Net liquidity vs S&P 500: historical tests
Study nl_study_v1 · computed 2026-10-20 · sample 2016-10-26 to 2026-10-14 (513 weekly observations)
Over non-overlapping 13-week windows, the correlation between the change in Fed net liquidity and the S&P 500 log return was -0.24 (95% interval -0.52 to 0.08, n = 39). A relationship measured over the same window does not show that liquidity leads prices; the lead tests below check that separately.
USD billions · Wednesday levels · snapshot of 2026-10-14 · S&P 500 on right axis (close on or up to 3 days before each Wednesday)
Contribution to change
| Component | Last week | Selected range |
|---|---|---|
| Fed assets | +8.1bn | −1,246.4bn |
| TGA (subtracted) | 0.0bn | −213.7bn |
| ON RRP (subtracted) | +2.1bn | +1,442.6bn |
| Net liquidity | +10.2bn | −17.5bn |
Last week: 2026-10-07 → 2026-10-14. Range: 2023-09-27 → 2026-10-14.
Scenario from 2026-10-14
Net liquidity $5,869.6bn → $5,869.6bn (0.0bn)
Holding other inputs constant. This is arithmetic on the proxy, not an asset-price forecast.
Keep this chart and your assumptions. Create a free account to reopen this analysis with its selected dates, inputs, and sources. No card required to save.
Underlying indicator · Net liquidity methodology
Results
| Test | Full r (95% CI) | n | Before 2023-01-01 | Held out (from 2023-01-01) | Same sign |
|---|---|---|---|---|---|
| 4-week same window | 0.10 (-0.08 to 0.27) | 128 | 0.18 (n 79) | -0.15 (-0.42 to 0.14, n 48) | 54% |
| 4-week liquidity change → next return | 0.08 (-0.09 to 0.25) | 127 | 0.09 (n 78) | 0.09 (-0.21 to 0.37, n 47) | 52% |
| 13-week same window | -0.24 (-0.52 to 0.08) | 39 | -0.21 (n 24) | -0.59 (-0.85 to -0.08, n 14) | 56% |
| 13-week liquidity change → next return | 0.34 (0.02 to 0.59) | 38 | 0.36 (n 23) | 0.43 (-0.16 to 0.79, n 13) | 63% |
Where the relationship failed
Largest 13-week windows where net liquidity and S&P 500 moved in opposite directions:
| Window | Δ net liquidity | S&P 500 return |
|---|---|---|
| 2020-02-19 → 2020-05-20 | +2,083.8bn | -12.2% |
| 2020-05-20 → 2020-08-19 | −460.1bn | 13.6% |
| 2017-10-25 → 2018-01-24 | −35.0bn | 11.0% |
| 2025-06-11 → 2025-09-10 | −286.5bn | 8.5% |
| 2023-05-24 → 2023-08-23 | −229.6bn | 7.8% |
Design
- Net liquidity: Wednesday WALCL − WDTGAL − RRPONTSYD (USD millions, exact same-date alignment).
- S&P 500: FRED SP500 close on the Wednesday, or the latest close up to 3 days earlier.
- Horizons fixed in advance at 4 and 13 weeks. No lag search was run; no lag is presented as a trading rule.
- Inference uses non-overlapping windows only. Intervals are Fisher-z 95%; with few windows they are wide.
- Held-out period fixed at 2023-01-01 and not tuned.
- Rolling 52-week correlations of weekly changes (461 points) are descriptive only: they overlap and are not used for inference. Latest: 0.11.