Research insight · Issue 005
When markets move together, shocks travel further
The same shock travels further when markets are already dominated by a common factor.
- Published
- Data through
- 25 August 2026
ContentsExecutive summary
Executive summary
Large market shocks propagate strongly when common-factor dominance is already high.
- 01Larger shocks are followed by greater cross-market propagation.
- 02Shock sensitivity rises substantially as pre-shock commonality increases.
- 03The state-dependent effect becomes more pronounced for larger shocks.
- Daily shock events
- 1,109
- Main interaction
- 1.48
- HAC p-value
- 0.012
- Positive interaction estimates
- 11/11
- 9 of 11 statistically significant at the 5% level.
- Period
- 2010–2026
Why it matters
The size of a market move does not determine its consequences alone. The same increase in shock magnitude is associated with greater propagation when markets are already dominated by a common factor.
Analysis 01
Larger shocks are followed by greater market propagation
Larger market shocks are associated with stronger subsequent propagation across the wider system. Across 1,109 daily shock events, the relationship is positive despite substantial dispersion in outcomes. Smaller shocks are usually followed by relatively contained cross-market responses, while the largest events are associated with much wider variation and a higher average level of propagation. Shock magnitude therefore matters, but it does not explain the full response on its own.
Most observations are concentrated between shock magnitudes of 2 and 4, while larger shocks show wider dispersion and a higher fitted level of subsequent propagation.
Analysis 02
Shock sensitivity rises sharply when markets already move together
The effect of shock magnitude changes materially with the market state. In low-commonality periods, the estimated coefficient is just 0.029 and is not statistically distinguishable from zero. It rises to 0.084 in the medium state and to 0.276 when pre-shock commonality is high. The result suggests that larger market moves become much more consequential when common-factor dominance is already elevated. Market structure therefore changes how strongly an initial shock translates into subsequent cross-market propagation.
The estimated coefficient rises from 0.029 in the low-commonality state to 0.084 in the medium state and 0.276 in the high state, with confidence intervals shown for each estimate.
Analysis 03
The propagation effect of a shock strengthens as commonality rises
The effect of shock magnitude increases steadily as pre-shock commonality rises. At lower levels of PC1 variance share, the estimated marginal effect is close to zero and the confidence interval includes no effect. As common-factor dominance increases, the relationship strengthens and becomes more clearly positive. By higher-commonality states, an additional increase in shock magnitude is associated with substantially greater subsequent propagation. The result reinforces the regime evidence. Market structure changes how strongly a shock travels through the wider system.
The estimated marginal effect is close to zero at the lowest commonality levels and rises steadily to approximately 0.38 at the highest displayed PC1 variance share; uncertainty widens at higher values.
Analysis 04
Structural state matters more as shocks become larger
The role of market structure becomes more pronounced as larger shocks are isolated. The estimated interaction rises from 1.00 at a 1.5σ threshold to 1.48 at 2.0σ, 1.74 at 2.5σ and 1.77 at 3.0σ. The increase is strongest between the lower thresholds before flattening at the most severe events. This suggests that pre-shock commonality becomes increasingly important as the initial disturbance grows. For larger market moves, the structure already in place plays a greater role in determining how strongly the shock propagates.
The estimated interaction increases from 1.00 at the 1.5-sigma threshold to 1.48 at 2.0 sigma, 1.74 at 2.5 sigma and 1.77 at 3.0 sigma, with confidence intervals shown.
Analysis 05
The state-dependent shock effect is robust across specifications
The state-dependent effect remains positive across all 11 alternative specifications. Nine estimates are statistically significant at the 5% level, including every forward-horizon test and the 2.0σ, 2.5σ and 3.0σ shock thresholds. The two weaker results, at a 1.5σ threshold and a 252-day state window, remain positive but are estimated less precisely. The consistency of the coefficient across horizons, thresholds and state definitions suggests that the relationship is not driven by a single modelling choice. Pre-shock commonality repeatedly conditions how strongly larger shocks propagate through the wider market.
All 11 interaction estimates are positive. Nine are statistically significant at the 5% level; the 1.5-sigma shock threshold and 252-day state window remain positive but their confidence intervals cross zero.
