The relationship between firm characteristics and greenwashing in the European energy sector; An NLP approach

Abstract

Greenwashing is an increasing problem in our society as it is hard to detect due to increasingly sophisticated greenwashing techniques and an asymmetry of information between firms and the general public. Firms make claims regarding their corporate sustainability in their CSR reports, which might be genuine or could be greenwashed. When firms engage in greenwashing their symbolic actions (“the green talk”) do not align with their substantive actions on environmental issues (“the green walk”). These substantive actions are measured by examining patents which are an indicator of technological innovations. We examine 134 firms in the European energy sector to determine if they are “walking the talk”. To get a better understanding of which firms are more likely to engage in greenwashing this master thesis investigates which firm characteristics influence the extent of greenwashing in CSR reports. We obtain this objective using three steps. First, we define the extent of greenwashing as a discrepancy between symbolic and substantive actions. Second, we find three theoretically informed firm characteristics, which were described as important determinants of greenwashing and we formulate corresponding hypotheses for them. Third, we propose a new measurement approach to test these hypotheses based on the discrepancy between symbolic and substantive actions using state-of-the-art machine learning techniques.

This thesis demonstrates that the measurement approach manages to detect discrepancies between the symbolic and substantive actions for firms operating in the European energy sector. We find that neither of the three formulated firm characteristics directly relate to the extent of greenwashing, indicating that greenwashing does not seem to be a systematic phenomenon for our dataset. We also find that firms mainly engage in technological innovation with regards to solar energy, wind energy, and the reduction of emissions and toxic gasses. Moreover, we discover that the majority of energy firms are “walking the talk” with some outliers, indicating that greenwashing can be seen as an exception instead of the norm. Lastly, we find that electricity firms engage less in technological innovation on the topics wind and solar energy in comparison to oil & gas firms that seem to engage in a diversification strategy. Altogether, this research demonstrates that measuring greenwashing determinants is feasible in an empirical setting, by presenting a proof-of-concept, which hopefully will inspire other researchers to apply machine learning techniques more often to innovation sciences problems and test more determinants of greenwashing in an empirical setting.