Methodology & Sources
This page documents where the data on RiskCurve comes from, how it’s transformed and how often it updates, so every number on the site can be traced to its source.
Primary sources. Most series are read through the Federal Reserve Economic Data (FRED) API. They come from: the Federal Reserve Board (H.15 interest rates, including Treasury constant-maturity yields, and the effective federal funds rate and target range); the Federal Reserve Board’s GSW zero-coupon yield curve, read directly from the Federal Reserve; the U.S. Department of the Treasury (High Quality Market corporate bond spot rates, and the daily real yield curve read directly from Treasury.gov); the U.S. Bureau of Labor Statistics (CPI, PPI final demand, payrolls, unemployment); the U.S. Bureau of Economic Analysis (GDP, PCE); the Federal Reserve Bank of Cleveland (expected inflation); the Chicago Fed (NFCICREDIT credit subindex); the St. Louis Fed (STLFSI4 financial stress index); FHFA (Purchase-Only House Price Index); Freddie Mac (Primary Mortgage Market Survey); and the National Association of Realtors (existing home sales). The Cboe Volatility Index (VIX) is published by Cboe and read through FRED.
Inflation. Headline and core CPI come from the U.S. Bureau of Labor Statistics and PCE and core PCE from the U.S. Bureau of Economic Analysis, all via FRED. PPI is the Bureau of Labor Statistics’ final demand index — the figure BLS headlines — not the older All Commodities series. The 12-month change for CPI, core CPI and PPI is computed from the unadjusted index, because that is the basis BLS publishes its own headline change on, so our arithmetic and theirs are the same; the month-over-month change is computed from the seasonally adjusted index. BEA publishes the PCE price index only seasonally adjusted, so both the month and the 12-month change for PCE come from that single series. Every comparison is made with the same month a year earlier rather than with a fixed number of observations, so a missing month cannot quietly turn a 12-month change into a 13-month one.
Contributions to inflation. The stacked bars split the headline rate into components weighted by BLS relative importance, applied by the year each observation belongs to: food, energy, core goods and core services for CPI, and foods, energy, goods less foods and energy, trade services, other services and construction for PPI — each of them the component series of the final demand index those weights describe. BLS recomputes the weights each December, so the published set for an observation’s own year is used rather than one table for all history; PCE contributions use BEA nominal expenditure shares instead. The total line is the published 12-month change of the same index, and any gap between it and the sum of the components is shown in the tooltip rather than absorbed into a component. BLS chains the final demand index, so the weighted pieces do not add up exactly; the gap, at most about 0.13 point, appears in the tooltip as “Rounding & weights.”
Inflation expectations. The 1-, 2- and 3-year expected inflation lines are model estimates from the Federal Reserve Bank of Cleveland, published monthly, not survey results. The market-implied expectations — the 5-year and 10-year breakevens and the 5-year, 5-year forward — are read from the Treasury’s nominal and inflation-protected curve.
Estimated 2Y/3Y real yields. Published Treasury real yields start at 5 years. Our 2- and 3-year real yields and their breakevens come from the Federal Reserve’s GSW zero-coupon yield curve, a daily model fitted to the TIPS market, because no TIPS series is published at those tenors. The 5-, 7-, 10-, 20- and 30-year real yields come from the U.S. Treasury’s daily real yield curve. The GSW curve publishes a few days behind the Treasury curve, so the 2- and 3-year tenors can lag the rest of the curve; every breakeven is computed from the nominal and real observation for the same date.
Macro Regime inputs. The Macro Regime tab uses daily prices for the SPY and UUP exchange-traded funds from Yahoo Finance.
Real Rates Regime. The Real Rates Regime tab and page split the 10-year Treasury yield into its real yield (the 10-year TIPS constant maturity, DFII10) and a breakeven, computed as the 10-year nominal yield (DGS10) minus that real yield on the same date — the same definition the Yields tab and the Overview use, so one breakeven is shown everywhere. Each part’s move over the last 20 trading sessions is divided by its own recent daily volatility, taken over the trailing 252 sessions and scaled by the square root of the window, giving a move in σ. The signs of the two σ moves place the market in one of four quadrants — Reflation, Rising yields, Tightening or Falling yields — or, when both are inside 0.5σ, Quiet. A Stress label overrides the quadrant when real yields rise while the dollar (UUP) and equities (SPY) are each more than one standard deviation down. Every window is anchored on the last session the model could label, never on today.
Rate Outlook. The Rate Outlook tab and the Fed Rate Expectations page are our own calculation of a market-implied federal funds path and the probability of each range after each FOMC meeting. The inputs are public-domain data: the U.S. Treasury par yield curve (read directly from Treasury.gov), the effective federal funds rate and target range (Federal Reserve H.15, via FRED), and the FOMC meeting calendar published by the Federal Reserve. A bill-versus-fed-funds adjustment lines the short end of the curve up with the effective rate. It uses 1-month bills that mature before the next meeting. In the four weeks before each meeting, every 1-month bill matures after it, so the adjustment is held at its last measured level and that meeting’s odds are marked lower confidence. If the gap between bills and fed funds shifts during those weeks, the odds for that meeting can be off in either direction. When the Fed announces a decision, the page moves to the next meeting and restarts the odds from the decision, using the most recent prices until that day’s Treasury data is in. A least-squares fit gives one rate per period between meetings, and the odds are chained into range probabilities. It’s market-implied, not a forecast, and it’s not CME FedWatch data.
Perpetuals panel. The perps ticker and its candlestick panel show live Hyperliquid market data, streamed to your browser over Hyperliquid’s public WebSocket, with a 30-second refresh over its public Info API as the fallback. Both go straight to Hyperliquid rather than through our servers, and RiskCurve does not store it. You can add any Hyperliquid perpetual to the table; the markets you add are kept in your browser and are never sent to us. Every instrument there is a perpetual futures contract, not a spot price or an official index level, so its price can differ from the official close and from the daily closes shown elsewhere on this site.
Economic calendar. The Overview and News tabs show the economic calendar through TradingView’s official widget: economic calendar widget by TradingView, loaded in your browser directly from TradingView. TradingView provides those events and times under its own terms, and RiskCurve does not fetch, store or reprocess that calendar data on its servers.
News headlines and voice squawk: widgets by FinancialJuice. RiskCurve doesn’t edit or verify them.
Series we no longer publish. The ICE BofA high-yield and investment-grade option-adjusted spread series (BAMLH0A0HYM2, BAMLC0A0CM) are marked as requiring pre-approval on FRED, and ICE’s terms don’t permit redistribution without written consent. We’ve removed them from every page, dataset and chart.
Derived series. Where a figure isn’t published directly, we derive it and say so on the page. The investment-grade credit spread is a proxy: the Treasury High Quality Market 10-year corporate spot rate (HQMCB10YR) minus the 10-year Treasury yield for the same month, on the same monthly-average basis. CPI and PCE are published as price levels, so we compute the year-over-year change for display. Payrolls are shown as the month-over-month change.
Update cadence. A scheduled job runs twice a day. It fetches each active series, stores the observations, and rebuilds the cached values and charts.
Limitations. Government data is revised after first publication; we show the latest revised values. Some series publish with a lag, shelter inflation and GDP in particular, so the most recent data point may describe a period that’s already passed. Chart images may lag the underlying data by one scheduled run.
Questions, corrections or data requests? Get in touch. See also our Disclaimer.