Abstract
Sustainable financial markets link carbon pricing, environmental infrastructure, low-carbon technology, fixed-income financing, and broad environmental, social, and governance investment. Whether these markets transmit shocks differently in the tails of the return distribution remains important for diversification and transition-risk management. This study examines return connectedness among European Union Allowance prices, a global water exchange-traded fund, a global clean-energy index, a U.S.-dollar green-bond exchange-traded fund, and an ESG equity index. The synchronized sample contains 1,441 daily level observations from 16 March 2020 to 31 December 2025, yielding 1,440 percentage log returns. A quantile vector autoregression is combined with generalized forecast-error variance decomposition to estimate total, directional, and net connectedness at the 0.05, 0.25, 0.50, 0.75, and 0.95 quantiles. Rolling estimates, stationary-bootstrap confidence intervals, volatility and interest-rate regimes, and forecast-horizon sensitivity provide additional inference. Full-sample total connectedness is moderate and unusually stable, ranging from 33.90% to 34.16%. Rolling evidence nevertheless reveals statistically supported upper-tail amplification: the 0.95-quantile total connectedness exceeds the median by 0.672 percentage points on average, with a 95% stationary-bootstrap interval of 0.372 to 0.961. The lower-tail difference is not statistically distinguishable from zero. The water-equity portfolio is the most persistent net transmitter, while green bonds are the principal net receiver. High-VIX windows exhibit lower rather than higher average connectedness. An exact cluster decomposition attributes this decline to weaker cross-cluster links, while connectedness within the CGW–CLEAN–SPESG equity cluster rises slightly. Extreme conditional returns and observable market stress are therefore related but distinct states. The findings support state-contingent diversification and caution against assuming that all forms of stress mechanically intensify integration among sustainable assets. A strictly timed out-of-sample extension shows that a state-contingent QVAR-adjusted minimum-variance portfolio reduces annualized volatility from 13.22% for equal weighting to 5.91% and maximum drawdown from 21.39% to 13.22%, although this protection is accompanied by lower realized returns. Connectedness therefore provides decision value primarily through risk containment rather than return enhancement in this sample.