How to read this site
Every page opens with its answer: one sentence with its number, the key figures beside the lead chart, up to five takeaways and what would change the view. This page shows where each of those sits, what the words mean, and how to check any number you doubt.
Read a page in two minutes
Every page runs in the same order: the answer, then the evidence, then the appendixThe parts of a page, in the order you meet them
- The head. The conclusion as the title — one sentence with its key number — and the page's question beneath it, then the date the data run to.
- Key data. The numbers that decide the reading, the lead one large; each with one comparison and the date it is for.
- The lead exhibit. The chart a report would open with, beside the numbers, with its comparison drawn on it.
- In brief. Up to five takeaways, one claim each; “→ Exhibit n” takes you to the card it was read from.
- What would change this view. The test, the level that would fail it and where it stands today; the date or release that decides it is printed under Decided by (on hover on a wide screen).
- A numbered section. Its title is the section's answer; its question is printed beneath.
- An exhibit. “Exhibit n” and a title that says what is plotted, the subtitle beneath it with the units and the period, the source line underneath and the card's code in the corner.
- A range chip. What a printed range is — the spread of the inputs, past misses, named paths and so on. Never a probability.
- The Appendix. Closed tabs at the foot: Model (the equations), Data, Checks and Sources.
The sections come in a fixed order
| Step | What is in it |
|---|---|
| In brief | The takeaways and what would change the view, with the reasoning behind each point in a fold. |
| The headline | The measure itself and its history; on the Drivers and Outlook pages, the page's lead estimate with what was observed beside it. |
| What is behind it | One level down: components, contributions, the cross-section. |
| The shape of it | The models, each with its record beside it. |
| The wider view | Cross-checks, published counterparts, and where the argument goes next. |
| Detail | Every series in a table, robustness, reconciliations. Closed. |
| Appendix | Model, Data, Checks and Sources, as tabs. Closed. |
What the Detail and Appendix folds contain
Everything a page holds is on the page; a fold opens where you stand. Detail holds what you would go looking for on purpose: every series in a table, the explorers, the reconciliation to the September report where a page has one. The Appendix opens with the model as equations, then the data and its dates, the checks, and the sources.
“Sources & attributions” in the footer opens the Appendix of whatever page you are on: the shortest route from a number you doubt to the check behind it.
The four things every page is made of
A tile, a takeaway, a row of what would change the view, and a range with its kind — each a real one from today's pagesA tile: one measure, one comparison, one date
- The label says what is measured, in plain words.
- The value is the latest reading, at the precision its publisher prints.
- One comparison and the date — the change, the 2015–19 average or where it sits in its own history — and the period the value is for.
- A chip, only when the page has a rule for one: a verdict in words (above breakeven, crowded), never a number.
A takeaway: one claim, a date, and where it came from
A takeaway makes one claim, carries at most two figures beside a date, and runs to twenty-five words at most. The link at its end names the exhibit the sentence was read from: follow it and the chart or table is there. The first takeaway answers the page's question; the reasoning behind each point is the fold beneath them.
What would change the view: a test you can look up
| What would change this view | Trigger | Today |
|---|---|---|
| Core PCE momentum falls | core, six-month annualised < 2.36% (halfway to the 2015–19 average) | +2.92% (Sep 2026)not crossed |
Each row is a condition, not a caveat: a measure, the level or change that would make the view wrong with its unit, and today's value with where it stands — not crossed, within one typical move (a month's ordinary change could cross it), crossed, pending or not tracked here. The thresholds are set from the data on every update, so the box says what is true now. If crossing it would not change your mind, the page is not telling you anything new.
A range and its kind: what the spread measures
S&P 500: the model's fair value runs from 2,912 to 6,386 spread of inputs
| The chip says | What the range is |
|---|---|
| spread of inputs | The spread of the estimated inputs (usually their 10th–90th percentiles) carried through the model — how much the answer moves when the inputs move. Not a probability. |
| estimation error | The model's own estimation error carried through — the uncertainty of the fitted coefficients. A confidence band, not a forecast range. |
| assumed values | The values declared for an assumed input, swept from one end to the other. A choice, not an estimate, and not a probability. |
| other models | The same question answered by different models or methods — the disagreement between them. |
| past misses | The model's past misses: how far its earlier readings were from what happened. A width, not a direction. |
| named paths, no odds | Named paths, each a stated set of inputs. No odds are attached to any of them. |
| out-of-sample record | The error of the rule over its out-of-sample record — what it missed by when it was not fitted to the data. |
| simulated shocks | Simulated shocks: the spread of many drawn paths from a fitted model (a volatility model's cone, a factor model's projection). Model-conditional; not a probability of any outcome. |
A card prints one chip at most, and a tile none. Whatever the kind, a range is not the chance that the outcome lands inside it: the site prints no odds without the model that makes them.
Find your way
Four ways through the site, the argument the pages make, the map, the search and the code on every cardFour readers, four ways through
- New to thisThe story in six pages.
Home, then Macro, PCE (the inflation measure the Fed targets), Payrolls, Policy rates and the Cross Asset overview — with the word list open beside them.
- EconomistOne release at a time.
Start from Macro and read block by block: what the release said, what is behind it, what this site's models make of it. The nowcasts in Outlook, from This quarter's growth, are the forward end of that read.
- StrategistTop-down, in order.
Home for the view in one screen, then the Cross Asset overview, Drivers, the Thematic views and Outlook, in the order of the argument beside this card.
- BankerThe valuation chain.
Earnings, then Valuation — the discount rate, the fair value and how sensitive it is — then Range and scenarios and What is priced.
The pages are six stages of one argument
State — where the economy and markets stand: Macro, Performance, What is priced. Drivers — what is moving prices, and in what regime: Correlations & regime, Risk sentiment, the Thematic views. Expectations — growth, inflation, hiring and earnings, and the discount rate: this site's nowcasts, Earnings, Long-run growth. Transmission — how a change in those reaches the economy and the assets: Transmission, Sources & rate sensitivity, The hike's transmission. Scenarios — what named paths would do: Range and scenarios. Outlook — what each asset is worth and how wide the range is: Valuation, Fair value board, Range record and stress tests, Playbook & ranges. Every forward-looking page — the nowcasts, the scenarios, the ranges and the playbook — sits in the Outlook section, in the order of this argument. The loop closes on the front page, which scores every view against what the data went on to do.
Each block index opens with the block's answer, its key numbers and one paragraph on how its pages connect; each page ends its takeaways with “Next:”, the next page of the argument and why.
The map: Home, the overviews, the blocks and their pages
| Where | What it holds | Pages, in reading order |
|---|---|---|
| Home | ||
| Home | Where the US economy, policy and markets stand today, in one screen. | |
| Macro | ||
| Macro overview | Where is the US economy against the Fed’s mandate, and which way is it moving? | |
| Prices | inflation, measured and expected, and which prices a rate can reach | CPI · PPI · PCE · Expectations · Sources & rate sensitivity |
| Growth | output, the quarter still open and trade | GDP & GDI · Trade balance |
| Consumer | spending, income, credit, balance sheets and stress | Retail sales · Income & outlays · Balance sheet · Credit · Asset quality |
| Business | output, orders, stocks and margins | Industrial production · Orders & shipments · Inventories · Profits & margins |
| Housing | the pipeline, new-home supply, prices, rents and financing | Construction pipeline · New-home supply · Prices & rents · Financing & affordability |
| Labour | jobs, slack, flows and pay | Payrolls · Unemployment · Claims · Openings & quits · Wages |
| Credit | lending, spreads and conditions | Bank lending · Corporate spreads · Financial conditions |
| Policy & rates | the policy decision, the curve, the balance sheet and money | Policy rates · Treasury curve · Inside the 10-year · Policy rules · Balance sheet · Money supply |
| Government | borrowing, receipts and policy rules | Budget & debt · Receipts & taxes · Treasury supply |
| Energy | oil and petroleum | Oil & petroleum |
| Thematic | ||
| Thematic overview | Ten views, each with the evidence for it and what would change it. | |
| Energy and conflict | Gulf chokepoints · Russia–Ukraine diesel war · Oil and diesel reserves | |
| Policy and politics | The decision vs the source · The hike's transmission · US midterms 2026 | |
| Structural shifts | Housing supply 2027–28 · Private credit & life insurers · AI and labour · AI capex | |
| Cross Asset | ||
| Cross Asset overview | Where are markets, what moves them, and what is each asset worth? | |
| State | where markets stand: performance, volatility, positioning and what is priced | Performance · Volatility · Positioning · What is priced |
| Volatility & positioning | Skew & term structure · Dealer positioning | |
| Drivers | why: the regime, the appetite for risk, what explains returns | Risk monitor · Correlations & regime · Risk sentiment |
| Return analysis | Sources of return · Factor analysis | |
| Fundamentals | the fundamentals: what the index earns, and how fast the economy and earnings have grown | Earnings · Payout & reinvestment · Profitability & balance sheet · Composition · Long-run growth |
| Fair value | what each asset is worth: valuation, the fair value board and the record | Valuation · Statements value · Fair value board · Valuation history |
| Sectors & styles | Sector earnings · Sector valuation · Growth against value · Method and record | |
| Outlook | ||
| Outlook overview | What do this site’s models expect, and how wide is the range around it? | |
| Ledger | the market, the Committee and this site side by side, with the gap and what would close it | Expectations ledger |
| Macro expectations | this site's nowcasts of growth, inflation and hiring, beside the published ones | This quarter's growth · Inflation nowcast · Labour nowcast |
| Thematic impact | each theme's branches as moves in growth, inflation and jobs | Thematic impact · Reserve drawdown simulation · Brent price simulation · Producers and refills to 2029 |
| Synthesis | the named paths for policy and oil, with no odds, and what each does to every asset | Range and scenarios · Range record and stress tests · Earnings outlook · Discount rate · Sector & style outlook |
| Transmission | how a move in oil, rates or credit reaches what the market expects | Transmission |
| Asset outlook | the range for each asset, with no view and under each named path | Co-movement outlook · Options outlook |
| Playbook & ranges | what followed states like this one, and each asset's range | Playbook & ranges |
| Companies | ||
| Companies overview | ||
Search, the Menu, and the code on every card
Search. The search button in the bar (on a phone, inside Menu) opens a box that finds a page by its name, a section by a word of its question, a term from the list below, or a card by its code; a result inside a closed fold opens the fold when you follow it. The Menu lists every section, block and page; on a phone the button at the left of the bar, which carries the page's name, opens the block's pages and this page's contents. The trail under the bar names where you are, with the block's tabs and the page's contents.
