Part I · How the dashboard works

The Global Macro Cockpit is a leading-indicator-first dashboard of the business cycle for the United States and the main global economies. It has 109 tiles on 11 tabs: 98 data series shown on the public site, 5 licensed series shown there as links, and 6 manual tiles. This guide explains how it is built, what each tab and tile shows, how every number is computed and how each indicator’s publisher and its history suggest reading it.

The guide is written for readers who know the basics of macroeconomics and markets. Terms are defined where they first matter; the formulas are collected in Appendix A. Nothing here is investment advice.

The idea: follow the transmission chain

The dashboard is organised by transmission rather than by country. Its design brief describes the chain along which a change in financial conditions works its way through the economy:

financial conditions and liquidity → credit and risk appetite → orders and sentiment → activity → labour → inflation → policy → back to financial conditions

The brief’s premise is that leads and lags live along this chain: money, credit and the yield curve lead by quarters; orders, surveys and jobless claims by weeks to months; activity and inflation are coincident to lagging; policy responds with a lag and feeds back into conditions. The tabs follow the chain from left to right. For how far each indicator has actually led in its own history, see its section in Part III.

Each tile carries an indicator class, shown in its header. In the brief’s words the emphasis is on leading indicators, with coincident and lagging ones included to confirm or give context. Of the 109 tiles, 73 are tagged leading, 23 coincident and 5 lagging; the remaining 8 carry the tags fragility (3), overlay (3) and derived (2).

The cockpit

The second organising idea in the brief is a car’s cockpit, which says how to read each group. The brief describes the economy, like a road, as forecastable but not deterministic, and the dashboard as a view that helps choose a safe speed for the next bend rather than a crystal ball. Each tab is assigned to a zone (R/views.R); the questions are the brief’s:

Zone	Question it answers	Tabs
Warning lights	Does something need attention now?	Recession Watch
Windscreen	Where are we heading; is there a bend ahead?	Conditions & Liquidity, Credit & Rates, Inflation, Cycle & Growth, Labor, Trade & Commodities (in part)
Weather	Conditions to adapt to but not control	Inflation (with the windscreen)
GPS	Where in the cycle are we?	Cycle & Growth (with the windscreen)
Mirrors	What has already happened, and what is catching up?	Trade & Commodities, Consumer & Housing
Side windows	What are the other drivers doing?	Markets & Uncertainty, China & Leaders
Speedometer	How fast are we going relative to a safe speed?	Bubbles / Froth

The brief’s reading rule for the speedometer: the speedometer is not a clock. A froth gauge tells you the speed, not the time of a crash, and the brief says to read it against the windscreen rather than alone.

The two editions

The same code produces two editions.

The local app (R Shiny) is the private, live edition. It fetches every series, including licensed ones, lets you choose the start of the history shown (“Show history since”, three years back by default) and refetches when you click Refresh data.
The public site is a static snapshot rebuilt by a scheduled job every day at 06:17 UTC and on every change to the code. It shows the last five years of each series with 1-, 3- and 5-year zoom buttons (3 years by default); the growth and inflation regime tiles carry their history from 1960. Series whose licences don’t allow republication appear as link-only tiles that point to the provider (see Data and licences).

Anatomy of a tile

Every data tile has the same parts, top to bottom:

Title in the form “name — zone” where a zone is needed (for example 10Y yield — Germany).
Badge (where one exists): a short verdict on the latest value. The sidebar’s legend gives the colours as good on track, warn caution and bad recession signal; a few badges are neutral grey. See Badges.
ⓘ opens a short note on how to read the tile; this guide is the long version.
Headline value in the tile’s unit, then the change versus the prior observation with an arrow. The arrow’s colour shows direction only (green up, red down), not whether the move is good news: a rising Sahm rule gets a green arrow.
The date of the latest observation with a freshness dot.
Mini-chart with a reference line where the dashboard sets one (for example 0 for the yield curves, 100 for the OECD leading indicators, 2 for most inflation rates).
Footer: in the app, the provider, series code and frequency; on the public site, the source credit with a link to the source.

The freshness dot turns amber when the latest observation is older than the series’ frequency would suggest:

Frequency	Amber after
Daily	7 days
Weekly	12 days
Monthly	45 days
Quarterly	130 days
Event (policy rates)	never

Each tile’s fact box in Part III gives an example of how soon after its period the latest value was available, so an amber dot can be checked against the series’ normal lag.

A few tiles look different:

Two-line tiles (Stocks vs bonds) draw two series that are compared, with the gap between them as the headline value.
Regime tiles (Growth & inflation) place an economy in one of four growth and inflation regimes with a quadrant chart and a strip of past regimes.
Manual tiles stand in for licensed indicators such as the regional PMIs: they explain the indicator and link to the publisher’s release.
Link-only tiles on the public site name a series the dashboard tracks but may not republish, with a link to it.

Display modes

The sidebar switches every chart between three modes. The badge always reads the level.

Level: the series in its own unit.
Z-score: how many standard deviations the value sits from its mean over the window shown. It makes series in different units comparable: “how stretched is this compared with its own recent past?”
Percentile: the share of the window’s observations at or below the value, from 0 to 100.

The window is the history the edition holds: on the public site the full five years whatever the zoom; in the app, the period since “Show history since”. A z-score of +2 over five years of a trending series means something different from +2 over thirty years, so read the modes as relative to that window. In the z-score and percentile modes the reference line moves to 0 and 50, and the change versus prior is measured in that mode’s units. The formulas are in Appendix A.2.

Badges and the recession count

A badge appears only where a threshold gives the level a meaning, such as a yield curve below zero or the Sahm rule at 0.5 points or above; the code’s own rule is that it doesn’t fabricate thresholds. Some thresholds come from the indicator’s publisher or inventor, others are the dashboard’s convention; each tile’s section in Part III gives the exact rules and says which is which.

The sidebar sums up the Recession Watch tab: Recession signals counts its tiles with a red badge (out of the tiles with a badge) and Cautions counts the amber ones. How each signal has behaved around past recessions is in its tile’s section.

Data and licences

The dashboard takes each statistic from the organisation that produces it, through its official API or download: FRED (St. Louis Fed) for most US series, the OECD, Eurostat, the ECB, the IMF, the Bank of England, the ONS, the Bank of Japan and the New York Fed. It avoids third-party benchmarks and vendor indices where an official series exists.

The public site republishes data only for series whose terms allow it: public-domain data, data free to reuse with a citation, the OECD’s and Eurostat’s CC BY 4.0 licence, the UK Open Government Licence and the reuse terms of the ECB, IMF, Bank of Japan and New York Fed. Everything else (ICE BofA credit spreads, the S&P 500 and S&P Cotality Case-Shiller house prices) is link-only there. Each tile’s fact box in Part III gives its licence, and the site’s footer carries the notices some providers require.

A reading routine

One way to work through the board, following the cockpit:

Warning lights. Check the recession count. Is anything red, and is it one of the tiles classed as leading (the curves) or as coincident (the Sahm rule, the recession probability)?
Windscreen. Are financial conditions tightening or easing, credit spreads widening, leading indicators rolling over, claims rising off their lows?
Weather and GPS. Which growth and inflation regime is each economy in, and how long has it been there?
Mirrors and side windows. Do consumers, housing, trade, markets and China confirm or contradict the windscreen?
Speedometer. How stretched are valuations, read against the windscreen?

Switching to z-scores shows which series are unusual against their own recent history. Percentiles rank each value within the window, so a single extreme observation moves them less than it moves z-scores.

Part II · The tabs

The eleven tabs follow the transmission chain from the warning lights to the speedometer. Each section below says what the tab is for, how its tiles relate and what to look for across them; Part III covers the tiles one by one.

Recession Watch

Cockpit zone: Warning lights
Tiles: 4: Yield curve 10Y–3M · Yield curve 10Y–2Y · Sahm rule (real-time) · Recession probability

Four US recession signals kept together so they can be read as a set. The dashboard classes the two yield curves as leading and the Sahm rule and the recession probability as coincident, and their histories bear that out:

The yield curves. Before each recession since the 10Y–3M series began in 1982, its monthly average turned negative 8 to 16 months before the NBER’s business-cycle peak. The 10Y–2Y did so 10 to 22 months before each peak from 1980 to 2007, but not before the February 2020 peak.
The Sahm rule and the recession probability. The Sahm rule crossed its 0.5 trigger between two months before and four months after each peak from 1969 to 2020. The recession probability reached 50% one to eight months after each peak in that period except 2001, when its highest value was 30%.

The tiles below give the dates for each recession. The NBER, which dates US business cycles, treats the months from a peak in economic activity to the following trough as a recession; its most recent peak is February 2020 and its most recent trough April 2020.

The sidebar’s Recession signals count (red badges on this tab) and Cautions count (amber badges) summarise the tab (Appendix A.3).

Conditions & Liquidity

Cockpit zone: Windscreen
Tiles: 8: Financial conditions (NFCI) · Adjusted NFCI · St. Louis Fed financial stress · Fed balance sheet · Treasury General Account · Overnight reverse repo · Net liquidity (Fed − TGA − RRP) · M2 growth (year on year)

The first link in the dashboard’s transmission chain. The tab holds two kinds of gauge:

Conditions indices combine many market and financial measures into one number: the Chicago Fed’s NFCI and its adjusted version, and the St. Louis Fed Financial Stress Index. Each is scaled so that zero is its historical average.
Liquidity measures track the dollar system’s plumbing: the Fed’s balance sheet, the Treasury General Account and the overnight reverse repo, the net liquidity that combines them, and the growth of the broad money stock M2.

The liquidity tiles are linked through the Fed’s weekly balance-sheet release (H.4.1). It lists the factors that supply reserve funds, mainly the Fed’s securities and loans, and the factors other than reserve balances that absorb them, among them currency, reverse repurchase agreements and the Treasury’s account. Banks’ reserve balances are the difference. In the release of 1 October 2026, for Wednesday 30 September, factors supplying reserve funds came to $6,790,836 million and absorbing factors to $3,842,746 million, leaving $2,948,090 million of reserve balances.

Credit & Rates

Cockpit zone: Windscreen
Tiles: 21: C&I lending standards (SLOOS) · HY credit spread (OAS) · IG credit spread (OAS) · HY − IG (decompression) · 10Y Treasury volatility (realized) · 3M Treasury yield · 2Y Treasury yield · 10Y Treasury yield · Fed funds rate (effective) · ECB deposit rate · 10Y yield — Euro area · 10Y yield — Germany · 10Y yield — France · 10Y yield — Italy · 10Y yield — UK · 10Y yield — Japan · 10Y yield — Canada · 3M govt yield — Euro area (AAA) · SONIA — UK · Overnight call rate — Japan · 3M rate — Canada

Credit spreads and the structure of interest rates: what borrowing costs, and how much extra investors demand to lend to riskier borrowers. The tab has four groups:

US credit: banks’ lending standards from the Fed’s loan officer survey, and the option-adjusted spreads of high-yield and investment-grade corporate bonds with the gap between them. The three ICE BofA spread tiles are licensed data: the local app shows them, the public site links to them.
US rates: the effective federal funds rate, Treasury yields at 3 months, 2 years and 10 years, and the realized volatility of the 10-year yield.
Other policy and short rates: the ECB deposit rate, the euro-area 3-month AAA yield, SONIA for the UK, the overnight call rate for Japan and a 3-month rate for Canada.
Other 10-year government yields: the euro area, Germany, France, Italy, the UK, Japan and Canada.

The two yield-curve spreads built from these Treasury yields are on the Recession Watch tab.

Inflation

Cockpit zone: Windscreen / Weather
Tiles: 14: 5Y inflation breakeven · 10Y inflation breakeven · 5y5y forward breakeven · 1Y expected inflation (UMich) · CPI inflation — US · Core CPI inflation — US · Core PCE inflation — US · PPI final demand inflation — US · HICP inflation — Euro area · HICP inflation — EU · HICP inflation — Germany · HICP inflation — France · CPI inflation — UK · CPI inflation — Japan

Inflation, measured two ways:

Expected inflation: market-implied 5-year, 10-year and 5-year, 5-year forward breakevens from Treasury yields, and consumers’ one-year expectations from the University of Michigan survey.
Price changes: US CPI, core CPI, core PCE and producer prices, and consumer-price inflation for the euro area, the EU, Germany, France, the UK and Japan. All are changes over twelve months, in %.

The central banks’ own targets:

Central bank	Target
Federal Reserve	2%, as measured by the annual change in the price index for personal consumption expenditures (FOMC Statement on Longer-Run Goals)
ECB	2% inflation over the medium term, measured by the HICP; deviations on either side are equally undesirable
Bank of England	2% CPI inflation, a target set by the Government
Bank of Japan	a price stability target of 2%

The consumer-price tiles draw a reference line at 2 and carry the same badge, whose bands are the dashboard’s own:

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

Cycle & Growth

Cockpit zone: Windscreen / GPS
Tiles: 19: Growth & inflation — US · Growth & inflation — Euro area · Growth & inflation — UK · Growth & inflation — Japan · Growth & inflation — China · OECD CLI — G20 · Core capital goods orders — US · OECD CLI — US · Empire State manufacturing · Philadelphia Fed manufacturing · Industrial production — US · Real GDP — US · OECD CLI — G7 · OECD CLI — UK · OECD CLI — Japan · OECD CLI — Germany · PMI — Euro area · PMI — UK · PMI — Japan

Where the cycle is heading and how fast, from three angles:

Growth and inflation regimes. The first five tiles place the US, the euro area, the UK, Japan and China in one of four regimes, combining the direction of a leading indicator with the direction of inflation.
Leading indicators. The OECD’s composite leading indicators for the G20, the US, the G7, the UK, Japan and Germany, with the licensed purchasing managers’ indices for the euro area, the UK and Japan as links.
US activity. Core capital goods orders, the Empire State and Philadelphia Fed manufacturing surveys, industrial production and real GDP. The dashboard classes the orders and surveys as leading, industrial production as coincident and GDP as lagging.

The OECD describes its leading indicators as built from component series that show a reasonably consistent relationship with a reference series (GDP since March 2012, industrial production before) at turning points. It says they give qualitative rather than quantitative information, so their main message is whether they rise or fall, and that turning points in the reference series have been found about 4 to 8 months, on average, after the indicator signalled them.

Labor

Cockpit zone: Windscreen
Tiles: 11: Initial jobless claims — US · Continuing claims — US · Quits rate (JOLTS) — US · Job openings (JOLTS) — US · Temp-help employment — US · Manufacturing hours — US · Nonfarm payrolls — US · Unemployment rate — US · Unemployment rate — UK · Unemployment rate — Japan · Unemployment rate — Germany

The US labour market from the weekly unemployment-insurance claims to the monthly jobs report, with unemployment rates for the UK, Japan and Germany. The dashboard classes the US tiles in three groups:

Leading: initial and continuing claims, the quits rate and job openings from the JOLTS survey, temporary-help employment and manufacturing hours.
Coincident: nonfarm payrolls, shown as the monthly change.
Lagging: the unemployment rate, which also feeds the Sahm rule on the Recession Watch tab.

The two claims tiles carry the dashboard’s own badge: amber when the latest week is more than 15% above the lowest of the past 52 weeks.

Condition	Badge
$100\,(x_n - \min_{52})/\min_{52} > 15$	Rising off lows
$100\,(x_n - \min_{52})/\min_{52} \le 15$	Near lows — firm

Here $\min_{52}$ is the lowest of the 52 latest weekly values, the current week included.

Trade & Commodities

Cockpit zone: Mirrors / Windscreen
Tiles: 5: Global supply chain pressure · Brent crude oil · WTI crude oil — US · Henry Hub natural gas — US · Copper price

Supply-chain pressure and four commodity prices:

The New York Fed’s Global Supply Chain Pressure Index, which combines shipping and airfreight costs with purchasing managers’ survey measures of supply delays for seven economies.
Oil: Brent and West Texas Intermediate crude, two benchmark prices the EIA describes as reference prices for other crude streams.
Henry Hub natural gas, the US benchmark gas price.
The IMF’s global copper price.

The dashboard classes all five as leading indicators. Only the supply-chain index and copper carry badges; their rules are the dashboard’s own.

Consumer & Housing

Cockpit zone: Mirrors
Tiles: 6: UMich consumer sentiment · Building permits · Housing starts · New home sales · 30Y mortgage rate · Real retail sales — US

US households and housing:

Sentiment: the University of Michigan’s consumer sentiment index.
Housing: building permits, housing starts and new home sales from the Census Bureau and HUD, and the 30-year mortgage rate from Freddie Mac.
Spending: real retail sales, the Census Bureau’s advance retail and food services sales deflated by CPI.

The dashboard classes all but retail sales as leading indicators and real retail sales as coincident. None of the tiles carries a badge. The three housing measures are seasonally adjusted annual rates, the form in which the Census Bureau reports them.

Markets & Uncertainty

Cockpit zone: Side windows
Tiles: 5: S&P 500 · VIX · Broad US dollar · Policy uncertainty — US (daily) · Policy uncertainty — Global

What markets and the news say about risk appetite and uncertainty:

Market prices: the S&P 500 (licensed: the public site links to it), the VIX measure of expected equity volatility, and the Federal Reserve’s broad dollar index.
News-based uncertainty: Baker, Bloom and Davis’s economic policy uncertainty indexes for the US, daily, and the world, monthly.

The dashboard classes the S&P 500 as coincident, the VIX and the dollar as leading, and the two uncertainty indexes as overlays. Only the VIX carries a badge, with bands that are the dashboard’s own.

China & Leaders

Cockpit zone: Side windows
Tiles: 8: OECD CLI — China · OECD CLI — India · OECD CLI — Korea · OECD CLI — Brazil · CPI inflation — China · CPI inflation — India · PMI — China · PMI — India

Four large economies outside the G7, through the same OECD leading indicators used on the Cycle & Growth tab (China, India, Korea, Brazil), plus consumer-price inflation for China and India from the IMF’s CPI dataset, and licensed purchasing managers’ indexes for China and India as links. China’s leading indicator is also the growth input of the China regime tile.

The leading-indicator tiles carry the same badge as the other OECD indicators. Of the two inflation tiles, only India’s has a badge, read against the Reserve Bank of India’s own target band.

Bubbles / Froth

Cockpit zone: Speedometer
Tiles: 8: Equity / GDP (Buffett) — US · Stocks vs bonds — US · Stocks vs bonds — Euro area · Stocks vs bonds — UK · Real 10Y yield — US · IPO & SPAC issuance — US · Case-Shiller price/rent — US · Household debt service — US

The speedometer of the cockpit: how stretched valuations and household leverage are. The design brief’s rule for reading it is that the speedometer is not a clock: these tiles say how little margin of safety is left, not when a fall will come, and they are meant to be read against the windscreen (the Recession Watch and Conditions tabs) rather than alone.

Valuation: equity to GDP and the house price to rent ratio (licensed, link-only on the public site).
Stocks against bonds: the gap between an earnings yield and a real bond yield for the US, the euro area and the UK, with the real 10-year Treasury yield.
Speculation and leverage: IPO and SPAC issuance (links to public trackers) and the household debt service ratio.

The badge bands on this tab are the dashboard’s own, except that the euro-area and UK stocks-vs-bonds badges rank the gap against its own history.

Part III · The tiles

One section per tile, grouped by tab and in the order the tab shows them. Each starts with a fact box filled in from the dashboard’s own records: the source and its licence, how often the series is published and how long after the period it covers, its unit and indicator class, who computes it, and what the tile draws. Then:

What it is: the concept and why it matters for the cycle.
Computation: how the number is calculated, by its publisher or by the dashboard.
Reading it: the badge’s exact rules and what levels and changes mean.
Use in practice: lead times, typical signals and how it combines with other tiles.
Caveats: revisions, breaks, coverage and known false signals.

Recession Watch

Yield curve 10Y–3M

Source: Federal Reserve Bank of St. Louis via FRED, series T10Y3M
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percentage points (pp)
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Recession Watch tab · reference line at 0 · badge

What it is

The 10-year Treasury yield minus the 3-month Treasury yield. A negative value, an inverted curve, means the 10-year yield is below the 3-month yield.

