Credit, largely the market's move, accounts for +7.7pp of S&P 500's +9.4% this year.
Which factors drove this return, how exposed is the asset to each, and how much of its risk do they carry?
In brief
One asset, one window, one model — the return taken apart factor by factor, the risk each factor carries and the exposures behind them- Credit, largely the market's move, accounts for +7.7pp of S&P 500's +9.4% this year.→ Exhibit 1
- Credit, largely the market's move, carries 50% of S&P 500's daily variance this year; the residual 34%.→ Exhibit 1
- The factors explain 66% of S&P 500's daily returns this year.→ Exhibit 2
- S&P 500's exposure to credit today is +2.71 (t 9.6).→ Exhibit 2
| What would change this view | Trigger | Today | Decided by |
|---|---|---|---|
| Macro prices stop tracking the days | 63-session (macro factor set) < 30% | 60% (16 Nov 2026)not crossed | the week to 23 Nov 2026 |
| The residual becomes the story | equity-factor set residual over 63 sessions beyond ±2pp | +0.3pp (16 Nov 2026)not crossed | the week to 23 Nov 2026 |
| Credit stops carrying the market on the macro factor set | credit's correlation with the market inside ±0.20 over the year and the year to date | +0.77 / +0.78 (16 Nov 2026)not crossed | the week to 23 Nov 2026 |
The reasoning behind each point
- The returnCredit, largely the market's move, accounts for +7.7pp of S&P 500's +9.4% this year.
Over 220 sessions to 16 Nov 2026, macro prices only: credit +7.7pp, 10-year yield −2.9pp, breakeven inflation −2.9pp, emerging − developed −1.6pp; the 3-month bill +3.5pp and a residual of +7.1pp (what the macro prices do not explain).
- The riskCredit, largely the market's move, carries 50% of S&P 500's daily variance this year; the residual 34%.
Each part's share of the return's day-to-day variance over the same sessions; the factors, the bill and the residual add to 100%, and a negative share offsets the rest.
- The fitThe factors explain 66% of S&P 500's daily returns this year.
Over the last 63 sessions they explain 60%. The rest is the residual.
- The exposureS&P 500's exposure to credit today is +2.71 (t 9.6).
The return per 1% move, estimated on the 252 sessions to 13 Nov 2026; a contribution over a window is that exposure, as it stood each day, times the factor's move.
Factor sensitivity · The read is that gold in dollars is responding to the dollar, not to real yields. It would change if real yields' t moved beyond −2, or the dollar's fell inside ±2.
Next: Sources of return — the same return, by accounting identity
The factor table
Each factor's exposure and its uncertainty, the exposure the window applied, the factor's move, its contribution to the return and its share of the riskExhibit 1 · Each factor's contribution to the S&P 500's return and risk this year
About this exhibit
Exhibit 2 · The S&P 500's factor exposures, moves and contributions this year
| Factor | Exposure today | 90% band | t | Average exposure in window | Factor's move | Contribution, pp | Share of variance |
|---|---|---|---|---|---|---|---|
| Core: market and rates | |||||||
| 10-year yield per +10bp | −0.20 | −0.36 to −0.04 | -2.1 | −0.29 | +94bp | −2.88 | 4.1% |
| Macro: credit, oil and the dollar | |||||||
| Credit per 1% | +2.71 | +2.24 to +3.17 | 9.6 | +2.85 | +2.6 | +7.68 | 50.3% |
| Oil per 1% | −0.03 | −0.05 to −0.01 | -2.7 | −0.02 | +61.3 | −1.18 | 2.6% |
| Dollar per 1% | −0.06 | −0.25 to +0.13 | -0.5 | ±0.00 | +6.0 | +0.03 | -0.7% |
| Subtotal | +6.52 | 52.2% | |||||
| Secondary macro | |||||||
| Breakeven inflation per +10bp | +0.51 | +0.16 to +0.86 | 2.4 | +0.61 | −45bp | −2.86 | 0.7% |
| Emerging − developed per 1% | +0.18 | +0.12 to +0.24 | 5.2 | +0.40 | −3.8 | −1.58 | 6.9% |
| Gold per 1% | +0.03 | −0.01 to +0.08 | 1.3 | +0.04 | −8.7 | −0.36 | 1.8% |
| Subtotal | −4.80 | 9.3% | |||||
| All factors | −1.15 | 65.6% | |||||
| Cash: the 3-month Treasury bill, accrued | +3.46 | 0.0% | |||||
| Residual · of which the regression constant +3.2% a year, t 0.4 | +7.05 | 34.4% | |||||
| Total return | +9.36 | 100% | |||||
The factors explain 66% of the variance of the window's daily returns: one less the residual's variance over the return's. The “All factors” row is their share of that variance counted part by part. Exposures from the 252 sessions 13 Nov 2025 – 13 Nov 2026, whose regression R² is 68%. The window: 2 Jan 2026 – 16 Nov 2026, 220 sessions; annualised volatility 14.1%.
About this exhibit
Through time
What made up the return over every trailing window, week by week, beside the return itselfExhibit 3 · Factor contributions to the S&P 500's rolling 3-month return, weekly
About this exhibit
Exposures
One factor's exposure across every asset with its 90% band, and how the selected asset's largest exposures have movedExhibit 4 · Each asset's exposure to the selected factor
Loading S&P 500…
About this exhibit
Exhibit 5 · S&P 500 exposure to its largest risk factors, week by week
Loading S&P 500…
About this exhibit
57 of the 58 assets are responding clearly to a factor this year
Which price is each asset responding to over the window chosen above, with the other factors held fixed?The factor table above splits the return in one order. This section asks which price each asset is following, with the other factors held fixed.
A factor's t says how clearly: beyond ±2 its response is distinguishable from zero. A factor is named only when that holds in at least 60% of 400 redraws of the window.
Every chart here follows the asset, window and factor set chosen at the top of the page. With the US market in, the market is held fixed and the cards rank what moves the asset beyond it.
The factors are raw prices, so they are not the table's: the rates are split into the real 10-year yield, breakeven inflation and the 2-year yield.
Exhibit 6 · How clearly the S&P 500 moves with each macro factor this year
About this exhibit
Exhibit 7 · The factor moving most clearly with the S&P 500, weekly since 2008
About this exhibit
Exhibit 8 · How clearly each factor moved with the S&P 500, weekly since 2008
About this exhibit
Across the universe
What carries each asset's risk — every asset's daily variance split among the factors, the bill and the residualExhibit 9 · Each asset's daily variance split by factor, this year
Loading S&P 500…
About this exhibit
Appendix
How the model works and what each number means, how well it fits, the checks it passes, and every source110 checks, all pass
One asset, one window, one model: the return taken apart into each factor's contribution — the asset's exposure to the factor each day times the factor's move that day — beside the 3-month bill and the residual; each part's share of the return's day-to-day variance; and the exposures behind them, for this asset and across the universe.