The code. Every chart and table sits in a card with a code of its own in the corner: the page's name, then the card's number on it — rates.policy-4. To cite an exhibit, give its code; the code is also a link that puts the card's own address in the browser bar. Codes stay put: a card added later takes the next free number, so a code written down today names the same card tomorrow.
The words, by topic
Every term the pages use, in plain words, grouped by what it is about — each one reachable from search and from its underlined first use on a pageValuation
| Term | In plain words | Where you meet it |
|---|---|---|
| bufferon the pages: “gap to fair value” | The move in the index that would take it to the site's fair value; negative means the index sits above it. | Valuation; Distributions; Cross-Asset Overview |
| priced lineon the pages: “what the price fixes” | The one number a price pins down: the premium less growth. | Valuation |
| same-regime premiumon the pages: “the premium in past months like this one” | The median implied equity risk premium in the past months when stocks and bonds moved together as they do now, read from the valuation history on its own rules. The Valuation page prints it only as a cross-check beside the yield-gap premium; the model’s value does not use it. | Valuation; Valuation history |
| bridge | The average gap between national-accounts profits and index earnings per share. | Earnings |
| wedge | How much faster profits (or EPS) have grown than nominal GDP, on average. | Earnings |
| CAPE, ERP, H-model, implied growthon the pages: “implied equity risk premium (ERP)” | Cyclically adjusted P/E / the equity premium over the 10-year / a two-stage growth model that fades near-term growth to a long-run rate / the growth today's price implies at an assumed premium. | Valuation |
| identity, the valuation identityon the pages: “the two-stage fair-value model (the valuation formula)” | The formula that turns earnings, growth and a required return into a price; two-stage because near-term growth fades over some years to a long-run rate. | Valuation; Scenarios; Asset outlook |
| grade A / B / Con the pages: “record: strong / fair / unproven” | How well a fair-value method has predicted returns out of sample: strong, fair or unproven, with the criterion in the Appendix. | Board; Playbook |
| out-of-sample R² (R²_OS)on the pages: “out-of-sample R²” | How much better a forecast did than a simple benchmark on data it had not seen; above zero beats the benchmark (here, a forecast of no change). | Inside the 10-year; Valuation history |
| implied equity risk premium (ERP) | The return over the 10-year Treasury that makes the valuation model return today’s price: what the market is paying for holding equity risk. The earnings yield less a bond yield is a spread, not this premium. | Valuation; Valuation history; Board |
| transition period | The years over which near-term earnings growth fades in a straight line to terminal growth. | Valuation |
| the model’s value | What the two-stage fair-value model makes the index worth on its stated inputs: a model’s output, not a target or a forecast. | Valuation; Board; Valuation history |
| football field | Fair values side by side, one horizontal bar per method, premium basis or asset, with today’s price as a line: the spread of the bars is how much the answer depends on the choice. | Valuation; Board; Playbook & targets |
| record unproven | A valuation measure whose record of predicting returns is too short to tell it apart from a simpler benchmark: neither shown to work nor shown to fail. | Valuation history; Board |
| operating and reported earningson the pages: “operating (consensus) · reported (GAAP)” | Reported earnings follow accounting rules and carry write-downs and one-offs; operating earnings, which analysts forecast, leave most of those out and run higher. The two are never spliced into one series. | Earnings |
| the bridge from profits to EPSon the pages: “the gap between EPS and profits growth” | How much faster or slower index earnings per share have grown than national-accounts profits on average; the fan of profits is shifted by it to reach EPS. | Earnings; Long-run growth |
| unobserved-components modelon the pages: “a trend-and-cycle model” | A model that splits a growth series into a slow trend, a cycle around it and noise, none of them observed directly, estimated together with the Kalman filter. | Earnings; Long-run growth |
| cost of equityon the pages: “cost of equity (on the valuation pages); required return elsewhere” | The return investors require to hold stocks — the 10-year Treasury yield plus the equity risk premium — and so the rate at which the two-stage model discounts future earnings. | Valuation; Fair value board |
| payout | The share of a year's earnings that companies hand to shareholders, as dividends and as buybacks of their own shares; the rest is retained. | Payout & reinvestment; Valuation |
| net buybacks | Shares bought back less shares issued — to employees, in acquisitions; what actually shrinks the share count and lifts earnings per share. | Payout & reinvestment |
| retention | The share of earnings a company keeps rather than paying out: 100% less the payout. | Payout & reinvestment; Long-run growth |
| reinvestment rate | The share of earnings put back into the business — capital spending and acquisitions net of depreciation — rather than paid out or held as cash. | Payout & reinvestment |
| DuPont | Return on equity written as three parts multiplied: the net margin on sales, asset turnover (sales per dollar of assets) and leverage (assets per dollar of equity). | Profitability & balance sheet |
| fundamental growth | Retention × return on equity: the growth of earnings that reinvesting what is kept, at the return the company earns, would carry (Damodaran's measure). | Payout & reinvestment; Long-run growth |
| sustainable payout | 1 − growth ÷ return on equity: the most a company can pay out and still grow at a given rate from reinvestment alone. | Payout & reinvestment |
| terminal growthon the pages: “terminal growth (the assumed long-run rate)” | The rate earnings are assumed to grow at for ever once near-term growth has faded — the second stage of the two-stage fair-value model; here the Committee’s longer-run nominal GDP growth, an assumption, not an estimate. | Earnings; Long-run growth; Valuation |
| yield-gap premiumon the pages: “the yield-gap premium” | The equity risk premium the Valuation page reads the model’s value at: today’s implied premium plus the distance from today’s yield gap (the earnings yield less the 10-year) to its median in past months when stocks and bonds moved together. An assumption that the gap reverts, not a median of past premia. | Valuation; Board; Playbook |
| yield gapon the pages: “the yield gap” | The trailing earnings yield less the 10-year Treasury yield: a spread between two yields, not an equity risk premium, because it leaves out growth. | Valuation |
| forward earnings over the five-year averageon the pages: “forward EPS ÷ five-year average” | A basket’s next-twelve-month earnings per share divided by the average of its last five years of trailing earnings. Above one, forward earnings run ahead of what the basket earned on average over five years. | Sector earnings; Method and record |
| normal level | The forward P/E a basket’s own model says it returns towards, set by rates and the regime; the gap to it is today’s multiple against that level. | Sector valuation; Method and record; Sector & style outlook |
Rates and the Fed
| Term | In plain words | Where you meet it |
|---|---|---|
| reach, reachable, beyond reach, perverse, only via asset priceson the pages: “rate-sensitive / partly rate-sensitive / priced off asset values / raised by a hike / not rate-sensitive” | Whether a higher funds rate can slow that price within about two years. | Sources & rate sensitivity; Policy reach; How a hike transmits |
| ACM, Kim–Wright, term premiumon the pages: “term premium” | The extra yield for holding a 10-year rather than rolling short bills, on the New York Fed's / the Board's model. | Treasury curve; What is priced; Valuation |
| GSWon the pages: “the Board's zero-coupon curve” | The Board's zero-coupon Treasury curve (Gürkaynak, Sack and Wright). | Treasury curve; How a hike transmits |
| 5y5yon the pages: “5y5y breakeven / 5y5y nominal forward” | Two meanings: the five-year breakeven inflation rate five years ahead, and the five-year nominal Treasury rate five years ahead. | Expectations; Treasury curve; What is priced |
| 1y1yon the pages: “the one-year rate one year ahead (1y1y)” | The one-year Treasury rate one year ahead, read from today's curve. | What is priced; Policy rates; How a hike transmits |
| IORB, ON RRP, SOFR, EFFR, TGA, DTS, SRFon the pages: “the rate the Fed pays on reserves (IORB); not the policy rate, which is the fed funds target range” | The rate the Fed pays on reserves / the Fed's reverse-repo rate / the overnight repo rate / the traded fed funds rate / the Treasury's cash account at the Fed / the Daily Treasury Statement / the Standing Repo Facility. | Policy rates; Fed balance sheet; What is priced |
| bid-to-cover, dealer share, take-down | Bids per dollar sold / the share dealers were left with / who took the auction. | Treasury supply |
| completion | The share of the move from a start date to a path's fair value that has already happened; flagged done · under way · overshot · moved away. | What is priced |
| fair value of a path | The 10-year a policy path justifies at today's term premium (the funds rate held at today's setting, or raised once, through 2028, then the Committee's longer-run rate). | What is priced; How a hike transmits |
| SEP, SEP pathon the pages: “the Committee’s September projections” | The Federal Open Market Committee’s Summary of Economic Projections: each participant’s view of the policy rate, growth, unemployment and inflation over the next three years and the longer run; the “path” is the median rate by year. | What is priced; Policy rules; Scenarios; Policy rates |
| hold pathon the pages: “the on-hold path (the policy rate kept where it is)” | The path on which the Fed keeps its policy rate where it is through 2028 and then moves to the Committee’s longer-run rate. | Scenarios; Valuation; What is priced |
| bear flattening | Yields rising with short rates rising faster than long ones, so the gap between them narrows; bull steepening is the opposite move. | Treasury curve |