Computation

FRED computes the spread from the Treasury Department’s constant-maturity yields; the dashboard shows it unchanged. With $y^{10y}_t$ and $y^{3m}_t$ the 10-year and 3-month constant-maturity yields on day $t$ , in percent:

$S_t = y^{10y}_t - y^{3m}_t .$

The FRED series starts in January 1982.

Reading it

Condition	Badge
$S_n < 0$	Inverted — recession lead
$0 \le S_n < 0.5$	Flat — late-cycle
$S_n \ge 0.5$	Positively sloped

The reference line is at zero. The 0.5-point boundary between “flat” and “positively sloped” is the dashboard’s own convention.

Use in practice

What the research found. Estrella and Trubin (2006), at the New York Fed, examined this spread as a recession predictor and reported that:

the monthly average of the 10-year minus 3-month spread turned negative before each recession from January 1968 to July 2006;
all six recessions since 1968 were preceded by at least three negative monthly averages in the twelve months before the recession began, and inversion on a monthly-average basis gave no false signals over that period;
inversions seen only in daily or intraday data often proved false signals, so the spread is best read on at least a monthly average.

The New York Fed publishes a monthly probability of recession twelve months ahead computed from this spread’s monthly average; its update of 6 September 2026 put the probability for August 2027 at 13.9%.

What the FRED series shows. Monthly averages of this tile’s series around each NBER peak since it began:

NBER peak	First negative month	Lead (months)	Negative months in the 12 before the peak	First month back above zero
Jul 1990	Jun 1989	13	4	Jan 1990 (6 months before the peak)
Mar 2001	Jul 2000	8	7	Feb 2001 (1 month before)
Dec 2007	Aug 2006	16	6	Jun 2007 (6 months before)
Feb 2020	May 2019	9	5	Mar 2020 (1 month after)

Every negative monthly average in this series up to 2020 came within 16 months before an NBER peak. Since then the monthly average was negative from November 2022 to November 2024 (25 months) and again in March–April and June–August 2025, and no NBER peak has been dated as of October 2026. In daily data the spread was below zero continuously from 25 October 2022 to 12 December 2024.

Caveats

Not the researchers’ exact measure. Estrella and Trubin obtained their best results with the secondary-market 3-month bill rate on a bond-equivalent basis; this tile, like FRED’s series, uses the constant-maturity 3-month yield.
US only. Other economies’ yields are on the Credit & Rates tab.

Yield curve 10Y–2Y

Source: Federal Reserve Bank of St. Louis via FRED, series T10Y2Y
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percentage points (pp)
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Recession Watch tab · reference line at 0 · badge

What it is

The 10-year Treasury yield minus the 2-year Treasury yield, the other common measure of the curve’s slope.

Computation

FRED computes the spread from the Treasury Department’s constant-maturity yields; the dashboard shows it unchanged:

$S_t = y^{10y}_t - y^{2y}_t .$

The FRED series starts in June 1976.

Reading it

The same badge rules as the 10Y–3M curve:

Condition	Badge
$S_n < 0$	Inverted — recession lead
$0 \le S_n < 0.5$	Flat — late-cycle
$S_n \ge 0.5$	Positively sloped

Estrella and Trubin (2006) note that this spread tends to turn negative earlier and more often than the 10-year minus 3-month spread, which is usually larger.

Use in practice

Monthly averages of this tile’s series around each NBER peak since it began:

NBER peak	First negative month	Lead (months)	Negative months in the 12 before the peak	First month back above zero
Jan 1980	Sep 1978	16	12	May 1980 (4 months after the peak)
Jul 1981	Sep 1980	10	10	Nov 1981 (4 months after)
Jul 1990	Jan 1989	18	3	Apr 1990 (3 months before)
Mar 2001	Feb 2000	13	10	Jan 2001 (2 months before)
Dec 2007	Feb 2006	22	5	Jun 2007 (6 months before)
Feb 2020	none	—	0	—

Before the February 2020 peak the monthly average did not turn negative; the daily spread was below zero on three days, 27–29 August 2019. The 10Y–3M monthly average, by contrast, was negative for five months before that peak.

Caveats

Inversions without a following recession. The monthly average was also negative in June 1998 and from July 2022 to August 2024 (26 months); no NBER peak followed within 24 months of either. It was negative from February to June 1982 too, inside the 1981–82 recession.
US only.

Sahm rule (real-time)

Source: Claudia Sahm via FRED, series SAHMREALTIME
Licence: Public domain; data shown on the public site
Frequency: monthly · the September 2026 value was available on 7 October 2026
Unit: percentage points (pp)
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Recession Watch tab · reference line at 0.5 · badge

What it is

A recession indicator from the unemployment rate alone. Claudia Sahm proposed it in a Hamilton Project policy proposal of 16 May 2019 as the trigger for stimulus payments to individuals, to be paid when the three-month average national unemployment rate rose by at least 0.50 percentage points above its low of the previous 12 months.

Computation

Published by Claudia Sahm on FRED. With $u_t$ the US unemployment rate (U-3, %) and $\bar u_t$ its three-month average,

$\bar u_t = \tfrac{1}{3}\left(u_t + u_{t-1} + u_{t-2}\right), \qquad \text{Sahm}_t = \bar u_t - \min\left(\bar u_{t-1}, \dots, \bar u_{t-12}\right),$

in percentage points. FRED’s notes describe the series as real-time: each month’s value uses the unemployment rates available in that month. They add that the BLS revises the unemployment rate each year in January, when it publishes the December rate, and that otherwise the rate isn’t revised. Because the low is taken over the previous twelve months, the value can be negative.

Reading it

Condition	Badge
$\text{Sahm}_n \ge 0.5$	Triggered — recession
$0.3 \le \text{Sahm}_n < 0.5$	Rising — watch
$\text{Sahm}_n < 0.3$	Below trigger

The reference line marks the rule’s 0.5-point trigger. The 0.3 “watch” level is the dashboard’s own.

Use in practice

The first month the series crossed 0.5 around each NBER peak (FRED data from December 1959):

NBER peak	First month at or above 0.5	Relative to the peak
Apr 1960	Dec 1959 (the series’ first month, already 0.77)	4 months before
Dec 1969	Oct 1969	2 months before
Nov 1973	Mar 1974	4 months after
Jan 1980	Apr 1980	3 months after
Jul 1981	Nov 1981	4 months after
Jul 1990	Nov 1990	4 months after
Mar 2001	Jun 2001	3 months after
Dec 2007	Apr 2008	4 months after
Feb 2020	Apr 2020	2 months after

It was at or above 0.5 at two other times: November 1976 (one month) and July to September 2024 (0.53, 0.57 and 0.50). No NBER peak followed either, as of October 2026.

Caveats

It confirms rather than predicts. From 1969 to 2020 it crossed the trigger between two months before and four months after the peak.
Monthly and US only.

Recession probability

Source: Marcelle Chauvet and Jeremy Piger via FRED, series RECPROUSM156N
Licence: Public domain; data shown on the public site
Frequency: monthly · the August 2026 value was released on 30 September 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Recession Watch tab · no reference line · badge

What it is

The estimated probability, in percent, that the US economy was in recession in a given month, from a statistical model of four monthly measures of activity.

Computation

Published by Jeremy Piger on FRED; the dashboard shows it unchanged. Piger’s FAQ and FRED’s notes describe it as a smoothed probability from a dynamic-factor Markov-switching model applied to four monthly coincident variables: non-farm payroll employment, the index of industrial production, real personal income excluding transfer payments, and real manufacturing and trade sales. The model was developed in Chauvet (1998); Piger estimates it with Bayesian methods. The value for month $t$ is

$p_t = 100 \cdot P\left(\text{recession in month } t \mid \text{data available at the latest release}\right).$

The FAQ says the probabilities are updated each time a full month of data for all four series is available, generally on or just before the first of each month, and that since the release of 23 December 2020 the model allows its parameters to change for March to July 2020. Releases are posted on Piger’s page.

Reading it

Condition	Badge
$p_n \ge 50$	High probability
$20 \le p_n < 50$	Elevated
$p_n < 20$	Low

The 20% and 50% boundaries are the dashboard’s own. Piger’s FAQ gives a different rule: three consecutive months above 80% has historically been a reliable signal that a recession has started, and three consecutive months below 20% that an expansion has started.

Use in practice

From the current FRED series (back to June 1967) around each NBER peak:

NBER peak	First month at or above 50%	First month at or above 80%	Highest value in the recession
Dec 1969	Jan 1970 (1 month after)	Oct 1970 (10 after)	80.5%
Nov 1973	Jul 1974 (8 after)	Aug 1974 (9 after)	100%
Jan 1980	Mar 1980 (2 after)	Mar 1980 (2 after)	99.6%
Jul 1981	Sep 1981 (2 after)	Sep 1981 (2 after)	94.3%
Jul 1990	Oct 1990 (3 after)	never	63.9%
Mar 2001	never	never	30.4%
Dec 2007	Mar 2008 (3 after)	May 2008 (5 after)	100%
Feb 2020	Mar 2020 (1 after)	Mar 2020 (1 after)	100%

In the 618 months outside recessions, the probability was below 2% in 93% of them and below 5% in 98%; its highest value outside a recession was 40.9%, in November 1969, the month before the December 1969 peak.

Caveats

Revisions. Piger’s FAQ explains that smoothed probabilities are revised as new data arrive and earlier data are revised, which smooths away some of the spikes seen in real time. The table uses today’s smoothed series, not the values published at the time; in it, the 1990 and 2001 recessions never reach 80%.
Coincident. It describes the current month rather than forecasting.

Conditions & Liquidity

Financial conditions (NFCI)

Source: Federal Reserve Bank of Chicago via FRED, series NFCI
Licence: Free to reuse with a citation; data shown on the public site
Frequency: weekly · week ending Friday; the week ending 2 October 2026 was available on 7 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · reference line at 0 · badge

What it is

The Chicago Fed’s National Financial Conditions Index, a weekly update on US financial conditions in money markets, debt and equity markets, and the traditional and “shadow” banking systems. The Chicago Fed also publishes three subindexes: risk (volatility and funding risk in the financial sector), credit (measures of credit conditions) and leverage (debt and equity measures).

Computation

Published by the Chicago Fed; the dashboard shows it unchanged. The index is a weighted average of 105 measures of financial activity, each expressed relative to its sample average and scaled by its sample standard deviation:

$\tilde x_{it} = \frac{x_{it} - \bar x_i}{s_i}, \qquad \text{NFCI}_t \propto \sum_{i=1}^{105} w_i\, \tilde x_{it},$

constructed to have an average of 0 and a standard deviation of 1 over a sample extending back to 1971. How the weights $w_i$ are chosen is described in Brave and Butters (2011), linked from the Chicago Fed’s NFCI page.

Reading it

Condition	Badge
$\text{NFCI}_n > 0.5$	Tight — headwind
$0 < \text{NFCI}_n \le 0.5$	Slightly tight
$\text{NFCI}_n \le 0$	Loose — tailwind

The Chicago Fed describes positive values as historically associated with tighter-than-average financial conditions and negative values with looser-than-average ones. The reference line is at zero; the 0.5 boundary is the dashboard’s own. The risk, credit and leverage subindexes rise with higher risk, tighter credit and falling leverage.

Use in practice

From the FRED series since January 1971:

the index was above zero in 29% of weeks, most recently in the week of 1 May 2020;
its highest reading was 5.22, in the week of 19 July 1974;
from October 2021 to October 2026, a period that includes the Fed’s rate increases of 2022–23, its highest weekly reading was −0.10 (7 October 2022).

Caveats

Relative to 1971 onward. Zero is the average of a history that includes the episodes above, so “looser than average” isn’t the same as “loose”.
US only.

Adjusted NFCI

Source: Federal Reserve Bank of Chicago via FRED, series ANFCI
Licence: Free to reuse with a citation; data shown on the public site
Frequency: weekly · week ending Friday; the week ending 2 October 2026 was available on 7 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · reference line at 0 · badge

What it is

The adjusted NFCI measures financial conditions relative to the state of the economy. The Chicago Fed describes positive values as historically associated with conditions tighter than prevailing macroeconomic conditions would typically suggest, and negative values with the opposite.

Computation

Published by the Chicago Fed. It is built like the NFCI from the same indicators, after removing the variation in each indicator attributable to economic activity and inflation. Like the NFCI it has an average of 0 and a standard deviation of 1 over a sample extending back to 1971. The method is set out in the Chicago Fed’s article “Introducing the Chicago Fed’s New Adjusted National Financial Conditions Index” (2017), linked from its NFCI page.

Reading it

The same badge rules as the NFCI:

Condition	Badge
$\text{ANFCI}_n > 0.5$	Tight — headwind
$0 < \text{ANFCI}_n \le 0.5$	Slightly tight
$\text{ANFCI}_n \le 0$	Loose — tailwind

Positive values mean conditions tighter than the economy would typically produce; negative values, looser.

Use in practice

From October 2021 to October 2026 the adjusted index stayed below zero; its highest weekly reading was −0.01 (1 July 2022).

Caveats

Model-based. The part of conditions attributed to economic activity and inflation is an estimate.
US only.

St. Louis Fed financial stress

Source: Federal Reserve Bank of St. Louis via FRED, series STLFSI4
Licence: Free to reuse with a citation; data shown on the public site
Frequency: weekly · week ending Friday; the week ending 25 September 2026 was available by 7 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · reference line at 0 · no badge

What it is

The St. Louis Fed Financial Stress Index, version 4: a weekly measure of the degree of financial stress in markets.

Computation

Published by the St. Louis Fed; the dashboard shows it unchanged. According to FRED’s notes it is constructed from 18 weekly data series (seven interest rates, six yield spreads and five other indicators), each capturing some aspect of financial stress, on the reasoning that the series move together as stress changes. The index’s average since it begins in late 1993 is designed to be zero. Version 4 is the third revision of the original index; it uses the forward-looking 90-day SOFR rate in two of its spreads, where version 3 used the backward-looking 90-day average. FRED’s series notes link the papers describing each version.

Reading it

FRED’s notes: zero represents normal financial market conditions, values below zero below-average stress and values above zero above-average stress. The tile’s reference line is at zero; there is no badge.

Use in practice

The highest weekly reading in each of the six highest years since 1993:

Week ending	Value
10 Oct 2008	9.66
20 Mar 2020	5.62
2 Jan 2009	4.97
30 Nov 2007	2.09
21 Sep 2001	2.02
9 Oct 1998	1.92

From October 2021 to October 2026 the highest reading was 1.14, in the week ending 17 March 2023.

Caveats

Earlier versions differ. Earlier versions are separate, discontinued series on FRED; the tile uses version 4.
US only.

Fed balance sheet

Source: Board of Governors of the Federal Reserve System (US) via FRED, series WALCL
Licence: Public domain; data shown on the public site
Frequency: weekly · Wednesday level; the H.4.1 release of Thursday 1 October 2026 covers Wednesday 30 September
Unit: US$ millions
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · no reference line · no badge

What it is

The total assets of the Federal Reserve System (less eliminations from consolidation), from its weekly H.4.1 release.

Computation

Published by the Federal Reserve Board via FRED (WALCL): the Wednesday level, in US$ millions. The dashboard shows it unchanged.

Reading it

There is no badge or reference line. Milestones in the FRED series and the FOMC’s decisions:

In February 2020 total assets were $4,159–4,183 billion.
They peaked at $8,965 billion on 13 April 2022.
Under plans the FOMC published on 4 May 2022, from 1 June 2022 principal payments from its securities were reinvested only to the extent that they exceeded monthly caps, so its holdings declined.
On 29 October 2025 the FOMC decided to conclude the reduction of its securities holdings on 1 December 2025. Total assets were $6,557 billion on 17 December 2025.
On 10 December 2025 it judged that reserve balances had declined to ample levels and directed purchases of Treasury bills (and, if needed, other Treasury securities with up to three years’ remaining maturity) to maintain an ample level of reserves. Total assets were $6,743 billion on 30 September 2026.

Use in practice

In the H.4.1 accounting, the Fed’s assets supply reserve funds; how much ends up as bank reserves depends also on the absorbing factors, which is why the dashboard also shows net liquidity.

Caveats

Total assets combine different items, securities held outright and loans among them; the H.4.1 tables give the split.
A weekly snapshot: the Wednesday level.

Treasury General Account

Source: Board of Governors of the Federal Reserve System (US) via FRED, series WTREGEN
Licence: Public domain; data shown on the public site
Frequency: weekly · week ending Wednesday, in the H.4.1 release published the next day
Unit: US$ millions
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · no reference line · no badge

What it is

The US Treasury’s general account at the Federal Reserve. In the H.4.1 release it is one of the deposits with Federal Reserve Banks other than reserve balances, which count among the factors absorbing reserve funds: other things equal, a higher balance means lower bank reserves.

Computation

Published by the Federal Reserve Board via FRED (WTREGEN): the average of daily figures for the week ending Wednesday, in US$ millions. The dashboard shows it unchanged. (Net liquidity uses the Wednesday level instead, WDTGAL.)

Reading it

There is no badge or reference line. From mid-2019 to October 2026 the week average was lowest at $45 billion (week ending 7 June 2023) and highest at $1,817 billion (week ending 29 July 2020). In 2025 it fell to $306 billion (week ending 9 April) and $323 billion (23 July), then rose to $941 billion by the week ending 5 November.

Use in practice

Because the account absorbs reserve funds, a rise in the balance drains bank reserves unless other factors offset it. Compare its swings with the Fed’s balance sheet and the reverse repo, or see the combined effect in net liquidity.

Caveats

A week average, unlike the Wednesday level of the Fed’s total assets.
It reflects the Treasury’s cash management, not the FOMC’s decisions.

Overnight reverse repo

Source: Federal Reserve Bank of New York via FRED, series RRPONTSYD
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the operation for 7 October 2026 was on FRED the same day
Unit: US$ billions
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Conditions & Liquidity tab · no reference line · no badge

What it is

The amount of the Fed’s overnight reverse repurchase (ON RRP) operations. In a reverse repo the New York Fed’s Open Market Desk sells a security to an eligible counterparty with an agreement to buy it back at a specified price at a specified time. Eligible counterparties are primary dealers and approved RRP counterparties, which include 2a-7 money market funds, banks and government-sponsored enterprises; the collateral is US Treasuries. According to the New York Fed, the transaction shifts some of the Fed’s liabilities from bank reserves to reverse repos while it is outstanding.

Computation

Published by the Federal Reserve Bank of New York via FRED (RRPONTSYD): the aggregated daily amount of overnight reverse repo transactions, in US$ billions. The dashboard shows it unchanged.

Reading it

The FOMC sets the offering rate; in its implementation note of 10 December 2025 it was 3.5%, the bottom of the 3.5–3.75% target range for the federal funds rate. There is no badge or reference line. In the FRED series:

usage peaked at $2,554 billion on 30 December 2022;
from 28 December 2022 to 25 December 2024 it fell by $2,112 billion, while the Fed’s total assets fell by $1,665 billion and bank reserve balances (week average) rose by $229 billion;
since 5 August 2025 it has been below $100 billion every day except 31 December 2025 ($106 billion); on 7 October 2026 it was $2.3 billion.

Use in practice

In the H.4.1 accounting a fall in reverse repos, other things equal, adds to bank reserves; with usage near zero there is little left to fall.

Caveats

Billions, not millions. Unlike the two balance-sheet tiles this series is in US$ billions; net liquidity converts it.

Net liquidity (Fed − TGA − RRP)

Source: Board of Governors of the Federal Reserve System (US); Federal Reserve Bank of New York via FRED, series WALCL, WDTGAL, RRPONTSYD
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the Fed’s balance sheet and the Treasury account carried forward between weekly releases
Unit: US$ millions
Indicator class: derived
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Conditions & Liquidity tab · no reference line · no badge

What it is

The Fed’s total assets minus the two absorbing factors shown on this tab, the Treasury General Account and the overnight reverse repo. The dashboard computes it from three Fed series.

