U.S. Small Business Capital Access Atlas

An open-source geospatial research platform for integrating, validating, mapping, and comparing U.S. small-business and capital-access context using authoritative public data.

The Atlas is designed for researchers, economic-development practitioners, educators, data scientists, and policy analysts who need a reproducible way to examine how business activity, employment, payroll, entrepreneurial scale, and capital-support context vary across U.S. geographies.

Latest versioned release

v0.3.0 — County, CDFI, Robustness & Archival Release

• GitHub release: https://github.com/sakera023/us-small-business-capital-access-atlas/releases/tag/v0.3.0
• Includes a Python wheel and source distribution as release assets
• Release was created only after automated linting, tests, and package validation passed

Install the published PyPI package:

pip install us-small-business-capital-access-atlas

Import it in Python:

import capital_access_atlas

PyPI project: https://pypi.org/project/us-small-business-capital-access-atlas/

For exact source reproducibility, the tagged GitHub version can also be installed with:

pip install "git+https://github.com/sakera023/us-small-business-capital-access-atlas.git@v0.3.0"

Python package

The Atlas is published on PyPI as:

us-small-business-capital-access-atlas

Install:

pip install us-small-business-capital-access-atlas

Python import namespace:

import capital_access_atlas

This makes the data-retrieval, geography, index, and validation utilities reusable outside the Streamlit application.

Live public Atlas

Application: https://capital-access-atlas-sakera.streamlit.app/

The public dashboard provides:

• official SBA state-level small-business measures;
• official U.S. Census County Business Patterns state and county data;
• state-to-county geographic drill-down with Census TIGERweb boundaries;
• county establishments, employment, payroll, and industry-concentration analysis;
• current Certified CDFI organization geography;
• state and county comparison workflows;
• downloadable map-ready data;
• an exploratory Capital Access Opportunity Index Lab;
• alternative normalization, weight, omission, and missing-data sensitivity diagnostics;
• data-quality reports and reproducible validation artifacts.

No login or API key is required for the currently integrated public-data layers.

Research objective

The project examines a national-scale question:

Where do small-business activity, economic opportunity, and capital-support context appear to diverge across U.S. geographies, and how robust are those patterns to data and methodological choices?

The Atlas focuses on place-based public-data measurement, not applicant-level underwriting.

Authoritative public data

See Data Sources and Data Governance.

Core research capabilities

Public-data engineering

• direct retrieval from official U.S. government distribution endpoints;
• ZIP/Excel ingestion;
• state and county FIPS normalization;
• numeric-field cleaning;
• state and county aggregation;
• Census ZIP, SBA Excel, and CDFI workbook ingestion;
• resilient CDFI workbook-link discovery;
• source metadata preservation; and
• downloadable clean extracts.

Certified CDFI validation guard

The live CDFI integration now explicitly prefers the institution-level certification worksheet over state-summary sheets and rejects suspiciously small institution lists. This prevents a state summary from being misreported as the full Certified CDFI roster.

Geographic analysis

• U.S. state choropleths;
• 2023 Census TIGERweb county choropleths;
• state-to-county drill-down;
• county industry-concentration HHI;
• interactive metric selection;
• state rankings and multi-state comparison;
• Certified CDFI state geography; and
• explicit source/vintage labels.

Transparent composite-index research

The Capital Access Opportunity Index Lab supports:

• percentile, z-score, winsorized z-score, and robust median/MAD normalization;
• user-defined non-negative weights;
• inverse-direction metrics;
• data-coverage reporting;
• leave-one-metric-out rank sensitivity;
• one-component weight-emphasis sensitivity; and
• deterministic missing-data stress testing.

A composite score is never presented without its component inputs and methodology.

Validation and reproducibility

The repository includes:

• Python 3.11 and 3.12 CI;
• unit tests for geography, public-data parsing, index logic, and validation;
• Streamlit smoke testing;
• a dedicated public-data validation workflow;
• machine-readable state, county, industry-concentration, and robustness artifacts;
• reproducible Jupyter examples;
• documented case studies;
• an independent-review package; and
• an external-use evidence guide.

See Validation and Robustness and Reproducible Public-Data Validation.

Architecture

Authoritative U.S. Public Sources
          |
          v
   Retrieval / Provenance
          |
          v
Geographic + Numeric Cleaning
       /          \
      v            v
Map-Ready Data   Index Components
      |            |
      +------v-----+
             |
             v
     Validation Diagnostics
             |
       +-----+-----+
       |           |
       v           v
Interactive Maps  Rankings / Exports
       |
       v
 Public Streamlit Atlas

See Architecture.

