Research insight · Issue 002
Volatility peaks before diversification recovers
Volatility recovers faster than market structure.
- Published
- Data through
- 31 July 2026
ContentsExecutive summary
Executive summary
Volatility recovers faster than market structure.
- 01Sector commonality rises beyond the quietest volatility range.
- 02Volatility peaked after 5.5 days. Market structure peaked after 43.
- 03Larger shocks left deeper, longer-lasting common-factor dominance.
- Complete VIX shock events
- 28
- Median VIX peak
- Day 5.5
- Median PC1 peak
- Day 43
- Strongest VIX association
- 60-day maximum
- 0.458
- Period
- 2000–2026
Why it matters
A declining VIX does not necessarily mean diversification has recovered. Stress can retreat while sector returns remain dominated by a common factor, leaving portfolios more structurally dependent than headline volatility suggests.
Analysis 01
When sector independence begins to weaken
Sector commonality rises sharply as volatility moves beyond its quietest range. Median PC1 variance share increases from around 42% at the lowest VIX levels to more than 60% as VIX approaches 15. The relationship then strengthens more gradually through the transition range before rising again at higher volatility levels. Above a VIX of 30, the dominant market factor typically explains more than 70% of standardised sector-return variation. The result suggests that diversification can begin weakening before volatility reaches extreme levels. Sector independence is already declining while headline stress still appears moderate.
Median PC1 variance share rises sharply from about 42% at the lowest VIX interval to above 60% before VIX reaches 15. It increases more gradually through the transition range from 15 to 30 and exceeds 70% at higher volatility levels.
Analysis 02
Market structure reflects more than today’s VIX
Current VIX is positively associated with common-factor dominance, but it is not the strongest measure. PC1 variance share is most closely associated with the maximum VIX observed over the previous 60 trading days, with a correlation of 0.458. The 60-day VIX range, mean and frequency of readings above 30 also show stronger relationships than the current VIX level. This indicates that realised market structure reflects recent stress exposure rather than only today’s volatility reading. A market can appear calmer while sector returns continue to carry the imprint of an earlier shock.
The 60-day maximum VIX has the strongest correlation with PC1 variance share at 0.458. The 60-day VIX range follows at 0.434, while current VIX is lower at 0.360 and the share of 60 days with VIX at or above 20 is lowest at 0.305.
Analysis 03
Volatility peaks before diversification recovers
Across 28 complete VIX shock events, volatility rises quickly around the threshold crossing and reaches its median peak after 5.5 trading days. Common-factor dominance follows a slower path. Median PC1 variance share continues rising after VIX begins to retreat and reaches its median peak after 43 trading days. The difference is substantial. Visible stress peaks within the first week, while the dependence structure across sector returns continues tightening for several additional weeks. A falling VIX therefore does not imply that diversification has normalised. The immediate shock fades first. Market structure recovers later.
Across 28 complete VIX shock events, the median standardised VIX path peaks 5.5 trading days after the threshold crossing. The median PC1 variance-share path continues rising and peaks on day 43 before declining more slowly.
Analysis 04
Larger shocks leave a deeper structural imprint
The observed structural response is stronger at higher volatility thresholds. VIX crossings of 25 are followed by a moderate increase in common-factor dominance. Crossings of 30 produce a larger and more persistent response. Events in which VIX crosses 40 generate the sharpest increase, with median PC1 variance share moving close to 80% shortly after the shock. The difference remains visible well beyond the initial event. Common-factor dominance stays higher for longer following the most severe volatility episodes. Larger shocks therefore do more than raise volatility. They leave a deeper and more persistent imprint on the dependence structure between sectors.
Median PC1 variance share rises after all three VIX thresholds. The path following a VIX crossing of 40 approaches 80% shortly after the event and remains above the responses following crossings of 30 and 25 for most of the post-event window.
Analysis 05
Market structure peaks much later than volatility
The distribution of peak timings confirms that the lag between volatility and market structure is not driven by a small number of unusual events. VIX peaks are concentrated near the start of the post-event window, with a median of 5.5 trading days. PC1 peaks occur much later, with a median of 43 days and substantially greater dispersion across events. Some volatility episodes continue developing for longer, but the broader separation remains clear. The typical volatility peak arrives early. The structural peak occurs several weeks later. Diversification recovery is therefore slower and less uniform than the initial volatility response.
VIX peak timings are concentrated near the start of the 120-day post-event window, with a median of 5.5 trading days. PC1 variance-share peaks have a median of 43 days and a much wider distribution across events.
Methodology
How the analysis is constructed
- Sample period
- 2000–31 July 2026
- Volatility measure
- CBOE Volatility Index, VIX
- Data source
- Yahoo Finance, accessed through yfinance. Daily auto-adjusted closing prices.
- Data through
- 31 July 2026
- Sector universe
- XLB Materials
- XLE Energy
- XLF Financials
- XLI Industrials
- XLK Technology
- XLP Consumer Staples
- XLU Utilities
- XLV Healthcare
- XLY Consumer Discretionary
- XLRE Real Estate, when sufficient history is available
- XLC Communication Services, when sufficient history is available
Data
Daily auto-adjusted closing prices are converted into logarithmic returns. The available sector universe expands as newer ETFs enter the sample.
