Measuring Bubbles via Put-Call Disparity: A Model-Free Approach
Robert A. Jarrow (Cornell U.) and Simon Kwok (U. of Sydney)
May 2026
This paper introduces simple, model-free lower and upper bounds for measuring the size of asset price bubbles. Assuming only that the market satisfies no-free-lunch-with-vanishing-risk and that all trading strategies are admissible, our framework avoids restrictive parametric models and the no-dominance assumption. We demonstrate that put-call disparity provides an observable lower bound, and is economically justified by short-sale constraints. Additionally, the lowest price of an out-of-the-money (OTM) call option determines the upper bound. To ensure empirical reliability, we modify these bounds using data-driven regularization and bootstrap methods to disentangle genuine bubble signals from market microstructure noise and to reduce reliance on thinly traded deep OTM options. Using S&P 500 index option prices from 1996 to 2025, we document a sustained bubble during the COVID-19 era and capture market exuberance preceding the 2000 dot-com and 2008 financial crashes. In addition, the empirical study suggests that the market violates no-dominance and is incomplete.
How Fear Beats Greed: The Impact of Positive and Negative Sentiment on Global Stock Markets
Jiye Ryu (Hongik University), et al.
February 2026
This paper investigates the impact of positive and negative sentiment on stock returns and volatilities across developed and emerging markets using the consumer confidence index as a proxy for sentiment. We conduct a comparative study of developed and emerging markets to assess whether sentiment-driven mispricing is due to overpricing or underpricing and to examine the effect of sentiment on return and risk dynamics.
Diversification Under Stress: Empirical Evidence of Correlation Breakdown Across Sectors and International Markets
Fabio Trachsler (ETH Zürich)
May 2026
We investigate whether portfolio diversification across U.S. sectors and international equity markets retains its risk-reducing properties under stress. Using daily return data from 1999 to 2025 spanning ten sector ETFs and eight international equity indices, we document a systematic correlation breakdown: mean pairwise correlations rise significantly during crisis episodes, precisely when diversification would be most valuable. Across nine crisis periods and 59 systematically identified stress events, correlations increase in 71.2 % of all cases, with a mean ∆ρ = 0.094 that is statistically highly significant (p < 0.0001). We show that the nature of a shock matters more than its magnitude: idiosyncratic events leave correlations intact, while systemic shocks produce strong co-movement across all sectors simultaneously. Sector and international diversification fail together in systemic crises but diverge in idiosyncratic ones, a distinction that, to our knowledge, has not been systematically documented across this breadth of episodes. We introduce the Trachsler Resilience Score, a formal criterion for selecting sectors that are robust to correlation breakdown under stress, and validate it out-of-sample: the Resilience Portfolio reduces maximum drawdown in 83 % of independent stress events (p = 0.003). Our findings suggest that naive diversification offers substantially less protection than classical portfolio theory implies, and that crisis-conditional correlation structure should be an explicit input to portfolio construction.
Industry Rotation Using Market-State Similarity
Valeriy Zakamulin (University of Agde)
May 2026
This paper studies whether lagged market returns contain useful information about subsequent industry returns and whether this information can be used for active industry rotation. Using monthly Kenneth French industry portfolios, we first show that conventional mean predictability is weak. Quantile predictive regressions, however, reveal that market-to-industry predictability is concentrated mainly in the tails of the conditional return distribution, especially in downside states. Motivated by this evidence, we propose a non-parametric strategy based on market-state similarity. At each portfolio-formation date, the strategy identifies historical months in which the standardized market excess return was closest to its current value, measures subsequent industry performance after those similar states, and goes long industries with positive historical subsequent returns and short industries with negative historical subsequent returns. Out-of-sample simulations show that the active strategy delivers a higher Sharpe ratio, lower volatility, and substantially smaller drawdowns than the passive equal-weighted industry benchmark. The results are robust to industry classification and model-parameter choices.
When Sector ETFs Pull Apart
Jackson Wang (independent)
May 2026
The SPDR Select-Sector ETFs are usually treated as positively correlated slices of one underlying market. Across the full 1999-2026 sample of daily returns the average pairwise correlation among the 11 sectors is well above 0.5, and almost no two sectors have ever been unconditionally negatively correlated. Conditioning on a rolling 60-day window, however, surfaces narrow but recurring divergence regimes: episodes in which the cross-sectional spread between the best-and worst-performing sector explodes, and pairs that are usually mildly positive flip to outright negative correlation. The XLK-XLE pair, for example, has a long-run average rolling correlation of 0.41 but reached-0.43 on 2000-10-17 during the dot-com unwind. We build a long history of these regimes and find that the most recent example-the post-ChatGPT AI boom from 2022-11-30 through 2026-05-08-is one of the largest in the sample: SMH returned 410.1% and XLK 164.9% while XLE managed 37.6% and XLP only 19.8%. During this window the average rolling 60-day correlation between XLK and XLE collapsed to 0.15. A naive cross-sectional sector momentum long/short (top-2 long, bottom-2 short, monthly rebalanced, three-month lookback) does not earn a positive risk premium over the full sample (Sharpe-0.14), but a long-only top-2 momentum sleeve captures most of the AI-boom rotation, returning 68.8% over the window vs SPY’s 95.0%. The lesson is that sector ETF divergence is real, identifiable, and useful for tilting equity exposure, but generic mean-reversion or momentum strategies do not naturally monetize it.
Regime-Based Portfolio Allocation Using Hidden Markov Models and Reinforcement Learning
Ajay Kumar Verma (independent), et al.
November 2025
This study develops a regime-aware portfolio allocation framework that integrates Markov switching models with Reinforcement Learning (RL) to dynamically allocate across equities (SPY), long-term Treasuries (TLT), and gold (GLD). Using daily ETF data from 2004-2025, we first characterize market behavior through a discrete Markov chain and then estimate a three-state Gaussian Hidden Markov Model (HMM) selected by the Bayesian Information Criterion (BIC). The estimated regimes-low-volatility, transitional, and high-volatility-exhibit strong persistence and state-dependent return dynamics consistent with recent findings on nonlinear market states (Ardia et al., 2024; Gupta & Pierdzioch, 2023). State-conditional analysis shows that SPY dominates in stable regimes, while TLT and GLD provide protection during stressed periods, motivating regime-conditioned allocation rules.
Learn To Use R For Portfolio Analysis
Quantitative Investment Portfolio Analytics In R:
An Introduction To R For Modeling Portfolio Risk and Return
By James Picerno
