HAFiscal Replication Package

Paper: Welfare and Spending Effects of Consumption Stimulus Policies
Authors: Christopher D. Carroll, Edmund Crawley, William Du, Ivan Frankovic, Håkon Tretvoll
Keywords: heterogeneous agents, fiscal policy, stimulus checks, iMPCs, HANK, consumption, welfare, QE replication Version: Development Version (HAFiscal-Latest)

---

Instant Results (No Installation Required)

Want to explore fiscal policy effects right now?

The interactive dashboard lets you:

• Compare stimulus checks, UI extensions, and tax cuts
• Adjust model parameters in real-time
• Visualize fiscal multipliers under different monetary policies
• See results in seconds (no 100+ hour computation needed)

No installation required — runs entirely in your browser via MyBinder.

For local installation, see dashboard/DASHBOARD_README.md or README/DASHBOARD.md.

---

Quick Start

New to HAFiscal? Start with the Getting Started Guide for navigation and workflow guidance.

For detailed documentation, see the README/ directory.

The README/ directory contains:

• GETTING-STARTED.md — Navigation guide and workflow overview (start here if new)
• Detailed README — Complete replication instructions and documentation
• INSTALLATION.md — Installation and setup instructions
• DOCKER.md — Docker container usage and setup
• DASHBOARD.md — Interactive dashboard documentation
• REPLICATION.md — Detailed replication instructions and data provenance
• QUICK-REFERENCE.md — Quick reference guide
• CONTRIBUTING.md — Contribution guidelines
• TROUBLESHOOTING.md — Common issues and solutions

---

Research Questions and Contributions

Primary Research Questions

1. What are the welfare and spending effects of different consumption stimulus policies (stimulus checks, tax cuts, unemployment insurance extensions) across the income and wealth distribution?

2. How do heterogeneous-agent mechanisms (liquidity constraints, sticky expectations, splurge behavior) affect the distributional and aggregate impacts of fiscal stimulus?

3. What is the optimal design of stimulus policies when accounting for household heterogeneity in marginal propensities to consume (MPCs)?

Key Contributions

1. Comprehensive HANK model calibration: Extends heterogeneous-agent New Keynesian (HANK) models to match both microeconomic evidence on intertemporal MPCs (iMPCs) and macroeconomic evidence on aggregate consumption dynamics, using Survey of Consumer Finances (SCF) 2004 data.

2. Novel behavioral mechanisms: Implements and quantifies the role of:

   • Sticky expectations (following Carroll et al. 2020, cAndCwithStickyE in bibliography)
   • Splurge behavior (lumpy consumption responses to windfalls)
   • Liquidity constraints and heterogeneous wealth distributions

3. Distributional welfare analysis: Provides systematic welfare comparisons across alternative stimulus designs, highlighting how policy effectiveness varies dramatically across households with different liquid wealth positions.

4. Methodological extension: Builds on the computational framework of Auclert et al. (2021, Auclert2021) and extends the two-asset HANK literature (Kaplan & Violante 2014, kaplan2014model; Fagereng et al. 2021, fagereng-mpc-2021) to incorporate additional behavioral frictions.

---

Literature Connections

Core Methodological Foundations

HANK Models and Computational Methods:

• Auclert et al. (2021) [Auclert2021]: Sequence-space Jacobian methods for solving heterogeneous-agent models (computational framework extended here)
• Kaplan & Violante (2014) [kaplan2014model]: Two-asset model with liquid/illiquid assets and high MPCs for hand-to-mouth households (calibration strategy extended)
• Carroll et al. (2017) [cstwMPC]: Distribution of wealth and MPCs in heterogeneous-agent models (empirical targets extended)

Sticky Expectations and Consumption Dynamics:

• Carroll et al. (2020) [cAndCwithStickyE]: Sticky expectations model explaining aggregate consumption persistence (mechanism implemented here)
• Lian (2023) [Lian2023-ca]: Future consumption mistakes and high MPCs (related behavioral mechanism)

Empirical Evidence on MPCs and Consumption Responses

Microeconomic MPC Estimates:

• Fagereng et al. (2021) [fagereng-mpc-2021]: Norwegian lottery data showing MPC heterogeneity by liquid assets (empirical target)
• Kotsogiannis & Sakellaris (2024) [kotsogiannisMPCs]: Tax lottery estimates of iMPCs (complementary evidence)
• Boehm et al. (2025) [boehm2025fivefacts]: Randomized experiment on MPCs (recent empirical evidence)
• Parker et al. (2013) [parker2013consumer]: Economic stimulus payments of 2008 (empirical benchmark)

Consumption During Unemployment:

• Ganong & Noel (2019) [ganongConsumer2019]: Consumer spending during unemployment (UI extension analysis relates)
• Graves (2024) [gravesUnemployment]: Unemployment risk and consumption dynamics (related mechanism)

Fiscal Multipliers and Policy Analysis

Fiscal Multipliers in HANK Models:

• Broer et al. (2023) [broer2023fiscalmultipliers]: Fiscal multipliers from heterogeneous-agent perspective (complementary analysis)
• Broer et al. (2025) [broer2025stimulus]: Stimulus effects of common fiscal policies (recent related work)
• Hagedorn et al. (2019) [hagedorn2019fiscal]: Fiscal multiplier in HANK models (methodological connection)

Automatic Stabilizers and Welfare:

• McKay & Reis (2016, 2021) [mckay2016role, mckay2021optimal]: Role of automatic stabilizers and optimal design (welfare analysis relates)
• Phan (2024) [phan2024welfare]: Welfare consequences of countercyclical fiscal transfers (related welfare analysis)

Behavioral Mechanisms

Near-Rationality and Bounded Rationality:

• Andre et al. (2025) [ansQuickfix]: Near-rationality in consumption and savings (related behavioral mechanism)
• Akerlof & Yellen (1985) [akerlof1985near]: Near-rational model of business cycle (foundational work)
• Ilut & Valchev (2022) [ilutEconomic]: Economic agents as imperfect problem solvers (related framework)

Present Bias and Mental Accounting:

• Laibson et al. (2024) [lmmPresentBias]: Present bias amplifies balance-sheet channels (related mechanism)
• Graham & McDowall (2024) [graham2024mental]: Mental accounts and consumption sensitivity (related behavioral feature)

Related HANK Literature

Unemployment and Business Cycles:

• Ravn & Sterk (2017, 2021) [Ravn2017, Ravn2021]: Job uncertainty, HANK & SAM models (related HANK extensions)
• Christiano et al. (2016) [Christiano2016]: Unemployment and business cycles (search-and-matching framework)
• Graves (2024) [gravesUnemployment]: Unemployment risk affects business cycle dynamics (related mechanism)

Distributional Effects of Monetary Policy:

• Gornemann et al. (2021) [Gornemann2021]: Distributional consequences of systematic monetary policy (related distributional analysis)

Data and Calibration

SCF Data and Wealth Distribution:

• SCF 2004 [SCF2004]: Survey of Consumer Finances 2004 (primary data source)
• Kaplan et al. (2014) [kaplan2014model]: Liquid wealth construction methodology (followed here)

Income Process Calibration:

• Crawley et al. (2024) [crawley2024parsimonious]: Parsimonious model of idiosyncratic income (income process specification)

---

What This Repository Provides (AI- and search-friendly summary)

• Replication code and data for the HAFiscal paper, built on Econ-ARK tools, with a Heterogeneous Agent New Keynesian (HANK) model calibrated to U.S. micro data.

• Consumption stimulus policy analysis: effects of stimulus checks, tax cuts, and UI extensions on spending, iMPCs, and welfare across the income and wealth distribution.

• Model artifacts: code for sticky expectations, splurge behavior, and robustness appendices (HTML/PDF links in appendices).

• Data: SCF-based liquid wealth and income moments (paper uses 2013-dollar SCF vintage; scripts document 2022→2013 inflation adjustment using CPI-U-RS and the 1.1587 factor).

• Outputs: paper PDFs, slides, tables, and figures for direct reuse in scholarly work or derivative projects.

---

5. Getting Started

For complete setup and reproduction instructions, see README/GETTING-STARTED.md.