- Approach. A returns-based factor model: each asset's daily return set against the factors' daily returns. There is no portfolio and no benchmark, so the page reports contributions, not attribution, the line the CFA Institute's 2025 exposure draft on attribution draws.
- Factors and their order. The market and the 10-year yield are taken as they are. Credit, oil and the dollar are measured after what the market and the 10-year explain, the secondary macro factors after those, and the styles after both, so no move is counted twice. Each factor's proxy and the factors it is measured after are listed under Data.
- Return basis. Daily log total returns less the 3-month Treasury bill accrued daily. The bill is named, as GIPS asks, and is the cash line.
- Exposures. Least squares on the 252 sessions ending the day before each session an exposure is applied to, so no estimate reads the day it explains. Newey–West standard errors with 4 lags; the band is ±1.645 standard errors, 90%. Funds that close abroad take each factor today and yesterday, the exposure being the sum (Dimson's design).
- Contribution to return. Over a window, each day's exposure times the factor's move that day, summed, beside the bill and the residual. The log parts are put on the simple return by one scale per window, (e^R − 1) ÷ R, so they add to the return printed; a style spread stays in log points.
- Contribution to risk. Each part's covariance with the return divided by the return's variance over the same sessions: exposure × volatility × correlation (the EIPC guidance). The shares of the factors, the bill and the residual add to 100%; a negative share offsets the rest.
- Residual. Shown on its own and never spread over the factors (the EIPC guidance). Its regression constant is printed a year with its t: alpha on the set with the market, the regression constant on macro prices only. No other number is called alpha.
- Significance. Rows whose exposure cannot be told from zero, |t| below 1.96, are faded. No exposure is described as a signal.
- Windows. One month, three months, the year to date, 12 months: the windows State › Performance prints. Cumulative, not annualised.
1 · Returns, the cash leg and the factors
| Symbol | Meaning | From | Assumed or estimated |
|---|---|---|---|
| the asset's daily log return, % (Yahoo Finance adjusted closes, NYSE sessions) | Yahoo Finance | – | |
| 3-month and 10-year Treasury par yields, % | U.S. Treasury | – | |
| the cash leg (zero for a self-financing style spread) | computed here | – | |
| factor k's raw proxy: returns in %, the yield change per +10bp; the full list is the spec's factor table | Yahoo Finance; U.S. Treasury; Federal Reserve Board (H.10) | – |
Two lenses: the equity factor set carries the market factor; the macro factor set does not.
2 · Residualised factors, point in time
| Symbol | Meaning | From | Assumed or estimated |
|---|---|---|---|
| factor k's parents in the tier order (the market and the 10-year are the core and are not residualised) | the spec | assumed | |
| the slopes fixed at the previous month-end | OLS | estimated | |
| the fit's intercept: kept in the factor (only the parents' slopes come out) | OLS | estimated |
Refreshed at each month-end on the trailing 756 sessions; a residualised parent enters residualised. The form that also removes the intercept is the printed sensitivity.
3 · Exposures, the daily parts and the residual
| Symbol | Meaning | From | Assumed or estimated |
|---|---|---|---|
| the asset's exposure to factor k applied on session t, estimated by OLS on the 252 sessions ending 1 session before t | OLS | estimated | |
| Dimson's lag-1 exposure, for assets whose exchange is shut when New York moves (EFA, EEM, EWJ, FXI, VGK, ACWI); their exposure is the sum of the two | OLS | estimated | |
| the residual: what the factors leave, the regression constant included | computed here | – |
An asset is never explained by a factor that has it, or the spread it is, as a leg. A factor outside its era carries no beta. Newey–West errors (4 lags) on the latest window, where they are printed; the exponentially weighted estimate (half-life 126 sessions) is the sensitivity.
4 · Windows, the display scale and the explained share
| Symbol | Meaning | From | Assumed or estimated |
|---|---|---|---|
| the scale that turns log parts into simple parts adding to the simple return (one for a style spread, printed in log points) | computed here | – | |
| explained share: a variance ratio, not a regression R², allowed to be negative; printed for windows of 21 sessions or more | computed here | – |
The parts, the cash leg and the residual add to the window's return exactly (a check on every window); nothing here is a forecast.
Exhibit 10 · The factors, their parents and their first dates
| Factor | Tier | Proxy | Residualised on | Its beta is | Lenses | First date |
|---|---|---|---|---|---|---|
| US total market | 1 | VTI log return − rf_d | equity: core (not residualised) | equity: total effect under the graph | equity | equity 2 Jan 2014 |
| 10-year yield | 1 | Δ 10-year par yield, bp ÷ 10 | equity: core (not residualised) · macro: core (not residualised) | equity: total effect under the graphmacro: total effect under the graph | equity, macro | equity 2 Jan 2014 · macro 2 Jan 2014 |
| Credit | 1 | HYG log returncomparison: HYG − LQD | equity: US total market, 10-year yield · macro: 10-year yield | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Jan 2017 · macro 3 Jan 2017 |
| Oil | 1 | USO log return | equity: US total market, 10-year yield · macro: 10-year yield | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Jan 2017 · macro 3 Jan 2017 |
| Dollar | 1 | UUP log return | equity: US total market, 10-year yield · macro: 10-year yield | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Jan 2017 · macro 3 Jan 2017 |
| Breakeven inflation | 2 | Δ (10-year par − 10-year TIPS par), bp ÷ 10 | equity: US total market, 10-year yield, Credit, Oil, Dollar · macro: 10-year yield, Credit, Oil, Dollar | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Feb 2020 · macro 3 Feb 2020 |
| Emerging − developed | 2 | EEM − EFA log return | equity: US total market, 10-year yield, Credit, Oil, Dollar · macro: 10-year yield, Credit, Oil, Dollar | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Feb 2020 · macro 3 Feb 2020 |
| Gold | 2 | GLD log return | equity: US total market, 10-year yield, Credit, Oil, Dollar · macro: 10-year yield, Credit, Oil, Dollar | equity: direct effect, holding its parents fixedmacro: direct effect, holding its parents fixed | equity, macro | equity 3 Feb 2020 · macro 3 Feb 2020 |
| Short volatility | 2 | SVXY log return | equity: US total market | equity: direct effect, holding its parents fixed | equity | equity 1 Apr 2021proxy starts 1 Mar 2018 |
| Size | 3 | IWM − SPY | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Value | 3 | VTV − VUG | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Momentum | 3 | MTUM − SPY | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Quality | 3 | QUAL − SPY | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Low volatility | 3 | USMV − SPY | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Breadth | 3 | RSP − SPY | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| Cyclicals − defensives | 3 | STYLE_DEF CYC-DEF | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
| AI complex − defensives | 3 | STYLE_DEF AI-DEF | equity: US total market, 10-year yield, Credit, Oil, Dollar, Breakeven inflation, Emerging − developed, Gold, Short volatility | equity: direct effect, holding its parents fixed | equity | equity 1 May 2024 |
- CreditThe credit factor's form, pending the author's decision.