Computation

Computed by the dashboard (Appendix A.4):

$\mathrm{NL}_\tau = W_\tau - T_\tau - 1000\, Q_\tau \quad (\text{US\$ millions}),$

with $W$ the Fed’s total assets and $T$ the Treasury General Account, both Wednesday levels in $ millions (WALCL, WDTGAL), and $Q$ the daily reverse repo in $ billions (RRPONTSYD). Each input is carried forward to every date any of them is published, so the series moves daily with the reverse repo.

Reading it

There is no badge or reference line. In the H.4.1 accounting, reserve balances are the factors supplying reserve funds minus all the absorbing factors; net liquidity starts from total assets and subtracts only two absorbing factors, so the others, currency among them, remain in it. A rise therefore means more of the Fed’s balance sheet is left as bank reserves, currency and its other liabilities; a fall, less.

Use in practice

From 28 December 2022 to 25 December 2024 the Fed’s total assets fell by $1,665 billion, but net liquidity rose by about $122 billion, because the reverse repo fell by $2,112 billion while the Treasury account (Wednesday level) rose by $325 billion.

Caveats

Partial. It leaves out the other absorbing factors, currency among them.
Timing. The two weekly inputs are Wednesday levels; the reverse repo is daily.

M2 growth (year on year)

Source: Board of Governors of the Federal Reserve System (US) via FRED, series M2SL
Licence: Public domain; data shown on the public site
Frequency: monthly · the H.6 release of 22 September 2026 carried August 2026
Unit: percent
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Conditions & Liquidity tab · reference line at 0 · no badge

What it is

The growth of the US money stock M2 over the previous twelve months. The Federal Reserve Board’s H.6 release defines M2 as M1 (currency, demand deposits and other liquid deposits) plus small-denomination time deposits (under $100,000) and balances in retail money market funds, less individual retirement account (IRA) and Keogh balances at depository institutions and money market funds.

Computation

The dashboard computes the annual growth rate from the seasonally adjusted monthly level (M2SL, US$ billions) published by the Federal Reserve Board (Appendix A.1):

$g^{M2}_t = 100 \left( \frac{M2_t}{M2_{t-12}} - 1 \right) \quad (\%).$

Reading it

The reference line is at zero: below it, the money stock is smaller than a year earlier. There is no badge. In the FRED series (from January 1959, so growth rates from January 1960):

the fastest growth was 26.8%, in February 2021;
the only months with negative growth were December 2022 to February 2024;
growth was 5.7% in August 2026.

Use in practice

Read the growth rate against its own history with the z-score or percentile mode; the zero line separates a growing money stock from a shrinking one.

Caveats

Definition change. FRED’s notes say savings deposits were summed separately into M2 before May 2020, and point to the H.6 announcements of 17 December 2020 on the regulatory change that created the “other liquid deposits” component of M1.
Seasonally adjusted, monthly.

Credit & Rates

C&I lending standards (SLOOS)

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DRTSCILM
Licence: Public domain; data shown on the public site
Frequency: quarterly · the July 2026 survey, covering the second quarter, was FRED’s latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

From the Federal Reserve Board’s Senior Loan Officer Opinion Survey on Bank Lending Practices (SLOOS): the net percentage of domestic banks that tightened their standards for commercial and industrial (C&I) loans to large and middle-market firms. The survey release defines large and middle-market firms as those with annual sales of $50 million or more.

Computation

Published by the Federal Reserve Board via FRED (DRTSCILM); the dashboard shows it unchanged. The SLOOS release defines the net percentage as the fraction of banks reporting tightened standards (“tightened considerably” or “tightened somewhat”) minus the fraction reporting eased standards:

$\text{Net}_q = 100 \left( \frac{n^{\text{tightened}}_q}{n_q} - \frac{n^{\text{eased}}_q}{n_q} \right) \quad (\%).$

The Fed surveys up to eighty large domestic banks and twenty-four US branches and agencies of foreign banks, generally quarterly, timed so that results are available for the FOMC meetings in January/February, April/May, August and October/November, and occasionally runs one or two extra surveys a year.

Reading it

Positive values mean more banks tightened than eased; negative values, more eased. The tile has no badge or reference line. In the FRED series (from the second quarter of 1990):

the highest reading was 83.6% (October 2008 survey) and the lowest −32.4% (July 2021 survey);
readings were positive in 49% of surveys;
the latest, the July 2026 survey (covering the second quarter), was 0%.

Use in practice

The survey readings nearest each NBER peak since 1990:

NBER peak	Latest survey at or before the peak	Highest reading in the year after
Jul 1990	46.7% (July 1990)	54.2%
Mar 2001	59.6% (January 2001)	50.9%
Dec 2007	19.2% (October 2007)	83.6%
Feb 2020	0% (January 2020)	71.2%

FRED dates each survey by the quarter in which it was taken (for example 1 July 2026 for the July 2026 survey).

Caveats

A survey of standards, not of loan volumes or prices; the release also covers loan demand and other loan categories, which this tile doesn’t show.
Quarterly, so it changes four times a year (occasionally more).

HY credit spread (OAS)

Source: ICE Data Indices, LLC via FRED, series BAMLH0A0HYM2
Licence: Republication needs the provider’s permission; link-only on the public site (data in the local app)
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · badge

What it is

The option-adjusted spread (OAS) of the ICE BofA US High Yield Index: the extra yield that US dollar, below-investment-grade corporate bonds (rated BB or below) pay over US Treasuries.

Computation

Published by ICE Data Indices via FRED; the dashboard shows it unchanged. FRED’s notes: an OAS index is built from each constituent bond’s OAS, weighted by market capitalisation, and the spread is measured between that index and a spot Treasury curve. To enter the index a bond must be rated below investment grade on an average of Moody’s, S&P and Fitch, have more than a year to maturity, a fixed coupon schedule and at least $100 million outstanding, among other rules.

Reading it

Condition	Badge
$\text{OAS}_n < 3$	Tight — risk-on
$3 \le \text{OAS}_n < 5$	Normal range
$5 \le \text{OAS}_n < 8$	Elevated stress
$\text{OAS}_n \ge 8$	Stressed — risk-off

These thresholds are the dashboard’s own. The tile has no reference line.

Use in practice

Read it with the investment-grade spread and the gap between them, and with the conditions indices on the Conditions & Liquidity tab.

Caveats

Licensed. ICE’s terms don’t allow republication, so the public site shows a link; the local app shows the data. This guide quotes no values.
Three years of history. FRED’s notes say that from April 2026 the series includes only three years of observations; longer history is available from ICE.

IG credit spread (OAS)

Source: ICE Data Indices, LLC via FRED, series BAMLC0A0CM
Licence: Republication needs the provider’s permission; link-only on the public site (data in the local app)
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The option-adjusted spread of the ICE BofA US Corporate Index, which FRED’s notes describe as an index of bonds considered investment grade (rated BBB or better).

Computation

Published by ICE Data Indices via FRED; the dashboard shows it unchanged. The spread is built the same way as the high-yield spread.

Reading it

There is no badge or reference line.

Use in practice

It covers the higher-rated part of the US corporate bond market; the high-yield spread covers the lower-rated part, and the HY − IG tile shows the gap between the two.

Caveats

Licensed (link-only on the public site) and three years of history on FRED since April 2026, as for the high-yield spread.

HY − IG (decompression)

Source: ICE Data Indices, LLC via FRED, series BAMLH0A0HYM2, BAMLC0A0CM
Licence: Republication needs the provider’s permission; link-only on the public site (data in the local app)
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percentage points (pp)
Indicator class: derived
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Credit & Rates tab · no reference line · badge

What it is

The high-yield spread minus the investment-grade spread: how much more the market charges below-investment-grade borrowers than investment-grade ones.

Computation

Computed by the dashboard from the two ICE BofA spreads (Appendix A.5):

$D_t = \text{OAS}^{HY}_t - \text{OAS}^{IG}_t \quad (\text{pp}).$

Reading it

Condition	Badge
$D_n < 2$	Compressed — complacent
$2 \le D_n < 3.5$	Normal range
$3.5 \le D_n < 5$	Decompressing
$D_n \ge 5$	Wide — risk-off

These thresholds are the dashboard’s own. The tile has no reference line.

Use in practice

A narrowing gap means high-yield spreads are falling relative to investment-grade spreads; a widening gap, the reverse.

Caveats

Licensed inputs: link-only on the public site; three years of history on FRED.

10Y Treasury volatility (realized)

Source: Board of Governors of the Federal Reserve System (US) via FRED; volatility computed by this site, series DGS10
Licence: Public domain; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: basis points (bp)
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

How much the 10-year Treasury yield has moved day to day over the last month, expressed as an annual rate in basis points. The dashboard uses it in place of the licensed ICE BofA MOVE index.

Computation

Computed by the dashboard from the 10-year constant-maturity yield (DGS10) (Appendix A.7):

$\sigma_t = \sqrt{252}\;\operatorname{sd}\left(\Delta y_{t-20}, \dots, \Delta y_t\right), \qquad \Delta y_t = 100\,(y_t - y_{t-1}) \ \text{bp}.$

Reading it

There is no badge or reference line. Computed over the whole DGS10 history (from January 1962):

the median was 77.8 bp; from October 2021 to October 2026 it was 91.5 bp;
the highest values were 436 bp (12 March 1980) and 408 bp (12 January 1981); since October 2021 the highest was 172 bp (12 July 2022);
on 6 October 2026 it was 85.5 bp.

Use in practice

A rising value means larger daily moves in the 10-year yield. Compare it with its own history using the z-score or percentile mode.

Caveats

Backward-looking: it measures moves that have already happened over 21 trading days.

3M Treasury yield

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DGS3MO
Licence: Public domain; data shown on the public site
Frequency: daily · the value for 5 October 2026 was available by 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The yield on US Treasury securities at a constant 3-month maturity.

Computation

Published by the Federal Reserve Board in its H.15 release, via FRED (DGS3MO); the dashboard shows it unchanged. The H.15 notes explain that constant-maturity yields are interpolated by the US Treasury from the daily yield curve for non-inflation-indexed Treasury securities, based on closing market bid yields on actively traded securities.

Reading it

There is no badge or reference line. The FRED series starts on 1 September 1981, when it stood at 17.01%, its highest; it was 0.00% on 10 December 2008, and 4.22% on 5 October 2026.

Use in practice

From December 2008 to October 2026 it tracked the effective federal funds rate closely: the daily difference averaged 0.02 percentage points (standard deviation 0.16) and the correlation was 0.997.

Caveats

Constant maturity: an interpolated point on the curve, not a specific bill.

2Y Treasury yield

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DGS2
Licence: Public domain; data shown on the public site
Frequency: daily · the value for 5 October 2026 was available by 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The yield on US Treasury securities at a constant 2-year maturity.

Computation

Published by the Federal Reserve Board in its H.15 release, via FRED (DGS2); interpolated by the Treasury as described for the 3-month yield. The dashboard shows it unchanged.

Reading it

There is no badge or reference line. The FRED series starts in June 1976; its highest value was 16.95% (8 September 1981), its lowest 0.09% (5 February 2021), and it was 4.84% on 5 October 2026.

Use in practice

It is one leg of the 10Y–2Y curve on the Recession Watch tab; comparing it with the 3-month yield shows how far yields two years out differ from those three months out.

Caveats

Constant maturity, as for the 3-month yield.

10Y Treasury yield

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DGS10
Licence: Public domain; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The yield on US Treasury securities at a constant 10-year maturity, the long end of the two yield curves on the Recession Watch tab.

Computation

Published by the Federal Reserve Board in its H.15 release, via FRED (DGS10); interpolated by the Treasury as described for the 3-month yield. The dashboard shows it unchanged.

Reading it

There is no badge or reference line. The FRED series starts in January 1962; its highest value was 15.84% (30 September 1981), its lowest 0.52% (4 August 2020), and it was 5.27% on 6 October 2026.

Use in practice

It is the input to the dashboard’s realized volatility tile and to the bond side of the US stocks-vs-bonds tile (Appendix A.8).

Caveats

Nominal: for the stocks-vs-bonds tile the dashboard subtracts the Cleveland Fed’s 10-year expected inflation to get a real yield.

Fed funds rate (effective)

Source: Federal Reserve Bank of New York via FRED, series EFFR
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The effective federal funds rate, published by the Federal Reserve Bank of New York. FRED’s notes describe the federal funds market as domestic unsecured borrowing in US dollars by depository institutions from other depository institutions and certain other entities, mainly government-sponsored enterprises.

Computation

Calculated by the New York Fed as a volume-weighted median of overnight federal funds transactions reported in the FR 2420 Report of Selected Money Market Rates; the dashboard shows it unchanged.

Reading it

There is no badge or reference line. Since 16 December 2008, when FRED’s series for the lower and upper limits of the FOMC’s federal funds target range begin, the rate has been inside the range on all but four of 4,474 days (16 and 31 December 2015, 14 December 2016 and 17 September 2019). On 6 October 2026 it was 3.88%, in a target range of 3.75–4.00%.

Use in practice

Because it has stayed inside the FOMC’s target range, its level shows where the FOMC has set policy.

Caveats

The FRED series starts in July 2000; FRED’s notes point to daily federal funds rate data for 1928–1954 elsewhere.

ECB deposit rate

Source: European Central Bank (ECB Data Portal), series ECB/FM/B.U2.EUR.4F.KR.DFR.LEV
Licence: ECB reuse terms; data shown on the public site
Frequency: event · changes on the dates the ECB sets; the latest change took effect on 16 September 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The ECB’s deposit facility rate: the rate on the deposit facility, which banks may use to make overnight deposits with the Eurosystem. The ECB’s key-rates page says the Governing Council decided in March 2024 to continue to steer the monetary policy stance through this rate.

Computation

From the ECB Data Portal (dataset FM, “Deposit facility - date of changes (raw data) - Level”); the dashboard shows it unchanged as a step series. The registry’s frequency for this tile is event, so it never shows the amber freshness dot.

Reading it

There is no badge or reference line. In the ECB’s data since 1999:

the rate was negative from 11 June 2014 (−0.10%) until 27 July 2022, when it rose to 0%; it was −0.50% from 18 September 2019;
it rose to 4.00% on 20 September 2023 and held there until 12 June 2024;
it was cut to 2.00% by 11 June 2025, and raised to 2.25% on 17 June 2026 and 2.50% on 16 September 2026.

Use in practice

Compare it with the 3-month AAA yield, which markets set from day to day.

Caveats

A step series: the value changes only on decision dates.

10Y yield — Euro area

Source: European Central Bank (ECB Data Portal), series M.U2.L.L40.CI.0000.EUR.N.Z
Licence: ECB reuse terms; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The ECB’s “long-term interest rate for convergence purposes” for the euro area: a 10-year government bond yield for the euro area in its changing composition, in euros.

Computation

From the ECB Data Portal (dataset IRS); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From December 1990 its highest monthly value was 10.16% (August 1992) and its lowest −0.22% (December 2020); it was 3.60% in August 2026.

Use in practice

Compare it with the German, French and Italian yields below.

Caveats

Changing composition: the euro area’s membership changes over the series’ history.

10Y yield — Germany

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01DEM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

Germany’s 10-year government bond yield, from the OECD’s long-term interest rate series (“Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark)”).

Computation

Published by the OECD via FRED (IRLTLT01DEM156N); monthly, in percent. The dashboard shows it unchanged.

Reading it

There is no badge or reference line. From May 1956 its highest value was 10.80% (July 1974) and its lowest −0.65% (August 2019); it was below zero in 38 months between June 2016 and January 2022, and was 3.18% in August 2026.

Use in practice

It is the base for the spreads of the French and Italian yields discussed below.

Caveats

Monthly averages via FRED rather than daily market yields.

10Y yield — France

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01FRM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

France’s 10-year government bond yield, from the same OECD series as Germany’s.

Computation

Published by the OECD via FRED (IRLTLT01FRM156N); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. Its spread over Germany’s yield, computed from the two monthly series since January 1999, was widest at 1.54 percentage points (November 2011) and narrowest at 0.02 (January 2005); in August 2026 it was 0.82, with the French yield at 4.00%.

Use in practice

The dashboard shows the levels; the spread over Germany’s yield can be read by comparing the two tiles.

Caveats

Monthly averages via FRED.

10Y yield — Italy

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01ITM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

Italy’s 10-year government bond yield, from the same OECD series.

Computation

Published by the OECD via FRED (IRLTLT01ITM156N); the dashboard shows it unchanged. The FRED series starts in March 1991.

Reading it

There is no badge or reference line. Its spread over Germany’s yield, computed from the two monthly series since January 1999, was widest at 5.18 percentage points (November 2011) and narrowest at 0.13 (February 2005); in August 2026 it was 0.81, with the Italian yield at 3.99%.

Use in practice

As for France, the spread over Germany’s yield can be read by comparing the tiles.

Caveats

Monthly averages via FRED.

10Y yield — UK

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01GBM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The UK’s 10-year government bond yield, from the same OECD series.

Computation

Published by the OECD via FRED (IRLTLT01GBM156N); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1960 its highest value was 16.34% (October 1981) and its lowest 0.21% (July 2020); it was 4.99% in August 2026.

Use in practice

Compare it with the UK’s short rate, SONIA.

Caveats

Monthly averages via FRED.

10Y yield — Japan

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01JPM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

Japan’s 10-year government bond yield, from the same OECD series.

Computation

Published by the OECD via FRED (IRLTLT01JPM156N); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1989 its highest value was 8.03% (September 1990) and its lowest −0.28% (August 2019); it was below zero in 24 months between February 2016 and April 2020, and was 2.94% in August 2026.

Use in practice

Compare it with Japan’s policy rate, the overnight call rate.

Caveats

Monthly averages via FRED.

10Y yield — Canada

Source: Organization for Economic Co-operation and Development via FRED, series IRLTLT01CAM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

Canada’s 10-year government bond yield, from the same OECD series.

Computation

Published by the OECD via FRED (IRLTLT01CAM156N); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1955 its highest value was 17.00% (September 1981) and its lowest 0.52% (July 2020); it was 3.67% in August 2026.

Use in practice

Compare it with Canada’s 3-month rate and the US 10-year yield.

Caveats

Monthly averages via FRED.

3M govt yield — Euro area (AAA)

Source: European Central Bank (ECB Data Portal), series B.U2.EUR.4F.G_N_A.SV_C_YM.SR_3M
Licence: ECB reuse terms; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The 3-month spot rate of the ECB’s AAA yield curve: the yield on euro-area government bonds of all issuers rated triple A, at a 3-month maturity, estimated each day.

Computation

From the ECB Data Portal (dataset YC; Svensson model, continuous compounding); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From September 2004 its highest value was 4.33% (14 August 2008) and its lowest −0.93% (29 December 2016); it was 2.54% on 6 October 2026.

Use in practice

Compared with the deposit facility rate in force each day since September 2004, it was on average 0.12 percentage points higher (standard deviation 0.48); over the last 250 observations to October 2026 the average difference was 0.05.

Caveats

Model-based: a point on a fitted curve, not a traded bill.

SONIA — UK

Source: Bank of England (SONIA, Open Government Licence v3.0), series IUDSOIA
Licence: Open Government Licence v3.0; data shown on the public site
Frequency: daily · the value for 5 October 2026 was available by 7 October
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

SONIA, the Sterling Overnight Index Average. The Bank of England, which administers it, describes it as based on actual transactions and reflecting the average of the interest rates that banks pay to borrow sterling overnight from other financial institutions and other institutional investors. It is published every London business day.

Computation

From the Bank of England’s database (IUDSOIA, “Daily Sterling overnight index average (SONIA) rate”); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From 23 April 2018 to October 2026 it was within 0.12 percentage points of Official Bank Rate (IUDBEDR) on every day, on average 0.05 below it; it was 3.73% on 5 October 2026, with Bank Rate at 3.75%.

Use in practice

Because it stays close to Bank Rate, its level shows where the Bank of England has set policy.

