Resarchers:
Manescu, Cristiana
Mougeot, Mathilde
Starica, Catalin
Teyssiere, Gilles

An in-depth literature review of the field that focuses on the study of the quantitative relationship between sustainability and profitability on a fund, portfolio and company level has been performed. The review revealed that on a fund level no statistical difference between SI and non-SI funds has been shown. At the portfolio level, the literature shows that returns of high SI-portfolio sizeably outperformed those of low ranked portfolio. At the company level there is an impact of SI: it pays to be green. But, the causality is unclear: it might be the case that it is profitable companies that indulge in SI-activities and that this is the reason for the statistical results presented above.

In January 2007 the data preparation stage of the project consisting in matching the company sustainability information (covering more than 1000 companies listed on the main financial markets of the world) from our sustainability data provider with specific accounting and market based data from Reuters data bases available at the Centre for Finance was finished. The data set that was created covers three years of economic activity, i.e. 2003, 2004 and 2005. The current challenge is the statistical study of the relationship between various measures of profitability at firm level and the sustainability criteria.

Manescu is currently investigating the relevance of a linear model. She has already implemented a couple of model validation methodologies commonly used in the statistical literature which seem to point towards the presence of strong non-linearities in the relationship of interest. She will continue the investigation by including interactions between the variables and enlarging the set of accounting measures that can possibly help explain firm's profitability.

Starica has been using techniques from the field of statistical learning to determine the sustainability variables that show relevance to profitability. Boosting methodology with both linear as well as smooth splines were tried. Both approaches yield improvements in the prediction error compared to linear regression of 5% and 10% respectively. The second method is a non-linear approach and the improvement in the prediction error represents a strong indication of the non-linear nature of the relationship between firm's profitability and sustainability variables. The next issue will be to analyze the relevance of the interactions in the framework of boosting with smooth splines. Later on a more sophisticated non-linear modelling approach involving first a dimension reduction followed by a non-parametric smoothing in a lower-dimensional space will be investigated.  

Mougeot has been studying the structure of the design space with the help of hierarchical clustering and self-organizing maps. This analysis helps quantify the amount of information contained in the SAM scores. The next step of this research direction is integrating the understanding of the design space gained by the careful analysis performed with the modelling of the relationship of the explanatory variables and the firm profitability.

Another direction of research that has not yet been explored is the panel dimension of the data set. Besides the extensions enumerated above a serious effort to include the time dimension is envisaged.