| 2s10s, 3m10yon the pages: “the 2-year to 10-year gap; the 3-month to 10-year gap” | The 10-year Treasury yield less the 2-year (or less the 3-month bill): the slope of the curve, negative when short rates sit above long ones. | Treasury curve |
| bootstrap (the curve) | Building forward rates one maturity at a time from the par yields Treasury publishes, so that each forward is consistent with the yields before it. | Treasury curve; What is priced |
| Taylor ruleon the pages: “a policy rule” | A formula that gives a policy rate from inflation’s distance from target and output’s (or unemployment’s) distance from normal; the site prints several versions and their prescriptions beside the actual rate. | Policy rules |
| r* (r-star)on the pages: “the neutral real rate (r*)” | The real short rate that neither stimulates nor restrains the economy once shocks fade; an estimate, never observed. | Policy rules; Treasury curve |
| corridoron the pages: “the Fed’s rate corridor” | The band the overnight rate trades in: the rate the Fed pays on reserves at the top, its reverse-repo rate at the bottom. | Policy rates |
| repo tail | The gap between the overnight repo rate and the Fed’s rate on reserves; a rising tail says cash is getting scarce. | Policy rates; Fed balance sheet |
| competitive award, take-down | At a Treasury auction, the share of the sale that went to bidders naming a yield, and who ended up holding it (dealers, funds, foreign buyers). | Treasury supply |
| expected policy path (10-year average) | The average short-term rate investors expect over the next ten years; the 10-year yield less the term premium. Nobody observes it, so each model estimates its own. | Inside the 10-year; Treasury curve; Policy rules |
| zero-coupon yield | The yield on a bond that pays only at maturity; the Board estimates a whole curve of them from traded Treasuries, and the models split those. | Inside the 10-year; Treasury curve |
| decision (FOMC)on the pages: “the decision” | What the Federal Open Market Committee set at its meeting: the target range for the federal funds rate, with the vote, as its statement gives them. | Policy rates |
| fed funds target rangeon the pages: “the fed funds target range: the policy rate” | The range the Committee sets for the overnight rate at which banks lend reserves to each other. It is what this site means by the Fed’s rate or the policy rate, and it is printed as a range. The rate the Fed pays on reserves (IORB) is set inside the range and holds the market rate there; it is a detail of the decision, not a second policy rate. | Policy rates; What is priced; Home |
| breakeven inflation | A nominal Treasury yield less the inflation-protected (TIPS) yield of the same maturity: the inflation rate at which the two would pay the same, read as expected inflation plus a premium for bearing it. | Expectations; Treasury curve |
| par yield | The yield at which a Treasury note or bond with a coupon would trade at exactly its face value — the yield Treasury publishes for each maturity every day. The site's 10-year is this daily yield unless another basis is named beside it (a monthly average, a zero-coupon yield). | Treasury curve; Home; Financing & affordability |
The economy's data
| Term | In plain words | Where you meet it |
|---|---|---|
| breakeven payrolls | The monthly job gain that holds the unemployment rate steady. | Payrolls |
| Sahm gaugeon the pages: “the Sahm gauge” | The three-month unemployment rate less its twelve-month low; +0.50pp has marked recessions. | Unemployment |
| control group | The part of retail sales that feeds GDP: excludes autos, gasoline, building materials and restaurants. | Retail sales |
| NDFI / nondepository financialson the pages: “non-bank lenders” | Finance companies, private-credit and private-equity funds, mortgage lenders. | Bank lending; Private credit |
| break-adjusted | Excluding reclassifications and bank charter changes. | Bank lending |
| HQM, GZ, EBP | Investment-grade spread (Treasury HQM) / all-ratings corporate spread (Gilchrist–Zakrajšek) / risk premium beyond default risk (excess bond premium). | Corporate spreads; Risk sentiment; What is priced |
| R-CPI-NTR / NTR, R-CPI-ATRon the pages: “the BLS New Tenant Rent index” | A BLS research series on rents for new leases, which leads CPI rent by about four quarters. | Shelter & supply; Policy reach |
| OERon the pages: “owners’ equivalent rent (OER)” | Owners' equivalent rent: what owner-occupiers would pay to rent their own homes. | CPI; Shelter & supply |
| ULC, ECI | Unit labour costs / the Employment Cost Index (pay free of job-mix shifts). | Wages; AI and labour |
| primary deficit | The budget deficit before interest payments: what the government spends over what it takes in, leaving aside the cost of its debt. | Federal budget; Treasury supply |
| SLOOSon the pages: “the Fed’s survey of loan officers (SLOOS)” | The Federal Reserve’s quarterly Senior Loan Officer Opinion Survey: whether banks are tightening or easing lending standards, and whether loan demand is rising or falling. | Bank lending; Credit conditions |
| trimmed mean | Inflation with the largest price rises and falls set aside each month, leaving the centre of the basket; a guide to the trend. | CPI; PCE |
| sticky, flexible prices | Prices that change rarely (rents, medical care, restaurant meals) against prices that change often (petrol, food at home, cars); sticky prices carry more of the trend. | CPI |
| final demand, intermediate demand, stage | Producer prices for goods and services sold to final users, against those sold to other producers; the stages order intermediate goods by how far they sit from the final product. | PPI |
| SCE, SPF, Michiganon the pages: “the household surveys (New York Fed, Michigan) and the forecasters’ survey (Philadelphia Fed)” | Three surveys of expected inflation: the New York Fed’s Survey of Consumer Expectations and the University of Michigan’s survey ask households, the Philadelphia Fed’s Survey of Professional Forecasters asks forecasters. | Expectations |
| GDI (gross domestic income) | Gross domestic income measures the economy from what is earned rather than what is spent; in principle it equals GDP, and the gap between the two is the statistical discrepancy. | GDP & GDI |
| final sales to private domestic purchaserson the pages: “final sales to private domestic buyers (core GDP)” | Consumer spending plus business investment: GDP without inventories, trade and government, the part that shows underlying demand. | GDP & GDI; Nowcasts |
| revolving, nonrevolving credit | Credit cards and other lines a borrower draws down and repays at will, against loans of a fixed amount and term (cars, education). | Consumer credit |
| flow into delinquency, share delinquent | The share of balances that became late in the quarter, against the share of all balances that are late; the flow turns first. | Consumer credit; Household credit |
| charge-off | A loan the lender has written off as a loss; the charge-off rate is the share of balances written off in a year. | Consumer credit; Bank lending |
| DPIon the pages: “disposable personal income” | Household income after taxes; the saving rate is the share of it not spent. | Income & spending |
| Beveridge curve | Job openings plotted against unemployment: down the curve means a cooling labour market; a shift of the whole curve means matching has changed. | Openings & quits |
| diffusion index | The share of industries (or firms) reporting a rise less the share reporting a fall, so 50 means as many up as down. | Payrolls; Production |
| months’ supply | How many months it would take to sell the homes now for sale at the current pace of sales; about six is a balanced market. | Shelter & supply |
| population controls | The Census Bureau’s yearly revision of the population estimates the household survey is scaled to; it moves the level of employment without any change in the labour market. | Household survey |
| SAARon the pages: “annual rate, seasonally adjusted” | A monthly or quarterly figure scaled to a year’s pace after seasonal patterns are removed, so months can be compared. | Construction; GDP & GDI; Retail sales |
| NMDBon the pages: “the mortgage database (NMDB)” | The National Mortgage Database, a sample of mortgages maintained by the housing-finance regulator, which shows the rates outstanding borrowers actually pay. | Mortgage |
| unit profits | Profit per unit of output: the price of the output less its cost per unit. | Profits |
| capacity utilisation | Output as a share of what industry could produce with its plant and equipment; the long-run average is near 80%. | Production |
| core capital goods | Orders for machinery and equipment excluding aircraft and defence: the cleanest monthly read on business investment. | Orders |
| inventories-to-sales | The stock of goods on hand divided by a month’s sales: how many months of sales are sitting in inventory. | Inventories |
| 5 units or more, multifamilyon the pages: “multifamily (buildings of five or more units)” | Housing in buildings of five or more units, as the Census Bureau classes it; single-family is one unit. | Construction; Shelter & supply |
| contribution (percentage points)on the pages: “contribution, in percentage points” | How much of a total's change one part accounts for: the part's weight times its own change, so the parts add to the total. | CPI; PPI; PCE; Sources; GDP; Income; Retail |
| trade balance, net exportson the pages: “trade balance (exports less imports)” | Exports less imports of goods and services; negative is a deficit. In GDP it enters as net exports, so a wider deficit subtracts from growth. | Trade balance; GDP & GDI |
| statistical discrepancy | The gap BEA publishes between GDP measured from spending and GDI measured from incomes; in principle the two are equal, so the discrepancy is measurement error on one side or the other. | GDP & GDI |
| implied retail deflator | The price index implied by dividing retail sales in dollars by retail sales in volume, so that growth in dollars is growth in volume plus growth in this price. | Retail sales |
| debt-to-income (DTI) | A borrower's monthly debt payments, the new mortgage included, as a share of monthly income when the loan is made; lenders set limits on it. | Financing & affordability |
Regimes and risk
| Term | In plain words | Where you meet it |
|---|---|---|