Caveats

Overnight: despite the tile’s id it is an overnight rate, not a 3-month rate.

Overnight call rate — Japan

Source: Bank of Japan, series STRDCLUCON
Licence: Bank of Japan terms of use; data shown on the public site
Frequency: daily · the value for 5 October 2026 was available by 7 October
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

The uncollateralized overnight call rate (average), published by the Bank of Japan. It is the rate the Bank’s guideline for money market operations targets: on 18 September 2026 the Bank raised its short-term policy rate to “around 1.25%” (uncollateralized overnight call rate), from “around 1.0%”.

Computation

From the Bank of Japan’s statistics (FM01, STRDCLUCON); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. In the Bank of Japan’s data from January 1998 it was below zero on every day from 17 March 2016 to 19 March 2024; it was 0.98% before the September 2026 decision and 1.23% from 24 September 2026.

Use in practice

It shows the Bank of Japan’s policy stance directly; compare it with Japan’s 10-year yield.

Caveats

Overnight, not a 3-month rate, despite the tile’s id.

3M rate — Canada

Source: Organization for Economic Co-operation and Development via FRED, series IR3TIB01CAM156N
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Credit & Rates tab · no reference line · no badge

What it is

Canada’s 3-month interbank rate, from the OECD’s short-term interest rate series (“3-Month or 90-Day Rates and Yields: Interbank Rates”).

Computation

Published by the OECD via FRED (IR3TIB01CAM156N); monthly, in percent. The dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1956 its highest value was 22.01% (August 1981) and its lowest 0.06% (January 2021); it was 2.28% in August 2026.

Use in practice

Compare it with Canada’s 10-year yield.

Caveats

Monthly, via FRED, rather than daily.

Inflation

5Y inflation breakeven

Source: Federal Reserve Bank of St. Louis via FRED, series T5YIE
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2.3 · badge

What it is

The 5-year breakeven inflation rate: the yield on 5-year Treasury securities minus the yield on 5-year Treasury Inflation-Indexed Securities (TIPS). FRED’s notes describe its latest value as implying what market participants expect inflation to be over the next five years, on average.

Computation

Published by FRED from the constant-maturity yields; the dashboard shows it unchanged:

$b^{5y}_t = y^{5y}_t - r^{5y}_t ,$

with $y^{5y}$ the nominal yield (DGS5) and $r^{5y}$ the inflation-indexed yield (DFII5). This identity holds exactly on every day of the FRED data since 2003.

Reading it

Condition	Badge
$b_n < 1.8$	Low — disinflation
$1.8 \le b_n \le 2.7$	Anchored
$2.7 < b_n \le 3.2$	Elevated
$b_n > 3.2$	High — inflation risk

The badge bands and the 2.3% reference line are the dashboard’s own. From January 2003 to October 2026 the series was in the “Anchored” band on 60% of days, “Low” on 36%, “Elevated” on 4% and “High” on 1%. Its highest value was 3.59% (25 March 2022) and its lowest −2.24% (28 November 2008); on 6 October 2026 it was 2.37%.

Use in practice

Compare it with the 10-year breakeven and the 5y5y forward, which cover longer horizons, and with realized inflation on this tab.

Caveats

A difference of two yields. Anything that moves the nominal and inflation-indexed yields differently moves the breakeven.
CPI-based. TreasuryDirect says it adjusts TIPS principal using a version of the BLS Consumer Price Index, while the Fed’s 2% goal is defined on PCE inflation.

10Y inflation breakeven

Source: Federal Reserve Bank of St. Louis via FRED, series T10YIE
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2.3 · badge

What it is

The 10-year breakeven inflation rate: the 10-year Treasury yield minus the 10-year TIPS yield, which FRED describes as implying the average inflation market participants expect over the next ten years.

Computation

Published by FRED; the dashboard shows it unchanged:

$b^{10y}_t = y^{10y}_t - r^{10y}_t ,$

with $y^{10y}$ the nominal yield (DGS10) and $r^{10y}$ the inflation-indexed yield (DFII10); the identity holds exactly in the FRED data since 2003.

Reading it

Condition	Badge
$b_n < 1.8$	Low — disinflation
$1.8 \le b_n \le 2.7$	Anchored
$2.7 < b_n \le 3.2$	Elevated
$b_n > 3.2$	High — inflation risk

The badge bands and the 2.3% reference line are the dashboard’s own. Since January 2003 the series was “Anchored” on 79% of days, “Low” on 20% and “Elevated” on 1%; it never reached the “High” band. Its highest value was 3.02% (21 April 2022), its lowest 0.04% (20 November 2008), and it was 2.36% on 6 October 2026.

Use in practice

Over a longer horizon than the 5-year breakeven; the gap between them shows whether markets expect inflation to be higher in the near term or later.

Caveats

A difference of two yields, and CPI-based, as for the 5-year breakeven.

5y5y forward breakeven

Source: Federal Reserve Bank of St. Louis via FRED, series T5YIFR
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · the value for 6 October 2026 was available on 7 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2.3 · badge

What it is

The 5-year, 5-year forward inflation expectation rate: a measure of expected average inflation over the five-year period that begins five years from today (FRED’s notes).

Computation

FRED constructs it from the 10-year and 5-year breakevens:

$f_t = 100\left[\left(\frac{(1 + b^{10y}_t/100)^{10}}{(1 + b^{5y}_t/100)^{5}}\right)^{1/5} - 1\right],$

where FRED’s formula writes each breakeven as the nominal minus the inflation-adjusted Treasury yield. Applied to the two FRED breakevens, the formula reproduces the published series to within 0.005 percentage points.

Reading it

Condition	Badge
$f_n < 1.8$	Low — disinflation
$1.8 \le f_n \le 2.7$	Anchored
$2.7 < f_n \le 3.2$	Elevated
$f_n > 3.2$	High — inflation risk

The badge bands and the 2.3% reference line are the dashboard’s own. Since January 2003 it was “Anchored” on 85% of days, “Low” on 9% and “Elevated” on 6%. Its highest and lowest values both came in late 2008 (3.05% on 12 November, 0.43% on 29 December); from October 2021 to October 2026 its highest value was 2.67% (21 April 2022), and it was 2.35% on 6 October 2026.

Use in practice

It leaves out the next five years, so it reads expectations further ahead than either breakeven alone.

Caveats

Derived from the two breakevens, so it inherits their caveats.

1Y expected inflation (UMich)

Source: University of Michigan via FRED, series MICH
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · FRED shows it a month late by agreement with the source (August 2026 was the latest on 7 October 2026)
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · no reference line · no badge

What it is

From the University of Michigan’s Surveys of Consumers: the median price change that consumers expect over the next 12 months.

Computation

Published by the University of Michigan via FRED (MICH); the dashboard shows it unchanged. FRED’s notes say the most recent value is not shown, by agreement with the source.

Reading it

There is no badge or reference line. From January 1978 the highest value was 10.4% (January 1980) and the lowest 0.4% (November 2001); since 2020 the highest was 6.6% (May 2025). In August 2026 it was 4.0%.

Use in practice

It is the only household survey on the tab; compare it with the market-implied breakevens and with realized CPI inflation.

Caveats

A month behind on FRED, as above.
A median of survey answers, not a price measure.

CPI inflation — US

Source: U.S. Bureau of Labor Statistics via FRED, series CPIAUCSL
Licence: Public domain; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

US consumer price inflation: the change over twelve months in the Consumer Price Index for All Urban Consumers, all items, US city average (BLS, seasonally adjusted).

Computation

Computed by the dashboard from the published index (CPIAUCSL, 1982–84 = 100) (Appendix A.1):

$\pi_t = 100\left(\frac{P_t}{P_{t-12}} - 1\right),$

for months where both $P_t$ and $P_{t-12}$ are published. FRED has no value for October 2025, so October 2025 and October 2026 have none.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

The reference line is at 2, the bands are the dashboard’s own, and the Fed’s 2% goal refers to PCE inflation, not CPI. From January 1948 the highest rate was 14.6% (March 1980) and the lowest −3.0% (August 1949); since 2019 the highest was 9.0% (June 2022). Since 1990 the badge would have read “Near target” in 38% of months, “Above target” in 41% and “High” in 14%. In August 2026 the rate was 3.35%.

Use in practice

Compare it with core CPI, which leaves out food and energy, and with core PCE, the core version of the measure the Fed’s goal is defined on.

Caveats

Seasonally adjusted index. The tile uses the seasonally adjusted index. Since 2021 its rates differed from those of the unadjusted index (CPIAUCNS) by up to 0.14 percentage points; in June 2022, 8.98% against 9.06%.
Missing month, as above.

Core CPI inflation — US

Source: U.S. Bureau of Labor Statistics via FRED, series CPILFESL
Licence: Public domain; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

US core CPI inflation: the change over twelve months in the CPI for all urban consumers, all items less food and energy (BLS, seasonally adjusted). FRED’s notes call it “core CPI” and say it is widely used because food and energy prices are volatile.

Computation

Computed by the dashboard from the published index (CPILFESL, 1982–84 = 100), as for headline CPI; FRED has no October 2025 value, so October 2025 and October 2026 have none.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1958 the highest rate was 13.6% (June 1980) and the lowest 0.6% (October 2010); since 2019 the highest was 6.6% (September 2022). In August 2026 it was 2.45%.

Use in practice

The gap between headline and core inflation reflects food and energy prices, which core leaves out.

Caveats

Missing month, as for headline CPI.

Core PCE inflation — US

Source: U.S. Bureau of Economic Analysis via FRED, series PCEPILFE
Licence: Public domain; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

US core PCE inflation: the change over twelve months in the chain-type price index for personal consumption expenditures excluding food and energy (BEA, seasonally adjusted). The FOMC’s 2% longer-run goal is defined on the annual change in the overall PCE price index; this tile shows the version without food and energy.

Computation

Computed by the dashboard from the published index (PCEPILFE, 2017 = 100), as for CPI. The PCE index has no missing months.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1960 the highest rate was 10.2% (February 1975) and the lowest 0.6% (July 2009); since 2019 the highest was 5.6% (September 2022). In August 2026 it was 3.01%.

Use in practice

Compare it with core CPI: FRED’s notes say the PCE and CPI indexes are constructed differently, resulting in different inflation rates.

Caveats

Revisions. FRED’s notes say BEA revises previously published PCE data to reflect updated information or new methodology.

PPI final demand inflation — US

Source: U.S. Bureau of Labor Statistics via FRED, series PPIFIS
Licence: Public domain; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Inflation tab · no reference line · no badge

What it is

US producer price inflation: the change over twelve months in the Producer Price Index by Commodity, Final Demand (BLS, seasonally adjusted).

Computation

Computed by the dashboard from the published index (PPIFIS, November 2009 = 100), as for CPI, so the rate starts in November 2010.

Reading it

There is no target, so no reference line or badge. From November 2010 the highest rate was 11.6% (March 2022) and the lowest −1.5% (October 2015); in August 2026 it was 5.41%.

Use in practice

Compare it with consumer-price inflation on this tab; the dashboard classes it as a leading indicator and the consumer-price measures as coincident.

Caveats

Short history: the final-demand index starts in November 2009.

HICP inflation — Euro area

Source: Eurostat (CC BY 4.0), series M.RCH_A.TOTAL.EA
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · September 2026 was available on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

Euro-area inflation: the annual rate of change of the Harmonised Index of Consumer Prices (HICP), from Eurostat, for the euro area in its changing composition. Eurostat’s label lists the compositions: 11 countries from 1999, 12 from 2001, rising to 20 from 2023 and 21 from 2026.

Computation

From Eurostat (dataset prc_hicp_minr, “HICP – ECOICOP ver.2 – indices and rates of change, monthly data”, unit “annual rate of change”, geo EA); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

The ECB aims for 2% inflation over the medium term, measured by the HICP; the reference line is at 2 and the bands are the dashboard’s own. From January 1997 the highest rate was 10.6% (October 2022) and the lowest −0.7% (July 2009); it was negative in 17 months between June 2009 and December 2020. In September 2026 it was 3.8%.

Use in practice

Compare it with the ECB deposit rate on the Credit & Rates tab and with the members’ rates below.

Caveats

Composition. The series follows the euro area’s membership; Eurostat also publishes fixed compositions (EA20 for 2023–2025, EA21 from 2026).

HICP inflation — EU

Source: Eurostat (CC BY 4.0), series M.RCH_A.TOTAL.EU27_2020
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

HICP inflation for the European Union, 27 countries (Eurostat’s “European Union – 27 countries (from 2020)”).

Computation

From Eurostat (prc_hicp_minr, annual rate of change, geo EU27_2020); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From December 2000 the highest rate was 11.5% (October 2022) and the lowest −0.6% (January 2015); it was negative in 8 months between December 2014 and May 2016. In August 2026 it was 3.2%.

Use in practice

It covers EU members outside the euro area too; compare it with the euro-area rate.

Caveats

Wider than the euro area: Eurostat’s EU series covers 27 countries, the euro area 21 (from 2026), so the ECB’s target applies to only part of it.

HICP inflation — Germany

Source: Eurostat (CC BY 4.0), series M.RCH_A.TOTAL.DE
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · September 2026 was available on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

HICP inflation for Germany (Eurostat).

Computation

From Eurostat (prc_hicp_minr, annual rate of change, geo DE); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1997 the highest rate was 11.6% (October 2022) and the lowest −0.7% (July 2009); it was negative in 15 months between May 2009 and December 2020. In September 2026 it was 3.3%.

Use in practice

Compare it with the euro-area rate and with France.

Caveats

Monthly annual rates as published by Eurostat.

HICP inflation — France

Source: Eurostat (CC BY 4.0), series M.RCH_A.TOTAL.FR
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · September 2026 was available on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

HICP inflation for France (Eurostat).

Computation

From Eurostat (prc_hicp_minr, annual rate of change, geo FR); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1997 the highest rate was 7.3% (February 2023) and the lowest −0.8% (July 2009); it was negative in 11 months between May 2009 and April 2016. In September 2026 it was 3.4%.

Use in practice

Compare it with the euro-area rate and with Germany.

Caveats

Monthly annual rates as published by Eurostat.

CPI inflation — UK

Source: Office for National Statistics (Open Government Licence v3.0), series ONS/MM23/D7G7
Licence: Open Government Licence v3.0; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

UK consumer price inflation: the ONS’s “CPI annual rate 00: all items” (series D7G7). The Government sets the Bank of England an inflation target of 2%, and the Bank says CPI is the measure of inflation it targets.

Computation

From the Office for National Statistics (D7G7); the dashboard shows it unchanged. The same series feeds the UK growth and inflation regime.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1989 the highest rate was 11.1% (October 2022) and the lowest −0.1% (April 2015); it was negative only in April, September and October 2015. In August 2026 it was 3.1%.

Use in practice

Compare it with SONIA on the Credit & Rates tab, which tracks Bank Rate.

Caveats

CPI, not CPIH. The ONS also publishes a CPIH annual rate (series L55O), a different measure; this tile shows CPI, the measure the Bank of England targets.

CPI inflation — Japan

Source: International Monetary Fund, Consumer Price Index (CPI), series JPN.CPI._T.YOY_PCH_PA_PT.M
Licence: IMF data terms; data shown on the public site
Frequency: monthly · August 2026 was the latest on 7 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Inflation tab · reference line at 2 · badge

What it is

Japanese consumer price inflation, % change over twelve months, from the IMF’s Consumer Price Index dataset, which the IMF describes as holding each economy’s national all-items headline index. The Bank of Japan has a price stability target of 2%.

Computation

From the IMF data API (dataset IMF.STA CPI, Japan, all items, % change over 12 months); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 0$	Deflation
$0 \le \pi_n < 1$	Below target
$1 \le \pi_n \le 2.5$	Near target
$2.5 < \pi_n \le 4$	Above target
$\pi_n > 4$	High

From January 1960 the highest rate was 24.8% (February 1974) and the lowest −2.5% (October 2009). Since 2000 it was negative in 133 months, most recently in August 2021, and at or above 2% every month from April 2022 to December 2025 (45 months), its longest such run in the data since 1990. In August 2026 it was 1.96%.

Use in practice

Compare it with the Bank of Japan’s overnight call rate on the Credit & Rates tab.

Caveats

A copy, not the producer’s release. The figures come from the IMF’s dataset rather than from Japan’s statistics bureau directly.

Cycle & Growth

Growth & inflation — US

Source: OECD Composite Leading Indicators (CC BY 4.0); U.S. Bureau of Labor Statistics via FRED; regime computed by this site, series DF_CLI:USA, CPIAUCSL
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest month on 9 October 2026
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · no reference line · badge

What it is

Which of four growth-and-inflation regimes the US economy is in, month by month. The idea comes from Bridgewater Associates, whose account of its All Weather strategy describes four economic environments: inflation rising, inflation falling, growth rising and growth falling, each relative to expectations. The tile measures direction instead of surprises, and the four names are the dashboard’s own.

Computation

Computed by the dashboard (Appendix A.10) from the OECD composite leading indicator for the US and CPI inflation:

$g_t = \text{CLI}_t - \text{CLI}_{t-3}, \qquad m_t = \pi_t - \frac{1}{12}\sum_{k=0}^{11}\pi_{t-k},$

with $\pi_t$ CPI inflation over twelve months. Growth is rising when $g_t > 0$ and inflation when $m_t > 0$ . FRED has no October 2025 CPI; the tile fills that month in log-linearly.

Reading it

Condition	Badge
$g_t > 0,\ m_t \le 0$	Goldilocks
$g_t > 0,\ m_t > 0$	Reflation
$g_t \le 0,\ m_t > 0$	Stagflation
$g_t \le 0,\ m_t \le 0$	Disinflationary slowdown

The tile shows the current regime, the month its spell began and the regime before it. Below it, a strip colours every month by regime; pick a period with the strip’s handles, by dragging across the strip or, on the public site, with the zoom, and the quadrant above traces that period, its last twelve months dark. On the public site the history starts in 1960. In August 2026 the US was in Reflation, since March 2026 and after Goldilocks: $g = +0.30$ and $m = +0.24$ (CPI inflation 3.35% against a 12-month average of 3.12%).

Use in practice

From January 1960 to August 2026 (800 months):

the regime changed 158 times, about 2.4 times a year, and the median spell lasted 3 months;
Goldilocks covered 30% of months, Stagflation 29%, Reflation 21% and Disinflationary slowdown 20%.

The regime in the month of each NBER business-cycle peak since 1969:

NBER peak	Regime
Dec 1969	Stagflation
Nov 1973	Stagflation
Jan 1980	Stagflation
Jul 1981	Disinflationary slowdown
Jul 1990	Stagflation
Mar 2001	Disinflationary slowdown
Dec 2007	Stagflation
Feb 2020	Reflation

The dashboard’s inflation rule changes regime less often than a common alternative: since 1970 the tile changes regime 2.28 times a year, while the same tile with inflation momentum measured as the 3-month change in CPI inflation would change 3.23 times a year.

Caveats

Direction only. A tiny positive $g_t$ or $m_t$ counts as rising; the quadrant shows how far from zero each month is.
Revisions. The OECD revises recent months of its leading indicators, so recent months’ regimes can change.

Growth & inflation — Euro area

Source: OECD Composite Leading Indicators (CC BY 4.0); Eurostat (CC BY 4.0); regime computed by this site, series 1, 2
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was the latest month on 9 October 2026
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · no reference line · badge

What it is

The four regimes for the euro area, built like the US tile.

Computation

Growth: the 3-month change in a euro-area leading indicator that the dashboard builds from the OECD indicators for Germany, France, Italy and Spain, weighted by nominal GDP; the OECD’s dataset has no euro-area series (Appendix A.10). Inflation: Eurostat’s HICP annual rate for the euro area in its changing composition, minus its average over the past 12 months.

Reading it

The same rules as the US tile. The history starts in December 1997. In September 2026 the euro area was in Reflation, since August 2026 and after Stagflation ( $g = +0.16$ , $m = +1.19$ ; HICP inflation 3.8% against a 12-month average of 2.61%).

