Conference Agenda

Natural disasters are increasingly affecting the financial system. While most of the literature on natural disasters and credit losses focuses on probability of default, very little is known about what happens after default. In this study, we combine two unique datasets to provide novel empirical evidence on the financial impact of wildfires through a loss given default channel. First, we determine Italian provinces’ exposure to wildfires using geospatial data on burned areas derived from satellite imagery. Second, we exploit a proprietary dataset on defaulted consumer credits obtained from a third-party collection agency in Italy. Our results reveal a robust negative relationship between debtors’ exposure to wildfires and the realized recovery rate. By focusing on wildfires that occur during the recovery process of already-defaulted consumer credits, we are able to isolate a loss given default channel, complementing existing evidence on default probabilities.

The increasing complexity of wildfire management necessitates accurate, timely, and frequently updated data from satellite-based Earth Observations (EOs) for rapid detection and monitoring. This paper assesses the impact of satellite imagery on US wildfire management costs, focusing on FIRMS Active Fire (AF) detection information and view angle dependencies from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument. Using a dataset that combines wildfire management costs with satellite and fire-related covariates, I find that fires detected by MODIS with a View Zenith Angle (VZA) below 30° prior to discovery are associated with a 9.74% reduction in final management costs compared to those without AF detection. Conversely, fires with a VZA above 30° do not exhibit consistent or statistically significant cost reductions. Causal random forest analysis shows heterogeneity in the effects of receiving a VZA below 30°, with open land covers and less rugged terrains benefiting the most. One potential underlying mechanism behind the results could be higher allocation of resources at the onset of the fire due to increased certainty, as fires detected with a VZA below 30° exhibit a positive and statistically significant relationship with initial fire management costs compared to those without AF detection.

As central banks increasingly rely on forward-looking models to estimate the financial risks associated with climate change, there is a pressing need for accurate assessments of physical risks such as wildfires. This paper introduces a refined methodology for evaluating wildfire risk, designed to inform the European System of Central Banks’ Expert Group in Climate Change and Statistics. It examines the relationship between the Fire Weather Index (FWI), land cover types, and fire risk across Europe, using data from 2001 to 2022 and a 2.5 x 2.5 km grid. We compare the performance of logistic regression with extreme gradient boosting models (xgboost), both unconstrained and constrained, to capture the complex, nonlinear dynamics influencing fire risk. The findings reveal that although the unconstrained version offers the highest predictive accuracy, the constrained version aligns more closely with the expected relationship between FWI and fire risk. Under the RCP 8.5 scenario and using the constrained xgboost model, the area at high risk is projected to increase from 569,000 square kilometers in 2022 to 635,000 square kilometers by 2050. This highlights the relevance of using advanced modeling techniques in improving the accuracy of financial risks assessments associated with climate change-driven wildfires.

Understanding the socioeconomic incidence of climate change impacts can inform the effective targeting of adaptation policies. Insight to date is limited by data availability; climate impacts and socioeconomic vulnerability are spatially heterogeneous and spatially-explicit profiles of social vulnerability are often unavailable. A method to overcome these data limitations can widen the scope with which one may estimate spatially heterogeneous climate impacts. This paper presents a novel spatial microsimulation estimation method to quantify social vulnerability at the small area level. We apply this method in the investigation of climate risk as a proof of concept for implementation, considering flood risk in Ireland as a case study. We demonstrate the utility of such insight for climate adaptation policy. The quantification of flood exposure provides a first-round approximation of locations which may require adaptation interventions. However, metrics of socioeconomic vulnerability allow for areas of highest priority to be identified, conditional on societal preferences towards equity.