The credit factor is HYG's own return, residualised on the equity market and the 10-year yield in the equity factor set and on the 10-year yield alone in the macro factor set. The macro factor set used to remove the market from it too, without the market being a regressor there: the month-end coefficients fixed HYG's equity beta, and when HYG's realised beta fell below the stored one the factor carried the market with the wrong sign — in that form S&P 500's macro credit beta is −2.21 (t −4.7), against −0.07 in the equity factor set. With only rates removed, the factor keeps HYG's equity-like component, the structural link between a firm's equity and its credit, and says so: its correlation with the market is printed beside it. The first form is printed as the sensitivity. This is pending the author's decision on the credit factor's form, and the page will say so until it is taken. HYG − LQD is the printed comparison.
- ThresholdWhen the parts are not read as drivers.
Every explained amount is printed with the explained share of its own window. Below 20% the parts are printed and greyed but not read as drivers: the headline says the factors do not explain the window, the waterfall greys, and in the cross-section the asset has no dominant factor. A window shorter than 21 sessions has no share, and the threshold is not assessed.
- VariantsA second point-in-time estimate.
Every window is also decomposed with a second point-in-time estimate of the betas: the same 252-session window ending the session before, with weights halving every 63 sessions (Faster-forgetting estimate (exponential weights, 63-session half-life)). It forgets a regime that has ended faster than the flat window, but it is estimated on the same past, so it cannot know a break before it happens. A part is robust when the two estimates give it the same sign, and marked not robust otherwise.
- BreaksWindows split at a dated break.
A window that contains a dated break — the Strait closed, 27 Feb 2026 — is also printed split at it, the halves on the window's own scale so that they add to it; a check asserts it part by part. The betas that applied in the first weeks after a break were estimated on the regime before it, which is what the split lets a reader see.
- InterceptsThe other residualisation convention.
The other convention — the earlier form — takes the fit's intercept out as well, so a factor loses its own trailing mean, net of its parents'. It is printed as a sensitivity for the page's default asset: each part split into the factor's move and its intercept part (−β·b0), with the asset's own regression constant, inside the residual, beside them. Over a long window the two means can outweigh everything the factors did, and in that form a part that is mostly its intercept is not read as a driver. The regression constant is reported inside the residual in both forms.
- LeakageWhether a factor carries the market.
Each factor's correlation with the market excess return (VTI less the 3-month rate) over the sessions it was applied to out of sample — the last year and the year to date. A factor whose correlation exceeds 0.20 in either window is flagged, in either factor set and whatever its construction: it moves with the market. On an asset that tracks the market more closely than it tracks that factor over a window, the factor's part is mostly the market's move under the factor's name — labelled so, and never called the top driver or the dominant factor. That is the S&P 500's credit part on the macro factor set, which has no equity factor; it is not Brent's oil part or the euro's dollar part, where the factor is the asset's own driver. The in-window orthogonality check cannot see any of this, because a fit is orthogonal to its parents on the window it was fitted on by construction.
- Oil signThe oil part against the stock–oil correlation.
For the page's default asset, the sign of the oil part since the Strait closed is compared with the sign the 63-day stock–oil correlation of the correlations page implies (a reference copy, re-read at each update): a negative correlation and a rise in oil imply a negative part. When they disagree the page says so.
- ConsistencyThe oil sign and the market in the factors, at this update.
S&P 500's oil part since 27 Feb 2026, on the macro factor set, is −0.7pp and the 63-day SPY–USO correlation is −0.36, with oil up: the signs agree; on the equity factor set 10 factors move with the market beyond ±0.20 (10-year yield −0.28, credit −0.49, oil −0.45, emerging − developed +0.47, gold +0.29, momentum +0.35, quality −0.24, breadth −0.21, cyclicals − defensives +0.29, AI complex − defensives +0.30); on the macro factor set 6 factors move with the market beyond ±0.20 (10-year yield −0.28, credit +0.78, oil −0.42, dollar −0.25, emerging − developed +0.50, gold +0.38). In the macro factor set the credit factor removes only rates until the author decides its form. With the market removed as well — printed as the sensitivity — S&P 500's macro-factor credit beta is −2.21 (t −4.7), its year-to-date credit part −3.4pp and its explained share 14.8%, against +7.7pp and 65.8% here. In that form credit's correlation with the market over the year is −0.49: the stale market coefficient carries the market into credit with the wrong sign.
- GraphThe declared factor graph.
The hierarchy the factors are residualised in is a directed acyclic graph, declared on 3 Oct 2026: in the equity factor set the market and the 10-year yield are its roots (94 edges over 17 factors), in the macro factor set the 10-year alone (15 edges over 7). Every edge carries the one-sentence claim behind it; the first recorded edges are credit's parents (rates → credit, mkt → credit), the author's decision of 3 Oct 2026 (D-2026-10-03-G). A declared graph is a hypothesis: the independences it implies are tested on the weekly factors and printed, and three discovery algorithms are asked whether the data contradict it — they review the graph, they never set it.
- KindsWhat a quoted beta is.
What a quoted beta is, in one word beside it. A root's beta is the total effect of that factor under the graph: its children are residualised on it, so what flows through them is counted on the root. A child's beta is the direct effect holding its parents fixed. Both are causal only if the graph is, which is why the chip says 'under the graph'. The weekly five-factor model on the Factor exposures page residualises nothing, so its betas are associations — the asset moved with the factor — and are labelled so. Only a response to an identified shock (a move on a day whose cause is known from its timing) earns a causal verb.
- TestsThe independences the graph implies, and discovery as a reviewer.
