[Link to Part 1]

Having introduced Hidden Markov Models (HMMs) earlier, I now apply them to study how ESG equity funds behave under different market regimes. The model identifies two latent states based purely on return dynamics – typically corresponding to calmer and more volatile market conditions.

To evaluate performance across these regimes, I estimate CAPM-style alphas separately within each regime for four prominent ESG ETFs, using the S&P 500 ETF SPY as the benchmark. This allows us to ask:

Do ESG funds consistently outperform or underperform the market once we account for different volatility regimes? Alternatively, do they tend to outperform during periods of high market volatility, as suggested by some researchers and practitioners?

To assess whether any observed outperformance is statistically meaningful, I use bootstrap methods to construct confidence intervals and test the significance of regime-specific alphas.

While VIX is not used to define the regimes, I include it post hoc to interpret the nature of each regime – verifying, for example, that one state corresponds to elevated market volatility. This ensures that performance differences are not only statistically grounded, but also economically interpretable.

In short, this approach combines data-driven regime discovery with risk-adjusted performance evaluation – offering a nuanced lens on ESG fund behavior under varying market stress.

I focus on four ESG ETFs—ESGE, ESGU, ESGD, and DSI—each with assets above $4.5 billion and at least eight years of history, to enable a stronger test. The same method applies to other ESG funds with sufficiently long records. ESGV also exceeds $4.5 billion but has only seven years of data, so I exclude it to preserve a longer sample. All analyses use monthly returns.

The findings are summarized as follows. It is important to note that although all ESG funds in our sample cover the same time period, the HMM treats each fund individually when estimating the regimes. So, “calm” and “volatile” are not fixed labels applied across all funds at the same time. Instead, each fund’s model identifies calm and volatile periods based on its own return patterns, which can lead to different regime counts even over the same date range.

To make the findings easier to follow, we show the results for calm and volatile regimes separately and offer brief interpretations, as follows:

Volatile Market Regime (High VIX):

Calm Market Regime (Low VIX):

The results in the above tables indicate that ESGE and ESGD exhibit statistically significant outperformance during volatile regimes, with alphas of 2.74% and 3.24%, respectively. In contrast, ESGU and DSI underperform in volatile markets, with alphas of –0.71% and –0.37%.

During calm regimes, all four funds record negative alphas. The alphas of ESGE and DSI are not statistically significant, suggesting that their performance does not differ materially from SPY. Conversely, the alphas of ESGD and ESGU are significantly negative (–0.66% and –0.12%), indicating underperformance relative to SPY.

Regime classification is derived from latent states inferred from price dynamics, with the ex post VIX analysis confirming higher volatility in the volatile regime.

Overall, the Hidden Markov Model framework offers valuable insights into how ESG fund performance varies across market regimes. By uncovering latent states of calm and volatility, it highlights that ESG excess returns are not uniform but regime-dependent. The evidence that ESGE and ESGD outperform during volatile periods while others lag suggests that ESG strategies may offer differentiated resilience under market stress. In contrast, the broadly negative and often insignificant alphas observed during calm markets indicate that ESG funds, as a group, tend to track or modestly lag the benchmark when volatility is subdued. This pattern suggests that the potential advantages of certain ESG strategies may emerge primarily during periods of heightened uncertainty rather than in stable conditions. These findings point to the need for incorporating regime-based modeling in both ESG performance assessment and portfolio construction. Future research could extend this approach to dynamic asset allocation, cross-market ESG comparisons, or the integration of macro-financial and sentiment indicators to better capture the evolving nature of sustainable investing.

Like (0)