Research in Progress
(In)Credibly Green: Why Investors are (Un)Willing to Pay a Premium for Green Bonds, with Christopher Scheins
We consider a large sample of over 3,000 green bonds issued between 2009 and 2018 in more than 30 different currencies worldwide. To investigate the credibility of green bonds we estimate the green bond premium using three different matching methods to find comparable conventional bonds. In contrast to the findings of previous studies on a smaller data set, our results reveal a significantly positive premium of 10-20 bps, implying that green bonds are trading at lower prices than their conventional counterparts. This premium is, however, declining over time and in particular, it is not significantly different from zero in the last two years. This indicates that the efforts of the Climate Bond Initiative and other ESG certification entities to promote integrity in the development of the Green Bond market are bearing fruits and the acceptance of green bonds increases. To further test this hypothesis, in a next step we consider only green bonds listed on an exchange with a dedicated green market segment. Though the analysis relies only on a small set of bonds, the resulting premium is significantly lower than that for the whole sample, pointing out that green bonds on dedicated markets appear to be more credible, and revealing the importance of exchanges for green bond market growth.
The impact of financial advice on investor portfolio performance and behaviour in social networks, with Ralf Conen
Focusing on a digital investment solution provided by the one of Europe’s largest social trading platforms, ayondo, we position our study between the evaluation of conventional forms of financial advisory and fully automated online investment solutions (i.e. robo advisors). Our main research question deals with the role of social connections and the associated quality of financial advice. Thereby, we aim to provide empirical evidence whether investors benefit from a novel, digital form of investment advisory. Second, we contribute to the understanding of the adoption of digital financial services models and underlying forces. Finally, our findings will offer starting points for value adding designs of online advisory models and the surrounding regulatory framework.
In this study, we make use of an exclusive data set on trades and social connections executed on ayondo’s social trading platform between 2011 and 2018. While previous studies in this area almost exclusively focus on users who publish their opinions or trades, i.e. financial advisors, our data allows us to observe portfolios of the advised party, i.e. users who copy other users‘ transactions in their own accounts. One particular feature of our dataset is that we can separate users trading in an isolated fashion and users explicitly following others actions or even coping other’s trades while being active traders themselves.
In a first analysis, we intend to apply an econometric difference in differences design. Therefore, we identify sub-groups of investors depending on their connection to other users on the platform: on the one end of the spectrum are „ivory tower“ traders who simply make trades without following other users or being observed by others, on the other end are those who are actively woven into the network. We then compute various performance and behavioral metrics before testing the robustness of our findings. For an intra-person analysis, we further intend to conduct event studies to evaluate whether individuals who first trade on their own and then decide to follow others, respectively the other way around, harvest higher, respectively lower, cumulated returns than before. Further, we can separate and compare trades of individual users made in isolation and advised trades copied from other users. In a very preliminary analysis, we used a shorter sample. We performed a 95 percent winsorization of daily returns and volatility and excluded users that were active for limited time and only completed few trades. A first glimpse at the results suggests significant differences in average performance between users that connect to others and those who do not .