Interdependence between Green Financial Instruments and Major Conventional Assets: A Wavelet-Based Network Analysis

Abstract

This paper examines the interdependence between green financial instruments, represented by green bonds and green stocks, and a set of major conventional assets, such as Treasury, investment-grade and high-yield corporate bonds, general stocks, crude oil, and gold. To that end, a novel wavelet-based network approach that allows for assessing the degree of interconnection between green financial products and traditional asset classes across different investment horizons is applied. The~empirical results show that green bonds are tightly linked to Treasury and investment-grade corporate bonds, while green stocks are strongly tied to general stocks, regardless of the specific time period and investment horizon considered. However, despite their common climate-friendly nature, there is no a remarkable association between green bonds and green stocks. This means that these green investments constitute basically two independent asset classes, with a distinct risk-return profile and aimed at a different type of investor. Furthermore, green financial products have a weak connection with high-yield corporate bonds and crude oil. These findings can have important implications for investors and policy makers in terms of investment decision, hedging strategies, and sustainability and energy policies.

Figures (4)

• Figure 1: Wavelet coherence-based networks over the full sample for the three frequency bands considered. (a) Short-term, between $2$ and $5$ days. (b) Medium-term, between $5$ and $22$ days. (c) Long-term, more than $22$ days. Source: own elaboration using the R statistical software Rbase together with packages wavScalogram wavscalogram, igraph igraph, and ggraph ggraph.
• Figure 2: Wavelet coherence-based networks during the European sovereign debt crisis sub-period (from October 13, 2010 to July 31, 2012) for the three frequency bands considered. (a) Short-term, between $2$ and $5$ days. (b) Medium-term, between $5$ and $22$ days. (c) Long-term, more than $22$ days. Source: own elaboration using the R statistical software Rbase together with packages wavScalogram wavscalogram, igraph igraph, and ggraph ggraph.
• Figure 3: Wavelet coherence-based networks during the oil price collapse sub-period (from June 20, 2014 to February 28, 2016) for the three frequency bands considered. (a) Short-term, between $2$ and $5$ days. (b) Medium-term, between $5$ and $22$ days. (c) Long-term, more than $22$ days. Source: own elaboration using the R statistical software Rbase together with packages wavScalogram wavscalogram, igraph igraph, and ggraph ggraph.
• Figure 4: Wavelet coherence-based networks during the COVID-19 sub-period (from January 1, 2020 to November 13, 2020) for the three frequency bands considered. (a) Short-term, between $2$ and $5$ days. (b) Medium-term, between $5$ and $22$ days. (c) Long-term, more than $22$ days. Source: own elaboration using the R statistical software Rbase together with packages wavScalogram wavscalogram, igraph igraph, and ggraph ggraph.