Curve flattens further as front end lags
Daily par SOFR curves stripped from the CFTC public swap tape, forward-start swap flow, and the derivative positions and collateral that life insurers disclose in NAIC Schedule DB. Everything here is derived and aggregated; no raw trade records or licensed data are republished.
The two-year sits at 4.608 percent, up 11 basis points on the week and 125 over the year, while the ten-year, at 4.5781, has managed only a single basis point weekly and sits at the 99th percentile of its one-year range. Two-tens has gone to minus 3 basis points from plus 23 a month ago. Deep forward prints ran 458 in the last five days against a twenty-day average of 289.2, even as overall swap prints fell to 4,025 from a 4,917 average.
Curve monitor
Par swap rates by tenor, one curve per trading day, built from spot-starting USD fixed-vs-SOFR prints on the DTCC public tape (trimmed medians, tenor-snapped). Scrub to any day; the ghost lines show one month and one year earlier where the series reaches.
IQR-trimmed median per tenor
Par curve, selected day
2y · 10y · 30y par history
2s10s and 10s30s, bp
Forward-start flow
Forward-starting USD swaps on the public tape (effective date more than 10 days after execution). The deep bucket is the one insurers and pension hedgers live in: swaps starting ten or more years forward. Daily bars with a 20-day mean; the table rolls by month. Prints matching the swap and option structures held in statutory filings are scored on the Options page.
Forward-start prints per day
≥10y-forward prints per day
Spot-start IRS notional traded, $bn/day
Monthly roll
notional in $bn; capped prints carry the DTCC cap, so notional is a floorInsurer tracker
Open derivative positions and net fair value from Schedule DB Part A, by filing period, for the insurers whose public PDFs carry position detail. Quarterly blanks exist for four of them; the rest report at year-end only. Net fair value is the sum of reported fair values across all products, in $m.
positions = open rows at period end
Positions and net fair value by period
sparkline = net FV, $mBook turnover, latest year pair
Rows matched on trade date, maturity, notional and counterparty between consecutive year-end blanks. Option-heavy books (Athene, Corebridge) turn over almost entirely inside a year; swap books mostly persist.
Counterparties named across FY2025 books
top 12 by open positionsInsurer roundup
Each day's swap prints scored against the swap structures in insurers' FY2025 Schedule DB filings: Genworth's decade-forward LTC hedges, the 20-year-plus receive-fixed books at MetLife and Pacific Life, Athene's bilateral jumbo hedges, and the uncleared long swaps that never reach a clearing house. A match says the structure is the kind an insurer holds; it does not say who traded it. Direction is not on the public tape, so both sides count. The same scoring runs on the Rates, Asset swaps, TIPS and Options pages.
Prints matching insurer-book structures, per day
Profiles, latest counts, and what the filings hold
Collateral regimes
Collateral pledged by each insurer to counterparties (Schedule DB Part D Section 2), and how it splits between cleared initial margin, cleared variation margin and dealer bilateral postings where the blank itemizes it. The regime shapes how a rate move turns into a margin call: cleared books move with a SIMM-like ladder, bilateral CSAs with a house multiplier.
I = initial, V = variation
Collateral pledged by insurer, year-end, $bn
Itemized split, FY2025
share of itemized rows; blank total may exceed itemized sumHow well public inputs predict the posted collateral
median absolute error, %; trained on 2024 periods, tested on 2025Where the regime is legible from the blank, an analytic model gets close: a constant IM plus deterministic VM explains MetLife's LCH postings to within 3%, and a CSA multiplier on underwater NPV explains Genworth's bilateral book to within 11%. Across mixed-product netting sets with no product split disclosed, learned models do worse than simply carrying forward last year's number. That gap is a disclosure limit, not a modelling one.
Filing coverage
Which insurers post their statutory annual statement publicly, and whether the posted PDF carries Schedule DB position detail. Groups that omit the E-pages disclose totals only; groups with no public posting require NAIC InsData.
Method and pipeline
What runs, when, and what each stage may publish.
pdftotext -layout, header-anchored column parsing for Schedule DB Parts A, B and D. Positions, fair values, counterparties, collateral by margin type.Definitions
- Par node
- Median fixed rate of spot-starting prints whose maturity snaps to the tenor (within tolerance), after dropping prints outside 1.5×IQR.
- Spot start
- Effective date 0–5 calendar days after execution.
- Forward start
- Effective date more than 10 days after execution; ≥10y-forward means at least 3,652 days.
- Capped print
- Notional reported at the DTCC dissemination cap and flagged with "+"; the true size is larger.
- Net fair value
- Sum of Schedule DB Part A fair values across all open positions at period end, $m, sign as reported.
- Pledged by
- Fair value of collateral the insurer has posted to counterparties (Part D Section 2), the blank's own total line.
Requests
Ask for a series, a cut of the data, or a correction. Posts are visible to everyone with the page; sign with a name or handle if you want a reply. Every section above has a "Section requests" link that lands here with the section preselected.