Invisible Inventory: How Federal Loan Pool Liquidation Programs Are Undermining U.S. Housing Affordability

Zenodo Data Description and Usage Guide for Google Colab This dataset supports the study "Invisible Inventory: How Federal Loan Pool Liquidation Programs Are Undermining U.S. Housing Affordability," providing essential data files and Google Colab notebooks for analysis and reproduction. The repository contains all necessary components to replicate key findings and perform further research into the dynamics of non-performing loan (NPL) dispositions and their relationship with investor activity in the U.S. housing market. Files Included in this Repository: 1. `NPL Quarterly Aggregates.xlsx`: Excel file containing aggregated quarterly data on Non-Performing Loan (NPL) disposition volumes, broken down by program type (SPO, EXPO) and total amounts from 2015-2025. This file forms the basis for time-series analysis of federal NPL activity.2. `NPL Transactions Detail.xlsx`: Excel file providing detailed records of individual NPL transactions, including dates, quarters, program types, and Unpaid Principal Balance (UPB) in millions, with additional notes. This granular data is used for calculating average pool sizes and specific transaction patterns.3. `download_investor_purchases_by_category_data.csv`: CSV file containing investor purchase data, likely categorized by property type or investor segment. This data is intended to be merged with NPL data for regression analysis examining the relationship between NPL dispositions and investor market activity.4. `NPL_Analysis_Direct_Load.ipynb`: A Google Colab notebook (Jupyter Notebook) that directly loads and processes the `NPL Quarterly Aggregates.xlsx` and `NPL Transactions Detail.xlsx` files. This notebook performs data cleaning, calculates key descriptive statistics, and generates Figures 1 and 2 (time series of NPL volumes and SPO/EXPO comparison).5. `FNMA.ipynb`: A Google Colab notebook (Jupyter Notebook) likely containing analysis specific to Fannie Mae (FNMA) data, potentially extending the NPL analysis or exploring different aspects of GSE disposition programs.6. `REGRESSION.ipynb`: A Google Colab notebook (Jupyter Notebook) specifically designed to run the regression analysis. This notebook is expected to integrate the `NPL_For_Regression.csv` output (from `NPL_Analysis_Direct_Load.ipynb`) with the `download_investor_purchases_by_category_data.csv` to model the relationship between NPL volumes and investor activity.7. `Regression_Full_Output.txt`: A text file containing the full output of a regression analysis, providing detailed statistical results, coefficients, standard errors, and model diagnostics. How to Use the Data and Notebooks in Google Colab: To fully utilize this dataset and replicate the analysis: Step 1: Download All Files from Zenodo1. Go to the Zenodo record page (https://doi.org/10.5281/zenodo.17644808).2. Download all seven files listed above to your local computer. Step 2: Access Google Colab and Upload Files1. Open [Google Colab](https://colab.research.google.com/).2. Start a new notebook, or open one of the provided `.ipynb` files (e.g., `NPL_Analysis_Direct_Load.ipynb`).3. Within your Colab notebook, run the following Python code in a cell: ```python from google.colab import files uploaded = files.upload() ```4. A file chooser dialog will appear. Select and upload all the files you downloaded from Zenodo. This will place them in the `/content/` directory of your Colab session. Step 3: Run the `NPL_Analysis_Direct_Load.ipynb` Notebook1. If you opened this notebook directly from Zenodo, ensure all the `.xlsx` files are uploaded as per Step 2.2. Execute each cell in the `NPL_Analysis_Direct_Load.ipynb` notebook sequentially (using Shift+Enter).3. This notebook will: Load the `NPL Quarterly Aggregates.xlsx` and `NPL Transactions Detail.xlsx` files. Clean and prepare the data. Calculate key descriptive statistics. Generate Figure 1 (Federal NPL Disposition Volumes time series) and Figure 2 (SPO vs. EXPO comparison). Create and download two CSV files: `Summary_Statistics_Table.csv` and `NPL_For_Regression.csv`. Ensure you download these generated files as they are inputs for the `REGRESSION.ipynb` notebook. Step 4: Run the `REGRESSION.ipynb` Notebook1. Open the `REGRESSION.ipynb` notebook in Google Colab.2. Before running, ensure you have uploaded `download_investor_purchases_by_category_data.csv` (from Zenodo) and the generated `NPL_For_Regression.csv` (from Step 3) to your Colab session.3. Execute each cell in the `REGRESSION.ipynb` notebook sequentially.4. This notebook will perform the time-series regression analysis described in the paper, merging the NPL data with investor data and producing the regression results. Step 5: Review `FNMA.ipynb` and `Regression_Full_Output.txt`1. The `FNMA.ipynb` notebook can be run similarly, after uploading the necessary data, to explore specific Fannie Mae (FNMA) program details or analysis.2. `Regression_Full_Output.txt` provides a pre-generated detailed output of the regression results, serving as a reference or a quick overview if you do not wish to run the `REGRESSION.ipynb` notebook immediately. --- Important Notes: File Paths: The notebooks are configured to load files directly from the `/content/` directory. Ensure your uploaded files are in this location. Reproducibility: By following these steps, researchers can fully reproduce the descriptive statistics, visualizations, and prepare the data for the regression analysis presented in the paper. Session Lifespan: Files uploaded directly to a Google Colab session are temporary and will be deleted when the session ends. Always download any generated output files you wish to keep.