The implied independences: for every pair of factors the graph does not join, the partial correlation given both factors' parents, with a Fisher-z p-value and a Newey–West p-value from the slope of one residual on the other. A failure is a missing edge or a wrong orientation — a finding the table prints, not a fix applied to the graph. Discovery: PC and FCI (constraint-based, Fisher-z), DirectLiNGAM (a non-Gaussian ordering) and PCMCI (lagged and contemporaneous partial correlations) on the weekly factors standardised by their GARCH(1,1)-t conditional volatility, because volatility clustering is otherwise read as the non-Gaussianity an ordering rests on. On daily financial data these algorithms need faithfulness and causal sufficiency, which co-moving factors violate; what is printed is the agreement with the declared graph, edge by edge. equity factor set: 18 of 42 implied independences hold at the 10% level on 454 weeks from 9 Mar 2018 to 13 Nov 2026; PC found 11 of 94 declared edges, orienting 2 the declared way and 4 the other; FCI found 10 of 94 declared edges, orienting 1 the declared way and 1 the other; DirectLiNGAM found 39 of 94 declared edges, orienting 25 the declared way and 14 the other; PCMCI found 74 of 94 declared edges, orienting 0 the declared way and 0 the other. macro factor set: 1 of 6 implied independences hold at the 10% level on 671 weeks from 10 Jan 2014 to 13 Nov 2026; PC found 6 of 15 declared edges, orienting 1 the declared way and 4 the other; FCI found 6 of 15 declared edges, orienting 0 the declared way and 1 the other; DirectLiNGAM found 7 of 15 declared edges, orienting 5 the declared way and 2 the other; PCMCI found 10 of 15 declared edges, orienting 0 the declared way and 0 the other.
- ShocksThe shock table by regime state.
The shock table by regime state: each identified shock is assigned the regime state of the week it falls in (the Correlations & regime page's model, the number of states chosen by BIC), and the table counts shocks per kind per state with their mean size. A description of when the identified news arrived, under the author's decision that the regime strip conditions nothing (D-2026-10-03-14).
About this exhibit
Exhibit 11 · Explained share, through time
About this exhibit
Exhibit 12 · Do the factors carry the market?
| Factor | Form | Residualised on | Reaches the market | Correlation, 1Y | Correlation, YTD | Flag |
|---|---|---|---|---|---|---|
| 10-year yield | this page | core — not residualised | no | −0.28 | −0.28 | moves against the market (ρ −0.28) |
| Credit | this page | 10-year yield | no | +0.77 | +0.78 | moves with the market (ρ +0.78) |
| Oil | this page | 10-year yield | no | −0.39 | −0.42 | moves against the market (ρ −0.42) |
| Dollar | this page | 10-year yield | no | −0.25 | −0.24 | moves against the market (ρ −0.25) |
| Breakeven inflation | this page | 10-year yield, Credit, Oil and Dollar | no | +0.06 | +0.07 | — |
| Emerging − developed | this page | 10-year yield, Credit, Oil and Dollar | no | +0.47 | +0.50 | moves with the market (ρ +0.50) |
| Gold | this page | 10-year yield, Credit, Oil and Dollar | no | +0.37 | +0.38 | moves with the market (ρ +0.38) |
| Credit | sensitivity: credit on the market and rates | US total market and 10-year yield | yes | −0.49 | −0.45 | moves against the market (ρ −0.49) |
| Breakeven inflation | sensitivity: credit on the market and rates | 10-year yield, Credit, Oil and Dollar | yes | +0.23 | +0.23 | moves with the market (ρ +0.23) |
| Emerging − developed | sensitivity: credit on the market and rates | 10-year yield, Credit, Oil and Dollar | yes | +0.46 | +0.50 | moves with the market (ρ +0.50) |
| Gold | sensitivity: credit on the market and rates | 10-year yield, Credit, Oil and Dollar | yes | +0.38 | +0.40 | moves with the market (ρ +0.40) |
The credit factor's other form. The macro factor set constructs credit on 10-year yield alone, pending the author's decision on the factor's form. Constructed on US total market and 10-year yield — the other form, printed as the sensitivity — S&P 500's macro-lens credit beta is −2.21 (t −4.7, 13 Nov 2025 to 13 Nov 2026), and year to date its credit part is −3.4pp and its oil part +9.3pp, with the factors accounting for 14.8% of the variance, below the threshold: in that form the year is not explained by the factors. In this page's form the same numbers are +2.71 (t +9.6), +7.7pp, −1.2pp and 65.8%.
About this exhibit
Exhibit 13 · S&P 500: the factors with their own means taken out, YTD
| Part | This page | With means taken out | of which the intercept | of which the move |
|---|---|---|---|---|
| US total market | +5.9 | +5.9 | ±0.0 | +5.9 |
| 10-year yield | +0.2 | +0.2 | ±0.0 | +0.2 |
| Credit | ±0.0 | +0.3 | +0.3 | ±0.0 |
| Oil | ±0.0 | +0.1 | ±0.0 | ±0.0 |
| Dollar | ±0.0 | ±0.0 | ±0.0 | ±0.0 |
| Breakeven inflation | ±0.0 | ±0.0 | ±0.0 | ±0.0 |
| Emerging − developed | −0.1 | −0.1 | ±0.0 | −0.1 |
| Gold | +0.1 | +0.1 | +0.1 | +0.1 |
| Short volatility | ±0.0 | +0.1 | ±0.0 | ±0.0 |
| Value | −1.1 | −1.1 | +0.1 | −1.2 |
| Cyclicals − defensives | ±0.0 | ±0.0 | ±0.0 | ±0.0 |
| AI complex − defensives | −0.1 | −0.1 | ±0.0 | ±0.0 |
| The factors' parts | +5.0 | +5.4 | +0.5 | +5.0 |
| Cash | +3.5 | +3.5 | ||
| Alpha, the residualof which the regression constant | +0.9+1.0 | +0.5+0.7 | ||
| Return | +9.4 | +9.4 | ||
| Explained share | 99% | 99% |
About this exhibit
Exhibit 14 · The declared graph against the data: 18 of 42 implied independences hold
| Method | Declared edges found | Oriented the declared way | The other way | Undecided |
|---|---|---|---|---|
| PC (constraint-based) | 11 / 94 | 2 | 4 | 5 |
| FCI (latent confounders allowed) | 10 / 94 | 1 | 1 | 8 |
| DirectLiNGAM (non-Gaussian ordering) | 39 / 94 | 25 | 14 | 0 |
| PCMCI (time-series, lag 0) | 74 / 94 | 0 | 0 | 74 |
Independences the data reject — pairs the graph does not join whose partial correlation, given both factors' parents, is not zero at the 10% level.
