RBA Historical Statistics

LICENSE

Fetch historical time series from the Reserve Bank of Australia's public CSV endpoints, in one Actor with two modes:

• fetch — pull a single RBA table and push its rows as uniform JSON records.
• discover — crawl the RBA's index page and push a list of every available CSV endpoint (slug, title, category, parent, URL).

Output records always have the same shape regardless of which table you pull, so a single downstream consumer can handle every table the RBA publishes.

{
  "table": "f11.1-data",
  "source_url": "https://www.rba.gov.au/statistics/tables/csv/f11.1-data.csv",
  "fetched_at": "2026-07-09T12:35:26Z",
  "date": "2026-07-09",
  "series": {
    "A$1=USD": 0.6941,
    "Trade-weighted Index May 1970 = 100": 65.1,
    "A$1=CNY": 4.7177
  }
}

Source: The RBA publishes ~200+ statistical tables as CSV files at predictable URLs — no auth, no rate limiting, no JS rendering. The Actor hits those URLs, parses the RBA's mixed-format CSVs (variable preamble + 01-Jan-1983 / 2026-07-08 dates), and normalises everything to ISO-8601 strings.

---

Why use this Actor?

• One shape, every table. All 213 RBA CSVs emit records with the same table / source_url / fetched_at / date / series keys. No per-table parsing downstream.
• Discoverable. Don't know the slug? Run mode: "discover" first and the Actor returns the full index.
• No browser, no proxy, no flakiness. Pure httpx + pandas. Runs in seconds.
• Bidirectional with AI agents. The discover mode lists slugs; the fetch mode takes a slug and returns records. That's a clean tool interface for MCP / LLM use.
• No scraping glue. The RBA's HTML index has nested <li>s up to three levels deep. The Actor handles that.

---

Quick start

# Run from the Apify CLI (you'll need an Apify account + token)
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"f11.1-data","limit":5}'

# Or invoke from Python via the Apify client
python - <<'PY'
from apify_client import ApifyClient
client = ApifyClient(token="YOUR_API_TOKEN")
run = client.actor("mrjerry/rba-statistics").call(
    run_input={"mode": "fetch", "table": "f1.1-data", "limit": 3},
)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["date"], item["series"].get("Cash Rate Target"))
PY

---

Modes

mode: "fetch" (default)

Pull a single RBA table. Returns one record per CSV row, sorted by date.

Field	Type	Required	Default	Description
table	string	yes	f11.1-data	The RBA table slug — get these from mode: "discover" or the TABLES.md. Short forms like f11.1 are accepted and rewritten to f11.1-data.
start_date	string	no	—	ISO YYYY-MM-DD; include only rows on or after this date.
end_date	string	no	—	ISO YYYY-MM-DD; include only rows on or before this date.
limit	integer	no	—	Cap the number of records pushed (after sorting & date filtering).
order	string	no	desc	desc (latest first) or asc (oldest first). RBA CSVs are oldest-first by default.

Output record: { table, source_url, fetched_at, date, series } where series is a {column_name: value} dict for that row.

Examples:

# Latest 5 daily AUD/USD + TWI observations
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"f11.1-data","limit":5}'

# All exchange rates from January 2026
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"f11.1-data","start_date":"2026-01-01"}'

# Last 3 cash-rate observations
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"f1.1-data","limit":3}'

# 10 most recent RBA balance-sheet rows
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"a1-data","limit":10}'

# Oldest 5 inflation-expectation observations
apify call mrjerry/rba-statistics --input '{"mode":"fetch","table":"g3-data","limit":5,"order":"asc"}'

mode: "discover"

Crawl the RBA's statistical-tables index and push one record per available CSV.

Field	Type	Required	Default	Description
category	string	no	—	Restrict results to one of the RBA's 11 H2 sections, e.g. Interest Rates, Exchange Rates, Reserve Bank of Australia, Market Economists' Forecasts. Case-insensitive.

Output record: { slug, title, category, parent_id, url, fetched_at } where slug is what you pass back into fetch mode.

Examples:

# All 213 tables
apify call mrjerry/rba-statistics --input '{"mode":"discover"}'

# Just the 18 interest-rate tables
apify call mrjerry/rba-statistics --input '{"mode":"discover","category":"Interest Rates"}'

# Just the 13 Market Economists' Forecasts tables
apify call mrjerry/rba-statistics --input '{"mode":"discover","category":"Market Economists\u2019 Forecasts"}'

The full 213-entry registry is auto-regenerated into TABLES.md and registry.json on every build.

