Asset driver fingerprints
What actually moves a stock, and what that reveals about its neighbours
Every stock is pushed around by a handful of forces — rates, oil, the dollar, credit, and a few style factors. Regressing 2,068 stocks against 10 drivers over 252 days gives each name a fingerprint: which forces move it, in which direction, and how hard. The headline number is humbling — median systematic R² is 0.22, so about 78% of the average name's movement is explained by nothing on the list. What the fingerprints are good for is relationships: twins in other sectors, names that have quietly left their sector, and holdings that are genuinely independent of each other.
The drivers
Ten forces, specified up front rather than discovered. A name whose real driver is missing from this list will look more idiosyncratic than it actually is — which is part of why the median R² is as low as it is.
Fingerprints
Standardised exposures for five names, sorted from most negative to most positive. Note the scale differs per card — NEM's gold beta of +4.0 would flatten every other card if they shared one axis. Bars are direct-labelled with signed values, so direction is never read from colour alone.
Air transportation
National commercial banks
Gold and silver ores
Petroleum refining
Semiconductors
Cross-sector twins
Names with almost the same fingerprint as the anchor, drawn from a different industry. JPM's closest relatives are insurers and brokers, which is unsurprising; DAL's are apparel and leather, which is not, until you notice they are all consumer-discretionary demand plays with the same dollar and fuel sensitivities.
Names that left their sector
Correlation to a name's own sector ETF, most recent 63 days against the 63 before it. A large fall means the name has stopped trading like its peers — sometimes a company-specific story, sometimes an early sign the sector definition was wrong for it.
| Ticker | Sector ETF | Prior 63d | Recent 63d | Change | Industry |
|---|---|---|---|---|---|
| ACIW | XLK | 0.44 | -0.15 | -0.58 | Prepackaged software |
| SNOW | XLK | 0.59 | +0.03 | -0.56 | Prepackaged software |
| JXN | XLF | 0.79 | +0.23 | -0.56 | Life insurance |
| DOCS | XLK | 0.51 | -0.02 | -0.52 | Computer programming |
| ARMK | XLY | 0.56 | +0.07 | -0.50 | Retail — eating places |
| FLUT | XLK | 0.40 | -0.07 | -0.48 | Data processing |
| PAYO | XLK | 0.47 | +0.00 | -0.47 | Business services |
| EA | XLK | 0.43 | +0.01 | -0.43 | Prepackaged software |
Genuine diversifiers
The inverse question, and the practically useful one: given a basket of AAPL, JPM, XOM, which names are least correlated to it? The answer is dominated by utilities, water and REITs — rate-sensitive, domestically driven, and indifferent to everything the basket cares about.
| Ticker | Corr to basket | R² | Industry | Top drivers |
|---|---|---|---|---|
| AWK | -0.16 | 0.19 | Water supply | rates+, size− |
| QGEN | -0.16 | 0.13 | — | gold+, dollar− |
| FTS | -0.14 | 0.11 | — | gold+, size− |
| HR | -0.13 | 0.15 | REIT | rates+, oil− |
| OHI | -0.13 | 0.12 | REIT | dollar+, size− |
| RELX | -0.13 | 0.10 | — | rates+, gold− |
Limitations
- A median systematic R² of 0.22 means roughly 78% of the average name's movement is idiosyncratic. These fingerprints describe the minority of variance the drivers explain, not the whole of it.
- Betas are measured over one 252-day window and are not stable. The decoupler screen exists precisely because these relationships move.
- Correlation is not mechanism. A gold beta of 4.0 on a name that does not mine gold usually means shared exposure to something else, not a causal link.
- Sector labels come from SIC codes, which are coarse and frequently missing — several twins show a blank sector because the source data has none.
- The driver set is chosen, not discovered. Ten factors were specified up front; a name whose real driver is absent from the list will look more idiosyncratic than it is.
2,068 stocks × 11 factors over 252 days, 2025-07-28 to 2026-06-26. Median systematic R² 0.22.