| Pair | Given | Partial correlation | p |
|---|---|---|---|
| 10-year yield · Short volatility | US total market | +0.20 | 0.008 |
| Credit · Dollar | US total market, 10-year yield | −0.22 | 0.002 |
| Dollar · Short volatility | US total market, 10-year yield | +0.16 | 0.006 |
| Breakeven inflation · Gold | Credit, US total market, Oil, 10-year yield, Dollar | +0.19 | 0.004 |
| Emerging − developed · Gold | Credit, US total market, Oil, 10-year yield, Dollar | +0.12 | 0.036 |
| Gold · Short volatility | Credit, US total market, Oil, 10-year yield, Dollar | −0.12 | 0.009 |
| Size · Value | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.33 | 0.000 |
| Size · Quality | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.29 | 0.000 |
| Size · Breadth | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.61 | 0.000 |
| Size · Cyclicals − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.21 | 0.000 |
| Value · Low volatility | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.50 | 0.000 |
| Value · Breadth | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.79 | 0.000 |
| Value · Cyclicals − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.33 | 0.000 |
| Value · AI complex − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.62 | 0.000 |
| Momentum · Quality | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.13 | 0.058 |
| Momentum · Breadth | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.26 | 0.000 |
| Momentum · AI complex − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.18 | 0.022 |
| Quality · Low volatility | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.19 | 0.007 |
| Quality · Breadth | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.37 | 0.000 |
| Low volatility · Breadth | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.42 | 0.000 |
| Low volatility · Cyclicals − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.68 | 0.000 |
| Low volatility · AI complex − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.67 | 0.000 |
| Breadth · AI complex − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | −0.39 | 0.000 |
| Cyclicals − defensives · AI complex − defensives | Credit, Emerging − developed, Gold, Breakeven inflation, US total market, Oil, 10-year yield, Short volatility, Dollar | +0.72 | 0.000 |
The identified shocks by regime state — each shock assigned the state of the week it fell in (6 states, the Correlations & regime page's model, chosen by BIC), 399 shocks 19 Sept 2014 – 13 Nov 2026. A description of when the identified news arrived; the strip conditions nothing.
| State | Weeks | Shock | n | Mean size | Mean |
|---|---|---|---|---|---|
| bonds neutral, oil with stocks | 56 | fomc_d1y1yMonetary policy | 8 | 13.24 bp | −4.49 |
| bonds neutral, oil with stocks | 56 | fomc_d2yMonetary policy | 8 | 9.88 bp | −3.88 |
| bonds neutral, oil neutral | 89 | gdp_newsGrowth news | 5 | 0.07 pp | −0.01 |
| bonds neutral, oil neutral | 89 | payrolls_newsGrowth news | 5 | 17.64 k | +16.95 |
| bonds neutral, oil neutral | 89 | cpi_newsInflation news | 4 | 0.06 pp | +0.06 |
| bonds neutral, oil neutral | 89 | oil_eventOil supply | 12 | 2.90 % | +0.47 |
| bonds neutral, oil neutral | 89 | oil_inventoryOil supply | 52 | 0.94 sd | −0.05 |
| bonds neutral, oil neutral | 89 | fomc_d1y1yMonetary policy | 13 | 8.15 bp | +2.16 |
| bonds neutral, oil neutral | 89 | fomc_d2yMonetary policy | 13 | 6.23 bp | +1.62 |
| bonds neutral, oil neutral | 89 | premium_dTerm premium | 28 | 4.48 bp | +1.35 |
| bond hedge weak, oil with stocks · dollar against gold · dispersed (R2) | 91 | fomc_d1y1yMonetary policy | 15 | 3.60 bp | +1.87 |
| bond hedge weak, oil with stocks · dollar against gold · dispersed (R2) | 91 | fomc_d2yMonetary policy | 15 | 2.13 bp | +1.20 |
| bond hedge weak, oil with stocksnow | 58 | oil_eventOil supply | 2 | 4.12 % | +4.12 |
| bond hedge weak, oil with stocksnow | 58 | oil_inventoryOil supply | 39 | 0.87 sd | +0.18 |
| bond hedge weak, oil with stocksnow | 58 | fomc_d1y1yMonetary policy | 10 | 7.90 bp | −0.71 |
| bond hedge weak, oil with stocksnow | 58 | fomc_d2yMonetary policy | 10 | 4.60 bp | +0.40 |
| bond hedge weak, oil with stocksnow | 58 | premium_dTerm premium | 6 | 4.75 bp | +0.06 |
| bond hedge weak, oil with stocks · dollar against gold · dispersed (R4) | 179 | fomc_d1y1yMonetary policy | 30 | 10.09 bp | −4.36 |
| bond hedge weak, oil with stocks · dollar against gold · dispersed (R4) | 179 | fomc_d2yMonetary policy | 30 | 6.77 bp | −3.30 |
| bond hedge weak, oil with stocks · dollar against gold · dispersed (R4) | 179 | premium_dTerm premium | 2 | 4.20 bp | +4.20 |
| bond hedge working, oil with stocks | 162 | fomc_d1y1yMonetary policy | 46 | 5.36 bp | −0.39 |
| bond hedge working, oil with stocks | 162 | fomc_d2yMonetary policy | 46 | 3.37 bp | −0.50 |
About this exhibit
Oil sign check · S&P 500's oil part since 27 Feb 2026, on the macro factor set, is −0.7pp and the 63-day SPY–USO correlation is −0.36 on 16 Nov 2026, with oil up over the window: the signs agree.
Exhibit 15 · Factor sensitivity: definitions
| Factor | The series behind it |
|---|---|
| Real 10-year yield | change in the 10-year TIPS par yield, bp |
| Breakeven inflation | change in the 10-year par yield less the 10-year TIPS par yield, bp |
| 2-year yield | change in the 2-year Treasury par yield, bp |
| Spot dollar | UUP less its trailing gap to the H.10 index, log % |
| Oil | USO, log % |
| High yield over investment grade | HYG less LQD, log % |
| Emerging − developed | EEM less EFA, log % |
| US total market | VTI less the 3-month bill accrued daily, log % |
- Windows
The page's window sets the sessions every card reads: one month is 21 sessions, three months 63, the year so far 219 on 13 Nov 2026, one year 252. The history charts roll the same number of sessions back from each Friday. Newey–West lags and redraw blocks grow with the window: 1M 2 lags, blocks of 5; 3M 4 lags, blocks of 10; YTD 4 lags, blocks of 15; 1Y 5 lags, blocks of 21. One month is 21 sessions for up to eight factors, so the redraws seldom hold a factor there.
- Response
The factor's beta in an OLS of the asset's daily log return on every factor of the set at once over the window, so each is read with the others held fixed. It is quoted per 10bp rise in a yield and per 1% rise in a price. None of the factors is adjusted for another; a factor made from the asset's own price is left out.