Methodology
How the analysis is constructed
- Sample period
- 4 January 2010–25 August 2026
- Market universe
- Cross-market instruments represented by eleven liquid ETFs
- Data source
- Yahoo Finance, accessed through yfinance. Daily adjusted prices converted to logarithmic returns.
- Data through
- 25 August 2026
- Sector universe
- SPY S&P 500
- QQQ NASDAQ
- IWM Russell 2000
- HYG High yield credit
- LQD Investment grade credit
- TLT U.S. Treasuries
- GLD Gold
- UUP U.S. dollar
- XLF Financials
- XLK Technology
- XLE Energy
Shock identification
Daily log returns are standardised using trailing 60-day volatility, lagged by one trading day. A shock event is identified when the largest absolute standardised move across the market universe exceeds 2.0σ. Only the largest shock on each trading day is retained. The base specification contains 1,109 daily shock events.
Pre-shock market state
Market commonality is measured over the preceding 120 trading days. Returns are standardised within each window before principal component analysis. The first principal component variance share, PC1, measures the proportion of cross-market variation explained by the dominant common factor.
- Higher PC1 variance share indicates greater pre-shock commonality.
Propagation measure
Subsequent propagation is measured across all instruments other than the shock origin. For each event, cumulative absolute returns over the following five trading days are standardised by pre-shock volatility and adjusted for the forward horizon. The event-level propagation measure is the mean standardised response across the remaining markets.
Estimation
The main specification estimates propagation from shock magnitude, pre-shock commonality and their interaction. Inference uses heteroscedasticity and autocorrelation consistent standard errors with 10 lags.
Propagation = Shock magnitude + Pre-shock commonality + Shock magnitude × Pre-shock commonality
The interaction term measures whether the effect of shock magnitude changes with the market state already in place.
Robustness
The interaction is re-estimated across forward horizons of 1, 3, 5 and 10 trading days; shock thresholds of 1.5σ, 2.0σ, 2.5σ and 3.0σ; and pre-shock state windows of 60, 120 and 252 trading days.
The estimated interaction remains positive in 11 of 11 specifications and statistically significant at the 5% level in 9 of 11.
Research workflow
- Data
- Daily adjusted prices across the 11-market universe
- Returns
- Daily logarithmic returns
- Volatility scaling
- 60-day trailing volatility, lagged one trading day
- Shock standardisation
- Daily return divided by lagged trailing volatility
- Shock selection
- Largest absolute standardised move across markets each day
- Primary event
- Daily shock exceeding 2.0σ
- Pre-shock state
- First principal component variance share
- State window
- 120 trading days before each shock
- Propagation horizon
- Five trading days after the shock
- Propagation measure
- Mean standardised cumulative response across all non-origin markets
Higher PC1 variance share indicates that cross-market returns are increasingly dominated by a common factor. A positive shock magnitude × commonality interaction therefore indicates that larger shocks are associated with stronger subsequent propagation when the market is already more structurally compressed.
Research implications
What the evidence shows
The evidence presented in this report shows that the consequences of a market shock depend on the structure already in place before it arrives.
- 01Larger shocks are followed by greater cross-market propagation.
- 02Shock sensitivity is weak when pre-shock commonality is low.
- 03The effect strengthens materially as common-factor dominance rises.
- 04High-commonality states produce substantially larger shock coefficients than low-commonality states.
- 05The role of market structure becomes more pronounced as increasingly large shocks are isolated.
- 06The state-dependent interaction remains positive across all 11 robustness specifications.
- 07Nine of the 11 alternative specifications are statistically significant at the 5% level.
Together, these results suggest that shock magnitude should not be interpreted independently of market structure. A large move entering a relatively independent system does not carry the same implications as a similar move entering a market already dominated by a common factor. Pre-shock commonality provides information about how strongly an initial disturbance may propagate across the wider system.
Publication record
Sources and disclosures
Data
- Yahoo Finance, accessed through yfinance.
- Market universe: SPY, QQQ, IWM, HYG, LQD, TLT, GLD, UUP, XLF, XLK and XLE.
- Data through 25 August 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.
- Newey, W. K. and West, K. D. (1987). “A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix.” Econometrica, 55(3), 703–708.
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 PDFIn development: extending this research into cross-market structure and risk intelligence.