Use in practice

From December 1997 to September 2026 (346 months) the regime changed 63 times, about 2.2 times a year, with a median spell of 4 months. Goldilocks covered 29% of months, Reflation 26%, Disinflationary slowdown 25% and Stagflation 21%.

Caveats

A constructed indicator. The growth input covers four countries, weighted by GDP.

Growth & inflation — UK

Source: OECD Composite Leading Indicators (CC BY 4.0); Office for National Statistics (Open Government Licence v3.0); regime computed by this site, series DF_CLI:GBR, ONS/MM23/D7G7
Licence: Open Government Licence v3.0; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest month on 9 October 2026
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · no reference line · badge

What it is

The four regimes for the UK, built like the US tile.

Computation

Growth: the 3-month change in the OECD composite leading indicator for the UK. Inflation: the ONS CPI annual rate (D7G7), minus its average over the past 12 months.

Reading it

The same rules as the US tile. The history starts in December 1989, as the ONS CPI annual rate begins in January 1989. In August 2026 the UK was in Goldilocks, since June 2026 and after Disinflationary slowdown ( $g = +0.70$ , $m = -0.03$ ).

Use in practice

From December 1989 to August 2026 (441 months) the regime changed 94 times, about 2.6 times a year, with a median spell of 3 months. Goldilocks covered 29% of months, Disinflationary slowdown 25%, Stagflation 24% and Reflation 22%.

Caveats

Close calls. In August 2026 inflation momentum was −0.03, close to the boundary.

Growth & inflation — Japan

Source: OECD Composite Leading Indicators (CC BY 4.0); International Monetary Fund, Consumer Price Index (CPI); regime computed by this site, series DF_CLI:JPN, IMF.STA/CPI/JPN.CPI._T.YOY_PCH_PA_PT.M
Licence: IMF data terms; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest month on 9 October 2026
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · no reference line · badge

What it is

The four regimes for Japan, built like the US tile.

Computation

Growth: the 3-month change in the OECD composite leading indicator for Japan. Inflation: CPI inflation from the IMF’s CPI dataset, minus its average over the past 12 months.

Reading it

The same rules as the US tile. The history starts in 1960. In August 2026 Japan was in Goldilocks, since August 2026 and after Reflation ( $g = +0.14$ , $m = -0.01$ ).

Use in practice

From January 1960 to August 2026 (800 months) the regime changed 174 times, about 2.6 times a year, with a median spell of 3 months. Goldilocks covered 29% of months, Stagflation 26%, Disinflationary slowdown 23% and Reflation 22%.

Caveats

Inflation from a copy. The IMF’s figures are not Japan’s statistics bureau’s own release.

Growth & inflation — China

Source: OECD Composite Leading Indicators (CC BY 4.0); International Monetary Fund, Consumer Price Index (CPI); regime computed by this site, series DF_CLI:CHN, IMF.STA/CPI/CHN.CPI._T.YOY_PCH_PA_PT.M
Licence: IMF data terms; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest month on 9 October 2026
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · no reference line · badge

What it is

The four regimes for China, built like the US tile. The rule reads direction, not level, so inflation rising from below 1% still counts as rising.

Computation

Growth: the 3-month change in the OECD composite leading indicator for China. Inflation: CPI inflation from the IMF’s CPI dataset, minus its average over the past 12 months.

Reading it

The same rules as the US tile. The history starts in December 1994. In August 2026 China was in Stagflation, since August 2026 and after Disinflationary slowdown ( $g = -0.78$ , $m = +0.07$ ; CPI inflation 0.80% against a 12-month average of 0.73%).

Use in practice

From December 1994 to August 2026 (381 months) the regime changed 78 times, about 2.5 times a year, with a median spell of 3 months. Goldilocks covered 29% of months, Stagflation 26%, Disinflationary slowdown 24% and Reflation 21%.

Caveats

Level versus direction. “Stagflation” here means inflation momentum is positive; in August 2026 inflation itself was 0.80%.
Inflation from a copy, as for Japan.

OECD CLI — G20

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:G20
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for the G20, amplitude adjusted, which the OECD names as its headline CLI.

Computation

From the OECD’s data API (dataset DF_CLI, area G20, amplitude adjusted, OECD harmonised methodology); the dashboard shows it unchanged. The OECD computes CLIs for the G20 countries, Spain and five zone aggregates.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

The badge rules and reading 100 as trend are the dashboard’s. From January 1961 the G20 indicator averaged 100.3 and was at or above 100 in 64% of months; its highest value was 104.9 (January 1973) and its lowest 89.4 (April 2020). In September 2026 it was 100.14, down from 100.16 the month before.

Use in practice

Read its direction, as the OECD advises, and compare it with the G7 and national indicators below.

Caveats

Revised. The OECD revises recent months.

Core capital goods orders — US

Source: U.S. Census Bureau via FRED, series NEWORDER
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest on 9 October 2026
Unit: percent
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · reference line at 0 · no badge

What it is

US manufacturers’ new orders for nondefense capital goods excluding aircraft, from the Census Bureau: the change over twelve months, in %.

Computation

Computed by the dashboard from the published level (NEWORDER, US$ millions, seasonally adjusted, from February 1992) (Appendix A.1):

$g_t = 100\left(\frac{O_t}{O_{t-12}} - 1\right).$

FRED’s notes say the data were reconstructed on 21 May 2001 to switch from the SIC to the NAICS classification.

Reading it

The reference line is at zero; there is no badge. From February 1993 the range was −32.1% (April 2009) to +24.7% (April 2021); growth was positive in 67% of months. In August 2026 it was 14.1%.

Use in practice

The dashboard classes it as a leading indicator; compare it with industrial production and the regional surveys.

Caveats

Nominal. Orders are in current dollars, so the rate includes price changes.

OECD CLI — US

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:USA
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for the US, amplitude adjusted: the growth input of the US regime tile.

Computation

From the OECD’s data API (DF_CLI, area USA); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1955 it averaged 100.0 and was at or above 100 in 54% of months; its highest value was 103.8 (December 1972) and its lowest 92.9 (April 2020). In September 2026 it was 101.00, up from 100.95.

Use in practice

In today’s series, its highest value in the 26 months before each NBER peak:

NBER peak	Highest value in the 26 months before	Months before the peak	Value at the peak
Aug 1957	Jul 1955	25	98.6
Apr 1960	Apr 1959	12	99.8
Dec 1969	Dec 1968	12	98.9
Nov 1973	Dec 1972	11	100.9
Jan 1980	Jul 1978	18	98.4
Jul 1981	May 1979	26	98.8
Jul 1990	Sep 1988	22	98.9
Mar 2001	Dec 1999	15	98.4
Dec 2007	Jun 2007	6	101.6
Feb 2020	Apr 2018	22	99.2

The indicator was below 100 at eight of these ten peaks.

Caveats

Today’s series, not real time. The OECD revises the indicator, so this history is not what was published at the time.

Empire State manufacturing

Source: Federal Reserve Bank of New York via FRED, series GACDISA066MSFRBNY
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 0 · no badge

What it is

The New York Fed’s Empire State Manufacturing Survey: the general business conditions index for New York State, the survey’s headline index (seasonally adjusted). The NY Fed sends the survey on the first day of each month to the same pool of about 200 manufacturing executives in the state, typically the president or CEO; about 100 reply.

Computation

Published by the New York Fed via FRED (GACDISA066MSFRBNY); the dashboard shows it unchanged. It is a diffusion index built from the shares of respondents reporting increases and decreases.

Reading it

The reference line is at zero; there is no badge. Above zero, more firms reported improving than worsening conditions; the NY Fed’s September 2026 report described a reading of 7.6 as business activity increasing modestly. From July 2001 the index ranged from −80.0 (April 2020) to 39.0 (April 2004) and was positive in 67% of months.

Use in practice

It describes the current month: the NY Fed says most responses are completed by the tenth. Compare it with the Philadelphia Fed survey.

Caveats

One state’s manufacturers. It covers New York manufacturing only.

Philadelphia Fed manufacturing

Source: Federal Reserve Bank of Philadelphia via FRED, series GACDFSA066MSFRBPHI
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 0 · no badge

What it is

The Philadelphia Fed’s Manufacturing Business Outlook Survey: the current general activity diffusion index for manufacturing firms in its district (seasonally adjusted), which FRED describes as the survey’s broadest measure.

Computation

Published by the Philadelphia Fed via FRED (GACDFSA066MSFRBPHI); the dashboard shows it unchanged. FRED’s notes: the percentage of firms reporting increases minus the percentage reporting decreases.

Reading it

The reference line is at zero; there is no badge. From May 1968 the index ranged from −58.7 (April 2020) to 58.5 (March 1973) and was positive in 73% of months; in September 2026 it was 37.8.

Use in practice

Compare it with the Empire State survey; together they cover two districts’ manufacturers.

Caveats

One district’s manufacturers.

Industrial production — US

Source: Board of Governors of the Federal Reserve System (US) via FRED, series INDPRO
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest on 9 October 2026
Unit: percent
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Cycle & Growth tab · reference line at 0 · no badge

What it is

US industrial production: the Federal Reserve’s index of the real output of manufacturing, mining, and electric and gas utilities, shown as the change over twelve months in %. FRED’s notes add that the industrial sector, together with construction, accounts for the bulk of the variation in national output over the business cycle.

Computation

Computed by the dashboard from the published index (INDPRO, 2017 = 100, seasonally adjusted, from January 1919) (Appendix A.1).

Reading it

The reference line is at zero; there is no badge. From January 1920 the range was −33.7% (February 1946) to +62.0% (July 1933); since 1960 the rate was negative in 24% of months. In August 2026 it was 1.4%.

Use in practice

The dashboard classes it as coincident. Compare it with the leading tiles above it.

Caveats

Industry only: services aren’t covered.

Real GDP — US

Source: U.S. Bureau of Economic Analysis via FRED, series A191RL1Q225SBEA
Licence: Public domain; data shown on the public site
Frequency: quarterly · quarterly; the third estimate for the second quarter of 2026 was the latest on 9 October 2026
Unit: percent
Indicator class: lagging
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 0 · no badge

What it is

US real GDP growth as BEA reports it in its headline: the percent change from the preceding quarter at a seasonally adjusted annual rate. BEA reported the second quarter of 2026 as “an annual rate of 2.2 percent” in its third estimate.

Computation

Published by BEA via FRED (A191RL1Q225SBEA); the dashboard shows it unchanged.

Reading it

The reference line is at zero; there is no badge. From the second quarter of 1947 the range was −28.0% (second quarter of 2020) to +34.9% (third quarter of 2020); growth was negative in 44 quarters, most recently the first quarter of 2022.

Use in practice

The dashboard classes it as lagging: it arrives after the quarter ends and is revised in later estimates. Compare it with industrial production and the leading indicators.

Caveats

Annualized: a quarter’s change is expressed as if it continued for a year.
Estimates are revised; the series shows the latest estimate for each quarter.

OECD CLI — G7

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:G7
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for the G7, amplitude adjusted.

Computation

From the OECD’s data API (DF_CLI, area G7); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1959 it averaged 100.0 and was at or above 100 in 53% of months; its highest value was 104.0 (January 1973) and its lowest 92.4 (April 2020). In September 2026 it was 100.94, up from 100.86.

Use in practice

Compare it with the G20 indicator and the national indicators.

Caveats

Revised, as for all CLIs.

OECD CLI — UK

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:GBR
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for the UK, amplitude adjusted: the growth input of the UK regime tile.

Computation

From the OECD’s data API (DF_CLI, area GBR); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From December 1957 it averaged 100.0 and was at or above 100 in 58% of months; its highest value was 106.8 (May 1972) and its lowest 87.0 (April 2020). In September 2026 it was 101.38, up from 101.14.

Use in practice

Compare it with the UK’s regime tile.

Caveats

Revised, as for all CLIs.

OECD CLI — Japan

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:JPN
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for Japan, amplitude adjusted: the growth input of the Japan regime tile.

Computation

From the OECD’s data API (DF_CLI, area JPN); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1959 it averaged 100.0 and was at or above 100 in 50% of months; its highest value was 106.0 (March 1973) and its lowest 94.5 (January 1975). In September 2026 it was 100.35, up from 100.32.

Use in practice

Compare it with Japan’s regime tile.

Caveats

Revised, as for all CLIs.

OECD CLI — Germany

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:DEU
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Cycle & Growth tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for Germany, amplitude adjusted: one of the four national indicators the dashboard combines into its euro-area indicator.

Computation

From the OECD’s data API (DF_CLI, area DEU); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1961 it averaged 100.0 and was at or above 100 in 51% of months; its highest value was 103.7 (May 1969) and its lowest 90.8 (April 2020). In September 2026 it was 101.01, up from 100.90.

Use in practice

Compare it with the euro-area regime tile.

Caveats

Revised, as for all CLIs.

PMI — Euro area

Source: The tile links to www.pmi.spglobal.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: Cycle & Growth tab · no reference line · no badge

What it is

S&P Global’s purchasing managers’ index (PMI) for the euro area. The data are licensed, so the tile shows no figures and links to S&P Global’s PMI site.

Reading it

The tile has no data, badge or reference line.

PMI — UK

Source: The tile links to www.pmi.spglobal.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: Cycle & Growth tab · no reference line · no badge

What it is

S&P Global’s purchasing managers’ index for the UK; licensed, so the tile links to S&P Global’s PMI site.

Reading it

The tile has no data, badge or reference line.

PMI — Japan

Source: The tile links to www.pmi.spglobal.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: Cycle & Growth tab · no reference line · no badge

What it is

S&P Global’s purchasing managers’ index for Japan; licensed, so the tile links to S&P Global’s PMI site.

Reading it

The tile has no data, badge or reference line.

Labor

Initial jobless claims — US

Source: U.S. Employment and Training Administration via FRED, series ICSA
Licence: Public domain; data shown on the public site
Frequency: weekly · week ending Saturday; the week ending 3 October 2026 was available by 9 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · badge

What it is

The number of initial claims for unemployment insurance in the US each week (Employment and Training Administration, seasonally adjusted). FRED’s notes: an initial claim is filed by an unemployed individual after a separation from an employer and requests a determination of basic eligibility for the unemployment insurance program.

Computation

Published by the Department of Labor via FRED (ICSA); the dashboard shows it unchanged.

Reading it

Condition	Badge
$100\,(x_n - \min_{52})/\min_{52} > 15$	Rising off lows
$100\,(x_n - \min_{52})/\min_{52} \le 15$	Near lows — firm

From January 1967 the weekly count ranged from 162,000 (week ending 30 November 1968) to 6,137,000 (4 April 2020). In the week ending 3 October 2026 it was 197,000, 4.2% above the lowest of the past 52 weeks.

Use in practice

Since 1970 the badge read “Rising off lows” in 31% of weeks. It did so for four weeks or more in 33 separate spells; 18 of them began outside the period from 13 months before an NBER peak to 2 months after the following trough. The first such spell in the 13 months before to 6 months after each peak:

NBER peak	Spell began	Weeks before (−) or after (+) the peak
Nov 1973	28 Jul 1973	−14
Jan 1980	23 Dec 1978	−53
Jul 1981	31 May 1980 (during the 1980 recession)	−57
Jul 1990	27 May 1989	−57
Mar 2001	15 Jul 2000	−33
Dec 2007	15 Dec 2007	+2
Feb 2020	30 Nov 2019	−9

Caveats

Frequent amber spells. More than half of the four-week spells since 1970 came away from recessions; compare the continuing claims tile.
Weekly noise: one week can move the badge.

Continuing claims — US

Source: U.S. Employment and Training Administration via FRED, series CCSA
Licence: Public domain; data shown on the public site
Frequency: weekly · week ending Saturday; the week ending 26 September 2026 was available by 9 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · badge

What it is

Continued claims, also called insured unemployment (seasonally adjusted). FRED’s notes: the number of people who have already filed an initial claim, experienced a week of unemployment and then filed a continued claim for that week; the data are based on the week of unemployment, not the week the initial claim was filed.

Computation

Published by the Department of Labor via FRED (CCSA); the dashboard shows it unchanged.

Reading it

Condition	Badge
$100\,(x_n - \min_{52})/\min_{52} > 15$	Rising off lows
$100\,(x_n - \min_{52})/\min_{52} \le 15$	Near lows — firm

From January 1967 the count ranged from 988,000 (31 May 1969) to 23,130,000 (9 May 2020). In the week ending 26 September 2026 it was 1,716,000, 1.0% above the lowest of the past 52 weeks.

Use in practice

Since 1970 the badge read “Rising off lows” in 24% of weeks, in 13 spells of four weeks or more; only one (from December 2022) began outside the period from 13 months before an NBER peak to 2 months after the following trough. The first such spell around each peak:

NBER peak	Spell began	Weeks before (−) or after (+) the peak
Nov 1973	15 Dec 1973	+6
Jan 1980	27 Oct 1979	−9
Jul 1981	31 May 1980 (during the 1980 recession)	−57
Jul 1990	5 Aug 1989	−47
Mar 2001	9 Dec 2000	−12
Dec 2007	26 Jan 2008	+8
Feb 2020	2 Feb 2019	−52

Caveats

A week later than initial claims, and it counts people claiming, not everyone unemployed.

Quits rate (JOLTS) — US

Source: U.S. Bureau of Labor Statistics via FRED, series JTSQUR
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; the August 2026 data were released on 29 September 2026
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

The quits rate from the BLS Job Openings and Labor Turnover Survey (JOLTS): quits in the month as a percentage of total nonfarm employment (seasonally adjusted).

Computation

Published by the BLS via FRED (JTSQUR); the dashboard shows it unchanged. In 99.7% of months it equals the JOLTS quits level (JTSQUL) divided by total nonfarm payroll employment (PAYEMS), times 100, to within 0.1.

Reading it

There is no badge or reference line. From December 2000 it ranged from 1.2% (August 2009) to 3.0% (November 2021); in August 2026 it was 1.9%.

Use in practice

Compare it with job openings, from the same survey.

Caveats

Starts in December 2000, so it covers only the 2001, 2007 and 2020 recessions.

Job openings (JOLTS) — US

Source: U.S. Bureau of Labor Statistics via FRED, series JTSJOL
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; the August 2026 data were released on 29 September 2026
Unit: thousands
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

Job openings from JOLTS, in thousands (seasonally adjusted). FRED’s notes: all jobs not filled on the last business day of the month, where a specific position exists, work is available and can start within 30 days, and the employer is actively recruiting.

Computation

Published by the BLS via FRED (JTSJOL); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From December 2000 openings ranged from 2.23 million (July 2009) to 12.30 million (March 2022); in August 2026 there were 7.08 million.

Use in practice

FRED’s notes describe openings as a measure of unmet demand for labour. Dividing by the number of unemployed (BLS, FRED series UNEMPLOY) gives openings per unemployed person: 2.04 at the March 2022 high and 1.01 in August 2026.

Caveats

Openings can fall without hiring: FRED’s notes say they decline when filled or when employers withdraw them.

Temp-help employment — US

Source: U.S. Bureau of Labor Statistics via FRED, series TEMPHELPS
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Unit: thousands
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

The number of employees in temporary help services, in thousands (seasonally adjusted), from the BLS Current Employment Statistics (establishment survey).

Computation

Published by the BLS via FRED (TEMPHELPS); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1990 it ranged from 1.11 million (July 1991) to 3.16 million (March 2022); in September 2026 it was 2.49 million.

Use in practice

Its highest point before each recession since the series began:

NBER peak	Highest month in the 36 before	Months before the peak
Mar 2001	Apr 2000	11
Dec 2007	May 2006	19
Feb 2020	Oct 2018	16

Caveats

Three recessions in the history, too few for a firm rule.

Manufacturing hours — US

Source: U.S. Bureau of Labor Statistics via FRED, series AWHMAN
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Unit: hours
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

Average weekly hours of production and nonsupervisory employees in manufacturing (BLS, seasonally adjusted). FRED’s notes: the hours they were paid for, which differ from standard or scheduled hours; unpaid absences, turnover, part-time work and stoppages make them lower than scheduled hours.

