FedPulse

FedPulse works with two different debt universes depending on the context:

• Total federal debt (GFDEBTN, ~$37.6T) — used for headline metrics including Total Debt Outstanding, Weighted Average Rate, and Annual Interest Expense. This includes both marketable securities and intragovernmental holdings (~$7.5T held by trust funds such as Social Security and Medicare).
• Marketable securities (~$28.9T from MSPD) — the five classes of tradeable Treasury instruments (Bills, Notes, Bonds, TIPS, and Floating Rate Notes) that form the basis of the roll-schedule analysis and scenario projections. These are the instruments that must be refinanced at prevailing market rates when they mature.

Although the roll-schedule analysis is centered on marketable securities, the scenario engine also models intragovernmental debt repricing. Intragov holdings (~$7.5T) are primarily short-term certificates of indebtedness that reprice at a statutory rate defined by 31 USC 3111(b) — the average market yield on outstanding marketable securities with 4+ years remaining maturity. The engine estimates this statutory rate from the scenario's yield curve weighted by outstanding amounts in each maturity bucket, then models ~59% annual turnover (20%/quarter) of the intragov portfolio. The resulting intragov interest expense is added to the marketable interest expense to produce total projected interest expense.

Sourced from FRED series GFDEBTN — total federal debt including intragovernmental holdings, reported quarterly in USD millions by the Treasury. Displayed in trillions. If GFDEBTN is unavailable, falls back to the sum of marketable outstanding amounts from MSPD (Bills + Notes + Bonds + TIPS + FRNs).

Primary source: Treasury's published average interest rate on total interest-bearing debt, which covers all federal obligations including intragovernmental holdings. This rate is sourced from Treasury's Fiscal Data API (avg_interest_rates endpoint).

Fallback: if the published rate is unavailable, WAR is computed from CUSIP-level MSPD data. Each security's effective rate is weighted by outstanding principal — Notes and Bonds use their stated coupon, Bills use actual auction discount rates from TreasuryDirect, TIPS use their stated real coupon, and FRNs use the current Fed Funds rate as a proxy. Securities with missing rate data are excluded to avoid understating the effective borrowing cost.

The average remaining time-to-maturity across all outstanding marketable securities (from MSPD), weighted by outstanding principal. Anchored to the MSPD record date. Expressed in years.

The headline figure is Treasury's net interest actually paid over the trailing twelve months, summed from the Monthly Treasury Statement (table 9, “Net Interest”), on the same cash basis and rolling window as the T12 deficit. On hover the header and the Fiscal Context card show the implied figure: total debt outstanding multiplied by the weighted average rate. Because that product includes the debt the government owes its own trust funds, it is larger than the outlay and is best read as what the whole debt costs at today's average rate rather than as a payment. If the newest statement is more than about ten weeks old the implied figure is shown alone.

Computed from the latest BEA quarterly data: federal interest payments divided by federal revenue (NIPA Table 3.2). Both values are Seasonally Adjusted Annual Rates (SAAR) in USD millions.

Total federal debt (GFDEBTN) divided by GDP (BEA Table 1.1.5). Both values are converted to a common unit before computing the ratio. Falls back to a static GDP estimate if BEA data is unavailable.

The trailing 12-month federal deficit: the sum of the twelve most recent monthly surplus/deficit figures from the Monthly Treasury Statement (FRED series MTSDS133FMS), on Treasury's cash basis. A rolling window is used deliberately. Treasury's own headline is the fiscal-year-to-date deficit, which resets every October and so shrinks to a partial-year figure each autumn; the rolling window is always a full year, comparable from one month to the next, and matches Treasury's figure only at the fiscal year-end in September. Both are built from the same monthly statements. The header shows the deficit as a share of the latest quarterly GDP, with the dollar amount on hover. If the newest statement is more than about ten weeks old the figure is withheld rather than replaced with a figure on another basis.

The scenario engine models debt refinancing over a chosen horizon (1Y, 3Y, 5Y, or 10Y). For each quarter, securities maturing in that period are removed from the portfolio and refinanced at the scenario's yield curve rates, matched by remaining maturity to the appropriate tenor bucket. The weighted average rate and annual interest expense are recalculated after each refinancing event.

