Where the data comes from

Ingestion pipeline

Every weekday at 18:00 SAST a systemd timer runs ingest_jse.py, which downloads the day's CSV (with a polite 1-hour buffer past the JSE 17:00 publish deadline), deduplicates rows by keeping the entry with the highest DaysDeal per instrument, classifies each row as fixed | frn | ilb via the instrument name, and upserts into the single rbonddata table using ON CONFLICT(report_date, bond) DO UPDATE. If the day's file isn't available yet the script walks back up to three weekdays. After ingest it rebuilds boot.json for first-paint.

RSB monthly rates and the RSB-vs-Rbond spread history come from two admin-maintained spreadsheets under /data/ and are loaded on demand via ingest_data_xlsx.py. SARB Policy Rate, ZARONIA, Retail Savings Bond rates and Treasury Bill auction results are scraped by the separate sa-rates-scraper subproject and stored in its own SQLite for audit, with HTML snapshots retained per run.

Spread convention

Every spread is long leg − short leg, quoted in basis points. So 2s10s_ZARGB = (ZARGB10 yield − ZARGB2 yield) × 100. A positive number means a normally-sloped curve at that point; a negative number means inversion. Week-on-week and month-on-month deltas compare the spread to the spread observed 5 and 22 weekdays ago respectively.

The time-series panel plots eight pairs: four generic-curve (2s5s, 2s10s, 5s30s, 10s30s), two R-bond (short-belly and belly-long), the repo-to-10y term premium, and the generic 2s30s. These cover the bits of the curve that move most.

RSB prediction rule

RSA Retail Savings Bond rates are reset at the start of each calendar month, anchored to the preceding month's end-of-month government bond yields plus a spread. We project next month's 2/3/5-year Fixed Rate by:

• Measuring the month-to-date change in ZARGB2, ZARGB3 and ZARGB5 yields (current level vs. last business day of the prior month).
• Rounding that change to the nearest 0.25% — the smallest quantum RSB uses when adjusting rates.
• Adding the rounded delta to the current RSB rate for each term.

The prediction is model-free (no regression, no scaling) and deliberately conservative: the quantum rounding absorbs small noise, and the spread column shows how the RSB-over-government-bond cushion is moving day-by-day. Treasury retains full discretion — this is an indicative projection, not a forecast.

Business-day calendar

Every internal "N business days ago" calculation respects South African public holidays (Good Friday, Family Day, Freedom Day, Workers' Day, Youth Day, Women's Day, Heritage Day, Day of Reconciliation, Christmas, Day of Goodwill) in addition to weekends. We use the holidays Python package with the ZA country code — it handles the Sunday-to-Monday observance shift mandated by the SA Public Holidays Act.

Consequences: the JSE backfill script skips known no-data dates; ingest_jse.py walks back through SA business days when the current day's file isn't up yet; and the RSB prediction panel reports "observed / expected SA business days month-to-date" so you can see at a glance whether the data pipeline has caught up.

Dedupe, idempotence, auditability

• Append-only DB. Every ingest is an upsert — no deletes. Re-running any ingest for the same date is a no-op.
• HTML snapshots. Every scrape saves its raw HTML to html_archive/ so any parsing regression can be replayed against the original page.
• Source runs log. source_runs / ingestion_log records every fetch with its HTTP code, row count and status; surfaced via /api/v1/ingest/status.
• Data caveats. Yields are close-of-day mark-to-market, not live. bond_maturity carries mixed date formats (legacy DD/MM/YYYY vs JSE YYYY-MM-DD) — parse defensively.

Licence

API data under CC BY 4.0 — please attribute "Data: rbond.co.za". Source code under MIT. No paywalls, no contracts, no rate-limit walls unless you're hammering it.

Data accuracy & disclaimer

All rates, spreads and derived metrics published on rbond.co.za are provided on an errors-and-omissions-excepted (E&OE) basis. Data is sourced in good faith from the JSE YieldX Daily Stats file, the South African Reserve Bank, and admin-curated spreadsheets; we make no warranty — express or implied — as to its accuracy, completeness, timeliness or fitness for any particular purpose.

Quotes are close-of-day mark-to-market, not live, and may differ from the prints your dealer gives you. Forward rates, breakevens and synthetic curves (ZARGB, ZARGBIL) are derived from observed instruments via the methods documented above and carry additional model risk.

Nothing published here constitutes investment, tax, accounting or legal advice. This site is a research and education resource. Do not trade, price, hedge, or settle any instrument solely on the basis of numbers displayed here. Use at your own risk; verify against your prime broker, custodian, or the primary source before taking action. Neither rbond.co.za nor its curator accepts liability for any loss, direct or consequential, arising from reliance on this information.