# Myrdal — Demography & Migration

Demographic, population and migration series compiled from named public
statistical sources. Named for Gunnar Myrdal (cumulative causation;
*An American Dilemma*, 1944).

Site: https://myrdal.heterodata.org  (Arcanum Research / Heterodata)
Package: Myrdal_Web_v1.0.0 · Anu Framework v12.2

## Counts — five DISTINCT measures, deliberately not equal
- 62 series registered
- 62 series published (nothing is withheld)
- 62 CSV + 62 Parquet + 62 XLSX files (one per series, identical rows)
- 60 series contain at least one data row (D018 and D019 ship a header only)
- 56 series can be charted (D057/D058/D059 are cross-sections with no year
  column; XS001 is a cross-project hand-off note, not observations)
- 1,149,344 observations across 7 source families

## Coverage and liveness
Every "through" date is the LAST REAL OBSERVATION in the file — never a
declared end year and never a projection.
- Long-run world/country population: year 1 CE to 2024 (D001, D023, D024)
- World Development Indicators panel: 1960 to 2023 or 2024 by indicator
- World Inequality Database population: 1800 to 2025
- US counties, American Community Survey 5-year: 2015-2022 (2017-2022 for the
  full 3,231-county panel)
- US county age structure, Population Estimates Program: 2023 vintage — a single
  cross-section with no time axis
- IRS Statistics of Income migration: flow periods 2018-2019 to 2022-2023. The
  key is a YEAR PAIR, because a migration flow spans two filing years.
- Census International Database: observed to 2024. The same files carry that
  publisher's projections to 2100, held separately and never counted as coverage.

## Some things the data says (each checkable in the file named)
- World population 2024: 8.14 billion on the World Bank's own world aggregate,
  8.12 billion summing individual countries. Both are in D001; they differ.
- US foreign-born share 2022: 13.6% nationally; 26.5% California, 1.6% West
  Virginia (D049).
- US dependency ratios, PEP 2023 vintage summed over counties: total 0.549,
  old-age 0.275 (D058).
- Net internal migration 2022-2023: FL, TX and NC gain; CA, NY and IL lose
  (D040, IRS exemptions as the person proxy).

## Downloads — CSV, XLSX, Parquet. Never JSON as a data format.
- /downloads/myrdal_data.zip        every series in all three formats + docs
- /downloads/BUNDLE_MANIFEST.csv    per-member bytes + SHA-256, and a TOTAL row
- /downloads/series/<ID>.csv        one series
- /downloads/series/<ID>.xlsx
- /downloads/series/<ID>.parquet
- /downloads/data_dictionary.csv    62 rows, one per series
- /downloads/PROVENANCE.csv         source, construction and liveness per series
- /downloads/TERMS.md               terms of use
- /downloads/CITATION.cff           how to cite
- /downloads/myrdal_code.zip        the site + package build code (MIT)
- /metadata/series_registry.json    METADATA only, not a data offering

## API (read-only, no key)
- GET /api/series-list              catalogue grouped by source family
- GET /api/series/{id}[?measure=]   chart-ready traces for one series
- GET /api/counts                   the manifest's counts block verbatim
- GET /healthz                      mount state + the live series count

Scripted access is welcome and is counted as real usage. If you are running a
first-party monitor rather than consuming the data, send `X-Myrdal-Synthetic: 1`
so your requests are recorded separately.

## Files inside the package
- chopped/<ID>/<ID>.csv             one CSV per series, long format
- chopped_parquet/<ID>/<ID>.parquet identical rows, columnar
- chopped_xlsx/<ID>/<ID>.xlsx       identical rows, Excel
- data_dictionary.csv, PROVENANCE.csv, TERMS.md, CITATION.cff
- series_registry.json              full Anu registry
- WEB_MANIFEST.json                 package manifest; asserted against the disk
                                    in both directions on every rebuild

## Sources
World Bank World Development Indicators · Maddison Project Database 2023 ·
Clio-Infra · World Inequality Database · US Census Bureau American Community
Survey 5-Year · US Census Bureau International Database · US Census Bureau
Population Estimates Program · IRS Statistics of Income migration data.

Every value traces to one of these. No synthetic and no estimated data. The
compilation is CC-BY-4.0 per CITATION.cff; each upstream publisher keeps its own
terms for its own data (see TERMS.md).

## Known gaps, stated plainly
- D018 and D019 are registered and published but their files contain zero data
  rows. They are kept, labelled, rather than deleted to tidy a count.
- The Anu construction pipeline (10 loaders, 6 processors, 2 validators) is NOT
  published: those scripts still contain absolute workstation paths and are
  withheld pending a publication scrub. Their inputs are named on /methodology.
- The project's knowledge-base extraction of 10 documents is preliminary text
  and tables only. No number on this site depends on it.