Computation

Published by the BLS via FRED (AWHMAN); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1939 the average ranged from 37.0 hours (May 1939) to 45.7 (November 1943); in September 2026 it was 42.0.

Use in practice

Read its direction against temporary-help employment and payrolls.

Caveats

Manufacturing only, and production and nonsupervisory employees only.

Nonfarm payrolls — US

Source: U.S. Bureau of Labor Statistics via FRED, series PAYEMS
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Unit: thousands
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Labor tab · reference line at 0 · no badge

What it is

The change in US total nonfarm payroll employment from the previous month, in thousands. FRED’s notes describe total nonfarm payrolls as excluding proprietors, private household employees, unpaid volunteers, farm employees and the unincorporated self-employed, and as accounting for about 80% of the workers who contribute to GDP.

Computation

Computed by the dashboard from the published level (PAYEMS, seasonally adjusted):

$\Delta E_t = E_t - E_{t-1} \quad (\text{thousands}),$

for consecutive published months.

Reading it

The reference line is at zero; there is no badge. From February 1939 the change ranged from −20.47 million (April 2020) to +4.63 million (June 2020). It was negative in seven months from January 2025 to September 2026 (January, June, August, October and December 2025, February and July 2026), and +29,000 in September 2026.

Use in practice

Compare it with the unemployment rate, which comes from the household survey (Current Population Survey).

Caveats

Small against the level: the September 2026 change of +29,000 compares with total payroll employment of 159.0 million.

Unemployment rate — US

Source: U.S. Bureau of Labor Statistics via FRED, series UNRATE
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Unit: percent
Indicator class: lagging
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

The US unemployment rate (U-3) from the Current Population Survey (household survey): the unemployed as a percentage of the labour force, people aged 16 and over who live in the 50 states or DC, outside institutions and not on active military duty (FRED’s notes).

Computation

Published by the BLS via FRED (UNRATE); the dashboard shows it unchanged. The Sahm rule is built on it.

Reading it

There is no badge or reference line. From January 1948 it ranged from 2.5% (May 1953) to 14.8% (April 2020); in September 2026 it was 4.2%.

Use in practice

Read its change through the Sahm rule on the Recession Watch tab, which turns a rise in the three-month average into a signal.

Caveats

A survey estimate, from a household sample.

Unemployment rate — UK

Source: Office for National Statistics (Open Government Licence v3.0), series ONS/LMS/MGSX
Licence: Open Government Licence v3.0; data shown on the public site
Frequency: monthly · monthly, three-month averages; the ONS release of 15 September 2026 covered May to July 2026, and the next is due on 20 October
Unit: percent
Indicator class: lagging
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

The UK unemployment rate for people aged 16 and over, seasonally adjusted, from the ONS Labour Force Survey (series MGSX). The ONS counts as unemployed people without a job who have been actively seeking work in the last four weeks and are available to start within the next two weeks, and expresses them as a share of the economically active population.

Computation

From the ONS (MGSX); the dashboard shows it unchanged. Each value covers three months and the series dates it at the middle month: the ONS’s 4.9% for May to July 2026 is the June 2026 point.

Reading it

There is no badge or reference line. From February 1971 it ranged from 3.4% (November 1973) to 11.9% (March 1984); for May to July 2026 it was 4.9%.

Use in practice

Compare it with the UK regime tile and the UK’s CPI inflation.

Caveats

Three-month averages, so it moves slowly and lags by about a month and a half.

Unemployment rate — Japan

Source: Organization for Economic Co-operation and Development via FRED, series LRHUTTTTJPM156S
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; July 2026 was the latest on FRED on 9 October 2026
Unit: percent
Indicator class: lagging
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

Japan’s unemployment rate for people aged 15 and over, seasonally adjusted, from the OECD’s infra-annual labour statistics (via FRED).

Computation

Published by the OECD via FRED (LRHUTTTTJPM156S); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1955 it ranged from 1.0% (November 1968) to 5.5% (June 2002); in July 2026 it was 2.4%.

Use in practice

Compare it with Japan’s regime tile and CPI inflation.

Caveats

An OECD copy via FRED, not Japan’s statistics bureau’s own release.

Unemployment rate — Germany

Source: Organization for Economic Co-operation and Development via FRED, series LRHUTTTTDEM156S
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; July 2026 was the latest on FRED on 9 October 2026
Unit: percent
Indicator class: lagging
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Labor tab · no reference line · no badge

What it is

Germany’s unemployment rate for people aged 15 and over, seasonally adjusted, from the OECD’s infra-annual labour statistics (via FRED).

Computation

Published by the OECD via FRED (LRHUTTTTDEM156S); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1991 it ranged from 2.9% (April 2019) to 11.2% (April 2005); in July 2026 it was 4.0%.

Use in practice

Compare it with Germany’s HICP inflation and leading indicator.

Caveats

An OECD copy via FRED, not Germany’s statistics office’s own release.

Trade & Commodities

Global supply chain pressure

Source: Federal Reserve Bank of New York, Global Supply Chain Pressure Index, series NYFed/GSCPI
Licence: New York Fed terms of use; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Unit: standard deviations
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Trade & Commodities tab · reference line at 0 · badge

What it is

The Global Supply Chain Pressure Index (GSCPI), published by the Federal Reserve Bank of New York and introduced in its Staff Report 1017 (Benigno, di Giovanni, Groen and Noble, May 2022). It summarises pressure in global supply chains in one number. The dashboard uses it in place of the Baltic Dry Index tile of the original design; the Baltic Dry Index is one of the GSCPI’s inputs.

Computation

Published by the New York Fed; the dashboard shows it unchanged. The staff report builds it in two steps from two sets of inputs:

Transport costs: the Baltic Dry Index (shipping costs for raw materials such as coal or steel), the Harpex index of container-ship charter rates, and the BLS price indices for airfreight to and from the US, for Asia and Europe.
Manufacturing PMI components for China, the euro area, Japan, South Korea, Taiwan, the UK and the US: delivery times, backlogs and purchased stocks.

First, each input is cleaned of demand effects by regressing it on PMI measures of demand (new orders, and for transport costs also quantities purchased) and keeping the residuals. Second, a principal component analysis extracts the inputs’ common component, which is the index:

$\text{GSCPI}_t = \frac{c_t - \bar c}{s_c},$

with $c_t$ the common component, expressed in standard deviations from its average since 1997.

Reading it

Condition	Badge
$\text{GSCPI}_n \ge 2$	Severe pressure
$1 \le \text{GSCPI}_n < 2$	Elevated pressure
$-1 \le \text{GSCPI}_n < 1$	Normal range
$\text{GSCPI}_n < -1$	Slack

The reference line is at zero, the index’s average. From September 1997 to September 2026 the index read “Normal range” in 84% of months, “Severe pressure” in 6%, “Elevated pressure” in 5% and “Slack” in 5%. Its highest value was 4.43 (December 2021) and its lowest −1.56 (May 2023); in September 2026 it was 1.28.

Use in practice

It was at 2 or above from March to July 2020 and from February 2021 to June 2022. The staff report found recent inflationary pressures closely related to the index, especially US producer price inflation; compare it with the PPI final demand tile on the Inflation tab.

Caveats

Relative to its own history: zero is the 1997–2026 average, and the scale is in standard deviations.

Brent crude oil

Source: U.S. Energy Information Administration via FRED, series DCOILBRENTEU
Licence: Public domain; data shown on the public site
Frequency: daily · daily; the price for 6 October 2026 was available by 9 October
Unit: US$ per barrel
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Trade & Commodities tab · no reference line · no badge

What it is

The Brent crude oil spot price in US dollars per barrel, from the US Energy Information Administration (EIA). The EIA defines Brent as a blended crude stream produced in the North Sea region that serves as a reference, or “marker”, for pricing a number of other crude streams.

Computation

Published by the EIA via FRED (DCOILBRENTEU); the dashboard shows it unchanged. The EIA names Refinitiv, an LSEG business, as the source of its spot prices.

Reading it

There is no badge or reference line. From 20 May 1987 the price ranged from $9.10 (10 December 1998) to $143.95 (3 July 2008); on 6 October 2026 it was $125.44.

Use in practice

The gap between Brent and WTI, computed from the two daily series:

Period	Average Brent minus WTI
1987–2010	−$1.29
2011–2014	+$12.48
2015–October 2026	+$4.37

Its widest point was +$54.34 on 20 April 2020, the day WTI closed below zero; on 6 October 2026 it was +$29.20.

Caveats

Nominal dollars, not adjusted for inflation.
Spot prices, which can differ from futures prices.

WTI crude oil — US

Source: U.S. Energy Information Administration via FRED, series DCOILWTICO
Licence: Public domain; data shown on the public site
Frequency: daily · daily; the price for 6 October 2026 was available by 9 October
Unit: US$ per barrel
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Trade & Commodities tab · no reference line · no badge

What it is

The West Texas Intermediate (WTI) crude oil spot price at Cushing, Oklahoma, in US dollars per barrel (EIA). The EIA defines WTI as a crude stream produced in Texas and southern Oklahoma that serves as a reference for pricing other crude streams and is traded in the domestic spot market at Cushing.

Computation

Published by the EIA via FRED (DCOILWTICO); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From 2 January 1986 the price ranged from −$36.98 (20 April 2020, the only negative day) to $145.31 (3 July 2008); on 6 October 2026 it was $96.24.

Use in practice

Compare it with Brent: the gap between the two is in the Brent section.

Caveats

Nominal dollars; spot prices, as for Brent.

Henry Hub natural gas — US

Source: U.S. Energy Information Administration via FRED, series DHHNGSP
Licence: Public domain; data shown on the public site
Frequency: daily · daily; the price for 6 October 2026 was available by 9 October
Unit: US$ per million BTU
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Trade & Commodities tab · no reference line · no badge

What it is

The Henry Hub natural gas spot price in US dollars per million British thermal units (EIA). The EIA’s notes say the prices are based on delivery at the Henry Hub in Louisiana, and define a spot price as the price of a one-time open-market transaction for immediate delivery.

Computation

Published by the EIA via FRED (DHHNGSP); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From 7 January 1997 the price ranged from $1.21 (8 November 2024) to $30.72 (23 January 2026); on 6 October 2026 it was $3.03.

Use in practice

Its range since 1997 spans a factor of about 25 ($1.21 to $30.72), so spikes dominate the level chart; the percentile mode shows where a price sits within the window instead.

Caveats

Daily spot prices can spike: in January 2026 the price rose from $3.06 on the 16th to $30.72 on the 23rd and was back at $4.40 on 2 February.

Copper price

Source: International Monetary Fund via FRED, series PCOPPUSDM
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; July 2026 was the latest on 9 October 2026
Unit: US$ per metric ton
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Trade & Commodities tab · no reference line · badge

What it is

The IMF’s global price of copper, in US dollars per metric ton (via FRED). FRED’s notes say the IMF’s commodity prices are benchmark prices representative of the global market, determined by the largest exporter of each commodity, and are period averages in nominal dollars.

Computation

Published by the IMF via FRED (PCOPPUSDM); the dashboard shows it unchanged.

Reading it

Condition	Badge
$100\,(P_n / P_{n-3} - 1) > 1$	Rising — demand firm
$100\,(P_n / P_{n-3} - 1) < -1$	Falling — softening
$-1 \le 100\,(P_n / P_{n-3} - 1) \le 1$	Flat

The badge compares the latest month with three months earlier. From January 1992 it read “Rising — demand firm” in 54% of months, “Falling — softening” in 39% and “Flat” in 7%. The price ranged from $1,377 (October 2001) to $13,552 (June 2026); in July 2026 it was $13,543, 5.1% above three months earlier.

Use in practice

The badge’s three-month window reads the recent direction; compare it with the leading indicators on the Cycle & Growth tab.

Caveats

Monthly averages, published with a lag: July 2026 was the latest month in early October.
Nominal dollars.

Consumer & Housing

UMich consumer sentiment

Source: University of Michigan via FRED, series UMCSENT
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; FRED shows it a month late at the source’s request (August 2026 was the latest on FRED on 9 October 2026)
Unit: index
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Consumer & Housing tab · no reference line · no badge

What it is

The University of Michigan’s Index of Consumer Sentiment, from its Surveys of Consumers (1966 first quarter = 100).

Computation

Published by the University of Michigan via FRED (UMCSENT); the dashboard shows it unchanged. FRED’s notes say the data are delayed by one month at the request of the source.

Reading it

There is no badge or reference line. FRED’s series begins in November 1952, with only 92 observations before 1978, and is monthly from 1978. Its highest value was 112.0 (January 2000) and its lowest 44.8 (May 2026); the latest on FRED was 51.7, for August 2026. The University’s own site gave a preliminary 46.3 for October 2026, against 48.1 for September and 53.6 for October 2025.

Use in practice

The University also publishes indexes of current economic conditions and of consumer expectations (44.7 and 47.3 in the preliminary October 2026 results); the tile shows the overall sentiment index.

Caveats

A month behind on FRED, as above.
A survey index, not a measure of spending; compare it with real retail sales.

Building permits

Source: U.S. Census Bureau; U.S. Department of Housing and Urban Development via FRED, series PERMIT
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; August 2026 was available by 9 October 2026
Unit: thousands
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Consumer & Housing tab · no reference line · no badge

What it is

New privately owned housing units authorized by building permits, in thousands at a seasonally adjusted annual rate (Census Bureau and HUD).

Computation

Published by the Census Bureau and HUD via FRED (PERMIT); the dashboard shows it unchanged. FRED’s notes say the universe of permit-issuing places rose from 19,000 to 20,000 places with the release of 16 February 2005.

Reading it

There is no badge or reference line. From January 1960 the rate ranged from 513,000 (March 2009) to 2,419,000 (December 1972). For August 2026 the Census release of 17 September gave 1,394,000; FRED’s series, updated on 24 September, shows 1,403,000.

Use in practice

Its highest month in the 36 months before each NBER peak since 1969:

NBER peak	Highest month	Months before the peak
Dec 1969	Feb 1969	10
Nov 1973	Dec 1972	11
Jan 1980	Jun 1978	19
Jul 1981	Dec 1978	31
Jul 1990	Jan 1990	6
Mar 2001	Dec 1998	27
Dec 2007	Sep 2005	27
Feb 2020	Nov 2019	3

Caveats

Figures change after release: the August 2026 example above.

Housing starts

Source: U.S. Census Bureau; U.S. Department of Housing and Urban Development via FRED, series HOUST
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; August 2026 was available by 9 October 2026
Unit: thousands
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Consumer & Housing tab · no reference line · no badge

What it is

New privately owned housing units started, in thousands at a seasonally adjusted annual rate (Census Bureau and HUD). FRED’s notes: a start occurs when excavation begins for the footings or foundation of a building, and all units in a multifamily building count as started at that point.

Computation

Published by the Census Bureau and HUD via FRED (HOUST); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1959 the rate ranged from 478,000 (April 2009) to 2,494,000 (January 1972); in August 2026 it was 1,275,000. The Census release of 17 September 2026 put August 2.6% below July, with a range of ±12.0% attached to that change.

Use in practice

Its highest month in the 36 months before each NBER peak since 1969:

NBER peak	Highest month	Months before the peak
Dec 1969	Jan 1969	11
Nov 1973	Jan 1972	22
Jan 1980	Apr 1978	21
Jul 1981	Nov 1978	32
Jul 1990	Sep 1987	34
Mar 2001	Dec 1998	27
Dec 2007	Jan 2006	23
Feb 2020	Jan 2020	1

Caveats

Wide sampling ranges: in August 2026 the monthly changes in starts and new home sales were smaller than the ranges Census attached to them; read the trend.

New home sales

Source: U.S. Census Bureau; U.S. Department of Housing and Urban Development via FRED, series HSN1F
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; August 2026 was available by 9 October 2026
Unit: thousands
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Consumer & Housing tab · no reference line · no badge

What it is

New single-family houses sold, in thousands at a seasonally adjusted annual rate (Census Bureau and HUD). The dashboard uses it in place of the licensed NAHB builder survey of the original design.

Computation

Published by the Census Bureau and HUD via FRED (HSN1F); the dashboard shows it unchanged.

Reading it

There is no badge or reference line. From January 1963 the rate ranged from 270,000 (February 2011) to 1,389,000 (July 2005); in August 2026 it was 684,000. The Census release of 24 September 2026 put August 6.4% above July, with a range of ±19.5% attached to that change, and the seasonally adjusted number of new houses for sale at the end of August at 483,000.

Use in practice

Its highest month in the 36 months before each NBER peak since 1969:

NBER peak	Highest month	Months before the peak
Dec 1969	Oct 1967	26
Nov 1973	Oct 1972	13
Jan 1980	Mar 1977	34
Jul 1981	Oct 1978	33
Jul 1990	Jul 1989	12
Mar 2001	Nov 1998	28
Dec 2007	Jul 2005	29
Feb 2020	Jun 2019	8

Caveats

Wide sampling ranges, as for starts.

30Y mortgage rate

Source: Freddie Mac via FRED, series MORTGAGE30US
Licence: Free to reuse with a citation; data shown on the public site
Frequency: weekly · weekly, as of Thursday; 8 October 2026 was available by 9 October
Unit: percent
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Consumer & Housing tab · no reference line · no badge

What it is

Freddie Mac’s 30-year fixed-rate mortgage average in the United States, from its Primary Mortgage Market Survey, weekly. The dashboard uses it in place of the licensed MBA mortgage-applications index of the original design.

Computation

Published by Freddie Mac via FRED (MORTGAGE30US); the dashboard shows it unchanged. FRED’s notes say Freddie Mac changed the survey’s method on 17 November 2022: the weekly rate is now based on applications lenders across the country submit to Freddie Mac. Freddie Mac’s page describes the earlier survey as covering first-lien prime conventional conforming home-purchase mortgages with a loan-to-value ratio of 80%.

Reading it

There is no badge or reference line. From 2 April 1971 the rate ranged from 2.65% (7 January 2021) to 18.63% (9 October 1981); on 8 October 2026 it was 7.40%.

Use in practice

Compare it with the 10-year Treasury yield on the Credit & Rates tab and with new home sales.

Caveats

Method change in November 2022, as above.

Real retail sales — US

Source: Federal Reserve Bank of St. Louis via FRED, series RRSFS
Licence: Free to reuse with a citation; data shown on the public site
Frequency: monthly · monthly; August 2026 was available by 9 October 2026
Unit: percent
Indicator class: coincident
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Consumer & Housing tab · reference line at 0 · no badge

What it is

Real retail and food services sales: the Census Bureau’s advance retail and food services sales, deflated by the consumer price index, shown as the change over twelve months. The St. Louis Fed computes the real series (FRED RRSFS, millions of 1982–84 dollars) from the Census series RSAFS and CPI (CPIAUCSL).

Computation

Computed by the dashboard from RRSFS (Appendix A.1):

$g_t = 100\left(\frac{S_t}{S_{t-12}} - 1\right),$

for months where both are published. FRED’s CPI has no October 2025 value, so RRSFS has none, and the tile has no value for October 2025 or October 2026. FRED’s notes add that the latest month is an advance estimate from a subsample of firms, later replaced by the full Monthly Retail Trade Survey.

Reading it

The reference line is at zero; there is no badge. From January 1993 the rate ranged from −19.9% (April 2020) to +45.8% (April 2021) and was negative in 16% of months; in August 2026 it was 2.6%. In nominal terms the Census Bureau reported August 2026 sales of $773.9 billion, 6.0% above a year earlier.

Use in practice

At the NBER peaks in its history the rate was −2.6% (March 2001), −1.2% (December 2007) and +2.4% (February 2020). It was also negative in many months of 2023 and 2024 without a recession.

Caveats

Not the control group. Despite the tile’s internal id, it covers all retail and food services sales.
Deflated by the overall CPI, which includes services that retail sales don’t cover.

