The project aims to investigate the effect of ESG scores on stock returns and risk measures. The research will explore different deep-learning tools to model complex multivariate time-series data. Deep Neural Networks can infer high-order correlations in complex data with large volumes and dimensionality. Researchers have developed several models to improve the performance of DNN-based methods, including CCN, TCN, LSTM, GRU, DeepAR, and Transformers. Their complexity grows as larger models are developed, requiring large training sample sizes and computational resources. So, it is crucial to determine if the complexity brought in by DNN-based methods is a necessary price to pay for a gain in performance. We need a general comparison covering all families of methods to allow us to answer this question within the specific application of the research. The novel strategy aims to model the relationship between ESG scores, returns and risk measures, addressing the following:1. Multi-horizon forecasting, i.e. the prediction at multiple future time steps2. Use various data sources, i.e. available information about the future, exogenous time series, and static metadata, without prior knowledge of how they interact.3. Interpretability, i.e. identify (i) globally-important variables for the prediction problem, (ii) persistent temporal patterns, (iii) significant events.4. Implementation of the overall strategy in Python and R language to facilitate dissemination of the results.

The candidate is expected to have a good background in Statistics, including a good knowledge of data management, data analysis, inference, statistical modelling, and statistical learning. Good knowledge of programming, algorithms and data structures, including a high-level programming language like Python and/or R, is also necessary. Knowledge of machine learning, neural network modelling, and optimisation will be a plus. A background in Econometrics with good programming and computational skills will also be considered.

The Department of Economics and Statistics research team has a tradition of research in neural networks and deep learning, time series analysis, and financial econometrics models (parametric and nonparametric, univariate and multivariate) for risk analysis and portfolio management. The research team has published in international journals, including the International Journal of Forecasting, International Journal of Approximate Reasoning, Annals of Statistics, JASA, IEEE Transactions on Engineering Management, Technological Forecasting and Social Change, Journal of Machine Learning Research, Journal of Nonparametric Statistics, Statistical Analysis and Data Mining, Soft Computing