| the regime | Which of four recurring patterns of co-movement the market is in, named by its legs: whether long Treasuries, the dollar, both or neither move against stocks there (a hedge is a leg whose correlation with US equity funds is below −0.15), and, when two regimes share their legs, whether stocks are volatile in it (“in stress”) or calm. The names come from each fit’s own profile. | Correlations & regime; Home |
| spine / webon the pages: “the ten core markets / the 27-market set” | The two sets of markets the regime engine is fitted on: ten core markets, and a 27-market set. | Correlations & regime; Distributions |
| engine, spine state, "R0 of 6"on the pages: “the regime model” | A model that sorts weeks into a handful of recurring co-movement patterns (the Appendix of Correlations & regime gives the number it runs and why) and names the closest. | Correlations & regime; Risk sentiment |
| quadranton the pages: “the market pattern” | The market state read from the signs of the stock–bond and stock–oil correlations. | Correlations & regime; Distributions; Cross-Asset Overview |
| absorption ratioon the pages: “how much markets move as one (the absorption ratio)” | How much of all co-movement runs in one or two common directions; higher means thinner diversification. | Correlations & regime; Return attribution |
| DCC | The model's day-by-day correlation, which weights recent days more. | Correlations & regime |
| HMM | The hidden Markov model behind the engine: see engine. | Correlations & regime |
| γ1, γ0 − r_fon the pages: “reward for beta (slope) / low-beta excess over cash (intercept)” | From the security market line fitted each week: the reward for a unit of beta (the slope) and what the lowest-beta end earned over cash (the intercept). | Risk sentiment; Return attribution |
| low / high turnover (presets)on the pages: “year-weighted / month-weighted” | Two weightings of the 21-, 63- and 252-session windows: 0.1/0.3/0.6 puts most weight on the year, 0.6/0.3/0.1 on the month. | Risk sentiment; Return attribution |
| sign rule, the two-correlation ruleon the pages: “the market pattern (stocks with or against bonds, with or against oil)” | The market state read from two correlations: whether stocks move with or against bonds, and with or against oil. Its four patterns are named by their legs: bonds with stocks, oil against; bonds against stocks, oil with; both with stocks; both against stocks. Each start date it prints names the rule that dated it (the daily sign rule or the weekly sticky rule). It is not the site's regime: “the regime” is the four-regime map of Correlations & regime. | Playbook; Asset outlook; Risk monitor |
| equity lens, macro lenson the pages: “with the market factor / macro prices only” | Two factor sets for one return: with the stock market itself as one factor — the default for every asset — or with macro prices (rates, credit, oil, the dollar) alone, printed as the sensitivity. | Factor analysis |
| coneon the pages: “the no-view range (volatility model)” | A range for an asset’s return over a horizon drawn from its own volatility alone, with no view on direction; a width, not a forecast. | Distributions; Asset outlook |
| engine bands, factor model rangeon the pages: “the factor model’s range” | The range of returns the factor model simulates for an asset over a horizon: shocks to the market-wide factors drawn from a volatility model with changing correlations (GARCH–DCC), passed to the asset through its estimated sensitivities with its own noise added — simulated shocks, not odds. | Distributions; Asset outlook; Scenarios |
| both negative, both positiveon the pages: “stocks fall with bonds / rise with bonds” | The signs of the two correlations that name the market pattern: both negative means stocks move against bonds and against oil; both positive, with both. | Correlations & regime; Playbook |
| TFFon the pages: “the Traders in Financial Futures report” | The weekly CFTC report of futures positions by type of trader (dealers, asset managers, leveraged funds), from which positioning is read. | Positioning |
| risk-on, risk-off | Which side of its own line a family of risk-appetite readings sits on: risk-on when risk is priced cheaply, risk-on spreads and beta are paying, and investors are confident and long; risk-off the other way. Risk sentiment and the risk monitor count the families on each side; the monitor names the state from two of them (see the state). | Risk sentiment; Risk monitor; Cross-Asset overview; Home |
| family counton the pages: “families reading risk-on” | How many of the six families of risk-appetite readings — the price of risk, the reward for beta, the risk-on spreads, the co-movement index, futures positioning and consumer sentiment — sit on the risk-on side of their own line this week. A count, not a score: each family keeps its own history. | Risk monitor; Risk sentiment; Home |
| price of risk | What it costs to hold risk, read before the fact: the average percentile of the readings priced in advance — the variance premium, put skew, credit premiums, financial stress, implied volatility and implied correlation — each against its own history. Above 50, risk is dear; below, cheap. | Risk sentiment; Risk monitor |
| variance premium | How much more the options market charges for the S&P 500’s volatility than the market has lately delivered: one-month implied volatility less the realised volatility of the last 21 sessions, in volatility points. High means protection is dear. | Options: expectations; Risk sentiment; Risk monitor |
| excess bond premium | The part of corporate bond spreads that expected defaults do not explain (Gilchrist and Zakrajšek): the extra yield lenders ask for bearing credit risk at all. It rises when their appetite for risk falls. | Credit conditions; Risk sentiment; Risk monitor |
| co-movement index | One weekly number for how far markets are trading as a single risk-on or risk-off bet: the common direction of six daily inputs — US and international stocks, high-yield against investment-grade bonds, the dollar, gold and the change in the S&P 500’s realised volatility — each signed so that a risk-off day counts up, in standard deviations. Above zero is a risk-off week. | Correlations & regime; Risk monitor |
| sticky patternon the pages: “the market pattern” | The market pattern read from the stock–bond and stock–oil correlations with a rule that holds it: a pattern begins when both readings clear two standard errors with its signs, and ends when the weaker has stayed below one standard error for two weeks or crossed zero. It changes about twice a year, not every few weeks. | Correlations & regime; Risk monitor; Risk sentiment |
| leg, the weaker leg | One of the two correlations the market pattern reads — stocks with bonds, and stocks with oil. The weaker leg is the one nearer zero in standard errors; the pattern holds while it stays above one standard error, and how far above is how firm the pattern is. | Correlations & regime; Risk monitor |
| hedge scorecardon the pages: “what each hedge did on the worst stock days” | For each candidate hedge — Treasuries, gold, the dollar and others — its average return on the worst tenth of S&P 500 sessions in the window, how often it rose on them, and how much it moved with stocks on the down days. | Correlations & regime; Risk monitor |
| the state (risk-on, wall of worry, complacent, risk-off) | The risk monitor’s two-by-two: the price of risk (cheap below 50, dear above) against what the risk-on spreads paid (paid above 50, punished below). Cheap and paid is risk-on; dear and paid, a wall of worry; cheap and punished, complacent; dear and punished, risk-off. It has changed cell about fifteen times a year, so a week within five points of a line is marked near it. | Risk monitor |
| closest past pattern | The recurring co-movement pattern, among those the regime model learned from weeks since 2014, that the latest weeks most resemble; its name is the pattern’s average, so today’s readings can sit outside it. | Correlations & regime |
| factor exposure (beta)on the pages: “exposure” | How much an asset’s return moves with a factor’s, the other factors held fixed: on Factor analysis the point-in-time beta of the daily model, printed with its 90% band and its t. An exposure times the factor’s move is that factor’s contribution to the return; an exposure holds when its sign and size survive three-year windows. | Factor analysis |
| sources of return | A return taken apart by accounting identity, by how the asset pays: income and carry (dividends, a bond’s carry), growth and roll (earnings growth, a bond’s roll-down) and revaluation (the multiple, the rate move, a spot price). No model of what moved it; what the identity cannot match is the tracking gap. | Sources of return |
| contribution to return | A factor’s share of a return in percentage points: the asset’s exposure to the factor times the factor’s move, summed day by day, so the contributions, the 3-month bill and the residual add to the return. With no benchmark and no decision being judged it is a contribution, not an attribution. | Factor analysis |
| contribution to risk (share of variance) | How much of a return’s ups and downs over a window each factor carries: its contribution’s covariance with the return, divided by the return’s variance. The shares, the bill’s and the residual’s included, add to 100%; a factor can carry much of the risk and little of the return, or the other way round. | Factor analysis |
| residual | What a factor model leaves of a return: the return less every factor’s contribution and the bill. On the set with the market it is alpha plus noise. | Factor analysis |
| tracking gap | What an accounting identity cannot match: the fund’s return less the identity’s, the fund’s holdings and fees against the model bond or index. Printed, never forced to zero. | Sources of return |
| not separated | A known block of a return that the data the site carries cannot yet split out: the oil fund’s roll and fees, a currency fund’s carry, a high-yield fund’s spread. Named, not called a residual. | Sources of return |
| implied volatility | The volatility an option's price implies for its market over the option's life, annualised and quoted in vol points: the price of insurance, read as a rate. It is a price, not a forecast — it includes what buyers pay to be insured. | Implied volatility; Skew & term structure |
| vol points | The unit of a volatility: one vol point is one percentage point of annualised standard deviation of returns. | Implied volatility; Skew & term structure |
