Methodology

Definitions, sources, sample windows, caveats, and paper-section references for every series the site exposes.

ρ_IBOR — money-market CIP deviation (IBOR-based)

Formula: ρ = cross-currency basis (bp)/100 − USD benchmark swap rate + foreign IRS, at tenors above 3m; ρ = NDS rate − USD benchmark for non-deliverable currencies; at 3m, ρ = annualized forward premium. ρ is the money-market (benchmark) CIP deviation — the paper's measure of dollar convenience. It enters the government CIP deviation as cip_govt = (y_FX − y_USD − ρ) × 100; negative cip_govt ⇒ the swapped foreign government bond yields less than the U.S. Treasury, i.e., the Treasury premium is negative.
Sources: USD LIBOR / IBOR pre-reform; same-tenor foreign IBOR
Cross-currency basis and NDS quotes (Bloomberg)
Interest-rate swap curves (Bloomberg)
FX spot and forward points (Bloomberg)
Sample window: 2000-01 – 2025-06
Caveats: IBORs carry bank credit risk and panel-quote-driven idiosyncrasies.
Coverage thins for some EME currencies pre-2007.
Paper: Du, Tepper, Verdelhan (2018); methodology §2
Citation: Du–Tepper–Verdelhan (2018), Du–Im–Schreger (2018)

ρ_SOFR — money-market CIP deviation (alternative-benchmark)

Formula: Same construction as ρ_IBOR with post-reform alternative benchmarks on both legs (SOFR vs SARON, SONIA, €STR, TONAR): cross-currency basis/100 − USD benchmark + foreign OIS rate above 3m; annualized forward premium at 3m.
Sources: SOFR (USD), SARON (CHF), SONIA (GBP), €STR (EUR), TONAR (JPY), etc.
OIS swap rates from Bloomberg
Sample window: 2000-01 – 2025-06 (at 3m, ρ is the forward premium and coincides with ρ_IBOR by construction; the OIS-based construction begins with each benchmark transition (from Feb 2021 at tenors of 1y and above))
Caveats: At the 3-month tenor the two bases coincide identically (ρ = annualized forward premium under either benchmark), so ρ_IBOR and ρ_SOFR are the same series there — the toggle distinguishes them only at tenors of 1y and above.
Coverage starts at different dates per currency, as alternative benchmarks were adopted unevenly.
For most currencies the rate is overnight; tenor structure built via OIS swaps.
Known source-data anomaly: PHP 3m values during 2012 violate the dataset's internal identity (cip_govt_sofr ≠ diff_y − ρ_SOFR) by more than 50 bp on 230 rows, up to ~177 bp; served as published in the source file (same count as the release manifest's warning).
Paper: Du, Keerati, Schreger (2025); §2
Citation: Du–Keerati–Schreger (2025)

Government-bond CIP (spliced IBOR/OIS)

Formula: Swapped foreign government bond yield (in USD via the same-currency benchmark swap) minus the US Treasury yield of the same tenor. Sign convention: positive values indicate the U.S. Treasury yields LESS than the swapped foreign bond — i.e., U.S. Treasury convenience is positive against that country. The series uses the IBOR-based swap leg up to each (currency, tenor)'s benchmark transition date and the OIS-based swap leg after — see the splice methodology.
Sources: Foreign government bond yields (Bloomberg generic yields)
US Treasury par yields (Bloomberg)
Currency-specific IBOR and OIS swap curves
Sample window: 2000-01 – 2025-06
Caveats: The IBOR→OIS transition happens on different dates per (currency, tenor); the spliced series respects each row's break date.
Negative values mean the swapped foreign bond yields LESS than the U.S. Treasury — i.e., the foreign sovereign is more convenient than the Treasury at that tenor (the decoupling finding).
For the euro, the foreign sovereign leg is German Bunds, matching the paper (which restricts the euro to Germany); other G10 sovereigns are the national government bond.
The U.S. Treasury premium shown on charts follows the paper's Figure 2 code: per-currency trailing means over a 14-calendar-day window (the paper's '13-day moving average', days −13…0) of the availability-spliced series (SOFR leg wherever present, IBOR otherwise), then the cross-country median per date. The per-currency lines use the break-date splice persisted as cip_govt; at 5y the two splices move the G10 median on ~200 dates in 2021–22 (max ~2 bp, mean ~1 bp).
Dates on which fewer than 7 of the 10 G10 economies report are suppressed from the U.S. Treasury premium (≈1.6% of dates at 5y). The paper has no such rule; without it, overlapping holidays leave 1–3 reporters whose one-day medians jump by tens of basis points.
Note on the paper's prose statistics: the text reports cross-country MEANS (e.g., '−26 bp from 2021 onward', matching Table 2), while Figure 2 — and the series here — show the cross-country MEDIAN. The two aggregations differ (median −28 bp vs mean −25 bp over 2021–2025 at 5y); cite the median when using this site's series.
Paper: Du, Keerati, Schreger (2025); §3
Citation: Du–Keerati–Schreger (2025), Du–Im–Schreger (2018)

Government-bond CIP (IBOR leg only)

Formula: Same form as cip_govt but uses the IBOR-based swap leg across the entire sample, including post-reform.
Sources: Foreign government bond yields
US Treasury par yields
IBOR swap curves
Sample window: 2000-01 – 2025-06
Caveats: Exposed deliberately only for the 'show both legs separately' toggle — not the default plotting series.
Post-reform IBOR quotes may not reflect liquid markets.
Paper: Du, Keerati, Schreger (2025); §3
Citation: Du–Keerati–Schreger (2025)

Government-bond CIP (OIS / SOFR leg only)

Formula: Same form as cip_govt but uses the alternative-benchmark swap leg across the entire sample, including pre-reform where data permits.
Sources: Foreign government bond yields
US Treasury par yields
OIS swap curves
Sample window: 2000-01 – 2025-06 (at 3m, identical to the IBOR leg by construction; the OIS-based leg begins with each benchmark transition (from Feb 2021 at tenors of 1y and above))
Caveats: Exposed deliberately only for the 'show both legs separately' toggle.
Pre-reform coverage requires OIS curves that may not have existed for some currencies.
Paper: Du, Keerati, Schreger (2025); §3
Citation: Du–Keerati–Schreger (2025)