# Alaska Campaign Finance Data (APOC)

> Every contribution, expenditure, debt and registration reported to the Alaska
> Public Offices Commission, rescraped from the state's site daily and
> republished as typed Parquet and as APOC's raw CSVs. No key, no rate limit,
> CORS enabled, MIT licensed.

Query the Parquet directly over HTTPS — DuckDB fetches only the columns and row
groups a query touches, so counting rows doesn't mean downloading the file:

    SELECT filer_name, sum(amount) AS raised
    FROM 'https://nickcrews.github.io/apoc-data/data/income.parquet'
    WHERE report_year = 2024
    GROUP BY 1
    ORDER BY raised DESC
    LIMIT 10;

If you can't fetch URLs, stop and ask the person you're working for to attach

    https://nickcrews.github.io/apoc-data/apoc-csvs.zip

to the conversation. Do not answer from memory and do not estimate anything.

## Tables

One Parquet file each, at the URL shown, plus a CSV of the same name in the zip.

- campaign_form — One row per filed report: totals for cash on hand, income, expenses and debt over a reporting period.
  https://nickcrews.github.io/apoc-data/data/campaign_form.parquet
- candidate_registration — Candidates who registered to run, with their committee, treasurer and contact details.
  https://nickcrews.github.io/apoc-data/data/candidate_registration.parquet
- debt — Outstanding debts a campaign or group reported owing.
  https://nickcrews.github.io/apoc-data/data/debt.parquet
- entity_registration — Registered entities, such as independent expenditure groups.
  https://nickcrews.github.io/apoc-data/data/entity_registration.parquet
- expenditures — Every expenditure reported to APOC: who was paid, how much, and what for.
  https://nickcrews.github.io/apoc-data/data/expenditures.parquet
- group_registration — Registered PACs, political parties and other groups.
  https://nickcrews.github.io/apoc-data/data/group_registration.parquet
- income — Every contribution reported to APOC: who gave, how much, and to which candidate or group.
  https://nickcrews.github.io/apoc-data/data/income.parquet
- letter_of_intent — Letters of intent filed by people considering a run for office.
  https://nickcrews.github.io/apoc-data/data/letter_of_intent.parquet

For columns, types, row counts, date ranges, and the CSV header each column was
parsed from, read https://nickcrews.github.io/apoc-data/data/manifest.json, or run
`DESCRIBE SELECT * FROM '<url>'` against any of the files above.

## Things to know before you trust an answer

- Money columns are decimal amounts. Negative amounts are refunds or
  corrections.
- filer_name is the candidate or group that filed the report. In income and
  expenditures, last_business_name/first_name are the other party: the donor
  for income, the payee for expenditures.
- The same person or business is often spelled several different ways, so
  grouping by name alone undercounts. Consider fuzzy matching or grouping on
  a normalized name.
- A few dates are data-entry typos (years like 1934 or 3030).
- report_year is the reporting year, which is not always the year the
  transaction happened — use the date column for that.
- Transaction in 24 hour reports are re-reported in the following report, so don't include both in a total.

## Working from the CSVs instead

They are APOC's raw exports, and need cleaning first:

- Every column is text, under APOC's own headers. Rename them to
  the lowercase names in the manifest, which is what everything here calls them.
- Money looks like "$1,234.56", with negatives in parentheses: "($1,234.56)".
  Strip the "$" and the commas, and turn a wrapping "(...)" into a minus sign.
- Dates are M/D/YYYY with no zero padding.
- Some files have a literal "--------" column separating the transaction's own
  fields from the fields describing the filer who reported it. Drop it. A
  header can repeat on either side of it — the one after the separator is the
  filer's.
- income.csv is far bigger than the rest. If it's too much to load whole,
  read it in chunks or with a library that streams (polars or duckdb rather
  than pandas defaults), or answer from the smaller tables and say which
  ones you used.

## More

- [manifest.json](https://nickcrews.github.io/apoc-data/data/manifest.json): every table's columns, types, row counts and date ranges, and when the data was scraped, as JSON
- [All CSVs, zipped](https://nickcrews.github.io/apoc-data/apoc-csvs.zip): for tools that can only read files attached to the chat
- [Interactive explorer](https://nickcrews.github.io/apoc-data/): the same data as a browser dashboard, with a SQL editor, over DuckDB-WASM
- [Source repository](https://github.com/NickCrews/apoc-data): the scraper, the CSV-to-Parquet converter, and a Python API and CLI for the releases
- [APOC's own site](https://aws.state.ak.us/ApocReports/Campaign/): where this is scraped from, and the authority if the two ever disagree