Abstract
Against the backdrop of the Paris Agreement’s target of a carbon-neutral economy,
advocating for the sustainability concept has become prominent and very urgent for
regulators and policy makers around the globe, but equally, for the investment industry.
This research critically reviews the specific recommendation related to the environmental,
social and governance (ESG) performance claim which advocates for the recognition of
ESG performance indicators when investment decisions are taken. Actively including nonfinancial preferences, such as ESG considerations, contradicts the current asset-pricing
theory and characterises a paradigm shift in the standards for fiduciary money
management. However, the advocated paradigm shift, and with it, the ESG performance
claim, is currently lagging theoretical attention, as well as facing empirical challenges
because researchers struggle to find the causal link between ESG and asset returns. Thus,
this research’s central question is whether investors’ ESG preferences drive a reliable and
quantifiable portion of the asset-pricing function. Specifically, this research aims to clarify if
investors’ ESG preference is one of the statistically and economically relevant asset price
factors and therefore, exhibits measurable effects on the predictability of asset returns.
Results from this research strongly suggest that current asset price theory, in particular
for bond-pricing, requires an extension by including investors’ non-financial preferences.
This research finds that ESG risks related to firm-level preferences are most important to
improve the predictability of bond returns regardless of their geographies. In sum, investors
price firm-level ESG risk between 32 and 55 per cent of total premia. Furthermore, this
research finds that investment portfolios constructed on the bases of ESG alignment exhibit
different risk structures compared to ESG-agnostic investments. Methodologically, it is the
first ESG asset-pricing inquiry triangulating insights and results from two distinctly different
research methods: classic empirical regression analysis and experimentally-inspired large
correlation matrix dependence structure statistic. Data and analytics for both methods
explore a novel and unresearched universe of single bond positions from two separate
geographical bond market domains recorded as risk and price return data from January
2015 to December 2020. This research’s underlying bond positions represent a list of over
6,000 frequently traded bonds originating from a corporate credit portfolio denominated in
EUR and USD.