As environmental, social, and governance, ESG, considerations gain prominence in the investment landscape, accurate forecasting models and investment strategies for ESG assets grow in significance and can incentivize private investors to revise their portfolios.

This thesis presents a comparative analysis of Long Short-Term Memory, LSTM, and Autoregressive Integrated Moving Average, ARIMA, forecasting models for ESG assets compared to the S&P 500 index. By leveraging a comprehensive dataset encompassing highly rated ESG assets from the S&P 500, the study explores the accuracy of predictions and effectiveness of these two forecasting techniques in the context of portfolio optimization.
The findings highlight the applicability of LSTM and ARIMA models in forecasting ESG assets and assessing their impact on mean-variance portfolio construction compared to the S&P 500 and simpler investments strategies. The ARIMA model demonstrate proficiency in capturing short-term fluctuations, in contrast, LSTM models exhibit superior ability in capturing non-linear relationships and long-term dependencies, allowing for two enhanced portfolio optimizations.
The study's results provide insights for investment professionals seeking to integrate ESG assets into their portfolio management strategies. By utilizing ARIMA and LSTM models, with further mean-variance portfolio optimization, results in profits exceeding those of the S&P 500 benchmark. This knowledge enables and incentivizes investors to constructs portfolios that align with ESG investment objectives while not compromising competitive risk-return profiles.