| at the money | An option struck at, or nearest to, today's price: the most traded and the least affected by how the two tails are priced. On this site's surface it is read from the put and the call at the middle of the delta grid, as the Appendix of Implied volatility states. | Implied volatility |
| volatility term structure | Implied volatility at each horizon, from the shortest-dated options to the longest. It usually slopes up when markets are calm; when the near term is dearer than the long, the curve is inverted. | Implied volatility; Skew & term structure |
| risk reversal | The implied volatility of an out-of-the-money put less that of the call equally far out of the money (here the puts and calls with a delta of −0.25 and +0.25). Positive when downside insurance costs more than upside: the usual skew of an equity market. | Skew & term structure |
| butterfly | The average implied volatility of the two wings — the put and the call equally far out of the money — less the at-the-money volatility: what both tails cost together, over the middle. | Skew & term structure |
| variance risk premium | What option buyers pay over the volatility markets go on to deliver: implied volatility less realised. Positive on average, because insurance carries a premium; read as what the market prices, it belongs with the month ahead on the Outlook ledger, not with the observed pages. | Implied volatility; the Outlook ledger |
| implied move | The move an option price implies over a horizon: implied volatility scaled to that horizon, one standard deviation either way. It is what the market prices, not this site's forecast, and it is printed on the Outlook ledger beside the site's own range. | the Outlook ledger |
| realised volatility | How much a market has actually moved: the standard deviation of its daily log returns over a window of sessions, annualised. The site uses one rule everywhere — the Volatility page's — so an implied and a realised figure can be set side by side. | Volatility; Implied volatility; Dealer positioning |
| kernel smoothing | How a smooth surface is drawn from scattered quotes: each point of the grid is a weighted average of the implied volatilities of the listed contracts, the weights falling off with distance in delta and in days to expiry. | Implied volatility; Skew & term structure |
| delta bucket | Options grouped by their delta — how much an option's price moves with its market — from deep puts (near −1) through the money (near ±0.5) to deep calls (near +1); with an expiry bucket it places where a position sits. | Dealer positioning |
| gamma exposure | How much dealers would have to buy or sell to stay hedged for each 1% move in the market, on an assumed side for each contract. Long gamma, their hedging sells into rises and buys into falls; short gamma, it does the opposite. | Dealer positioning |
| dealer positioning | Where the banks and market makers who sell options stand once their books are hedged. Their side of each contract is not published, so the site assumes it — long the calls and short the puts customers trade — and shows the result as state, with a test of what has followed it. | Dealer positioning |
| open interest | The number of contracts — options or futures — outstanding at the close: positions opened and not yet closed or expired. | Dealer positioning; Positioning |
| put/call ratio | Open interest in puts divided by open interest in calls: above one, more downside insurance is held than upside. | Dealer positioning |
| trailing percentile | Where today's reading sits among the market's own readings over a trailing window, today's included: causal, so a past day's percentile never changes when a later day arrives. | Implied volatility; Skew & term structure; Dealer positioning |
| horizon (of a correlation)on the pages: “1M · 3M · 6M · 12M” | The memory a weighted correlation is read over: its half-life is a third of the window it stands for — 7, 21, 42 and 84 sessions for one, three, six and twelve months. Three months is the level the page reads; the others show whether a shorter or a longer memory agrees. | Correlations & regime |
| rank gaugeon the pages: “the rank correlation of beta with alpha” | Each asset ranked twice — by the beta it had when the window opened and by its alpha over the window, its return less what the CAPM line gave that beta — and the two rankings correlated: +1 when every alpha ranked as its beta did, −1 when the order was reversed. The rank form of the reward for beta: one extreme return moves a rank by one place, not the line by its whole distance. | Risk sentiment |
| betas as borne | Each fund's beta as the daily model carried it through a window, session by session, rather than the one it had when the window opened; its alpha is what is left after those betas. | Risk sentiment |
| turned (a fund) | A fund has turned once it sits at least a month's typical move — its three-month daily standard deviation × √21 — below its three-month high. | Risk sentiment |
| order gaugeon the pages: “the rank correlation of own volatility with sessions since each fund turned” | Across the funds, each one's own volatility ranked against how many sessions ago it turned: above zero the most volatile turned first, below zero last. A description of the order of the current move, not a signal. | Risk sentiment |
| the four tenors together | The reward for beta read over one, three, six and twelve months at once. Within each tenor recent days weigh more, and each tenor is put on a yearly scale. The four are averaged so that each has about an equal say in how the reading moves. | Risk sentiment |
| an equal say (of the tenors) | Weights on the four tenors in proportion to the square root of each tenor's length, which gives each tenor about a quarter of the reading's weekly movement. Equal weights would let the one-month tenor, the noisiest once put on a yearly scale, make most of it. | Risk sentiment |
| tenor | The span a reading covers: one, three, six or twelve months of sessions (21, 63, 126 or 252) to the latest Friday. The page's tenor setting chooses the span of the cross-section, the leadership cards and the spreads. | Risk sentiment |
| out of line with its beta | With the market up, an asset in the top quarter of beta whose alpha sits in the bottom quarter, or one in the bottom quarter of beta whose alpha sits in the top. With the market down, the reverse. A description of the tenor, not a signal. | Risk sentiment |
Statistics
| Term | In plain words | Where you meet it |
|---|---|---|
| fanon the pages: “the simulated range” | A simulated range of a series along a horizon; its centre reverts to an assumed rate. | Earnings; Distributions |
| filtered / smoothed probability | Real time: what the model would have said that week; smoothed: with hindsight. | Correlations & regime |
| HAC, Newey–West, t (HAC), block bootstrap, naive SE | Ways of measuring how uncertain an estimate is; above the folds the site shows one ±, the rest sits in Method. | Factor analysis; Risk sentiment |
| EWMA, half-life 63 | A faster-forgetting estimate, which weights recent sessions more; printed as "robust" or "not robust" beside the fixed-window one. | Factor analysis |
| log points | Roughly per cent for small moves; above the folds the site prints simple per cent, log points stay in Method. | Performance; Factor analysis; Earnings |
| zero drift | The range assumes no average move: it measures width, not direction. | Distributions |
| coverage | How often past outcomes fell inside the band drawn at the time. | Distributions; Scenarios |
| pseudo-real-timeon the pages: “as it would have read at the time” | A model re-run using only the data that were available on each past date, so its record is not flattered by hindsight. | Nowcasts; Correlations & regime; Valuation history |
| bridge equation | A regression that maps the monthly indicators already published for a quarter onto the quarterly total not yet published; each nowcast combines several. | GDP nowcast; Inflation nowcast; Labour nowcast |
| RMSE, biason the pages: “typical error; average error” | How far a model’s past estimates fell from the outcome on average (root mean squared error), and whether they leaned one way (bias). | Nowcasts; Expectations |
| Hodrick, Newey–West, encompassing, BICon the pages: “the error and comparison tests (in the Appendix)” | Ways of measuring the uncertainty of a long-horizon regression (Hodrick, Newey–West), of asking whether one forecast adds to another (encompassing, Clark–West), and of choosing a model’s size (BIC); above the folds the site prints one ± and its record. | Valuation history; Long-run growth |
| z-score (σ, standard deviations)on the pages: “standard deviations from the average” | How far a reading sits from its own average, in units of its usual variation: 2σ is a move seen about one month in twenty. | CPI; PCE; Performance; the heat maps |
| percentile sinceon the pages: “percentile since <year>” | The share of readings since the named year that were lower than today’s: the 81st percentile since 2010 means four readings in five were lower. | every tile row |
| typical moveon the pages: “one typical move (one standard deviation of the monthly change)” | One standard deviation of the series’ monthly change; a trigger “within one typical move” could be crossed by an ordinary month. | What would change this view; Home |
| annualised, 3m point-to-pointon the pages: “three-month rate, annualised” | A short-period change scaled to a year’s pace; point to point compares the latest month with the month three months earlier, an average of the three monthly changes smooths it. | CPI; PCE; PPI; Retail sales |
| abnormal return | What a stock or basket did beyond what the market, the research factors, Treasuries and oil would have predicted that day; summed over the days around an event. | US midterms 2026 |
| first stage, first-stage Fon the pages: “how strongly the instrument moves the variable (first stage)” | In a two-step regression, the first step shows how strongly the instrument moves the variable of interest; a weak first stage (a low F) means the second step’s estimate cannot be trusted. | How a hike transmits |
| Hodrick standard errors | Standard errors for a regression of returns over long, overlapping horizons that allow for the overlap (Hodrick, 1992); with few independent windows they are the honest measure of how uncertain a predictive slope is. | Valuation history (Appendix) |
| encompassing test | A test of whether one forecast adds anything to another: if, given the first, the second gets no weight in explaining the outcome, the first encompasses it. | Valuation history (Appendix) |