Markets & Uncertainty

S&P 500

Source: S&P Dow Jones Indices LLC via FRED, series SP500
Licence: Republication needs the provider’s permission; link-only on the public site (data in the local app)
Frequency: daily · daily, at the close
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Markets & Uncertainty tab · no reference line · no badge

What it is

The S&P 500 price index from S&P Dow Jones Indices. FRED’s notes: the daily index value at market close; 500 leading companies listed on the NYSE or Nasdaq, covering 75% of US equities; a price index, so it does not include dividends.

Computation

Published by S&P Dow Jones Indices via FRED (SP500); the local app shows it unchanged.

Reading it

There is no badge or reference line. The data are licensed: S&P’s terms require prior written permission for reproduction, so the public site shows only a link, and this guide quotes no values.

Use in practice

Compare it with the VIX, whose option prices are on stock indexes.

Caveats

Ten years of history. FRED’s notes say an agreement with S&P Dow Jones Indices limits FRED to 10 years of daily history for S&P series.
No dividends: as a price index it understates total returns.

VIX

Source: Chicago Board Options Exchange via FRED, series VIXCLS
Licence: Free to reuse with a citation; data shown on the public site
Frequency: daily · daily, at the close; 8 October 2026 was available by 9 October
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Markets & Uncertainty tab · no reference line · badge

What it is

The Cboe Volatility Index (VIX). FRED’s notes: it measures the market’s expectation of near-term volatility conveyed by stock index option prices.

Computation

Published by the Chicago Board Options Exchange via FRED (VIXCLS); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{VIX}_n \ge 30$	Fear — stress
$20 \le \text{VIX}_n < 30$	Elevated
$\text{VIX}_n < 20$	Calm

From 2 January 1990 the index ranged from 9.14 (3 November 2017) to 82.69 (16 March 2020), with a median of 17.6. It was “Calm” on 63% of days, “Elevated” on 29% and at 30 or above on 8%. On 8 October 2026 it closed at 15.41.

Use in practice

The highest close in each of the six highest years since 1990:

Date	Close
16 Mar 2020	82.69
20 Nov 2008	80.86
20 Jan 2009	56.65
8 Apr 2025	52.33
8 Aug 2011	48.00
20 May 2010	45.79

Around the NBER peaks since 1990 it first closed at 30 or above on 6 August 1990 (a month after the July 1990 peak), 14 April 2000 (11 months before the March 2001 peak), 15 August 2007 (4 months before the December 2007 peak) and 27 February 2020 (the month of the February 2020 peak). It also closed above 30 in years without a recession, among them 2010, 2011 and 2025.

Caveats

Expectations, not outcomes: it reads option prices, not realised volatility; the 10-year Treasury tile on the Credit & Rates tab is the realised measure for bonds.

Broad US dollar

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DTWEXBGS
Licence: Public domain; data shown on the public site
Frequency: daily · daily, released weekly on Mondays (H.10); 2 October 2026 was the latest on 9 October
Unit: index
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Markets & Uncertainty tab · no reference line · no badge

What it is

The Federal Reserve’s nominal broad US dollar index (January 2006 = 100). The Federal Reserve describes its dollar indexes as designed to help estimate the overall effects of dollar exchange-rate movements on US international trade; the broad index covers the currencies of the 26 economies whose bilateral trade with the US is at least 0.5% of the total.

Computation

Published by the Federal Reserve Board (H.10 release) via FRED (DTWEXBGS); the dashboard shows it unchanged. Following the Fed’s note of January 2019, each day’s index chains the previous day’s value by a geometrically weighted average of the changes in bilateral exchange rates,

$I_t = I_{t-1} \prod_{j=1}^{N(t)} \left(\frac{e_{j,t}}{e_{j,t-1}}\right)^{w_{j,t}},$

with $e_{j,t}$ the price of the dollar in currency $j$ and $w_{j,t}$ that economy’s share of US bilateral trade in goods and services. In the note’s 2017 weights the largest were the euro area (18.6%), China (16.2%), Canada (13.6%) and Mexico (13.3%).

Reading it

There is no badge or reference line. A rise means a stronger dollar. From January 2006 the index ranged from 85.47 (26 July 2011) to 130.04 (13 January 2025); on 2 October 2026 it was 121.38, 0.8% above a year earlier.

Use in practice

Compare it with commodity prices on the Trade & Commodities tab, which are quoted in dollars.

Caveats

Nominal: the Fed also publishes real (inflation-adjusted) indexes, monthly; this tile shows the daily nominal index.

Policy uncertainty — US (daily)

Source: Scott R. Baker, Nick Bloom and Steven J. Davis via FRED, series USEPUINDXD
Licence: Public domain; data shown on the public site
Frequency: daily · daily; 8 October 2026 was available by 9 October
Unit: index
Indicator class: overlay
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Markets & Uncertainty tab · no reference line · no badge

What it is

The daily US Economic Policy Uncertainty index of Scott Baker, Nicholas Bloom and Steven Davis, built from US newspapers.

Computation

Published by its authors via FRED (USEPUINDXD); the dashboard shows it unchanged. Their site describes the method: from US newspaper sources in the NewsBank service, count the articles that contain at least one term from each of three sets: “economic” or “economy”; “uncertain” or “uncertainty”; and “legislation”, “deficit”, “regulation”, “congress”, “federal reserve” or “white house”. Because NewsBank’s coverage grew from under 100 newspapers in 1985 to over 2,000 by 2020, the counts are normalised by the total number of articles. The method is set out in Baker, Bloom and Davis (2016).

Reading it

There is no badge or reference line. From 1 January 1985 the daily index ranged from 3.32 (7 August 2015) to 1,048.95 (2 May 2026), with a median of 92. Its 30-day average peaked at 586 (4 May 2025) and was 276 on 8 October 2026, when the daily value was 353.79.

Use in practice

Daily values jump from day to day; read the level over several weeks, or the global index, which is monthly.

Caveats

Revised: the authors update the previous 30 days each day.

Policy uncertainty — Global

Source: Scott R. Baker, Nick Bloom and Steven J. Davis via FRED, series GEPUCURRENT
Licence: Public domain; data shown on the public site
Frequency: monthly · monthly; July 2026 was the latest on 9 October 2026
Unit: index
Indicator class: overlay
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Markets & Uncertainty tab · no reference line · no badge

What it is

The Global Economic Policy Uncertainty index of Baker, Bloom and Davis, the version weighted by current-price GDP.

Computation

Published by its authors via FRED (GEPUCURRENT); the dashboard shows it unchanged. Their site: each national index reflects the share of the country’s own newspaper articles that contain terms on the economy, policy and uncertainty. To build the global index they re-normalise each national index to a mean of 100 over 1997 (or its first year) to 2015, impute missing values to get a balanced panel of 18 countries from January 1997, and take the GDP-weighted average using IMF World Economic Outlook data. The 18 countries are Australia, Brazil, Canada, Chile, China, France, Germany, Greece, India, Ireland, Italy, Japan, Russia, South Korea, Spain, Sweden, the UK and the US. (FRED’s notes still list 20 countries, adding Mexico and the Netherlands.)

Reading it

There is no badge or reference line. From January 1997 the index ranged from 47.6 (July 2007) to 620.9 (April 2025), with a median of 123. Its five highest months were April, May and March 2025, March 2026 and May 2020. In July 2026 it was 241.7.

Use in practice

Compare it with the US daily index and with the VIX.

Caveats

Newspaper-based: it counts articles, not policy events.

China & Leaders

OECD CLI — China

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:CHN
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for China, amplitude adjusted.

Computation

From the OECD’s data API (DF_CLI, area CHN); the dashboard shows it unchanged. The OECD’s description of its CLIs is in the Cycle & Growth section.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From May 1992 it averaged 100.0 and was at or above 100 in 57% of months; its highest value was 106.4 (September 2007) and its lowest 85.6 (March 2020). In September 2026 it was 97.41, down from 97.64.

Use in practice

Its 3-month change is the growth input of the China regime tile, which placed China in Stagflation in August 2026.

Caveats

Revised, as for all CLIs.

OECD CLI — India

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:IND
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for India, amplitude adjusted.

Computation

From the OECD’s data API (DF_CLI, area IND); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From April 1994 it averaged 100.0 and was at or above 100 in 49% of months; its highest value was 104.4 (February 2000) and its lowest 67.1 (April 2020). In September 2026 it was 102.00, up from 101.82.

Use in practice

Compare it with India’s CPI inflation.

Caveats

Revised, as for all CLIs.

OECD CLI — Korea

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:KOR
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for Korea, amplitude adjusted.

Computation

From the OECD’s data API (DF_CLI, area KOR); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1990 it averaged 100.0 and was at or above 100 in 49% of months; its highest value was 105.5 (September 1999) and its lowest 94.8 (March 1998). In September 2026 it was 102.53, down from 102.66.

Use in practice

Compare it with the G20 and China indicators.

Caveats

Revised, as for all CLIs.

OECD CLI — Brazil

Source: OECD Composite Leading Indicators (CC BY 4.0), series DF_CLI:BRA
Licence: CC BY 4.0; data shown on the public site
Frequency: monthly · monthly; September 2026 was available by 9 October 2026
Indicator class: leading
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · reference line at 100 · badge

What it is

The OECD’s composite leading indicator for Brazil, amplitude adjusted.

Computation

From the OECD’s data API (DF_CLI, area BRA); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\text{CLI}_n \ge 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Above trend & rising
$\text{CLI}_n \ge 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Above trend, slowing
$\text{CLI}_n < 100,\ \text{CLI}_n \ge \text{CLI}_{n-1}$	Below trend, recovering
$\text{CLI}_n < 100,\ \text{CLI}_n < \text{CLI}_{n-1}$	Below trend & falling

From January 1989 it averaged 100.0 and was at or above 100 in 50% of months; its highest value was 104.2 (January 2021) and its lowest 93.3 (April 2020). In September 2026 it was 102.36, down from 102.44.

Use in practice

Compare it with the G20 indicator.

Caveats

Revised, as for all CLIs.

CPI inflation — China

Source: International Monetary Fund, Consumer Price Index (CPI), series CHN.CPI._T.YOY_PCH_PA_PT.M
Licence: IMF data terms; data shown on the public site
Frequency: monthly · monthly; August 2026 was the latest on 9 October 2026 (the NBS released August on 10 September)
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · no reference line · no badge

What it is

Chinese consumer price inflation, % change over twelve months, from the IMF’s CPI dataset, which carries each economy’s national all-items headline index.

Computation

From the IMF data API (dataset IMF.STA CPI, China, all items, % change over 12 months); the dashboard shows it unchanged. For June, July and August 2026 it matches the rates the National Bureau of Statistics of China published: 1.0%, 0.5% and 0.8%.

Reading it

There is no badge or reference line: no official inflation target for China has been verified for this guide. From January 1994 the rate ranged from −2.2% (May 1999) to 27.7% (October 1994). Since 2000 it was negative in 41 months, most recently in September 2025, and since 2023 it has been below 1% in 86% of months. In August 2026 it was 0.8%.

Use in practice

It is the inflation input of the China regime tile, which reads its direction, not its level.

Caveats

A copy of the national figure: the NBS released August 2026 on 10 September; the IMF’s copy had it by 9 October.

CPI inflation — India

Source: International Monetary Fund, Consumer Price Index (CPI), series IND.CPI._T.YOY_PCH_PA_PT.M
Licence: IMF data terms; data shown on the public site
Frequency: monthly · monthly; July 2026 was the latest on 9 October 2026
Unit: percent
Indicator class: coincident
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: China & Leaders tab · reference line at 4 · badge

What it is

Indian consumer price inflation, % change over twelve months, from the IMF’s CPI dataset. The Reserve Bank of India’s framework: under the RBI Act, the central government, in consultation with the RBI, sets the inflation target in terms of CPI every five years. It was set at 4% with a tolerance band of 2% to 6% in August 2016 and retained in the reviews of March 2021 and of 25 March 2026, the latter for April 2026 to March 2031.

Computation

From the IMF data API (dataset IMF.STA CPI, India, all items, % change over 12 months); the dashboard shows it unchanged.

Reading it

Condition	Badge
$\pi_n < 2$	Below the RBI band
$2 \le \pi_n \le 6$	Within the RBI band
$\pi_n > 6$	Above the RBI band

The reference line is at the 4% target; the badge reads the RBI’s band. Since August 2016 the rate was within the band in 71% of months, above 6% in 23% and below 2% in 6%. In July 2026 it was 4.44%.

Use in practice

The RBI’s framework names failure as average inflation outside the band for three consecutive quarters; the badge reads single months.

Caveats

A copy, not the national release: the figures have not been checked against India’s statistics ministry for this guide.

PMI — China

Source: The tile links to www.pmi.spglobal.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: China & Leaders tab · no reference line · no badge

What it is

S&P Global’s purchasing managers’ index for China. The data are licensed, so the tile shows no figures and links to S&P Global’s PMI releases.

Reading it

The tile has no data, badge or reference line.

PMI — India

Source: The tile links to www.pmi.spglobal.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: China & Leaders tab · no reference line · no badge

What it is

S&P Global’s purchasing managers’ index for India; licensed, so the tile links to S&P Global’s PMI site.

Reading it

The tile has no data, badge or reference line.

Bubbles / Froth

Equity / GDP (Buffett) — US

Source: Board of Governors of the Federal Reserve System (US); U.S. Bureau of Economic Analysis via FRED, series NCBEILQ027S, GDP
Licence: Public domain; data shown on the public site
Frequency: monthly · quarterly; the second quarter of 2026 was available by 9 October 2026
Unit: ratio (times)
Indicator class: fragility
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Bubbles / Froth tab · reference line at 1 · badge

What it is

US nonfinancial corporations’ equity relative to the size of the economy, which the dashboard calls the Buffett indicator: the Federal Reserve’s Financial Accounts series for the corporate equities of nonfinancial corporate business (a liability, level), divided by GDP.

Computation

Computed by the dashboard (Appendix A.9):

$B_q = \frac{E_q}{1000\, Y_q},$

with $E_q$ in US$ millions (NCBEILQ027S) and $Y_q$ nominal GDP in US$ billions at an annual rate (GDP). A value of 1 means the equity equals one year of GDP.

Reading it

Condition	Badge
$B_n \ge 2$	Very stretched
$1.5 \le B_n < 2$	Elevated
$B_n < 1.5$	Moderate

The reference line is at 1. From the first quarter of 1947 the ratio had a median of 0.69 and a mean of 0.84; it ranged from 0.32 (second quarter of 1982) to 2.55 (second quarter of 2026, the latest value). It first exceeded 1.5 in the fourth quarter of 1999 (1.55) and 2 in the first quarter of 2021; it was 1.15 at the end of 2007 and 2.19 at the end of 2021. The badge would have read “Moderate” in 90% of quarters, “Elevated” in 7% and “Very stretched” in 4%.

Use in practice

Read it against its own history and with the stocks-vs-bonds tile, which sets the same equity against profits and bond yields.

Caveats

Nonfinancial corporations only: financial corporations’ equity is not in the numerator.
Quarterly and revised: FRED’s notes say the Financial Accounts may revise data and structure with each quarterly release.

Stocks vs bonds — US

Source: U.S. Bureau of Economic Analysis; Federal Reserve Board; Federal Reserve Bank of Cleveland via FRED; gap computed by this site, series NFCPATAX, NCBEILQ027S, DGS10, EXPINF10YR
Licence: Public domain; data shown on the public site
Frequency: daily · daily bond yield; quarterly profits, the second quarter of 2026 the latest on 9 October 2026
Unit: percentage points (pp)
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Bubbles / Froth tab · no reference line · badge

What it is

A comparison of two yields that the dashboard, after Ray Dalio, reads as expected real returns: the earnings yield of US nonfinancial corporations, and the real yield on 10-year Treasuries. The tile draws both lines and shows the gap between them as its value.

Computation

Computed by the dashboard (Appendix A.8):

$G_t = \mathrm{EY}_{q(t)} - \left(y^{10y}_t - \pi^{e,10y}_{m(t)}\right),$

with the earnings yield $\mathrm{EY}_q$ = after-tax profits of nonfinancial corporations (BEA, annual rate) over their corporate equities (Financial Accounts), the 10-year Treasury yield $y^{10y}$ and the Cleveland Fed’s 10-year expected inflation $\pi^{e,10y}$ . The history starts on 4 January 1982.

Reading it

Condition	Badge
$G_N < 0$	Crossed — bonds out-yield stocks
$0 \le G_N < 1$	Close to crossing
$G_N \ge 1$	Stocks out-yield bonds

In level mode the tile shows both lines, the stocks line labelled with its quarter and the bond line daily, and shades any stretch where stocks yield less than bonds; the z-score and percentile modes show the gap. From January 1982 the gap ranged from −1.64 (28 March 2002) to 6.34 (25 July 2012), with a median of 2.36. The badge read “Stocks out-yield bonds” on 78% of days, “Close to crossing” on 11% and “Crossed” on 11%. On 7 October 2026 the gap was 0.66: stocks 3.37% (second quarter of 2026), bonds 2.71%.

Use in practice

The gap was negative for spells of 20 days or more in 1987, 1990–92 and from May 1999 to June 2002 (with a break in late 2001); its last negative day was 28 June 2002. Its narrowest point since then was 0.63, on 5 October 2026.

Caveats

Dates of the profits. Each quarter’s earnings yield is dated at the quarter’s end although it is published later (see the note in Appendix A.8).
Quarterly steps: between quarterly updates only the bond line moves the gap.

Stocks vs bonds — Euro area

Source: European Central Bank (ECB Data Portal); gap computed by this site, series 1
Licence: ECB reuse terms; data shown on the public site
Frequency: daily · daily bond yield; quarterly accounts, the second quarter of 2026 the latest on 9 October 2026
Unit: percentage points (pp)
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Bubbles / Froth tab · no reference line · badge

What it is

The same comparison for the euro area, from ECB data: the after-tax earnings yield of euro-area non-financial corporations against a real 10-year AAA government yield.

Computation

Computed by the dashboard (Appendix A.8): net entrepreneurial income minus current taxes over four quarters, over the equity the corporations have issued, from the ECB’s quarterly sector accounts for the euro area of 21 countries; against the ECB’s AAA yield-curve 10-year spot rate minus the longer-term HICP expectation of the ECB Survey of Professional Forecasters. The history starts on 6 September 2004.

Reading it

Condition	Badge
$G_N < 0$	Crossed — bonds out-yield stocks
$G_N \ge 0,\ R \le 2$	Near its narrowest since year
$2 < R \le 10$	Narrow: bottom 10% since year
$10 < R \le 25$	Narrower than usual
$R > 25$	Usual range or wider

The badge ranks the latest gap among all days since the history starts. The gap has never been negative; it ranged from 4.54 (28 September 2026, the narrowest) to 9.25 (28 September 2016). On 8 October 2026 it was 4.65, narrower than on 99.8% of days since 2004: stocks 6.14% (second quarter of 2026), bonds 1.48%.

Use in practice

Read it against its own history rather than against the US tile: the dashboard’s code treats the zones’ levels as not comparable.

Caveats

Quarterly steps and dating, as for the US tile.

Stocks vs bonds — UK

Source: Office for National Statistics (Open Government Licence v3.0); Bank of England (Open Government Licence v3.0); gap computed by this site, series 1, 2
Licence: Open Government Licence v3.0; data shown on the public site
Frequency: daily · daily bond yield; quarterly accounts, the second quarter of 2026 the latest on 9 October 2026
Unit: percentage points (pp)
Indicator class: leading
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Bubbles / Froth tab · no reference line · badge

What it is

The same comparison for the UK, from ONS and Bank of England data: the after-tax earnings yield of UK private non-financial corporations against the Bank of England’s 10-year real zero-coupon gilt yield.

Computation

Computed by the dashboard (Appendix A.8), with depreciation from the ONS profitability data, which end in the second quarter of 2025; for later quarters the dashboard continues depreciation at its growth over the year before. The history starts on 31 December 1997.

