Title: Forecasting Italian firms' default probability using Prophet: A cycle-informed approach Authors:  Eugenio Cangiano - Sogei (Italy) [presenting]
Andrea Rollin - Ministry of Economy and Finance Italy (Italy)
Daria Scacciatelli - Sogei S.p.A. (Italy)
Abstract: The Italian Ministry of Economy and Finance plays a crucial role in monitoring public loan guarantee programs, which serve as key financial interventions to support enterprise liquidity and capitalisation. A critical aspect of this activity is the estimation and monitoring of default probabilities, essential for evaluating expected losses in guarantee portfolios and ensuring the efficient allocation of public resources. Default probabilities are estimated conditional on the economic cycle, using models that incorporate sectoral and macroeconomic variables to provide a forward-looking assessment of credit risk. Traditional econometric models often struggle to capture major economic shocks, including the pandemic crisis and financial downturns, due to their limited ability to handle structural breaks and nonlinear trends or dynamics. In contrast, the Prophet model offers a more flexible approach, effectively managing macroeconomic shocks and regime shifts. The analysis compares Prophet and traditional models across the most representative industrial sectors of a major Italian guarantee fund, evaluating the model's contribution to improving default probability projections. The results highlight the advantages of integrating machine learning-based time series models into the econometric toolkit of financial institutions to enhance credit risk analysis and the strategic management of the public guarantees framework.