| BIC (Bayesian information criterion) | A score for choosing among models of different sizes that rewards fit and charges for every added parameter; the model with the lowest score is preferred. | Correlations & regime (Appendix); Long-run growth |
| out-of-sample | Judged only on data a model was not fitted to: each estimate made from the data available at the time and scored against what followed. | Valuation history; Nowcasts; Range of outcomes |
| ratio of sums | An index-wide ratio taken by adding every company's numerator and every company's denominator first and dividing once, so each company counts by its size and a loss stays in. | Payout & reinvestment; Profitability & balance sheet; Composition |
| nowcaston the pages: “this site's nowcast; published nowcasts (NY Fed, Atlanta Fed)” | An estimate of a figure for a period already under way, made from the data released so far; this site's nowcasts are its own models, printed beside the published ones. | GDP nowcast; Inflation nowcast; Labour nowcast; Nowcasts |
| alpha (the regression constant) | The constant of a factor regression, printed with its t: the average return the factors do not account for over the estimation window. It is one part of the residual, not all of it. | Factor analysis; Risk sentiment |
| explained share, explained amount | Two different things. The explained share is how closely the factors track an asset’s daily ups and downs over a window (1 − the residual’s variance ÷ the return’s); the explained amount is how many percentage points their contributions add up to. A model can track the days and miss the window’s trend, which is then the residual. | Factor analysis |
| Diebold–Mariano test | A test of whether one forecast's squared errors are smaller than another's over the same periods, with standard errors that allow for correlation from one period to the next; a positive statistic here means the site's model missed by less than the benchmark. | the nowcast pages (Appendix) |
| variance inflation factor | How far two or more factors in one regression move together: 1 means a factor is unrelated to the others; the larger it is, the less a beta on that factor can be told apart from the betas on the factors it moves with. | Factor exposures |
| bins of beta | The S&P 500's members sorted by their beta into twenty groups of equal size, each drawn as one point at its average beta. Its level is fitted together with one level per sector and read at the average mix of sectors, so a point shows what members of that beta earned beside companies in the same business. | Risk sentiment |
| median slopeon the pages: “the median slope (the typical member)” | The slope of a line drawn through the middle of the returns at each beta rather than their average (a quantile regression at the 50th percentile): what a unit of beta paid the typical company, which a few large winners cannot lift. | Risk sentiment |
| percentile slopeon the pages: “the 10th and 90th percentile slopes” | The slope of a line drawn through a chosen percentile of the returns at each beta, by quantile regression: the 10th runs near the worst outcomes and the 90th near the best, so the two show how each end of the outcomes moved with beta. | Risk sentiment |
| spread of the slopeson the pages: “the spread (90th less 10th percentile slope)” | The 90th percentile slope less the 10th: how much faster outcomes fan out as beta rises. It is almost always positive, because the high-beta companies are the volatile ones; its level against its own history is the reading. | Risk sentiment |
| half-life | How fast a weighted estimate forgets: a session that many sessions old counts half as much as today's, one twice as old a quarter as much. Nothing drops out of a window on a date, so a quiet day does not move the reading. | Correlations & regime; Risk sentiment |
| effective sample | How many equally weighted sessions a set of declining weights is worth: (1 + λ)/(1 − λ) for weights that fall by λ a session, about three half-lives — 20, 61, 121 and 242 sessions at the half-lives of one, three, six and twelve months. | Correlations & regime |
| compounding gap | Over a window the average simple return of a volatile stock sits above its typical return, because gains compound on gains; on log returns, which add up day by day, that lift is absent. The gap between the mean and median slopes is printed both ways, so the part that compounding explains can be seen. | Risk sentiment |
| median lowon the pages: “the median low” | Each simulated path's lowest weekly level, taken across all the paths: half the paths bottom below it. It is the middle of the lows, not the low of one middle path. | Reserve drawdown simulation |
| stability rule | A level is trusted for direction only if the gap to it has closed the way theory says with a t of 2 or more, in both halves of the sample, and its one-year target re-estimated each quarter beats a random walk. A basket that fails prints a model level, not a direction. | Fair value board; Method and record; Sector & style outlook |
| central case | The middle of a model’s simulated paths for a year ahead, under its stated inputs. It is model-conditional and carries no odds; on the Sector & style outlook its width is the factor model’s range, which is scored a quarter ahead. | Sector & style outlook |
The themes
| Term | In plain words | Where you meet it |
|---|---|---|
| the issue, the September reporton the pages: “the September report (11 Sep 2026)” | The written macro review of September 2026, whose data date was 11 September 2026; the theme pages carry its logged evidence and compare today with it in a fold. | Themes; Reconciliation folds |
| logged, evidence logon the pages: “the dated evidence log” | A dated list of the facts and judgements behind a theme, written once and never revised; a page prints what was logged and when. | Themes |
| hollow deal | A reopening of the Strait by regional consent — a corridor the Gulf states recognise — without an agreement between the United States and Iran: ships move, the blockade stays. | Gulf chokepoints |
| tenure channel | How a rate rise pushes would-be buyers into renting: dearer mortgages keep them as tenants, adding to rental demand six to eighteen months on (Dias and Duarte). | Housing supply; The decision vs the source |
| BDCon the pages: “business development company (BDC)” | A listed fund that lends to mid-sized private companies; its share price against the value of its loans is a market read on private credit. | Private credit |
| NAV, book value per shareon the pages: “net asset value per share” | The value of a fund’s holdings less its debts, per share; a price below it says the market doubts the holdings’ stated value. | Private credit |
| general account | The assets an insurer holds against its own promises to policyholders, as distinct from assets it manages for others. | Private credit |
| sidecar, cession, recapture | An insurer passes (cedes) part of its liabilities to a reinsurance vehicle it sponsors offshore (the sidecar); recapture is taking them back. | Private credit |
| canarieson the pages: “the flagged managers” | The five listed private-credit managers the June report chose in advance as the first to show stress: Blue Owl, KKR, Ares, Blackstone and Apollo. | Private credit |
| normalisation | Each response is divided by how much the surprise moved the 1-year yield, so it reads as the effect of a +25bp move; when that divisor is weakly estimated, levels are less certain than their bands show. | The hike's transmission |
| ex-ante basket | Every listed name a Democratic Senate would change, long the names the policy helps and short the ones it hurts, chosen from the policy alone before looking at any prices. | US midterms 2026 |
| tuned basket | The names of the ex-ante basket whose measured move with the Senate odds has the sign the policy implies. Chosen with the data, so its record before the choice flatters it. | US midterms 2026 |
| lever (midterms) | A policy the election outcome decides — a confirmation, a bill, an oversight power — and the names it reaches; a lever that differs between a Democratic sweep and a split Congress is a Senate lever. | US midterms 2026 |
| Senate surprise | How far a vote's outcome landed from what the market priced on its eve: the outcome (one if Democrats won the Senate, zero if not) less the eve-of-vote odds, in points. | US midterms 2026 |
| thesison the pages: “view” | A standing argument about one question — a theme’s claim, the evidence for it and what would change it; the pages call it a view. | Thematic; Home |
| attrition (hiring) | Jobs lost by not replacing people who leave — fewer openings and hires, fewer quits — rather than by laying people off. | AI and labour |
| the lock (11 Sep)on the pages: “the September report's data date (11 Sep 2026)” | The date the September report stopped taking data: every figure it printed is as of 11 September 2026, and the theme pages compare today with it. | Themes |
| verdict (holding · tested · under pressure · resolved) | The word a theme page gives its own view, set by rule from the dated log and the tests: holding while no test has crossed, tested once one has, under pressure while one is closing, resolved when the log closes the theme. | Themes |
| the author's assessment | Odds the author wrote down on a dated day, kept as written in the record of each set and never revised; unlike a market price or a model's output, they are a judgement. | Themes |
| mb (million barrels) | Million barrels, the unit EIA reports crude oil and product stocks in; kb/d, thousand barrels a day, is the unit of a flow. | Gulf chokepoints; Oil |
| satellite count (of ships)on the pages: “the satellite count” | IMF PortWatch's daily count of the ships that cross a strait, worked out from the position signals (AIS) ships broadcast. Ships that switch their transponders off are missed, so it reads below trackers that add confirmed dark transits; the page draws its 7-day average, the measure the traded contracts settle on. | Gulf chokepoints |
| the contracts' lines | The levels at which the Polymarket contracts on the straits settle, read from their own rules: Hormuz traffic is "normal" when PortWatch's 7-day average reaches 60 ships a day, and Bab el-Mandeb is "effectively closed" when it falls to 10 or fewer. | Gulf chokepoints |
| daily strike risk | The average daily chance of a first US strike on Iran between two contract dates, read from the prices of the contracts that pay if no strike has landed by each date: one minus the ratio of the later price to the earlier, spread evenly over the days between them. | Gulf chokepoints |
