Econometric modelling for the estimation of direct flood damage to enterprises: a local-scale approach from post-event records in Italy

Flood-related damage has increased dramatically in recent decades with direct and indirect economic impacts accounting for a large share of gross national products. Therefore, there is an urgent need to acquire more quantitative knowledge about flood damage to mitigate economic losses and reduce exposure to flood risk.

Firms are especially affected in case of flood. Still, flood damage assessment to businesses is hindered by the paucity of available data to characterize the enterprises, the lack of high-quality damage data to derive new models or validate existing ones, and the high variability of activity types which hampers generalization. This study contributes at improving knowledge about types and extent of damage of flood events on economic activities through the analysis of empirical data, focusing on direct damage and with specific reference to the Italian context.

In detail, the investigated dataset is composed by around a thousand of observed damage records collected after four flood events in Italy, along with additional information on the dimension (i.e., surface and number of employees) and the typology of the affected firms (i.e., NACE category) as well as on local water depth levels. Damage data are further classified in damage to the building structure, the stock, and the equipment.

Several econometric models have been implemented to better understand the links among the damage, the characteristics of the economic activities and the water depth. Since the heterogeneity of the affected firms is very high, in terms of surface, water depth levels, and number of employees and this might have had influence on the firm’s damage reporting, data has been analyzed with Heckman's selection bias model.

Obtained results show the absence of a constant return scale relationship, therefore, the total damage increases less than proportionally to the firm’s surface; the water depth plays an important role to explain the damage to the stock that results the more vulnerable asset.  Information on the NACE category made it possible to quantify the differences in damage by economic sector. The results reveal as the most vulnerable sectors for building structure, stock and equipment, respectively, human health, commercial, and manufacture. The accuracy of the prediction models represented by adjusted R² varies between 0.25, 0.36 depending on the damage component.

Despite characterized by significant uncertainty, obtained results supply a first model for the prediction of flood damage to firms for the Italian context, in the support of more effective risk mitigation actions. In fact, the model identifies the more vulnerable elements within the business sectors orienting modelers and decision-makers choices.

Acknowledgements:

Authors acknowledge with gratitude: Francesca Carisi, Alessio Domeneghetti and Armando Brath (from University of Bologna), Giovanni Menduni, Giulia Pesaro and Guido Minucci (from Politecnico di Milano), Simone Sterlacchini and Marco Zazzeri (from the Italian National Research Council) for their collaboration in collecting the observed damage records analysed in the research. A special thanks to Marta Galliani (from Politecnico di Milano) for providing the refined dataset used in this study.