Reading it

Condition	Badge
$G_N < 0$	Crossed — bonds out-yield stocks
$G_N \ge 0,\ R \le 2$	Near its narrowest since year
$2 < R \le 10$	Narrow: bottom 10% since year
$10 < R \le 25$	Narrower than usual
$R > 25$	Usual range or wider

The gap has never been negative; it ranged from 3.79 (30 March 2001) to 11.90 (29 May 2020). On 7 October 2026 it was 5.18, narrower than on 90.7% of days since 1997: stocks 7.23% (second quarter of 2026), bonds 2.05%.

Use in practice

As for the euro area, read it against its own history.

Caveats

Estimated depreciation after the second quarter of 2025, as above.
Quarterly steps and dating, as for the US tile.

Real 10Y yield — US

Source: Board of Governors of the Federal Reserve System (US) via FRED, series DFII10
Licence: Public domain; data shown on the public site
Frequency: daily · daily; 7 October 2026 was available by 9 October
Unit: percent
Indicator class: overlay
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Bubbles / Froth tab · reference line at 0 · no badge

What it is

The yield on 10-year US Treasury inflation-indexed securities (TIPS) at constant maturity, from the Federal Reserve’s H.15 release.

Computation

Published by the Federal Reserve Board via FRED (DFII10); the dashboard shows it unchanged. It is also the real leg of the 10-year breakeven: the nominal 10-year yield minus this yield.

Reading it

The reference line is at zero; there is no badge. From 2 January 2003 it ranged from −1.19% (3 August 2021) to 3.15% (21 November 2008); it was negative on 16% of days, most recently on 28 April 2022. On 7 October 2026 it was 2.92%.

Use in practice

Compare it with the bond line of the US stocks-vs-bonds tile, which uses a survey-and-model estimate of expected inflation instead of TIPS.

Caveats

Starts in 2003 on FRED.

IPO & SPAC issuance — US

Source: The tile links to www.renaissancecapital.com, www.spacinsider.com
Licence: Licensed; no data in either edition
Frequency: monthly
Indicator class: leading
On the dashboard: Bubbles / Froth tab · no reference line · no badge

What it is

A manual tile for US IPO and SPAC issuance. The dashboard has no free, republishable feed for it, so the tile links to two public trackers: Renaissance Capital’s US IPO market statistics and SPACInsider’s SPAC market statistics.

Reading it

The tile has no data, badge or reference line.

Case-Shiller price/rent — US

Source: S&P Dow Jones Indices LLC; U.S. Bureau of Labor Statistics via FRED, series CSUSHPINSA, CUSR0000SEHA
Licence: Republication needs the provider’s permission; link-only on the public site (data in the local app)
Frequency: monthly · monthly
Indicator class: fragility
Computed by: The dashboard, from published inputs (Appendix A)
On the dashboard: Bubbles / Froth tab · no reference line · no badge

What it is

The ratio of the S&P Cotality Case-Shiller US national home price index to the CPI for rent of primary residence: a measure of house prices relative to rents.

Computation

Computed by the dashboard in the local app (Appendix A.6).

Reading it

There is no badge or reference line. The home price index is licensed, so the public site shows only a link and this guide quotes no values. The two indexes have different base periods (January 2000 and 1982–84), so read the ratio’s changes and its position in its own history rather than its level.

Caveats

Licensed: link-only on the public site.

Household debt service — US

Source: Board of Governors of the Federal Reserve System (US) via FRED, series TDSP
Licence: Public domain; data shown on the public site
Frequency: quarterly · quarterly; the second quarter of 2026 was released on 22 September 2026
Unit: percent
Indicator class: fragility
Computed by: The publisher; the dashboard shows it unchanged
On the dashboard: Bubbles / Froth tab · no reference line · no badge

What it is

The Federal Reserve’s household debt service ratio: required household debt payments as a percentage of disposable personal income, seasonally adjusted. It has two parts, the mortgage ratio and the consumer-debt ratio, which sum to the total.

Computation

Published by the Federal Reserve Board via FRED (TDSP); the dashboard shows it unchanged. Starting with the publication for the second quarter of 2024, the Board moved to a new method based on credit-bureau data; the new series starts in the first quarter of 2005, and FRED’s series matches it.

Reading it

There is no badge or reference line. From the first quarter of 2005 it ranged from 9.05% (first quarter of 2021) to 15.85% (fourth quarter of 2007); in the second quarter of 2026 it was 11.11%.

Use in practice

Compare it with the 30-year mortgage rate and the real retail sales tile.

Caveats

New method: the Board’s earlier series, from 1980, used a different method; FRED now carries only the new series, from 2005.

Part IV · Appendices

Appendix A · Formulas

Every calculation the dashboard performs itself, written out. Series computed by their publishers (spreads and breakevens on FRED, the Sahm rule, the NFCI, the OECD leading indicators and so on) are described in their tiles’ sections in Part III. Each formula names the function that implements it.

A.1 Notation and alignment

$t$ is the date of an observation: a business day, week, month or quarter, as the series is published. $n$ is the latest observation in the window shown. $x_t$ is a series in its own unit; “pp” means percentage points and “bp” basis points (0.01 pp).

Carrying the last observation forward. Series published at different frequencies are combined on the union of their dates. A series observed at dates $d_1 < d_2 < \dots < d_m$ takes, at any date $\tau \ge d_1$ , its latest value published for a date on or before $\tau$ :

$x(\tau) = x_{d_j}, \qquad j = \max\{\, i : d_i \le \tau \,\}.$

The dates here are the periods the data refer to, not the days they were released (see the note at the end of A.8). (.align_locf in R/transforms.R.)

Annual rates. From a monthly index $P_t$ , $\pi_t = 100\,(P_t / P_{t-12} - 1)$ in %. How a missing month is treated depends on the tile:

US price tiles (CPI, core CPI, core PCE, PPI final demand): published months only. A month whose index, or whose index a year earlier, is missing gets no value. FRED’s CPI and core CPI have no value for October 2025, so those tiles have none for October 2025 and October 2026. (.yoy_published, .yoy_tile in R/fetch.R.)
Regime tiles: a missing month of a price index is filled in log-linearly from its neighbours before the rate is taken, and where the source publishes the annual rate itself, an isolated missing month is filled in linearly. M2 growth uses the same function; its index has no gaps. (.yoy_from_index, .regime_series in R/fetch.R.)

A.2 Display modes

Over the $N$ observations $x_1, \dots, x_N$ in the window (the public site: the last five years; the app: since “Show history since”), the z-score and percentile of $x_t$ are

$z_t = \frac{x_t - \bar x}{s}, \qquad \bar x = \frac{1}{N}\sum_{i=1}^{N} x_i, \qquad s = \sqrt{\frac{1}{N-1}\sum_{i=1}^{N} (x_i - \bar x)^2},$

with $s$ set to 1 if it is zero or undefined, and

$p_t = \frac{100}{N}\left( \#\{i : x_i < x_t\} + \frac{\#\{i : x_i = x_t\} + 1}{2} \right),$

the rank of $x_t$ among the window’s values (ties share their average rank) as a share of $N$ . The highest value has $p = 100$ . Reference lines become 0 for z-scores and 50 for percentiles. The headline’s change versus prior is $v_n - v_{n-1}$ in the mode shown, where $v$ is $x$ , $z$ or $p$ . Badges always read the level $x_n$ . (apply_mode, mode_ref in R/display.R; change_tag in R/tiles.R.)

A.3 Freshness and the recession count

The freshness dot is amber when $\text{today} - d_n > L_f$ days, where $d_n$ is the date of the latest observation and $L_f$ is 7 for daily series, 12 weekly, 45 monthly, 130 quarterly, 45 for any other frequency, and never for event series such as policy rates. (is_stale in R/display.R.)

The sidebar’s recession count: with $R$ the Recession Watch tiles that have a badge and data,

$\text{signals} = \#\{\, i \in R : \text{badge}_i \text{ is red} \,\} \text{ of } |R|, \qquad \text{cautions} = \#\{\, i \in R : \text{badge}_i \text{ is amber} \,\}.$

A.4 Net liquidity

The Federal Reserve’s total assets less two of the factors that absorb reserve funds in its H.4.1 release: the Treasury’s general account and overnight reverse repos.

$\mathrm{NL}_\tau = W_\tau - T_\tau - 1000\, Q_\tau \quad (\text{US\$ millions}),$

where $W$ is the Fed’s total assets (WALCL, Wednesday level, $ millions), $T$ the Treasury General Account (WDTGAL, Wednesday level, $ millions) and $Q$ overnight reverse repurchase agreements (RRPONTSYD, daily, $ billions, hence the factor 1000). Each is carried forward to every date any of them is published, so the series moves daily with the reverse repo although its other two inputs are weekly.

A.5 HY − IG

$D_t = \text{OAS}^{HY}_t - \text{OAS}^{IG}_t \quad (\text{pp}),$

the option-adjusted spread of the ICE BofA US High Yield Index (BAMLH0A0HYM2; bonds rated BB or below) minus that of the ICE BofA US Corporate Index (BAMLC0A0CM; bonds rated BBB or better, investment grade), both daily in %, per FRED’s notes. Licensed data: computed in the local app, link-only on the public site.

A.6 Case-Shiller price to rent

$\mathrm{PR}_t = \frac{H_t}{K_t},$

with $H$ the S&P Cotality Case-Shiller U.S. National Home Price Index (CSUSHPINSA, January 2000 = 100, not seasonally adjusted) and $K$ the CPI for rent of primary residence (CUSR0000SEHA, 1982–84 = 100, seasonally adjusted), both monthly. The two indices have different base periods (January 2000 and 1982–84), so the ratio’s level depends on those choices; its percentage changes and its position in its own history don’t. Licensed data: link-only on the public site.

A.7 Realized 10-year Treasury volatility

A free stand-in for the licensed ICE BofA MOVE index. With $y_t$ the 10-year constant-maturity Treasury yield (DGS10, %) on business day $t$ and $\Delta y_t = 100\,(y_t - y_{t-1})$ its daily change in basis points (between consecutive days with a published yield),

$\sigma_t = \sqrt{252}\;\operatorname{sd}\left(\Delta y_{t-20}, \dots, \Delta y_t\right) \quad (\text{bp, annualised}),$

the sample standard deviation ( $N - 1$ in the denominator) of the last 21 daily changes, scaled to a year of 252 trading days. It measures how much the yield has actually moved over the last 21 trading days.

A.8 Stocks vs bonds

The dashboard’s code describes these tiles as comparing expected real returns, after Ray Dalio: the earnings yield of the corporate sector against a real 10-year government bond yield. For each day $t$ with a bond yield,

$G_t = \mathrm{EY}_{q(t)} - r_t \quad (\text{pp}),$

where $\mathrm{EY}_q$ is the earnings yield for quarter $q$ (in %), dated at the quarter’s last day, $q(t)$ the latest quarter that ended on or before $t$ , and $r_t$ the real 10-year yield in %. Where the real yield is a nominal yield $y_t$ less expected inflation $\pi^e$ , the latest expectation dated on or before $t$ is used: $r_t = y_t - \pi^e_{j(t)}$ .

US (stocks_vs_bonds_real), from FRED:

$\mathrm{EY}_q = 100 \cdot \frac{1000\, \Pi_q}{E_q}, \qquad r_t = y^{10y}_t - \pi^{e,10y}_{m(t)},$

with $\Pi_q$ BEA’s nonfinancial corporate business profits after tax, without inventory valuation and capital consumption adjustments (NFCPATAX, a seasonally adjusted annual rate in $ billions), $E_q$ the Federal Reserve Board’s nonfinancial corporate business corporate equities liability, level (NCBEILQ027S, $ millions, Financial Accounts), $y^{10y}$ the 10-year Treasury constant-maturity yield (DGS10) and $\pi^{e,10y}_{m(t)}$ the Federal Reserve Bank of Cleveland’s 10-year expected inflation for the latest month $m(t)$ (EXPINF10YR; FRED’s notes say the Cleveland Fed estimates it with a model that uses Treasury yields, inflation data, inflation swaps and survey-based measures of inflation expectations).

Euro area (stocks_vs_bonds_euro_area), from the ECB’s quarterly sector accounts for non-financial corporations in the euro area (area code I10, “Euro area 21 (fixed composition) as of 1 January 2026”):

$\mathrm{EY}_q = 100 \cdot \frac{\sum_{k=0}^{3} \left(\mathrm{B4N}_{q-k} - \mathrm{D5}_{q-k}\right)}{\mathrm{F51}_q}, \qquad r_t = y^{AAA,10y}_t - \pi^{e,LT}_{s(t)},$

with $\mathrm{B4N}$ the net entrepreneurial income and $\mathrm{D5}$ the current taxes payable of non-financial corporations (flows flagged by the ECB as neither seasonally nor working-day adjusted; the dashboard sums four quarters), $\mathrm{F51}$ the equity they have issued (end of quarter), $y^{AAA,10y}$ the ECB’s AAA yield curve 10-year spot rate (daily; government bonds of all issuers rated triple A, Svensson model) and $\pi^{e,LT}_{s(t)}$ the average point forecast for longer-term HICP inflation in the latest ECB Survey of Professional Forecasters (five calendar years ahead in the third- and fourth-quarter rounds, four in the first and second).

UK (stocks_vs_bonds_uk), from ONS series for private non-financial corporations and the Bank of England:

$P_q = \mathrm{BPI}_q + \mathrm{DIV}_q + \mathrm{REI}_q - \mathrm{CFC}_q - \mathrm{TAX}_q, \qquad \mathrm{EY}_q = 100 \cdot \frac{\sum_{k=0}^{3} P_{q-k}}{\mathrm{EQ}_q},$

where, in the ONS titles, $\mathrm{BPI}$ is the gross balance of primary income (B.5g, RPBO), $\mathrm{DIV}$ distributed income of corporations paid (D.42, ROCH), $\mathrm{REI}$ reinvested earnings on foreign direct investment paid (D.43, ROCI) and $\mathrm{TAX}$ taxes on income (D.51, RPLA), all seasonally adjusted in £ million; $\mathrm{CFC} = \mathrm{GOS} - \mathrm{NOS}$ is gross minus net operating surplus (LRWL − LRWM, from the ONS profitability data), which the dashboard uses as depreciation; and $\mathrm{EQ} = \mathrm{NLBZ} + \mathrm{NLCA} + \mathrm{NLCB}$ is listed UK shares, unlisted UK shares and other UK equity issued (levels, not seasonally adjusted). The profitability data end at 2025 Q2 ( $q^*$ ; as of October 2026), so for later quarters the dashboard continues depreciation at its growth over the year before:

$\mathrm{CFC}_{q^*+h} = \mathrm{CFC}_{q^*} \left(\frac{\mathrm{CFC}_{q^*}}{\mathrm{CFC}_{q^*-4}}\right)^{h/4}, \qquad h = 1, 2, \dots$

The bond side is the Bank of England series IUDMRZC, “Yield from British Government Securities - 10 year Real Zero Coupon”; no expectation is subtracted.

Reading against history (euro area and UK). Their levels aren’t comparable with the US tile’s, so their badges rank the latest gap among all days since each history starts (1999 for the euro area; for the UK, when the equity data begin):

$R = \frac{100}{N} \sum_{s=1}^{N} \mathbf{1}\left[G_s \le G_N\right],$

the share of days with a gap as narrow as today’s or narrower. (.gap_history.)

US badge:

Condition	Badge
$G_N < 0$	Crossed — bonds out-yield stocks
$0 \le G_N < 1$	Close to crossing
$G_N \ge 1$	Stocks out-yield bonds

Euro-area and UK badge (the year is when the history starts):

Condition	Badge
$G_N < 0$	Crossed — bonds out-yield stocks
$G_N \ge 0,\ R \le 2$	Near its narrowest since year
$2 < R \le 10$	Narrow: bottom 10% since year
$10 < R \le 25$	Narrower than usual
$R > 25$	Usual range or wider

Dating note: earnings yields enter the history from the last day of the quarter they describe, although they are published only after the quarter has ended. The chart therefore shows each quarter’s value earlier than it was known at the time; the latest value uses the latest published quarter, whose date the tile shows.

A.9 Equity to GDP (Buffett indicator)

$B_q = \frac{E_q}{1000\, Y_q} \quad (\text{times}),$

with $E_q$ the Federal Reserve Board’s nonfinancial corporate business corporate equities liability, level (NCBEILQ027S, $ millions, Financial Accounts) and $Y_q$ gross domestic product (GDP, a seasonally adjusted annual rate in $ billions, BEA). A reading of 1 means the equity equals one year of GDP. Only nonfinancial corporations’ equity is included.

A.10 Growth and inflation regimes

After Ray Dalio’s four economic environments, as the dashboard’s code describes it: growth and inflation each either rising or falling. For each month $t$ , with $\text{CLI}_t$ the OECD composite leading indicator (amplitude adjusted) and $\pi_t$ headline inflation year on year in %:

$g_t = \text{CLI}_t - \text{CLI}_{t-3}, \qquad m_t = \pi_t - \frac{1}{12} \sum_{k=0}^{11} \pi_{t-k},$

growth momentum $g_t$ in index points and inflation momentum $m_t$ in pp (inflation against its own average over the past twelve months, the current month included). Above zero counts as rising:

Condition	Badge
$g_t > 0,\ m_t \le 0$	Goldilocks
$g_t > 0,\ m_t > 0$	Reflation
$g_t \le 0,\ m_t > 0$	Stagflation
$g_t \le 0,\ m_t \le 0$	Disinflationary slowdown

“Since” is the first month of the current regime’s unbroken spell.

Inputs by economy.

Tile	Leading indicator	Inflation
US	OECD CLI, United States	CPI, all items (BLS via FRED, CPIAUCSL), annual rate per A.1
Euro area	GDP-weighted CLI of Germany, France, Italy, Spain (below)	HICP annual rate, euro area in its changing composition (Eurostat, EA)
UK	OECD CLI, United Kingdom	CPI annual rate (ONS, D7G7)
Japan	OECD CLI, Japan	CPI year on year (IMF CPI dataset)
China	OECD CLI, China	CPI year on year (IMF CPI dataset)

Euro-area leading indicator. The OECD’s leading-indicator dataset (DF_CLI) has no euro-area series (queries for the areas EA20, EA19 and EA return no data, October 2026), so the dashboard weights the indicators for Germany, France, Italy and Spain by nominal GDP:

$\text{CLI}^{EA}_t = \sum_{c \in \{DE, FR, IT, ES\}} w_c\, \text{CLI}^{c}_t, \qquad w_c = \frac{\text{GDP}_c}{\sum_{c'} \text{GDP}_{c'}},$

with GDP at current prices for the latest year Eurostat has for all four, over the months all four indicators cover.

The quadrant. It plots $(g_t, m_t)$ for the months in the period the regime strip shows, the last twelve dark. Its axes are symmetric so the origin stays centred: $\pm 1.3 \max(0.3, \max_t |g_t|)$ across and $\pm 1.3 \max(0.3, \max_t |m_t|)$ up, over the months plotted.

Appendix B · References

Baker, S. R., Bloom, N. and Davis, S. J. (2016). Measuring economic policy uncertainty. Quarterly Journal of Economics 131(4), 1593–1636.
Benigno, G., di Giovanni, J., Groen, J. J. J. and Noble, A. I. (2022). The GSCPI: a new barometer of global supply chain pressures. Federal Reserve Bank of New York Staff Reports, no. 1017.
Brave, S. A. and Butters, R. A. (2011). Monitoring financial stability: a financial conditions index approach. Federal Reserve Bank of Chicago, Economic Perspectives 35(1).
Chauvet, M. (1998). An econometric characterization of business cycle dynamics with factor structure and regime switching. International Economic Review 39(4), 969–996.
Chauvet, M. and Piger, J. (2008). A comparison of the real-time performance of business cycle dating methods. Journal of Business & Economic Statistics 26, 42–49.
Estrella, A. and Trubin, M. R. (2006). The yield curve as a leading indicator: some practical issues. Federal Reserve Bank of New York, Current Issues in Economics and Finance 12(5).
Sahm, C. (2019). Direct stimulus payments to individuals. The Hamilton Project, policy proposal, 16 May 2019, part of Recession Ready: Fiscal Policies to Stabilize the American Economy.