- t
The response over its Newey–West standard error (Bartlett, no small-sample correction, as statsmodels computes it). Beyond ±2 the response is distinguishable from zero at about the 5% level. A t ranks how clearly the asset moves with a factor, not how much of its return the factor made.
- Redraws
400 moving-block resamples of the window's sessions, each refitted: how often each factor's |t| is at least 2 ("clear in"), and how often its |t| is the largest of the ranked factors ("first in").
- Named factors
A title names a factor when its |t| is at least 2 and it is clear in at least 60% of the redraws, in order of |t|. When the top two are within 1 of each other, the one first in more redraws goes first. "Responding to" describes co-movement over the window, not cause.
- With the US market
The market is a control: it is in every regression, so each other response is read beyond it, but it is never ranked, named, a leader or drawn. An index fund is mostly the market itself; the table prints how much the market alone explains, and how much the other factors add.
- The rates together
A Wald test that the responses to the real 10-year yield, breakeven inflation and the 2-year yield are all zero, on the same Newey–West covariance (χ² with 3 degrees of freedom); distinguishable when p is under 0.05. Rates that move together can be clear jointly when none is clear alone.
- Change against the months before
One OLS over the window and the 189 sessions before it, in which every response may differ in the window: the earlier sessions' response, the window's (equal to the window's own OLS) and the t of the difference (5 lags). A change is named when that t is beyond ±2.
- Leader
Each Friday, the ranked factor with the largest |t| over the window's length of sessions to it, when that |t| is at least 2; otherwise no factor leads that week.
- Unit change and the rest
Gold and silver are priced in dollars, so a 1% rise in the dollar takes 1% off their dollar price by itself. With the spot dollar among the factors, pricing the metal in the dollar basket moves only the dollar's response, by exactly +1; every other response and t is unchanged. The table's second dollar line is that basket response: the part of the dollar's pull that is not the unit change.
- Order-free share
Each factor's average addition to the explained share over every order in which the factors could enter, divided by the explained share of all of them (LMG, the Shapley value of R²). The shares add to one.
- Dominant
The factors explain at least 20% of the daily variance over 252 sessions; the leader ranks first in at least 80% of 400 moving-block resamples of the window (blocks of 21 sessions) and is their most frequent leader; and it adds more than the runner-up on average at every number of factors (Budescu's conditional dominance) or to every combination of the others (complete).
- Leading
The explained share is at least 20% and the leader is the resamples' most frequent leader, short of the first-place share or the level.
- No dominant factor
The explained share is under the floor, or the resamples most often rank another factor first.
- A window's parts
Each day's fitted move, with exposures from the 252 sessions before it, split over every order of the factors: a subset is worth the move its observed factors imply for the rest through their covariance. Summed over the window and put on its simple-return scale, (e^R − 1)/R, so the parts add to the return. The band is 90% from 400 draws of the exposures from their Newey–West covariance at the window's end (4 lags); the shaded range is the part over every order; the diamond is the part in the declared order: the real 10-year yield, breakeven inflation, the 2-year yield, the spot dollar, oil, high yield over investment grade, the emerging-over-developed spread, the US market.
- A window's verdict
Dominant when the largest part is the largest in at least 80% of the draws and the declared order agrees; leading otherwise; no dominant factor when the factors explain under 20% of the asset's daily variance on the latest week (gold in dollars reads gold in the basket's).
- The spot dollar
UUP's return carries its collateral income, futures roll and fee. The spot dollar is UUP less the trailing 252-day mean of its daily gap to ICE's six-currency index, made from the Federal Reserve's H.10 noon rates, lagged a day. H.10 runs to 6 Nov 2026; the gap is held at its last value after that. UUP less the index, by calendar year (log points): 2020 +0.1, 2021 −1.1, 2022 +2.5, 2023 +6.2, 2024 +4.8, 2025 +4.5, 2026 +3.5.
- Gold or silver in the dollar basket
The metal's log return plus the spot dollar's: the metal priced in the currencies the dollar is measured against. The metal in dollars is the metal in the basket less the dollar every day, so in a window's parts the dollar's move is a row of its own and only the basket is split.
Exhibit 16 · Each factor's share of the S&P 500's explained variance, weekly since 2008
The strip beneath: a dominant factor that weeka leading factor, short of dominanceno dominant factor
Week to 13 Nov 2026: The emerging-over-developed spread leads with 29% of what the factors explain, short of dominant. It ranks first in 62% of 400 resampled years, and it adds more than oil on average however many factors are in.
About this exhibit
Exhibit 17 · Each factor's part of the S&P 500's return this year, with ranges
The bar is the factor's part, averaged over every orderthe whisker its 90% bandthe shaded span its lowest and highest part over every orderthe diamond its part in the declared order
Dominant. The real 10-year yield is the largest part at −4.4pp in 92% of 400 redraws of the exposures: dominant. Over every order of the factors its part runs from −12.1 to +1.2pp.
About this exhibit
Exhibit 18 · More of the reasoning: which factor each asset is responding to
- Gold nowGold in dollars fell 2.23% for each 1% rise in the spot dollar over the past three months to 13 Nov 2026.
Gold is responding to the dollar (t −4.0) beyond the market. That response stays distinguishable from zero in 90% of 400 redraws of the window. One point of it is the unit change: priced in the dollar basket, gold moved −1.23% per 1%.
Gold → Exhibit 6
- Gold and real yieldsGold in dollars rose 0.70% for each 10bp rise in the real 10-year yield over the past three months to 13 Nov 2026.
That response is not distinguishable from zero (t +0.8). The real yield's t was beyond ±2 in 37% of weeks from 2008 to 2021, and in 19% since.
Gold → Exhibit 8
- Clear responses57 of the 58 assets respond clearly to at least one factor this year, on macro prices only.
Clearly means a t beyond ±2 that stays there in at least 60% of 400 redraws of the window. Every chart in the section follows the asset, window and factor set chosen at the top.
- The S&P 500On macro prices only, the S&P 500 fell 0.08% for each 1% rise in oil this year to 13 Nov 2026.
The S&P 500 responds to oil (t −6.2), credit and five more. With the market in, the market alone explains 99% of the index's daily moves, so the index is read on macro prices only.