---

The most useful tables (curated)

The RBA publishes 213 CSVs. Here are the ones you'll reach for most often:

Slug	What it is	Frequency	Records since
f1.1-data	Cash rate target, interbank overnight, BAB/NCD, OIS	Monthly	1969
f11.1-data	Exchange rates — daily (TWI + ~22 bilaterals)	Daily	2023
f1-data	Money-market rates & yields	Daily	varies
f2-data	Capital market yields — government bonds	Daily	varies
f16-data	Indicative mid rates of Australian Government Securities	Daily	varies
a1-data	RBA balance sheet	Weekly	1969
a2-data	Changes in monetary policy & administered rates	As changes occur	varies
a3-daily-open-market-operations	Daily open market operations	Daily (when active)	varies
a3-omo-outright-transaction-detail	OMO outright transactions	As executed	varies
d1-data	Credit & broad money aggregates	Monthly	varies
g1-data	Consumer price inflation	Quarterly	1922
g3-data	Inflation expectations	Quarterly	1993
h1-data	GDP and gross domestic income	Quarterly	varies
j1-gdp-growth	Market economists' GDP growth forecasts	Quarterly	1990
j1-cash-rate	Market economists' cash-rate forecasts	As surveyed	1990
j1-headline-inflation	Market economists' inflation forecasts	Quarterly	1990
j1-unemployment-rate	Market economists' unemployment forecasts	Quarterly	1990

Note on "records since": the RBA trims some history from the live index. For the full long-run series (e.g. f11.1 back to 1983, or monthly exchange rates back to 1969), use the RBA's Historical Data page or pass the slug directly — the Actor will resolve it through the URL pattern fallback.

For the full 213-table registry, see TABLES.md.

---

Categories

discover mode groups tables under the RBA's own 11 H2 sections:

Category	Tables
Assets and Liabilities	61
Payments System	33
Reserve Bank of Australia	30
Household and Business Finances	23
Interest Rates	18
Money and Credit Statistics	16
Market Economists' Forecasts	13
Exchange Rates	5
Output and Labour	5
International Trade and External Finance	5
Inflation and Inflation Expectations	4

---

How the data is parsed

RBA CSVs are not standard. They start with a small preamble (table title, units, source, publication date) before the real header row. The date column is in mixed formats: older tables use 01-Jan-1983; newer ones use 2026-07-08. The Actor:

1. Fetches the CSV via httpx (60s timeout, follow redirects).
2. Scans the first 20 lines for the row whose first cell is the literal token Title — that's the header row.
3. Parses the rest with pandas, normalising the date column to YYYY-MM-DD strings.
4. Sorts by date (descending by default), applies date filters, then limit.
5. Coerces each cell to a number when possible, otherwise keeps the string. Blank cells are dropped from the series dict.
6. Pushes one record per row to the default Apify Dataset.

The discover mode uses httpx + BeautifulSoup to crawl the index page. It handles three different HTML shapes (leaf with title-div, leaf with bare link, container with nested <ul>s) and recurses up to three levels deep to find every CSV endpoint.

---

Use as an LLM tool (MCP)

This Actor is a clean fit for an LLM agent that needs to answer questions about the Australian economy. Two tool calls work together:

1. discover(category: "Interest Rates") → the agent sees the 18 interest-rate tables and their slugs.
2. fetch(table: "f1.1-data", limit: 12) → the agent gets 12 months of cash-rate history.

You can wrap it as an MCP server with one extra Actor using the python-mcp-empty template, or expose it directly to Claude via Apify's MCP integration.

Prompt example

"What has happened to the AUD/USD exchange rate over the last month?"

The agent runs discover(category="Exchange Rates") to find the slug, then fetch(table="f11.1-data", start_date="<30 days ago>") and walks the series dict for the A$1=USD key.

---

Local development

# Install
pip install -r requirements.txt

# Run with a test input
echo '{"mode":"discover","category":"Interest Rates"}' \
    > storage/key_value_stores/default/INPUT.json
apify run

# Inspect the dataset
ls storage/datasets/default/
cat storage/datasets/default/000000001.json

To regenerate the full registry after the RBA adds new tables:

python -m scripts.generate_registry
# writes registry.json and TABLES.md

---

Limits & gotchas

• RBA CSVs are old. f11.1-data has ~10k rows. The Actor reads the whole CSV into memory before filtering; consider start_date / end_date if you only need a slice.
• Free plan = zero monetization. On Apify's free tier, you pay for compute with monthly free credits. This Actor's runs are tiny (a few seconds each) but heavy users should be on a paid plan.
• No retry on transient errors. Single httpx.get with no backoff. The RBA is reliable; if you need retries, add tenacity.
• Series breaks. Some tables publish a separate series-breaks CSV (e.g. b11.1-series-breaks). The Actor doesn't merge these into the data — call them separately if you need them.

---

License

MIT.