ESG Risk for Banks Research Project

Lead Researcher: Guillaume Bolivard | Faculty Advisor: Dr. Abol Jalilvand | Institution: Loyola University Chicago

Abstract: This research develops and validates a comprehensive framework for assessing default risk in banking institutions through both market-based (Merton model) and accounting-based (Bharath-Shumway naive) approaches. The study incorporates Environmental, Social, and Governance (ESG) factors into credit risk analysis for U.S. banks over the period 2016-2023. Our improved equity volatility methodology with hierarchical fallbacks and complete provenance tracking achieves 100% convergence rates and 90.6% data coverage across 244 unique banking institutions. All methodological specifications, validation procedures, and econometric designs presented herein were developed under the supervision of Dr. Abol Jalilvand.

Research Methodology

Validation & Quality Assurance

Current Status: PRODUCTION READY - Version 3.0 | Research Review Ready

Metric	Value	Status
Version	3.0 (October 11, 2025)	Latest
Convergence Rate	100% (1,290/1,290)	Perfect
Sample Coverage	1,290/1,424 (90.6%)	Excellent
Total Observations	1,424 bank-years	Complete
Period Coverage	2016-2023 (8 years)	Comprehensive
Unique Banks	244 institutions	Robust
Data Quality	Tier 1: 92%	High
Time Integrity	No lookahead bias	Verified
Documentation	>2,000 lines	Publication-ready

Key Achievements (Version 3.0):

100% convergence rate achieved through improved equity volatility methodology
90.6% coverage (1,290/1,424 observations) with hierarchical volatility estimation
Comprehensive methodology documented in 247-line academic paper section
Complete provenance tracking for all volatility estimates (method, window, obs_count)
Time integrity guaranteed with strict t-1 window endings (no lookahead bias)
Clean, reproducible pipeline from CRSP data to final DD/PD estimates

Validation Reports:

Complete validation: logs/INSTRUCTION_3_COMPLETE.md (358 lines)
Methodology paper: docs/writing/dd_and_pd.md (970 lines, v3.0)
Dataset alignment: logs/PAPER_DATASET_ALIGNMENT_COMPLETE.md (318 lines)

Research Overview

Motivation and Objectives

The banking sector's exposure to ESG-related risks necessitates rigorous quantitative frameworks for default probability estimation. This study addresses this need by implementing two complementary methodologies for credit risk measurement, each validated through comprehensive diagnostic procedures:

Market-Based Approach (Merton Model)

Solves the Merton KMV model using equity market data
Estimates asset value (V_t) and volatility (σ_V,t) through Newton-Raphson optimization
Computes market-implied DD (DD_m) and PD (PD_m)
Uses risk-free rate as drift parameter (μ = r_f) under Q-measure
100% convergence rate on observations with sufficient data

Accounting-Based Approach (Bharath-Shumway Naive DD)

Implements the naive distance-to-default methodology
Uses accounting proxies without numerical solvers
Computes accounting-based DD (DD_a) and PD (PD_a)
Leverages balance sheet and market equity data
90.6% coverage across full sample

Equity Volatility Methodology (New in v3.0)

Primary method: 36-month rolling standard deviation (~98% coverage)
Fallback 1: EWMA for limited data (λ=0.94, ~1-2% coverage)
Fallback 2: Peer median by size bucket (<1% coverage)
Complete provenance tracking: method, window months, start/end years
Time integrity: All windows end strictly at t-1 (no lookahead bias)
Quality tiers: Tier 1 (92%), Tier 2 (0%), Tier 3 excluded (8%)

Research Objectives

Develop numerically stable implementations of market-based (Merton KMV) and accounting-based (Bharath-Shumway) distance-to-default models
Establish comprehensive validation framework ensuring time-series integrity and preventing lookahead bias
Estimate default risk metrics for U.S. commercial banking institutions over 2016-2023
Investigate the empirical relationship between ESG performance metrics and credit risk indicators
Conduct comparative analysis of market-implied versus accounting-based risk measures
Perform rigorous sensitivity analysis examining solver parameters and alternative methodologies
Execute econometric specifications including instrumental variables, fixed effects, and robust standard errors

Methodological Framework

The computational implementation employs a structured development approach:

Algorithm Specification: All numerical procedures, optimization frameworks, and econometric models specified according to established academic literature with novel adaptations for numerical stability
Implementation Strategy: Code generation leverages AI-assisted development tools under direct researcher supervision, ensuring rapid prototyping while maintaining reproducibility standards
Validation Protocol: Comprehensive testing framework validates (i) numerical convergence, (ii) time-series integrity, (iii) parameter sensitivity, and (iv) comparison with alternative methodologies
Documentation Standards: Automated generation of technical reports ensures transparency, reproducibility, and adherence to computational finance best practices

This approach maintains rigorous academic standards while enhancing development efficiency, enabling focus on methodological innovation and empirical interpretation.

