以稀疏性模型估計綠色金融資產之報酬溢出網路

Abstract

本研究使用自適應彈性網路向量自我迴歸模型 (Adaptive Elastic-Net VAR)，以及基於廣義預測誤差變異數分解的Diebold-Yilmaz鍵結指數，針對2015年1月至2024年11月期間綠色債券、綠色能源類股、碳權，以及債券市場、主要國家之股市、能源市場之間的報酬溢出效果進行研究。本研究首先針對全樣本以及子樣本透過ForceAtlas2引力導向模型進行靜態溢出效果，並使用滾動窗口法在成對層面進行動態溢出效果分析。針對估計結果，本研究同時透過最小鍵結法建立投資組合，與傳統的最小變異數法與最小相關法進行比較。研究結果顯示，1. 無論是綠色債券、碳權或是各種綠色能源類股，彼此間透過網路傳遞的方向性淨溢出效果，多數情形皆源自於各資產類型間彼此影響，並且不同資產類型間的互相影響通常由特殊事件 (如上證股災、英國脫歐、新冠疫情、烏俄戰事等) 所驅動。2. 綠色債券主要受美國中長期公債、投資等級債券市場的淨溢出效果影響，能夠對美元指數、黃金傳遞報酬淨溢出效果3. 碳權市場主要受到能源市場的影響，並且受到整體股市較微弱的負向淨溢出效果影響，且其與歐洲股市的鍵結並未明顯異於美國股市 4. 綠能類股表現與科技類股相似，且主要受到美國市場的淨溢出效果影響，並向亞洲市場傳遞淨溢出效果5. 最小鍵結法有著最高的報酬率以及夏普比率，而投資組合中碳權、太陽能類股以及核能類股有較高的資產權重。

This study employs the Adaptive Elastic-Net Vector Autoregression (Adaptive Elastic-Net VAR) model and the Diebold-Yilmaz connectedness index based on Generalized Forecast Error Variance Decomposition (GFEVD) to investigate the return spillover effects among green bonds, green energy stocks, carbon credits, as well as bond markets, major national stock markets, and energy markets during the period from January 2015 to November 2024. The analysis first examines the static spillover effects for the full sample and sub-sample periods using the ForceAtlas2 force-directed layout model, and then conducts dynamic spillover analysis on a pairwise level using a rolling-window approach. Based on the estimation results, this study constructs investment portfolios using the Minimum Connectedness approach and compares their performance with the traditional Minimum Variance and Minimum Correlation approaches. The empirical results show that (1) for green bonds, carbon credits, and various green energy stocks, directional net spillover effects within the network mostly originate from mutual influences among asset types, and are typically driven by specific events such as the 2015 Chinese stock market crash, Brexit, the COVID-19 pandemic, and the Russia-Ukraine war; (2) green bonds are primarily influenced by net spillovers from U.S. medium- to long-term Treasury bonds and investment-grade bond markets, while also transmitting net spillover effects to the U.S. dollar index and gold; (3) the carbon credit market is mainly influenced by the energy market, and experiences relatively weak negative net spillover effects from the overall stock market, with its connectedness with the European stock market not significantly different from that with the U.S. market; (4) green energy stocks are primarily affected by net spillovers from the U.S. market and transmit net spillover effects to Asian markets; (5) the Minimum Connectedness Portfolio achieves the highest return and Sharpe ratio within the sample period, with carbon credits, solar energy stocks, and nuclear energy stocks receiving higher portfolio weights.