The engine runs two parallel projections:

• Static Debt — total marketable outstanding remains constant. Maturing securities are replaced one-for-one at the scenario's rates, with no net new issuance.
• Projected Debt Issuance — total outstanding grows according to CBO's projected “Debt Held by the Public” trajectory (from Table 1-1 of the Budget and Economic Outlook). CBO's annual end-of-fiscal-year debt levels are linearly interpolated to quarterly targets. Each quarter, net new issuance equals the difference between the CBO target and current outstanding, distributed across security types (Bills, Notes, Bonds, TIPS, FRNs) in proportion to the current MSPD composition. If CBO debt data is not available, falls back to accumulating CBO projected deficits each quarter.

Both projections include intragovernmental interest expense modeled via the statutory rate mechanism described above.

When CBO data is available, scenario charts include projected revenue and deficit trajectories from CBO as reference lines. The interest-to-revenue ratio at the projection horizon uses CBO's projected revenue for the corresponding fiscal year.

Pre-computed scenarios use three yield curve sources:

• Current Curve — the latest Treasury yield curve from FRED (DGS1MO–DGS30).
• Market Forwards — the one-year forward curve implied by today's curve. Par yields are bootstrapped to discount factors, the discount curve is rolled forward one year, and the result is re-expressed as par yields so it is directly comparable to the spot curve. Tenors of a year or less are quoted as forward zero rates, matching how bills are quoted.
• Historical — rate environments from 2000, 2006, and 2019, representing different interest rate regimes.

Pre-computed results are cached and refreshed after each data update. Custom scenarios allow users to input arbitrary rate shifts and are computed on-the-fly with rate limiting (10 requests/minute per IP).

Revenue, expenditure, deficit, debt-to-GDP, and interest-to-revenue charts show quarterly BEA actuals. When CBO data is available, projected values extend 3 years beyond the latest historical quarter as lighter-shaded bars or dashed lines.

This section compares CBO's projected net interest expense against the interest expense implied by the current yield curve and the Treasury's maturity schedule (from the Projected Debt Issuance scenario). Because CBO's “net interest” excludes intragovernmental interest (the government paying itself), the market-implied figure also subtracts the projected intragov component for an apples-to-apples comparison. The market-implied deficit holds CBO's spending and revenue projections constant, substituting only the interest expense component.

The comparison is truncated at the last fiscal year with full quarterly coverage from the scenario projection to avoid partial-year artifacts.

Each of the 18 regularly auctioned tenors (4-, 6-, 8-, 13-, 17-, 26- and 52-week bills; 2-, 3-, 5-, 7- and 10-year notes; 20- and 30-year bonds; 5-, 10- and 30-year TIPS; 2-year FRN) gets a 0–100 score after every auction:

score = clip(50 + 50 × z̄ / 3, 0, 100), where z̄ is the weighted mean of the signed component z-scores, each measured against the tenor's trailing 12 auctions (new issues and reopenings pooled; at least 6 prior auctions are required for a z-score). 50 is typical for that tenor; a composite three standard deviations from normal hits 0 or 100. Components with no data are dropped and the weights renormalised; fewer than three available components gives no score.

Weights are equal (1/7 each) by default and configurable through the DEMAND_SCORE_WEIGHTS environment variable; the weights in force are reported in the API payload (methodology.weights), and this table shows the defaults. The score is linear rather than squashed so it reads as “the average z-score”.

Rank orders the tenors by score, 1 = weakest. Trend is the ordinary-least-squares slope of the score over the tenor's last 3 and 6 auctions, reported as the total move over the window (slope × (n − 1)); a move of −5 points or more is labelled “weakening”, +5 or more “strengthening”, otherwise “stable”. The page's two callouts are the lowest score and the most negative 3-auction slope.

From Treasury's investor-class allotments, the marginal buyer is the class with the largest share over the tenor's trailing four issues. HHI is the sum of squared class shares on the standard 0–10,000 scale. A tenor is flagged fragile when its HHI is more than one standard deviation above its own trailing-12 history and the top class's share has fallen by 3 percentage points or more versus the prior four issues — relative thresholds, because allotments are always fund-concentrated. The heatmap shows each class's latest share as a z-score against that tenor's own history: colour means above or below its usual share, not good or bad.

Auction results, allotments and every cross-source series are stored with the date they were published; revisions are inserted as new vintages rather than overwriting. The page shows the latest vintage, and the Data Freshness view reports each source's cadence, expected lag, last event and revision count.

A true auction tail compares the high yield to the 1pm when-issued yield, which is not published. FedPulse uses the prior day's par yield at the tenor instead and labels the result a proxy everywhere it appears. It captures the same direction but also absorbs any move in the market between the previous close and the auction.