Skip to content

RBNZ data via eolas

The Reserve Bank of New Zealand publishes the canonical NZ money-and-banking statistics. eolas serves 32 datasets from RBNZ, organised by the RBNZ's own letter-numbered table system (B1, B2, …, C series, M series).

If you do anything involving mortgages, exchange rates, NZ monetary aggregates, or balance-of-payments — RBNZ is the source.

What's in the catalogue

The 32 datasets cover RBNZ's statistical-release tables. The naming convention is rbnz_<table_code>_<topic>, e.g. rbnz_b1_exchange_rates_daily.

Interest rates

Dataset Description
rbnz_b1_exchange_rates_daily Daily TWI and bilateral exchange rates (NZD vs major currencies).
rbnz_b1_exchange_rates_monthly Monthly averages of the same.
rbnz_b2_wholesale_rates_daily Wholesale rates, end-of-day close — OCR, bank bills, government bond yields, swap rates.
rbnz_b2_wholesale_rates_monthly Monthly averages of the same (close basis).
rbnz_b3_retail_rates Retail / household rates (term deposits, lending).
rbnz_b20_mortgage_rates Mortgage rates by fixed term and lender.

Money supply, credit, banking

Dataset Description
rbnz_c5_credit_extended Credit extended to households, business, and agriculture.
rbnz_c31_new_mortgage_lending New mortgage lending by LVR band and borrower type — useful for housing-market analysis.
rbnz_c33_mortgage_by_purpose New mortgage lending by purpose (first-home, investor, owner-occupier).
rbnz_c40_monetary_aggregates New residential mortgage lending by debt-to-income (DTI) band. (The monetary_aggregates id is a historical misnomer — the content is DTI lending.)
rbnz_c71_residential_mortgage_lending Residential mortgage lending by rate type (fixed/floating mix).

Macro + external

Dataset Description
rbnz_m7_balance_of_payments Current account, capital account, NIIP.
rbnz_m8_overseas_trade Goods and services exports / imports.
rbnz_m9_labour_market Employment, wages, productivity proxies (RBNZ's view; differs slightly from Stats NZ).

For the full list, browse eolas.fyi/datasets?source=RBNZ.

Refresh schedule

Daily for the rate/FX tables (B1 exchange rates, B2 wholesale rates), daily checks with monthly/quarterly source cadence for the rest (the pipeline's change detection skips unchanged source files). One exception: rbnz_m13_inflation_expectations is archived — RBNZ discontinued its source file in June 2022, so the dataset is the 1995–2022 historical series and no longer updates.

License

All RBNZ statistical data is published under CC-BY 4.0 per the RBNZ Open Data policy. Commercial use is fine; attribution is required.

Recommended attribution: "Source: Reserve Bank of New Zealand, via eolas (eolas.fyi). CC-BY 4.0."

Common patterns

OCR + 90-day rate

The two interest rates every analyst tracks:

import pandas as pd
import matplotlib.pyplot as plt

rates = client.rbnz("rbnz_b2_wholesale_rates_daily", start="2018-01-01")

# Tables are WIDE: one column per series, snake_cased from RBNZ's headers.
rates.plot(x="date",
           y=["cash_rate_official_cash_rate_ocr", "bank_bill_yields_90_days"],
           figsize=(10, 4), title="OCR vs 90-day bank bill")
plt.show()

library(tidyr)
library(ggplot2)

rates <- eolas_get_rbnz("rbnz_b2_wholesale_rates_daily", start = "2018-01-01")

# Tables are wide: one column per series, snake_cased from RBNZ's headers.
ggplot(rates, aes(date)) +
  geom_line(aes(y = cash_rate_official_cash_rate_ocr, colour = "OCR")) +
  geom_line(aes(y = bank_bill_yields_90_days, colour = "90-day")) +
  labs(y = "Rate (%)", colour = NULL)

Mortgage rates by term

mort = client.rbnz("rbnz_b20_mortgage_rates", start="2020-01-01")
# One column per fixed term (wide format)
mort.plot(x="date",
          y="new_standard_residential_mortgage_interest_rates_average_pct_end_of_month_2_years",
          title="NZ 2-year fixed mortgage rate")

LVR composition of new lending

lvr = client.rbnz("rbnz_c31_new_mortgage_lending", start="2018-01-01")
# High-LVR (>80%) share of all new lending — the RBNZ's macroprudential target
lvr["high_lvr_share_pct"] = (
    lvr["higher_than_80_pct_lvr_lending_b1_all_borrower_types_b2_b3_b4_b5"]
    / lvr["total_lending_a1_all_borrower_types_a2_a3_a4_a5"] * 100
)
lvr.plot(x="date", y="high_lvr_share_pct",
         title="High-LVR share of new mortgage lending (%)")

Pipeline use

RBNZ datasets are full-snapshot — each RBNZ release replaces the whole table. Most RBNZ tables are small (exchange-rate series are a few hundred KB; the larger balance-sheet tables top out around 5 MB), so full re-downloads are fast.

Use sync_bulk / eolas_sync_bulk to keep a local file fresh — it issues a lightweight HEAD check and only re-downloads when the snapshot has changed. In weeks where RBNZ hasn't published a new release, the call returns "unchanged" and transfers zero bytes.

result = client.sync_bulk("rbnz_b1_exchange_rates_daily",
                          path="/data/rbnz_b1_exchange_rates_daily.parquet")
print(result.status)  # "downloaded" (first run) or "unchanged"

import pandas as pd
df = pd.read_parquet("/data/rbnz_b1_exchange_rates_daily.parquet")

result <- eolas_sync_bulk("rbnz_b1_exchange_rates_daily",
                          path = "/data/rbnz_b1_exchange_rates_daily.parquet")
result$status  # "downloaded" or "unchanged"

eolas sync rbnz_b1_exchange_rates_daily --out /data/rbnz_b1_exchange_rates_daily.parquet
# → downloaded (first run) or unchanged (snapshot 5503437…)

See the Bulk downloads guide for cron and Airflow recipes.

Source-specific notes

  • Naming: dataset names follow rbnz_<rbnz_table_code>_<topic> so they cross-reference cleanly to RBNZ's own Statistics page — if you read a "B2" release on RBNZ's website, the eolas dataset is rbnz_b2_*.
  • Wide format: tables are one row per date with one column per series, snake_cased from RBNZ's multi-row Excel headers (e.g. cash_rate_official_cash_rate_ocr). No pivoting needed for charting; use df.melt() / tidyr::pivot_longer() if you want long format.
  • Cloudflare: RBNZ's site is Cloudflare-protected. Our pipeline uses residential-proxy fallback to fetch reliably; you don't see this — just clean data. If you're scraping RBNZ directly without proxies you'll see 403s from Fargate-IP-range addresses.
  • Recent retirement: RBNZ retired 8 legacy tables in May 2026 (404s on the source). We removed those from our schedule rather than report broken pipelines; if you have older code referencing them, the dataset names will 404 from us too.

Where to find more