS_USERNAME="your_username"

3. Run the Full Pipeline#

doit

4. Other commands#

Unit Tests and Doc Tests#

You can run the unit test, including doctests, with the following command:

pytest --doctest-modules

The full doit also ends with a run_pytest task that executes this same suite as its final step, after the data pulls and the Table I / II replications. This means the integration tests that check the replicated coefficients against the published paper run automatically at the end of the pipeline; they skip on their own if the processed data has not been built.

You can build the documentation site (also run automatically by doit) with:

chartbook build -f

Setting Environment Variables#

You can export your environment variables from your .env files like so, if you wish. This can be done easily in a Linux or Mac terminal with the following command:

set -a  # automatically export all variables
source .env
set +a

On Windows (PowerShell):

Get-Content .env | ForEach-Object { if ($_ -match '^([^=]+)=(.*)$') { [Environment]::SetEnvironmentVariable($matches[1], $matches[2], 'Process') } }

Formatting#

This project uses Ruff for linting and formatting Python code.

# Auto-fix linting issues (e.g., unused imports, undefined names)
ruff check . --fix

# Format code (consistent style, spacing, line length)
ruff format .

# Sort imports, then fix linting issues, then format
ruff format . && ruff check --select I --fix . && ruff check --fix .

  • ruff check --fix applies safe auto-fixes for linting violations

  • ruff format formats code similar to Black

  • --select I targets only import sorting rules (isort-compatible)

General Directory Structure#

p03_lopez_salido_stein_zakrajsek_2017/
├── chartbook.toml       # the manifest — configuration for the published ChartBook site
├── dodo.py              # PyDoit task runner — defines and runs the full pipeline
├── environment.yml      # conda environment spec
├── requirements.txt     # pip requirements (alternative to the conda environment)
├── README.md            # this file — also acts as the site's front page
├── .env.example         # sample .env for private paths & WRDS credentials (not tracked in Git)
├── assets/              # hand-drawn figures/logo not generated from code
├── data_manual/         # manually-collected data that can't be recreated (tracked)
├── _data/               # data pulled/processed by the pipeline (gitignored, regenerable)
│   ├── raw_data/        
│   ├── processed_data/          
│   └── data_dictionaries/       
├── _output/             # chart HTML/PDF and tables generated by the pipeline (gitignored, regenerable)
├── docs_src/
│   └── site/            # extra site pages merged into the published site
├── docs/                # built ChartBook/Jupyter Book site (gitignored)
├── reports/             # LaTeX report and bibliography
└── src/                 # code that produces the artifacts: pulls, processing, replication, notebooks, tests

Additional Notes#

  • We are using the doit Python module as a task runner. It works like make and the associated Makefiles. To rerun the code, install doit (https://pydoit.org/) and execute the command doit from the project’s root directory (where dodo.py lives). Note that doit is very flexible and can be used to run code commands from the command prompt, thus making it suitable for projects that use scripts written in multiple different programming languages.

  • We are using the .env file as a container for absolute paths that are private to each collaborator in the project. You can also use it for private credentials, if needed. It should not be tracked in Git.

Data and Output Storage#

We’ll often use a separate folder for storing data. Any data in the data folder can be deleted and recreated by rerunning the PyDoit command (the pulls are in the dodo.py file). Any data that cannot be automatically recreated should be stored in the “data_manual” folder. Because of the risk of manually-created data getting changed or lost, we keep it under version control where we can. Thus, data in the “_data” folder is excluded from Git (see the .gitignore file), while the “data_manual” folder is tracked by Git.

Output is stored in the “_output” directory. This includes dataframes, charts, and rendered notebooks.