Abstract

Purpose – This paper represents the first attempt to examine investor behavior for green stocks through the lens of return co-movement, and provides evidence indicating that green investment practices have gained traction after 2012.

Design/methodology/approach – We empirically test the hypotheses that the stock returns of firms with similar carbon dioxide emissions levels move together and, if so, whether this co-movement has increased over time as people become more “carbon-conscious.” Our baseline sample, based on carbon emissions data from public company disclosures, suffers from limited coverage, particularly before 2016, leading to low statistical power and sample selection bias. To address this, we employ machine learning methodologies to forecast the carbon emissions of firms that do not disclose such information, nearly quadrupling the sample size. Our findings indicate that stocks with similar carbon emissions exhibit higher co-movement in stock returns in both the baseline and augmented data samples. Furthermore, this co-movement has increased during the 2012–2020 period compared to the 2004–2011 period, suggesting that green investment has gained traction over time.

Findings – We find that stocks with similar carbon emissions exhibit higher co-movement in stock returns in both the baseline sample and the augmented data sample, and the co-movement has increased in the 2012–2020 period compared to the 2004–2011 years, suggesting that green investment has gained traction over time.

Originality/value – (1) We use machine learning methodology to augment carbon emissions sample which goes back to 2004. Our approach almost quadruples the original data, enabling large-sample testing. (2) We are the first paper to examine how green companies' stock returns co-move and thus provide complementary results on the research on expected returns and carbon emissions.

In their research on enhancing LoRA-based fine-tuning for LLMs, the research team introduced an Enhanced Matrix Decomposition for single-task scenarios and a routing mechanism for multi-task learning, improving flexibility and performance without increasing computational complexity.

Time-consuming testing and computer simulations are bottlenecks in the design of new materials. A thesis from the University of Gothenburg aims to develop an AI model that can efficiently determine the durability and strength of woven composite materials.

Stanford's Eric Sun has created sophisticated machine learning frameworks that measure biological aging at single-cell resolution, discovering cell types that can dramatically accelerate or slow aging in surrounding brain tissue. His recent Nature publication establishes new computational tools for understanding brain aging and developing targeted interventions against age-related cognitive decline.

Research team from Shandong University introduced MetaGIN, a lightweight deep learning framework for molecular property prediction. MetaGIN achieves SOTA performance with fewer parameters, using a “3-hop convolution” technique and MetaFormer architecture to balance efficiency and accuracy.