Climate financial risk evaluation using a conditional VAR model: an application to the European electricity sector

The importance of the financial sector outcomes for the low-carbon transition is starting to be widely recognized by international institutions. Different global average temperature pathways give rise to different societal net impact of the transition, as measured by the main macroeconomic variables. The drastically different effects of an orderly transition and a disorderly one over the financial sector and back to the real economy is now qualitatively understood but there is a lack of a quantitative setup to study the phenomenon. The IAM-CFR framework recently developed by Battiston et. Al gives general guidelines to implement models which blend together the Integrated Assessment Models (IAM), which describe the “physical” sectors of the economy and the the financial sector, whose relevant variables are modeled by the use of Climate Financial Risk (CFR) methodology. In order to provide a quantitative estimation of Climate Financial Risk it is necessary to have a forecast of asset prices in the future, conditional on various climate scenarios (e.g. business as usual vs. 2°C scenario). Within the literature on financial market models a stream of work on VAR (Vector Autoregressive models) is focused on the calibration of the relation between past observation of variables, in order to obtain unconditional forecasts for the main financial variables. Another strand, apparently disconnected from the previous one, is related to Dividend Discount Model (DDM), which instead focuses on predicting prices using future dividend flows. The methodological contribution of this thesis is trying to reconcile these two approaches, justifying the use of VAR from DDM and developing a general framework that allows the evaluation of the impact on prices of long term structural breaks, such as in the case of low-carbon transition. The link between the two approaches builds also on suggestions of previous literature and on previous work of Waggoner and Zha on conditional forecasting in VAR. A theoretical prediction of this DDM-VAR blended approach is the presence of Granger causality from stock prices towards electricity production. The empirical contribution of the thesis consists instead in the application of the model to forecast stock prices for portfolios with different exposures to renewable sources in the European electricity sector. These forecasts are conditional on the different pathways for electricity production in different NGFS scenarios. This approach allows to compute the change in prices for different scenarios with respect to the baseline business as usual climate pathway. These forecasts open the door to the quantification of climate financial risk and to the possibility of studying how it varies with different portfolio exposure to renewables. A statistically significant indication of Granger causality (confirming the theoretical prediction) from stock prices to electricity production is obtained within a time range which spans over approximately a year. This result is robust over all the portfolios and gives reliability to the forecasts of the model. The strong quantitative difference between portfolios, as far as stock prices and climate risk are concerned, is also highlighted.