Market Structure Weekly — Week of March 16, 2026

The VIX term structure re-steepened modestly this week. The front-month / second-month spread widened to 1.1 points from 0.4 last week, driven more by a decline in the front month than by any move in the back. Realized volatility (10-day) on the S&P 500 ticked up to 12.8% from 11.2%, while 30-day implied held steady near 15.5%.

The implied-realized spread narrowed from 4.6 to 2.7 points. That compression came from the realized side catching up, not from implied declining. In our observation, when the spread narrows via rising realized vol rather than falling implied, the market is adjusting to actual movement rather than repricing expectations. The two have different follow-through profiles.

Single-stock IV dispersion remained elevated. The interquartile range of 30-day IV across the S&P 500 widened to 8.4 points, up from 7.1 last week. The market continues to differentiate at the name level even as index-level vol stays range-bound.

Breadth stabilized but did not meaningfully recover. The percentage of S&P 500 constituents above their 50-day moving average held at 49%, essentially flat from 48% last week. New 52-week highs ticked up to 18 from 12, while new lows contracted to 22 from 31.

The advance/decline ratio improved to 0.94 from 0.72. That is no longer a deterioration signal, but it is not confirmation of broad participation either. The tape moved from actively narrowing to neutral. Whether this is a pause before further deterioration or the beginning of a recovery is not something breadth alone can answer at this stage.

The defensive tilt from last week partially reversed. Industrials and Materials led this week, outperforming the index by 1.4% and 1.1% respectively. Utilities gave back most of last week’s relative gain, underperforming by 0.6%.

Technology was mixed at the sub-industry level. Semiconductors outperformed by 0.8%, while Software underperformed by 0.5%. That internal divergence within Technology is worth noting because it suggests the sector is not moving as a single unit. When sub-industries within a sector rotate in opposite directions, it usually reflects fundamental differentiation rather than macro-driven risk-on/risk-off behavior.

Energy showed quiet relative strength for the second consecutive week, outperforming by 0.4%. Not large enough to call a trend, but consistent enough to register.

Intra-sector correlation within Technology pulled back to 0.53 from 0.61, consistent with the sub-industry divergence noted above. When names within a sector start decorrelating, it often means stock-specific catalysts are re-emerging as the dominant driver.

Cross-sector correlation between Industrials and Materials rose to 0.48 from 0.33, suggesting these two are beginning to move more in tandem again. Historically, rising cross-sector correlation between cyclicals tends to reflect a shared macro catalyst — in this case, likely the same growth-expectation repricing.

The average pairwise correlation across all S&P 500 names declined slightly to 0.28 from 0.31. Declining average correlation with stable breadth is a dispersion environment: index-level movement compresses while individual name movement expands. That is generally favorable for strategies that rely on cross-sectional differentiation.

Our momentum scanner flagged an unusual cluster in Energy: 9 names in the sector triggered the 20-day breakout filter within the same 3-session window. The trailing 12-week average for this cluster count is 3. The z-score was 1.9.

Separately, the mean-reversion scanner’s mid-cap industrials cluster from last week (14 observations, z-score 2.1) partially unwound. This week the count dropped to 7, closer to the baseline average of 6. That normalization is consistent with the sector’s relative outperformance this week — oversold names reverted.

As always, scanner output describes statistical clustering, not opportunity. The observation is structural: when a sector shows coordinated momentum breakouts at this frequency, it typically reflects a sector-level re-rating rather than coincidental name-level strength.