| long shot (a traded price) | A contract priced under 10% or over 90%. Prediction markets are known to overprice the unlikely side at these extremes, so such a price is read as "near zero" or "near certain" rather than at face value. | Gulf chokepoints |
| thin market | A contract with less than $100,000 traded. Its price can move on a few trades, so the page shows it but never bases a risk or an anchor on it. | Gulf chokepoints |
| the reserve's reporting line (252.4 mb) | The level of the Strategic Petroleum Reserve below which the limited-drawdown authority of the Energy Policy and Conservation Act (§161(h)) may not draw it. Exchanges and sales under a presidential finding are not bound by it, so it is a line to report against, not a hard stop. | Gulf chokepoints; Oil and diesel reserves |
| strike oddson the pages: “the market's chance of a US strike by each date” | The chance of a US strike on Iran by each date that traders price in the dated “ceasefire continues through …” contracts, read at the mid of each contract's bid and ask and joined by a constant daily risk between dates. It is a market price, not this site's judgement. | Reserve drawdown simulation |
| diesel crackon the pages: “the crack (diesel less Brent)” | What a barrel of diesel sells for above a barrel of crude. On this site it is NY Harbor ULSD futures, in dollars a gallon times 42, less Brent futures, from two front-month closes; it is not a refiner’s margin. | Russia–Ukraine diesel war |
| General License 135on the pages: “General License 135 (the licence)” | The US Treasury’s licence of 9 October 2026 allowing dealings in diesel of Russian origin, imports into the United States included, to 7 April 2027. It sets no volume, so Russia’s refineries, not the licence, decide how much can move. | Russia–Ukraine diesel war |
| energy truce | A pause in strikes on energy targets only — refineries, terminals, power plants and the grid — while the rest of the war goes on. It is not a ceasefire, and a contract on a ceasefire does not settle on it. | Russia–Ukraine diesel war |
| ceasefire-agreement contract | A prediction-market contract that settles yes if both sides agree a general suspension of fighting by its date, announced by both governments or confirmed by credible reporting. An energy-only truce or a unilateral pause does not count; its price is a traded probability, not a forecast. | Russia–Ukraine diesel war |
| distillate fuel oil | The middle cut of the refined barrel: diesel for trucks, trains, farms and ships, and No. 2 heating oil for homes. EIA reports its stocks, output, use and trade every week. | Oil and diesel reserves; Russia–Ukraine diesel war |
| days of supply | How long stocks would last at the recent pace of use: stocks divided by the average of the last four weeks of product supplied, which is EIA's measure of use. | Oil and diesel reserves |
| range for the week | The lowest and highest reading for the same week of the year in each of the five years before, with their average. A stock that rises and falls with the seasons is read against the time of year, not against last month. | Oil and diesel reserves |
| Northeast Home Heating Oil Reserveon the pages: “the heating-oil reserve” | A federal stock of about one million barrels of ultra-low-sulphur diesel held at terminals in New England and New Jersey. The President may sell it on a finding of a severe supply interruption; the statute deems a heating-oil market dislocation to occur when the gap between crude and Northeast heating oil runs well above its heating-season average and is still rising. | Oil and diesel reserves |
Site words
| Term | In plain words | Where you meet it |
|---|---|---|
| falsifieron the pages: “what would change this view” | The dated, observable condition that would make the page's reading wrong; the box is labelled "What would change this". | every page |
| takeaway / read | The page's answer in one sentence each; the full read is the fold beneath the takeaways. | every page |
| anchoron the pages: “start date (What is priced) / assumed long-run rate (the earnings fan)” | Two meanings, now split: the date a move is measured from, and the rate a fan's centre reverts to. | What is priced; Earnings; Return attribution |
| the 2015–19 normon the pages: “the 2015–19 average” | The pre-pandemic average, the default comparison; a second norm is printed where the first is structurally wrong. | US Macro |
| model-conditional | What a named model implies under stated inputs; not a forecast. | Outlook; Earnings; Valuation |
| the lock / "since the lock"on the pages: “since the September issue (11 Sep)” | The September report's data date, 11 September 2026; comparisons against it now live only in a fold headed "Reconciliation to the issue". | Reconciliation folds |
| the author | Who writes this site and the research practice behind it; Notices says who that is and what the author owns. | Notices |
| Macro Review calculations | Whose number it is: a publisher’s own figure is shown as published, with its name; a vendor’s price or estimate is shown with its attribution, under its terms; a figure this site produced from them — a model’s estimate, a decomposition, a fair value — says “Macro Review calculations” in its source line, and the Appendix names its inputs. | every source line |
| data to / updated / update due | A page's head says when its data run to ("Data to Aug 2026"), the day its figures were last computed ("updated"), and, once its next update has fallen due and not yet been made, "Update due since" that day. | every page |
| pin, branchon the pages: “sets the 10-year at … / scenario” | A scenario “pins” the 10-year yield at a level and works out what follows; each pinned path is one branch of the scenario set. | Scenarios; Range of outcomes |
| kind (of a range)on the pages: “what this range is” | What a printed range measures: the spread of the inputs (input), the estimation error of a parameter (parameter), assumed values (assumption), other models (model), past misses (residual), named paths with no odds (scenario), the out-of-sample record (backtest) or simulated shocks (simulated); none is a probability of the outcome. | Valuation; Board; Long-run growth; Scenarios; Asset outlook; Distributions |
| real term premium, inflation risk premium | The two parts of the term premium: extra yield for tying money up at a fixed real rate, and extra yield for the risk that inflation surprises. | Inside the 10-year |
| TIPS liquidity premium | The extra yield inflation-protected Treasuries (TIPS) pay because they trade less easily than ordinary Treasuries; it pulls the breakeven below expected inflation. | Inside the 10-year |
| DKWon the pages: “the Board's DKW model” | A Federal Reserve Board staff model (D'Amico, Kim and Wei) that splits nominal and inflation-protected yields into expected rates, expected inflation and premia. | Inside the 10-year |
| κ (kappa)on the pages: “κ” | How much of a move in today's one-year real rate is assumed to carry into the ten-year average expected real rate; a stated parameter with a range, not an estimate of the future. | Inside the 10-year |
| local projection | A regression of how far a series has moved h months after a shock on the shock itself, run once for each horizon; it describes what followed past shocks, not what a new one must do. | Transmission |
| pass-through shareon the pages: “the share of a move passed through” | How much of a sustained move in one rate or price has reached another after a given number of months: 1 means all of it. | Transmission |
| range kinds: input · parameter · assumption · model · residual · scenario · backtest · simulatedon the pages: “spread of inputs · estimation error · assumed values · other models · past misses · named paths, no odds · out-of-sample record · simulated shocks” | The eight things a printed range can be, each named on its card in these words; none is a probability that the outcome lands inside it. | every page that prints a range |
| triggeron the pages: “what would change it” | The level or change, with its unit and the date or release that decides it, that would make a view wrong; the pages print it in the box headed “What would change this view”. | every page; the Thematic pages |
| rate-sensitive | A price a higher federal funds rate can slow within about two years, through demand and credit (shelter and market services, mainly). | Sources & rate sensitivity |
| household debt 90+ days lateon the pages: “90+ days late (seriously delinquent)” | Balances on which a payment is at least 90 days overdue, as a share of all balances — the stock of problem debt, where the flow into delinquency is the new problems each quarter. | Asset quality |
| volatility model | A fitted model of how much each market moves day to day and how markets move together (GARCH with a dynamic correlation), run forward many times from today to give a range with no view on direction. | Range of outcomes; Asset outlook |
| factor model | A model that simulates nine market-wide factors (stocks, rates, credit, oil, the dollar and others) forward and maps them to each asset through its estimated sensitivities, adding the asset's own noise; its bands are simulated shocks, not odds. | Scenarios; Asset outlook; Range of outcomes |
| full pass-through | The valuation formula's assumption on a rate path: the whole move in the 10-year passes to the discount rate while the premium and growth stay where they are — the extreme case, not an estimate of what would happen. | Scenarios; Asset outlook |
| rate pathon the pages: “rate path (scenario)” | A named path for the Fed and the 10-year yield, set at a value a dated source publishes; the site attaches no odds to any path. | Scenarios; Asset outlook; Playbook |
| attrition (Hormuz) | Both sides hold out: traffic through the Strait stays at a trickle and neither Iran nor the United States concedes before the 3 November election. | Gulf chokepoints |
| Brent–Dubai spread | The price of crude delivered outside the Gulf (Brent) less crude priced inside it (Dubai): while the Strait is shut a barrel that can leave carries a premium and the spread widens; a narrowing spread says Gulf barrels are moving again. | Gulf chokepoints |
What the numbers claim, and what they do not
No probability, forecast or target without a printed model: the rule, in plain wordsFour limits on what the site's numbers claim
- ForecastsNo probability, forecast or target without a printed model.