Project Structure

risk_bank/
├── dd_pd_market.ipynb          # Market-based DD/PD calculation (Merton model)
├── dd_pd_accounting.ipynb      # Accounting-based DD/PD calculation (Naive DD)
├── merging.ipynb               # Merge accounting and market datasets
├── analysis.ipynb              # Comprehensive analysis and visualizations
│
├── data/
│   ├── clean/                  # Cleaned input datasets
│   ├── outputs/
│   │   ├── datasheet/         # Timestamped DD/PD outputs
│   │   └── analysis/          # Analysis outputs and outliers
│   └── logs/                   # Execution logs
│
├── archive/
│   └── datasets/              # Archived outputs (max 5 per type)
│
├── docs/
│   ├── reference/             # Technical documentation
│   ├── literature_review/     # Research papers
│   └── writting/              # Working papers and notes
│
└── scripts/                   # Utility scripts

Workflow

1. Calculate Distance to Default and Probability of Default

Market Approach

# Run dd_pd_market.ipynb
# Outputs: market_YYYYMMDD_HHMMSS.csv

Loads market cap, equity volatility, debt, and risk-free rate data
Solves Merton model for asset value (V) and asset volatility (σ_V)
Computes DD_m and PD_m
Archives old files automatically

Accounting Approach

# Run dd_pd_accounting.ipynb
# Outputs: accounting_YYYYMMDD_HHMMSS.csv

Loads balance sheet and market data
Computes equity proxies (price-to-book, D/E, WACC weights)
Calculates naive DD_a and PD_a using Bharath-Shumway methodology
Archives old files automatically

2. Merge Datasets

# Run merging.ipynb
# Outputs: merged_YYYYMMDD_HHMMSS.csv

Loads latest accounting and market datasets
Merges on [instrument, year] with outer join
Applies prefixes: a_* (accounting), m_* (market)
Keeps DD_a, PD_a, DD_m, PD_m unprefixed

3. Comprehensive Analysis

# Run analysis.ipynb
# Outputs: Various analysis files and visualizations

Includes:

Visualizations: Distributions, box plots, scatter plots, time series
Correlation Analysis: Correlation matrices and heatmaps
OLS Regression: With year fixed effects
2SLS Analysis: Instrumental variable estimation
ESG Regression Analysis:
- ESG raw scores (environmental, social, governance)
- ESG combined score
- Dummy variables (size, year)
- Control variables (lnta, td/ta, price_to_book, capital_adequacy)
- Sequential model building
Outlier Detection: Categorizes outliers by data quality issues
- Zero-cost debt
- Low debt (≤1M)
- Low leverage (D/E ≤ 0.05)
- Outputs: outliers_accounting_*.csv, outliers_market_*.csv

Key Features

Numerical Stability & Accuracy

Log-space parameterization ensures non-negativity constraints for asset values and volatilities
Bounded d1/d2 calculations prevent numerical overflow in cumulative normal distribution evaluations
Robust loss function (soft_l1) provides resistance to outlier observations
Convergence criterion achieves machine precision (residuals < 10^-10)
Comprehensive diagnostic framework validates solver performance across parameter space

Time-Series Integrity Protocol

Equity volatility (σ_{E,t-1}): Estimated using returns from period [t-252, t-1] exclusively
Market equity (E_t), face value of debt (F_t), risk-free rate (r_{f,t}): Values as of time t
Asset value (V_t) and asset volatility (σ_{V,t}): Solved at time t conditional on information available through t-1
Distance-to-default (DD_{m,t}) and default probability (PD_{m,t}): Computed at time t
Automated test suite validates absence of lookahead bias (8/8 specifications verified)

Timestamped Outputs

All outputs use format: {type}_YYYYMMDD_HHMMSS.csv (CDT timezone)
Enables version tracking and reproducibility
Automatic archiving (max 5 files per type)

Column Naming Convention

Accounting variables: Prefixed with a_*
Market variables: Prefixed with m_*
Final metrics: DD_a, PD_a, DD_m, PD_m (unprefixed)
Provenance columns track data sources and calculation methods

Data Quality & Validation

Solver status tracking (converged, no_sigma_E, validation failures)
Pre-solve validation filters invalid inputs
Outlier categorization (zero-cost debt, low leverage, etc.)
Missing data indicators and coverage reporting
Comprehensive diagnostic reports

Variables

Distance to Default (DD)

DD_a: Accounting-based distance to default
DD_m: Market-based distance to default

Probability of Default (PD)

PD_a: Accounting-based probability of default (Φ(-DD_a))
PD_m: Market-based probability of default (Φ(-DD_m))

ESG Variables

environmental_pillar_score
social_pillar_score
governance_pillar_score
esg_combined_score

Control Variables

lnta: Log total assets
td/ta: Total debt to total assets ratio
price_to_book_value_per_share
capital_adequacy_total_(%): Capital adequacy ratio