Quick Summary:

• Installation: See README/INSTALLATION.md for detailed setup instructions
• Docker: See README/DOCKER.md for containerized setup (alternative to local installation)
• Reproduction: Run ./reproduce.sh --help to see available modes
• Timing Estimates: See reproduce/benchmarks/TIMING-ESTIMATES.md for hardware-specific timing data
• Troubleshooting: See README/TROUBLESHOOTING.md for common issues

Quick Commands:

# View all reproduction options
./reproduce.sh --help

# Quick document generation (5-10 minutes)
./reproduce.sh --docs

# Minimal computational validation (~1 hour)
./reproduce.sh --comp min

# Full computational replication (4-5 days)
./reproduce.sh --comp full

For detailed documentation on each mode, see README/GETTING-STARTED.md.

---

6. Data Availability

This research uses publicly available data from the Survey of Consumer Finances (SCF) 2004. All data can be downloaded automatically via provided scripts.

Data Sources:

• SCF 2004: Board of Governors of the Federal Reserve System

  • Summary Extract: rscfp2004.dta
  • Full Public Data Set: p04i6.dta
  • Download: Code/Empirical/download_scf_data.sh
  • URL: https://www.federalreserve.gov/econres/scf_2004.htm

• Norwegian Population Data: Fagereng, Holm, and Natvik (2021)

  • Summary statistics and moments (published in paper)
  • Individual-level data not publicly available

Data Files Included: Data files are included in Code/Empirical/ for convenience. See README/REPLICATION.md for detailed data provenance and processing information.

Citation: Data sources are cited in HAFiscal-Add-Refs.bib and in the paper text (see Subfiles/Parameterization.tex).

---

7. Computational Requirements

Hardware: Minimum 4 cores, 8GB RAM; Recommended 8+ cores, 16GB RAM
Software: Python 3.9+, LaTeX (TeX Live 2021+), Unix-like environment (macOS/Linux/WSL2)
Package Manager: uv (recommended) or conda
Alternative: Docker container (see README/DOCKER.md)

For detailed requirements, platform support, and dependency information, see README/INSTALLATION.md.

---

8. Reproduction

The primary reproduction script is ./reproduce.sh, which provides multiple modes:

• --docs: Document generation only (5-10 minutes)
• --comp min: Minimal computational validation (~1 hour)
• --comp full: Full computational replication (4-5 days)
• --all: Complete reproduction pipeline

Timing Estimates: See reproduce/benchmarks/TIMING-ESTIMATES.md for detailed timing information and hardware scaling data.

For detailed reproduction instructions, see README/GETTING-STARTED.md.

---

9. Results Mapping

This repository generates 6 main figures and 8 main tables, plus additional appendix figures and tables. All figures and tables are defined as LaTeX subfiles that include generated content from computational Python scripts.

For complete details on figure/table provenance, including:

• Exact LaTeX subfile locations
• Source PDF/LTX files and their paths
• Python scripts that generate each figure/table
• Captions and figure/table numbers
• Appendix figures and tables

See: README/REPLICATION.md - Section 6: Results Mapping

Quick Summary

Figures: Generated by Python scripts in Code/HA-Models/ and compiled as LaTeX subfiles in Figures/ directory. Main figures include:

• Splurge factor estimation (Figure 1)
• Wealth distribution fit (Figure 2)
• Non-targeted moments validation (Figure 3)
• Policy effectiveness during recessions (Figure 4)
• HANK-SAM model IRFs and multipliers (Figure 5)
• PE vs HANK multiplier comparison (Figure 6)

Tables: Generated by Python scripts and pulled into LaTeX via \fetchgeneratedtabular{}. Main tables include:

• MPC by wealth quartile (Table 1)
• Model calibration parameters (Table 2)
• Recession parameters (Table 3)
• Estimated discount factors (Table 4)
• Non-targeted moments (Table 5)
• Policy multipliers (Table 6)
• Welfare effectiveness (Table 7)
• Welfare comparison with splurge (Table 8)

Parameter Values: Model parameters are defined in:

• Code/HA-Models/FromPandemicCode/Parameters.py - Main parameter definitions
• Code/HA-Models/FromPandemicCode/EstimParameters.py - Estimation parameters

Note: Each figure and table .tex file can be compiled standalone. See README/REPLICATION.md for detailed compilation instructions and complete provenance information.