S&P 500 → Exhibit 6
About this exhibit
How the numbers are made
Yahoo Finance chart API, daily adjusted closes (ETF prices) ──▶ the return attribution: daily log returns, excess of the accrued 3-month par yield ──▶ excess returns excess returns + U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates and Par Real Yield Curve Rates + Board of Governors of the Federal Reserve System, H.10 Foreign Exchange Rates (nominal broad dollar index), Data Download Program ──▶ factor proxies (each listed under Sources) ──▶ raw factors raw factors ──▶ the factors made independent of each other: month-end OLS on the parents, applied out of sample ──▶ residualised factors excess returns + residualised factors ──▶ rolling sensitivities: 252 sessions ending t − 1 (Dimson lags for foreign closes) ──▶ point-in-time betas point-in-time betas + residualised factors ──▶ the contributions: β × factor, cash, residual ──▶ daily decomposition daily decomposition ──▶ the attribution over each window and the simple-return parts: sum over each tenor, scaled to the simple return ──▶ the page's numbers excess returns + residualised factors ──▶ rolling sensitivities with a 63-session half-life, over the sessions the page's windows read (the second point-in-time estimate) ──▶ variant betas variant betas + daily decomposition ──▶ the split at a dated break, the same-sign rule and the explained-share threshold: the split at the dated break, the robustness marks and the explained-share threshold ──▶ the page's marks residualised factors + raw factors ──▶ each factor's correlation with the market: each residualised factor against the market over the sessions it was applied to ──▶ the leakage test
Sources
- licensedYahoo Finance chart API, daily adjusted closes (ETF prices)
Prices: Yahoo Finance. · https://query2.finance.yahoo.com/v8/finance/chart/
- publicU.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates and Par Real Yield Curve Rates
Source: U.S. Department of the Treasury · https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?type=daily_treasury_yield_curve
- publicBoard of Governors of the Federal Reserve System, H.10 Foreign Exchange Rates (nominal broad dollar index), Data Download Program
Source: Board of Governors of the Federal Reserve System · https://www.federalreserve.gov/datadownload/
- publicKenneth R. French, Data Library: daily Fama–French 5 factors and momentum
Kenneth R. French Data Library · https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
- computed hereThe site's own analysis (Macro Review)
Computed here on the sources listed.
Reproduction checks
| Check | Published | Computed | Tolerance | Result |
|---|---|---|---|---|
| Fundamental split: for every window, earnings plus the multiple add to the log price change, and the 10-year and the premium add to the multiple, within the three decimals the valuation page publishes (5 windows, worst residual) | 0 | 0.001 | 0.0015 | pass |
| Causal layer (equity): the partial correlations of the implied independences equal the correlation of the two residuals on the conditioning set, recomputed with numpy (first 8 tests; max abs difference) | 0 | 0 | 0 | pass |
| Causal layer (macro): the partial correlations of the implied independences equal the correlation of the two residuals on the conditioning set, recomputed with numpy (first 6 tests; max abs difference) | 0 | 0 | 0 | pass |
| Causal layer: every node's parents in the declared graphs equal the parents the residualisation ran on, both factor sets (109 edges; nodes that differ) | 0 | 0 | 0 | pass |
| Causal layer: every row of the shock table has a source dated on or before the shock (rows that look ahead) | 0 | 0 | 0 | pass |
| Causal layer: the shocks assigned to a regime state plus those after the strip's last week equal the shock table's rows (399 + 0 against 399) | 399 | 399 | 0 | pass |
| Every window's total log return equals ln(P_end ÷ P_before-start) from the adjusted closes (worst: BTC-USD 1W) | 0 | 0 | 0 | pass |
| Additivity in logs: the factor parts, the cash leg and the residual add to the total log return, for every asset, factor set and window | 0 | 0 | 0 | pass |
| Additivity on the display scale: the same parts add to the printed total | 0 | 0 | 0 | pass |
| Point in time: the beta applied on the last session equals an OLS on the 252 sessions ending the day before it (SPY, XLK, TLT) | 0 | 0 | 0 | pass |
| Statsmodels: the latest betas of SPY (macro), XLK (equity) and EEM (macro, Dimson) reproduce | 0 | 0 | 0 | pass |
| Statsmodels: their Newey–West standard errors reproduce | 0 | 0 | 0 | pass |
| Residualisation: at the latest month-end every residualised factor's in-window residual is orthogonal to its parents (worst |corr|: equity.quality~gold) | 0 | 0 | 0 | pass |
| Residualisation: the published coefficients reproduce the published residualised factor on the last session, within the rounding of the six decimals all of them are printed to (21 factors; worst equity.usd at 8.79e-07 against a bound of 1.174e-06; share of that bound) | 0 | 0.7493 | 1 | pass |
| Residualisation: the published coefficients cover every factor that has parents in both factor sets (factors compared) | 21 | 21 | 0 | pass |
| Self-exclusion equals the set derived from the style definitions and the factor proxies (assets that differ) | 0 | 0 | 0 | pass |
| No asset carries a factor it is a leg of (occurrences) | 0 | 0 | 0 | pass |
| The value factor is the negative of the growth-minus-value style spread on every session (max abs difference) | 0 | 0 | 0 | pass |
| Sanity: weekly log changes of UUP and the Board's H.10 broad index correlate at least 0.7 since 2014 (actual 0.835) | 0 | pass | ||
| Legacy bridge: cross.factors' own attribution, recomputed on this page, equals its published snapshot to the three decimals that snapshot carries (72 values, vintage 2026-11-17) | 0 | 0 | 0 | pass |
| The Fama–French correlations are computed from the Data Library, or the panel says the library was not fetched for this date | 0 | pass | ||
| Explained share: every stored share equals 1 − Var(u) ÷ Var(r − cash) recomputed from the window's days (812 windows; worst none differs) | 0 | 0 | 0 | pass |
| Explained share: the gate flag is the share below 0.20 on every window that has a share (windows that differ) | 0 | 0 | 0 | pass |
| Break split: before and after the break add to the window, part by part, in logs, for every window that contains a break under every beta estimate (464 splits) | 0 | 0 | 0 | pass |
| Break split: the two halves, printed on the window's own scale, add to the window's printed parts | 0 | 0 | 0 | pass |
| Beta variant: every window's parts under the second estimate add to the same total | 0 | 0 | 0 | pass |
| Robustness: a part is marked robust exactly when every beta estimate gives it the same sign (marks that differ) | 0 | 0 | 0 | pass |
| Beta variant, point in time: the exponentially weighted beta applied on the last session equals statsmodels WLS on the sessions ending the day before it (SPY, XLK, EEM) | 0 | 0 | 0 | pass |