Dummy Variables

size_dummy: 0 if total_assets ≤ 1M, 1 if > 1M
year_dummy: 0 if 2016-2019, 1 if 2020-2023

Technical Stack

Core Dependencies

pip install pandas numpy scipy statsmodels matplotlib seaborn pytz linearmodels

Package Versions (Tested)

Python 3.11+
pandas 2.2+
numpy 2.0+
scipy 1.16+
statsmodels 0.14.5
linearmodels 6.1
matplotlib 3.10+
seaborn (latest)

Development Tools

Jupyter Notebook/Lab for interactive analysis
Git for version control
pytest for automated testing (optional)

Usage

Clone the repository

git clone https://github.com/GuillaumeBld/risk_bank.git
cd risk_bank

Run the workflow

# Step 1: Calculate market DD/PD
jupyter notebook dd_pd_market.ipynb

# Step 2: Calculate accounting DD/PD
jupyter notebook dd_pd_accounting.ipynb

# Step 3: Merge datasets
jupyter notebook merging.ipynb

# Step 4: Run analysis
jupyter notebook analysis.ipynb

Outputs will be in:

data/outputs/datasheet/ - DD/PD datasets
data/outputs/analysis/ - Analysis results and outliers
archive/datasets/ - Archived files

Research Attribution and Methodology

Principal Investigator: Guillaume Bolivard
Quinlan School of Business, Loyola University Chicago

Faculty Supervisor: Dr. Abol Jalilvand
Quinlan School of Business, Loyola University Chicago

Intellectual Contributions

All research design elements, including algorithm specifications, validation frameworks, econometric model designs, and empirical interpretations, were developed by the principal investigator under faculty supervision. The computational implementation leverages AI-assisted code generation as a development tool, analogous to the use of specialized software libraries or computational platforms in quantitative finance research.

Development Protocol

Research Design Phase: Specification of numerical algorithms, optimization procedures, validation protocols, and econometric models according to established literature with methodological enhancements

Implementation Phase: Translation of specifications into executable code using AI-assisted development tools under direct researcher oversight, ensuring adherence to design specifications

Validation Phase: Comprehensive testing including convergence analysis, numerical stability verification, time-series integrity checks, and comparison with alternative methodologies

Documentation Phase: Generation of technical reports, diagnostic summaries, and reproducibility documentation

This workflow maintains rigorous academic standards and full intellectual attribution while employing modern computational tools to enhance research efficiency and code maintainability.

Documentation

Technical Reports

docs/SOLVER_DIAGNOSTIC_REPORT.md - Comprehensive solver validation
PROJECT_STATUS.md - Current project status and progress
tests/test_time_tags.py - Automated time-integrity tests

Key Files

dd_pd_market.ipynb - Market-based Merton model implementation
dd_pd_accounting.ipynb - Accounting-based naive DD implementation
merging.ipynb - Dataset integration pipeline
analysis.ipynb - Econometric analysis (OLS, 2SLS, fixed effects)
solver_diagnostics.ipynb - Solver sensitivity and validation analysis

Academic References

Primary Methodologies

Merton, R. C. (1974). "On the Pricing of Corporate Debt: The Risk Structure of Interest Rates." Journal of Finance, 29(2), 449-470.
Bharath, S. T., & Shumway, T. (2008). "Forecasting Default with the Merton Distance to Default Model." Review of Financial Studies, 21(3), 1339-1369.
Vassalou, M., & Xing, Y. (2004). "Default Risk in Equity Returns." Journal of Finance, 59(2), 831-868.

Numerical Methods

Crosbie, P., & Bohn, J. (2003). "Modeling Default Risk." KMV Corporation.
Powell, M. J. D. (2009). "The BOBYQA Algorithm for Bound Constrained Optimization Without Derivatives." Cambridge NA Report.

ESG and Banking

Chiaramonte, L., & Casu, B. (2017). "Capital and Liquidity Ratios and Financial Distress." British Accounting Review, 49(2), 138-161.
Shakil, M. H., et al. (2019). "Do Environmental, Social and Governance Performance Affect the Financial Performance of Banks?" Business Strategy and the Environment, 28(8), 1446-1459.

Contact

For inquiries or collaboration:

Email: gbolivard@luc.edu
GitHub: @GuillaumeBld

License

This project is licensed under the MIT License - see the LICENSE file for details.

Name		Name	Last commit message	Last commit date
archive		archive
data		data
docs		docs
logs		logs
scripts		scripts
tests		tests
utils		utils
.gitignore		.gitignore
README.md		README.md
analysis.ipynb		analysis.ipynb
dd_pd_accounting.ipynb		dd_pd_accounting.ipynb
dd_pd_market.ipynb		dd_pd_market.ipynb
merging.ipynb		merging.ipynb
solver_diagnostics.ipynb		solver_diagnostics.ipynb

Latest commit

History

Folders and files

Repository files navigation