A cone, a fan, a scenario set, a projection or a policy-rule path is printed beside the model that produced it and the inputs it was given, and its note says first that it is model-conditional: what the model implies on those inputs, not a statement about what will happen.
- RankingsA ranking describes; it never signals.
A table ordered by return, by beta or by distance from a norm says what has happened. Where a page shows what followed similar states, the count of episodes beside each number says how thin that history is.
- Fair valuesA fair value comes with its model, inputs, range and record.
It is what a stated model implies on stated inputs, with the range around it, the condition that would change it and how well the method has done before. It is not a recommendation: this is a personal research site, and nothing on it is investment advice.
- RevisionsNothing moves quietly.
When a number changes because the method changed rather than the data, the page says what moved and why, and fixes any earlier claim the change makes wrong. See Revisions and corrections.
Three ways to read a return
What a return is made of, how exposed the asset is, and what each factor contributed: three questions, two pages, one set of wordsSources, exposures and contributions
- SourcesWhat the return is made of, by accounting identity.
A bond fund's return is its carry, its roll-down and the rate move; an index's is its dividends, its earnings growth and the change in its multiple. Sources of return groups these into income and carry, growth and roll, and revaluation — one colour each on every asset — with no model of what moved them. What the identity cannot match is the tracking gap; a block the data cannot yet split, such as an oil fund's roll, is “not separated”. Sources of return.
- ExposureHow much the asset moves with a factor.
A beta, estimated from history: printed with its 90% band and its t, and faded where it cannot be told from zero. An exposure is a sensitivity, not a share of the return.
- ContributionWhat each factor added to the return, and how much of the risk it carries.
The exposure times the factor's move, day by day: the contributions, the 3-month bill and the residual add to the return. Where there is no benchmark and no decision being judged this is a contribution, not an attribution. Its share of the return's variance says how much of the ups and downs the factor carries, which can differ a lot from its share of the return. Factor analysis.
- What is leftThree words, one meaning each.
The residual is what a factor model leaves; alpha is only that model's constant, printed with its t. The tracking gap is what an identity leaves. “Not separated” names a known block the data cannot yet split.
Reading the risk monitor
What the five pillars are, what the family count counts, what the state names, and what none of them claimsHow the risk monitor groups, counts and labels risk appetite
- PillarsFive ways the industry reads risk appetite.
What risk costs to hold, priced in advance (option prices, credit premiums, financial stress); what risk paid (the reward for beta, risk-on spreads); how markets move together (the co-movement index, the market pattern, the hedge on the worst stock days); what investors hold (futures positions, dealers' options, margin debt); and what they say (surveys). The Risk monitor sets their readings side by side, each copied from the page that owns it and linked to it.
- HistoryEvery reading against its own past.
A percentile says where today sits among the reading's own weeks or months, on the basis its page states; a composite is the mean of its members' percentiles, and its members are printed beneath it. A percentile has no units: the same rank in a calm decade and in a violent one are different amounts.
- CountHow many families agree.
Six families are each read against their own line, and the page counts those on the risk-on side. They disagree in most weeks, so the count, not any one family, is the reading. It treats the six as equals, which they are not, and two of them share the market's days.
- StateFour names from two families.
The price of risk across and what risk-on spreads paid up give four cells: risk-on (risk cheap to hold, and paid), wall of worry (dear, and paid all the same), complacent (cheap, but not paid) and risk-off (dear, and not paid). Close to a line one Friday can change the cell, so the page marks a week near a line and says how often the state has changed. It describes the weeks it names; it is not a regime, a signal or a forecast.
- Not shownGaps are named, not filled.
Where a source the industry uses is closed to this site — a members-only survey, a subscription index, a publisher that refuses the site's requests — the monitor names it with its reason rather than putting another number in its place.
Check a number
Every figure is reproduced from its publisher before it is printed, and the page shows you the testHow each figure is checked against its publisher before printing
Every printed figure is reproduced from the publisher's own release within a stated tolerance before the page is published; a page whose checks fail is not published at all. The Checks tab of each page's Appendix lists every test, the publisher's figure, the figure produced here, the tolerance and the result. Read the tolerance before you trust a derived figure — reproducing a rate to a hundredth of a point is a stronger claim than to two tenths — and where a page copies a number from another page, its check compares the copy with the original every time the site is updated.
How the numbers are made
From the publisher's file to the printed figure, whose figure it is, and where the rules are kept
From the publisher's file to the printed figure
- CollectThe publisher's own file.
Each source has its own code that downloads the publisher's release into a dated folder with a record of what was taken and when. Nothing enters any other way: a source is never swapped for a more convenient one, and a missing number is never filled in by hand.
- ComputeOne module per page.
It reads those files, computes what the page shows and writes one data file: the series, the key data, the takeaways, the notes and the checks.
- CheckPublished against computed.
Each check names what it tests, the publisher's figure, the figure computed here, the tolerance the two must agree within, and the result.
- RefuseA failing page is not published.
The step that writes a page's data refuses it when a check fails; the page keeps its last good version and the new numbers never reach the site. Being blocked is acceptable; a quiet repair is not.
- ReadWhere to find them.
The Checks tab of any page's Appendix holds every check on that page, not a sample.
Today 110 pages' data files pass between them 8,989 reproduction checks — counted from the published files when the site was last put together, not written into this page.
Whose number it is
- AgenciesShown as published.
Statistical agencies and central banks publish their files for anyone to use; their figures appear as published, with the publisher named underneath.
- VendorsShown as levels, with attribution.
Market prices, index levels, quoted earnings and option data reach the site from their vendors and are shown under the vendor's terms, with the vendor's name under the card.
- HereMade here, inputs named.
A model's estimate, a decomposition or a fair value made on this site says it was computed here and names its inputs in the source line.
- Cross-checksCompared, not drawn.
A few series carry a personal-use notice from their publisher (the Atlanta Fed's sticky-price CPI and GDPNow among them): the checks compare against them, and at most one dated figure is quoted with attribution.
Where the rules are kept
- StandardThe page guide.
File layout, how sources are collected, the data contract, the section order, and what must pass before a page may be published.
- ContractOne step writes a page's data.
It refuses data whose checks do not pass.
- TolerancesOne specification per page.
A page's sources, windows, display choices and the tolerance of every check it runs sit side by side, where a change to one is visible.
Why the How to read page has no checks table
This page has no data of its own, which is why its Appendix carries no checks table. The examples above are copied from the pages they name each time the site is updated; the map is read from the site's navigation and the word list from its glossary, so none of them can drift from what the pages print.