---

10. File Organization (simplified)

/
├── README.md							 # This file
├── environment.yml						 # Conda environment specification
├── pyproject.toml						 # Python dependencies (uv format)
├── HAFiscal.tex						 # Main LaTeX document
├── HAFiscal.bib						 # Bibliography
├── HAFiscal-Abstract.txt				 # Abstract text
├── HAFiscal-Slides.tex					 # Presentation slides
├── reproduce.sh						 # Main reproduction script
├── reproduce.py						 # Python mirror (cross-platform)
├── reproduce_min.sh					 # Quick validation test
├── reproduce/							 # Additional reproduction scripts
│   ├── reproduce_computed.sh			 # Run all computations
│   ├── reproduce_computed_min.sh		 # Minimal computation test
│   ├── reproduce_documents.sh			 # Generate LaTeX documents
│   ├── reproduce_environment_comp_uv.sh # Set up Python environment (uv)
│   ├── reproduce_environment_texlive.sh # Set up LaTeX environment
│   └── [other reproduction scripts]
├── Code/								 # All computational code
│   ├── HA-Models/						 # Heterogeneous agent models
│   │   ├── FromPandemicCode/			 # Core model implementation
│   │   ├── Results/					 # Model output files
│   │   └── [model-specific directories]
│   └── Empirical/						 # Empirical data processing
│       ├── download_scf_data.sh		 # Download SCF data
│       ├── make_liquid_wealth.py		 # Construct liquid wealth measure
│       ├── adjust_scf_inflation.py		 # Inflation adjustments
│       └── compare_scf_datasets.py		 # Dataset comparisons
├── Figures/							 # Figure LaTeX files (*.tex, *.pdf)
├── Tables/								 # Table LaTeX files (*.tex, *.pdf)
├── Subfiles/							 # Paper section files
│   ├── Appendix-*.tex					 # Appendix sections
│   ├── Conclusion.tex					 # Conclusion section
│   └── [other section files]
├── Data/								 # Data files directory
├── dashboard/							 # Interactive dashboard (Jupyter/Streamlit)
├── binder/								 # Binder configuration for cloud execution
├── Equations/							 # Equation definitions
├── @local/								 # Local LaTeX packages and configuration
└── @resources/							 # LaTeX resources and utilities

---

11. Known Issues and Workarounds

For detailed troubleshooting information, see README/TROUBLESHOOTING.md.

Common Issues:

• Windows Native: Not supported; use WSL2 (see README/INSTALLATION.md)
• Long Computation Times: Use ./reproduce.sh --comp min for quick validation
• LaTeX Issues: See README/TROUBLESHOOTING.md for platform-specific solutions

---

12. Contact Information

Technical Issues

For technical issues with replication:

• Open an issue: https://github.com/llorracc/HAFiscal-Latest/issues
• Email: ccarroll@jhu.edu (Christopher Carroll)

Data Questions

For questions about SCF data:

• Federal Reserve SCF page: https://www.federalreserve.gov/econres/scfindex.htm
• Email: scf@frb.gov

Paper Content

For questions about the paper content:

• See author emails in paper
• Christopher Carroll: ccarroll@jhu.edu
• Edmund Crawley: edmund.s.crawley@frb.gov

---

13. Citation

If you use this replication package, please cite:

@misc{carroll2025hafiscal,
  title={Welfare and Spending Effects of Consumption Stimulus Policies},
  author={Carroll, Christopher D. and Crawley, Edmund and Du, William and Frankovic, Ivan and Tretvoll, H{\aa}kon},
  year={2025},
  howpublished={Development version},
  note={Available at \url{https://github.com/llorracc/HAFiscal-Latest}}
}

---

Last Updated: January 09, 2026
README Version: 1.1
Replication Package Version: 1.0

Version 1.1 Changes:

• Added comprehensive reproduce.sh documentation with all modes
• Updated timing data to use benchmark system measurements (not placeholders)
• Added hardware scaling examples (minimum, mid-range, high-performance)
• Integrated benchmark system references and instructions
• Added timing variability factors and explanations

Note: This is the development version. For public release, see HAFiscal-Public. For journal submission, see HAFiscal-QE.