Empty pledges and powerless conventions: How transition climate risks are disrupting financial markets?

Abstract

We propose mention volume index (MVI) as a novel alternative measure of attention to Google's search volume index (SVI). We construct several physical and transition climate risk indexes as shocks to corresponding climate MVIs that we construct by employing textual analysis on climate change narratives on social media between July 2010 and 2022. Using predictive regressions and several test assets, we investigate the response of asset prices to our climate risk indexes at the market level and the asset level. The predictions of our physical climate risk indexes support three stylized findings in the climate finance literature: (i) the carbon premium hypothesis, (ii) the rise in investors' awareness after the Paris Agreement, and (iii) the prediction that green firms outperform brown firms when concerns about weather risk increase unexpectedly. The predictions of our transition climate risk indexes provide new evidence documenting noise trading behavior in the form of return reversal an excess volatility in aggregate market level indexes following unexpected increases in attention to climate pledges and conventions. Moreover, at the asset-level, a rise in attention to climate pledges today is associated with an initial increase (decrease) in the returns of green (brown) firms on the second day. These responses are reversed on the fourth day. Peak and sentiment analyses of climate mentions around these events show that the return reversal is due to the backtracking and the lack of credibility of these promises. Finally, we find that the ineffectiveness of the U.S. carbon policy triggers flight to safety to the bond mutual fund market and disrupts the performance of green (but not brown) firms' stock prices. Unexpected increases in concerns about carbon policy risk on social media today is associated with an initial increase in green returns on the third day that is almost entirely reversed on the fourth day.