Reproducible examples

Four notebooks demonstrate the research workflow:

1. SBA State Atlas
2. Census CBP State Context
3. Index Sensitivity
4. Data Quality and Provenance

See the examples guide.

Documented case studies

The repository includes methodological case studies that demonstrate how to use the Atlas without making unsupported causal or applicant-level claims:

1. Mid-Atlantic State Comparison
2. Rural and Appalachian Research Workflow
3. High-Growth State Business Context

See the case study index.

Quick start

Clone and install:

git clone https://github.com/sakera023/us-small-business-capital-access-atlas.git
cd us-small-business-capital-access-atlas
python -m venv .venv
pip install -r requirements.txt

Run the public-data dashboard locally:

streamlit run app.py

Development setup:

pip install -r requirements-dev.txt
ruff check src tests scripts app.py
python -m pytest -q

Run the reproducible Census validation workflow locally:

python scripts/run_cbp_validation.py

Research and responsible-use boundary

The Atlas analyzes aggregate geographic data.

It does not:

• determine whether an individual or business should receive credit;
• estimate a real applicant's approval probability;
• infer protected characteristics;
• prove discrimination or causation from map differences;
• establish policy effectiveness from descriptive correlations; or
• treat a composite score as a legal, regulatory, or underwriting finding.

Geographic and index results are only as reliable as their source definitions, vintage, coverage, transformations, and sensitivity to analytical choices.

Academic citation and research metadata

The repository includes:

• CITATION.cff for GitHub's Cite this repository feature;
• CITATION.md with a recommended citation and BibTeX;
• codemeta.json for machine-readable research-software metadata; and
• explicit instructions to cite the underlying SBA and Census datasets separately.

No DOI is claimed unless a verified DOI is minted through an external research-software archive. The repository is prepared for Zenodo archival; see Software DOI and Archival Plan.

External milestones that cannot be self-certified

Two research milestones intentionally remain external:

• Independent methodology review: the repository now provides a structured review package, but completion requires a genuine third-party reviewer.
• Verified outside use: the repository provides an evidence guide, but adoption is recorded only when real users, organizations, courses, or researchers provide verifiable evidence.

The maintainer does not self-create independent reviews or fabricate adoption evidence.

External use and research impact

Genuine outside use is encouraged through:

• public forks and derivative repositories;
• research/adoption issue reports;
• external pull requests;
• independent replication;
• citations;
• public teaching use; and
• documented organizational testing.

See ADOPTION.md, Independent Replication Protocol, Verified External Use Log, Research Outreach Kit, and Research Impact and Adoption Framework.

The project does not manufacture stars, users, citations, testimonials, or adoption claims.

Research roadmap

The roadmap progresses from the current state-level atlas toward:

1. county-level Census business structure;
2. CDFI/community-finance geography;
3. rural and metropolitan comparisons;
4. a validated state/county Capital Access Opportunity Index; and
5. independent replication and public research dissemination.

See Research Roadmap.

Repository structure

.
├── app.py
├── src/capital_access_atlas/
│   ├── analysis.py
│   ├── cdfi.py
│   ├── census_cbp.py
│   ├── geography.py
│   ├── indicators.py
│   └── public_data.py
├── scripts/
│   └── run_cbp_validation.py
├── examples/
├── tests/
├── docs/
│   ├── case-studies/
│   ├── ARCHITECTURE.md
│   ├── DATA_GOVERNANCE.md
│   ├── DATA_SOURCES.md
│   ├── INDEX_METHODOLOGY.md
│   ├── RESEARCH_IMPACT.md
│   ├── RESEARCH_ROADMAP.md
│   ├── INDEPENDENT_REVIEW.md
│   ├── EXTERNAL_USE_EVIDENCE.md
│   ├── REPLICATION_PROTOCOL.md
│   ├── ADOPTION_LOG.md
│   ├── OUTREACH_KIT.md
│   ├── DOI.md
│   ├── VALIDATION.md
│   └── VALIDATION_REPORT.md
├── .github/workflows/
├── ADOPTION.md
├── CITATION.cff
├── CITATION.md
├── codemeta.json
├── CODE_OF_CONDUCT.md
├── CONTRIBUTING.md
├── SECURITY.md
└── LICENSE

Related research software

This Atlas complements Equitable Capital Optimization AI, which focuses on predictive modeling, explainability, fairness auditing, and simulated capital allocation. The Atlas remains methodologically separate and focuses on public, place-based geographic evidence.

Contributing

Contributions are welcome in public-data integration, geographic validation, methodology, documentation, accessibility, and independent replication.

See CONTRIBUTING.md and CODE_OF_CONDUCT.md.

License

MIT License. See LICENSE.

Maintainer

Sakera Begum