Financial Development and Inequality: Evidence from Panel Data

Desenvolvimento Financeiro e Desigualdade: Evidências em Dados em Painel

Note: The figure below shows the within-country relationship between lagged domestic credit and inequality after removing country-specific fixed effects.

Overview

This project investigates whether financial development — measured by private credit as a share of GDP — is associated with reduced income inequality, measured by the Gini index, in developing countries.

Research Question

Does greater access to private credit reduce income inequality in developing countries?

Hypothesis

Based on the finance-growth-inequality literature, financial deepening may reduce inequality by expanding access to capital for lower-income households and small firms. However, in the short run, credit may primarily benefit higher-income groups, potentially increasing inequality.

Data

• Source: World Development Indicators (World Bank), accessed via the WDI R package
• Period: 2010–2022
• Sample: Developing countries (high-income countries and regional aggregates excluded)
• No raw data file is stored in this repository — the dataset is downloaded directly at runtime (see data/README.md)

Variables

Dependent variable:

• gini: Gini index of income inequality (SI.POV.GINI)

Main variable of interest:

• credito_lag: Private credit to the private sector (% GDP), one-period lag (FS.AST.PRVT.GD.ZS)

Controls:

• pib_pc: GDP per capita growth (NY.GDP.PCAP.KD.ZG)
• inflacao: Inflation, consumer prices (FP.CPI.TOTL.ZG)
• urbano: Urban population (% of total) (SP.URB.TOTL.IN.ZS)

Methods

The following panel data estimators are implemented:

Model	Description
Pooled OLS	Baseline, ignores unobserved heterogeneity
Fixed Effects (one-way)	Controls for country-specific effects
Fixed Effects (two-way)	Controls for country and time effects
Random Effects (Swamy-Arora)	GLS with random country effects
GMM Difference (Arellano-Bond)	Addresses endogeneity and dynamic bias
GMM System (Blundell-Bond)	Adds level equations for efficiency
Local Projections (panel)	Traces the dynamic response path of Gini

All static models use Arellano-robust standard errors clustered at the country level. GMM results use two-step robust standard errors.

Main Results

Os resultados indicam evidência parcial de que maior crédito ao setor privado está associado à redução da desigualdade de renda em países em desenvolvimento. A relação aparece nos modelos de efeitos fixos, efeitos aleatórios, GMM Difference e Local Projections. No entanto, o GMM System sugere que o efeito do crédito perde significância quando se controla de forma mais intensa a persistência do Gini, indicando sensibilidade à especificação dinâmica.

Limitations

• Gini data coverage is unbalanced across countries and years, reducing effective sample size
• The exclusion of high-income countries limits external validity
• Financial development is proxied by a single credit variable; broader measures (e.g., financial inclusion indices) are not used
• Instrument proliferation in GMM is partially mitigated by restricting lag depth to 2–3 periods

Outputs

Tables

File	Description
outputs/tables/static_models.html	Pooled OLS, FE (1-way), FE (2-way), RE — with robust standard errors
outputs/tables/dynamic_gmm_models.html	GMM Difference and GMM System — two-step robust SEs
outputs/tables/gmm_simple_table.html	Simplified GMM results (coefficients, SEs, p-values)
outputs/tables/static_models.png	Static models table (image)
outputs/tables/dynamic_gmm_models.png	GMM models table (image)

Figures

File	Description
outputs/figures/raw_correlation_credit_gini.png	Raw scatter plot: lagged credit vs. Gini
outputs/figures/fixed_effects_credit_gini.png	Within-country variation: demeaned credit vs. demeaned Gini
outputs/figures/static_model_coefficients.png	Comparison of credito_lag coefficients across static models
outputs/figures/local_projections_irf.png	Cumulative IRF: Gini response to credit shock via Local Projections

How to Reproduce

1. Install R (>= 4.0) and the required packages (see below)
2. Open scripts/panel_credit_gini_analysis.R
3. Run the script from top to bottom — tables are exported to outputs/tables/ and figures are rendered in the R graphics device

No data download is required in advance — the WDI package fetches the data automatically.

Required R Packages

install.packages(c(
  "WDI",           # World Development Indicators API
  "plm",           # Panel data models
  "dplyr",         # Data manipulation
  "lmtest",        # Hypothesis tests
  "sandwich",      # Robust standard errors
  "ggplot2",       # Graphics
  "lpirfs",        # Local Projections
  "modelsummary"   # Regression tables
))

Repository Structure

.
├── README.md
├── .gitignore
├── scripts/
│   └── panel_credit_gini_analysis.R
├── outputs/
│   ├── tables/
│   │   ├── static_models.html
│   │   ├── dynamic_gmm_models.html
│   │   ├── gmm_simple_table.html
│   │   ├── static_models.png
│   │   └── dynamic_gmm_models.png
│   └── figures/
│       ├── raw_correlation_credit_gini.png
│       ├── fixed_effects_credit_gini.png
│       ├── static_model_coefficients.png
│       ├── local_projections_irf.png
│       └── README.md
├── slides/
│   └── financial_development_inequality_panel.pdf
└── data/
    └── README.md

Citation

World Bank (2024). World Development Indicators. Retrieved via the WDI R package.