| Intercepts: in the main form each factor residualised on core factors equals the other form's factor plus its intercept (4 factors); in the sensitivity every part is its move plus its intercept part (12 parts, year to date); and the residual's constant part is the regression constant summed over the window | 0 | 0 | 0 | pass |
| Cross-section: the largest part is the largest in size, and no asset is given a dominant factor that is below the gate, is its own factor's proxy, whose largest part is mostly the market's move, runs against its own co-movement with that factor, or is mostly that factor's own trailing mean (rows that differ) | 0 | 0 | 0 | pass |
| Cross-section: each asset's correlation with its largest factor over the window recomputes, within the rounding of the three decimals it is stored to (464 rows) | 0 | 0.0005 | 0.0005 | pass |
| Cross-section: the co-movement flag is the sign of the largest part against the sign the asset's correlation with that factor and the factor's move imply (flags that differ) | 0 | 0 | 0 | pass |
| Leakage: every stored correlation of a factor with the market recomputes over its window (27 factors × 2 windows) | 0 | 0 | 0 | pass |
| Leakage: a factor is flagged exactly when |corr| with the market exceeds 0.20 in either window, in either factor set and whatever its construction (flags that differ) | 0 | 0 | 0 | pass |
| Market-like parts: every window's correlation of the asset with the market and with each flagged factor recomputes (812 windows) | 0 | 0 | 0 | pass |
| Market-like parts: a part is marked exactly when its factor moves with the market, the asset moves with the market beyond the same threshold, and the asset tracks the market more closely than that factor (windows that differ) | 0 | 0 | 0 | pass |
| Credit factor, macro factor set: residualised on rates as the spec's default (pending G8), with the pack's D3 form printed as the sensitivity | rates | rates | 0 | pass |
| Oil sign: the flag is the sign of the oil part since 2026-02-27 against the sign the stock–oil correlation (-0.36) and oil's move imply (agrees: True) | 0 | pass | ||
| Alpha cards: the asset they open on when the page's asset is not in cross.sml's universe has the largest |t| of the 252-session alphas copied from it (XLE) | 1.758 | 1.758 | 0 | pass |
| Share of variance: every part's Cov(part, r) ÷ Var(r) adds to one across the factors, the bill and the residual (464 asset × factor set × window; worst UUP equity 1Y) | 0 | 0 | 0 | pass |
| Share of variance: the factors' shares add to the covariance of their summed daily part with the return, recomputed from the daily decomposition (worst TLT macro 1Y) | 0 | 0 | 0 | pass |
| Average exposure applied × the factor's move × the window's scale = the printed contribution (5252 values; worst USO equity 1Y rates) | 0 | 0 | 0 | pass |
| The factor table's group subtotals add to the factors' contributions (worst XLE equity 1Y) and, with the bill and the residual, their shares to one | 0 | 0 | 0 | pass |
| Exhibit 1's rows — the largest parts, the rest as one row, the bill and the residual — add to the window's return and their shares to one (worst SLV macro 1Y) | 0 | 0 | 0 | pass |
| Contributions through time: each rolling window's last point equals the window the factor table prints, series by series (116 asset × factor set; worst BNO equity 1Y r) | 0 | 0 | 0 | pass |
| Contributions through time: at every point the series, the bill and the residual add to the window's return | 0 | 0 | 0 | pass |
| Exposures through time: the last point is the exposure the factor table prints today (the regression ending the day before the last session; worst None) | 0 | 0 | 0 | pass |
| One factor across the universe (Exhibit 4): every exposure, 90% band and t is the asset's own factor table's (mismatches) | 0 | 0 | 0 | pass |
| Risk across the universe (Exhibit 6): every row adds to one (232 asset × window rows, each asset on its default factor set) | 0 | 0 | 0 | pass |
| Risk across the universe (Exhibit 6): every cell is the sum of the asset's own published shares (mismatches) | 0 | 0 | 0 | pass |
| Each exposure's standard error is its column's Newey–West error where the design has one column per factor (the Dimson sum's comes from the lag block of the same covariance) | 0 | 0 | 0 | pass |
| Opener: the first takeaway names S&P 500's largest part on the macro factor set as its side file publishes it (Credit +7.7pp of 7 parts; parts that are larger) | 0 | 0 | 0 | pass |
| Opener: S&P 500's parts, cash and residual add to its year-to-date return in the side file it opens on (abs difference, pp; the additivity identity the other windows are held to) | 0 | 0 | 0 | pass |
| Cross.correlations: the source page is published with all its checks passing (dated 17 Nov 2026, 57 checks) | 1 | 1 | 0 | pass |
| Cross.correlations: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Risk sentiment page: the source page is published with all its checks passing (dated 17 Nov 2026, 112 checks) | 1 | 1 | 0 | pass |
| The Risk sentiment page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Performance page: the source page is published with all its checks passing (dated 17 Nov 2026, 23 checks) | 1 | 1 | 0 | pass |
| The Performance page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Valuation page: the source page is published with all its checks passing (dated 17 Nov 2026, 278 checks) | 1 | 1 | 0 | pass |
| The Valuation page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| Cross.regime: the source page is published with all its checks passing (dated 17 Nov 2026, 80 checks) | 1 | 1 | 0 | pass |
| Cross.regime: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| Cross.correlations: every copied observation equals the source page, matched by date (186 values in 1 arrays; mismatches) | 0 | 0 | 0 | pass |
| The Risk sentiment page: every copied single value equals the source page (19 values; mismatches) | 0 | 0 | 0 | pass |
| The Performance page: every copied single value equals the source page (3 values; mismatches) | 0 | 0 | 0 | pass |
| The Valuation page: every copied single value equals the source page (25 values; mismatches) | 0 | 0 | 0 | pass |
| Cross.regime: every copied single value equals the source page (1 values; mismatches) | 0 | 0 | 0 | pass |
| The page file is within the 400,000-byte limit (347,521 bytes) | 0 | pass | ||
| Every side file is within the 400,000-byte limit (121 files; largest residuals.capm at 278,195 bytes) | 0 | pass | ||
| Decisions: every desk decision this page's specification names is in the decisions file (D-2026-10-03-G; ids missing) | 0 | 0 | 0 | pass |
- Yahoo Finance chart API, daily adjusted closes (ETF prices) — Prices: Yahoo Finance.
- U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates and Par Real Yield Curve Rates — Source: U.S. Department of the Treasury
- Board of Governors of the Federal Reserve System, H.10 Foreign Exchange Rates (nominal broad dollar index), Data Download Program — Source: Board of Governors of the Federal Reserve System
- Kenneth R. French, Data Library: daily Fama–French 5 factors and momentum — Kenneth R. French Data Library
- The site's own analysis (Macro Review) — Computed here on the sources listed.