Principal component analysis
Sector returns are standardised within each 60-trading-day rolling window. Principal component analysis is then applied to the rolling sector-return matrix. The variance share explained by the first principal component measures how much of total standardised sector-return variation is attributable to the dominant common factor. Higher PC1 variance share indicates that sectors are increasingly responding to the same underlying market force. Lower values indicate greater independence across sector returns.
Average sector correlation
Pairwise sector-return correlations are estimated within the same rolling window and averaged across the available sector universe. Average correlation is used as a supporting measure of commonality. It is closely related to PC1 variance share, although PC1 remains the primary structural measure throughout the analysis.
Rolling estimation
PC1 variance share and average sector correlation are estimated using a 60-trading-day rolling window, recalculated daily. Windows require sufficient observations and at least three available sector series. The number of sectors therefore varies from nine to eleven through the sample.
Volatility measures
The analysis compares PC1 variance share with several measures of current and recent volatility:
- Current VIX
- 20-day median VIX
- 60-day median VIX
- 60-day mean VIX
- 60-day maximum VIX
- 60-day VIX range
- Share of the previous 60 days with VIX at or above 20
- Share of the previous 60 days with VIX at or above 30
Pearson and Spearman correlations are calculated between each volatility measure and rolling PC1 variance share. This comparison distinguishes between the market’s current stress level and its recent exposure to elevated volatility.
Primary event definition
An event occurs when VIX crosses 30 from below. To reduce overlap between closely spaced episodes, events must be separated by at least 60 trading days.
Event window
40 trading days before the threshold crossing. 120 trading days after the threshold crossing.
Thirty threshold-crossing events met the 60-trading-day separation rule. Twenty-eight contained complete pre-event and post-event windows and were retained for the main analysis.
Standardisation
For each event, VIX and PC1 variance share are standardised relative to their own pre-event distribution from day −40 to day −1. The reported event paths show the median standardised value across complete events for each relative trading day.
Peak timing
For every event, the timing of the maximum VIX and maximum PC1 variance share is recorded within the 120-day post-event period. The median and distribution of these peak dates are then compared across events.
Alternative thresholds
The event study is repeated using VIX thresholds of 25, 30 and 40. This tests whether the depth and persistence of common-factor dominance change with shock severity.
Research workflow
- Data
- Daily sector ETF prices and VIX
- Returns
- Daily logarithmic sector returns
- Standardisation
- Rolling z-scores within each PCA window
- Structural measure
- First principal component variance share
- Supporting measure
- Average pairwise sector correlation
- Rolling window
- 60 trading days
- Primary event
- VIX crossing 30 from below
- Event separation
- Minimum 60 trading days
- Event window
- Day −40 to day +120
- Complete events
- 28
Higher PC1 variance share indicates that sector returns are increasingly dominated by a common market factor. A decline in VIX alongside elevated PC1 therefore represents a market in which visible volatility has eased, but diversification remains structurally weakened.
Research implications
What the evidence shows
The evidence presented in this report shows that volatility and diversification recover on different timelines.
- 01Sector commonality begins increasing before volatility reaches extreme levels.
- 02Current VIX does not fully capture the effect of recent stress on market structure.
- 03Volatility typically peaks within the first week after a shock.
- 04Common-factor dominance continues rising for several additional weeks.
- 05Larger volatility shocks produce deeper and more persistent structural compression.
- 06PC1 peak timing varies across events but remains substantially later than the typical VIX peak.
- 07A falling VIX is not equivalent to restored diversification.
Together, these results suggest that headline volatility provides only a partial view of market recovery. VIX captures the immediate intensity of stress. Rolling PC1 variance share captures how strongly sector returns remain tied to a common source of variation. Monitoring both helps distinguish between a market that appears calmer and one whose underlying diversification structure has genuinely normalised.
Publication record
Sources and disclosures
Data
- Yahoo Finance, accessed through yfinance.
- SPDR sector ETFs: XLB, XLE, XLF, XLI, XLK, XLP, XLU, XLV, XLY, XLRE and XLC.
- CBOE Volatility Index, VIX.
- Data through 31 July 2026.
Methodology reference
- Jolliffe, I. T. and Cadima, J. (2016). “Principal component analysis: a review and recent developments.” Philosophical Transactions of the Royal Society A, 374(2065).
- Pearson, K. (1901). “On lines and planes of closest fit to systems of points in space.” Philosophical Magazine, 2(11), 559–572.
Software
Python, pandas, NumPy, scikit-learn, statsmodels, matplotlib and yfinance.
This publication has been prepared by LB Research for informational and educational purposes only. It does not constitute investment advice, investment research as defined under applicable regulation, a recommendation, or an offer to buy or sell any financial instrument.
The analysis is based on publicly available market data and the methodologies described within this publication. While reasonable care has been taken in preparing this report, no representation or warranty is made regarding the accuracy, completeness or timeliness of the information presented.
Any opinions expressed reflect the author's judgement at the publication date and may change without notice. Past performance is not indicative of future results.
Readers remain solely responsible for their own investment decisions.
© 2026 LB Research. All rights reserved. Questions or feedback: Contact LB Research
Download canonical PDF