latin-america-financial-development-lab

Replication Guide

This repository now has two replication modes.

Mode A: Reconstructed Public-Data Pipeline

Use this mode in the current environment. It does not require the missing legacy backup files.

From the repository root:

powershell.exe -NoProfile -ExecutionPolicy Bypass -File .\src\reconstruct_public_data.ps1 -Root (Resolve-Path .).Path

This command:

• redownloads official public sources where the legacy URLs still respond;
• downloads World Bank WDI JSON files;
• rebuilds PanelCompleto.reconstructed.csv, CreditType.reconstructed.csv, and EconomicSector.reconstructed.csv;
• regenerates figures, tables, OLS model outputs, dashboard HTML, paper, and executive report.

Expected core outputs:

• data/processed/PanelCompleto.reconstructed.csv
• data/processed/CreditType.reconstructed.csv
• data/processed/EconomicSector.reconstructed.csv
• data/metadata/source_download_status.csv
• outputs/figures/figure_01_*.png through outputs/figures/figure_15_*.png
• outputs/tables/*.csv
• outputs/models/model_results.csv
• dashboard/index.html
• report/financial_development_report.html
• report/executive_report.html

Mode B: Exact Legacy Pipeline

Use this mode only if the original historical files are restored.

Place raw or cleaned data in data/raw/, Data/, Data.ES/, or the legacy subfolders using names such as:

• PanelCompleto.final.csv or PanelCompleto.final.xlsx
• CreditType.final.csv or CreditType.final.xlsx
• EconomicSector.final.csv or EconomicSector.final.xlsx
• IPC.xlsx
• old_tc*.csv, old_*.xls, FinancialSystemBolivia.xlsx

Then install R 4.2 or newer, Quarto, and the packages listed in requirements.txt, and run:

source("src/99_run_all.R")

or:

source("replication/run_project.R")

Dashboard

The current reconstructed dashboard is static and can be opened directly:

dashboard/index.html

The Shiny scaffold remains available for an R-enabled environment:

shiny::runApp("dashboard")

Reproducibility Statement

The reconstructed outputs use real official public data. CreditType.reconstructed and EconomicSector.reconstructed are annual official equivalents, not the exact lost monthly product-credit and sector-credit panels. This limitation is documented in docs/DATA_RECONSTRUCTION_LOG.md and data/metadata/source_download_status.csv.

This site is open source. Improve this page.