In the last two years the European economy grew at a remarkable pace, sustained by a synchronized global expansion, low financing costs, improving private balance sheets and labour market conditions. In this context the role played by financial institutions and banking sector on one side and the demographic changing pattern on the other represent a crucial element to estimate future rate of growth of the European Economy and in particular of Italy. The research project aims at dealing with the issues of systemic risk connected with these two topics. In fact, both of them represent the major source of uncertainty for the entire community when facing the issues of Non Performing Exposures (Public and private institutions are involved) or the changing mortality pattern which impact the welfare cost.
Non Performing Exposure (NPE) of financial institutions have reached dimensions which were simply unthinkable before the 2009 financial crisis. The complex nature of the NPE components needs a specific approach for the restructuring process. The industrialisation of the NPE management, i.e. recovery process, bad loans disposals, securitization, has to be set in order to guarantee stability of the financial sector and the availability of financial resources which foster investments and economic growth.
The worldwide progressive increase in life expectancy causes a change in the structure of older population and represents a serious challenge for the sustainability of pension and healthcare systems. On a methodological level, we need to reconsider the problem of estimating and forecasting the risk of mortality (i.e. mortality rates). We aim to explore the potential of fully general factor models using empirical structures and functional data analysis.
Understanding and modelling the main risk drivers which affect the stability of the financial system is a crucial task for Regulators, Policy makers and Researchers. In this research project we focus on credit risk management and mortality risk modelling.
a) The credit risk management is analysed focusing on the description of NPE in the Italian financial system and providing a stochastic framework to model their dynamics in the next decade. The accurate understanding of the structure and statistical features of Italian NPE is a first innovative contribution which represents the pillar to set up successful strategies to manage and reduce the bulk of NPE in the next decade. One main aim for regulators, whose attention usually focuses on the implementation and the forward-looking nature of systemic risk indicators, is to bring the NPE ratio (NPE/Total Assets) below a target value of 5%. In Italy currently this ratio is equal to 10%, so a crucial task of this research is to identify the optimal actions in term of workouts, disposals or securitization of NPEs, which will bring this ratio to the required target. A stochastic optimization framework will be implemented to identify the optimal level of NPEs which will allow the NPE/ratio to reach the target level in a specific amount of time. It will also be possible to identify the optimal time needed to reach the target level under specific macroeconomic scenarios. We identify the optimal dynamics for this ratios and the optimal required time to reach the target zone. In addition, to provide financial institutions with a robust framework we aim to identify the optimal NPE's coverage ratio which guarantees stability of financial institutions.
The research is innovative in the retrieval of the data, their analysis and the built of simulation scenarios as well as in proposing approaches for management of NPEs and dynamic recovery plan, definition of an efficient reduction plan for NPEs in view of the European Bank Authority (EBA) threshold, assessment of the impact of the reduction of NPEs on economic growth, unemployment rate, investment growth, analysis of NPE sustainable ratio in the long term.
b) On the mortality risk, the research aims at innovate in two different ways. On one hand, new methodologies will be applied to the empirical problem of forecasting mortality rates. On the other, this empirical application will provide a challenging test-bed for methodological advances which are still in full development. Let us see some details.
(i) Factor structures: Forecasting performances can be expected to crucially depend on the ability to impose valid exclusion or homogeneity restrictions on the factor loadings. However, this is still an open problem, as standard test procedures will need to handle a number of constraints increasing with the sample size. We shall tackle the issue exploiting a recent contribution by Di Iorio and Fachin (2017), who circumvent the problem by defining a model selection approach based on bootstrap.
(ii) Cointegration with big datasets: The simulations in Onatski and Wang (2018) show that their "double asymptotic" theory, where both the number of variables and the time span of the data go to infinity, delivers a much better small sample approximation when the dimension of the system is large, even in the presence of moderate time span. This opens important perspectives for the development of cointegrated models of mortality rates.
(iii) FDA: most existing concepts and methods in FDA have been developed for independently and identically distributed observations, or at most for stationary AutoRegressive (AR) processes. However, high persistence and nonstationary (random walk type) behavior has been documented in numerous applications (see e.g. Chang et al., 2016). This evidence calls for a framework that is able to accommodate and address jointly functional time series with strong persistence as well as random walks. One of the members of the research group (Franchi) has worked in this direction (see Franchi and Paruolo, 2018b).
References
- Chang, Y., C. Kim, and J. Park (2016b) Nonstationarity in time series of state densities. Journal of Econometrics, 192, 152 -167.
- Di Iorio and Fachin (2017) Evaluating restricted Common Factor models fornon-stationary data. DSS-E3 WP 2017/2
- Franchi, M. and Paruolo, P. (2018b) Cointegration in functional autoregressive processes, Arxiv e-print, arXiv:1712.07522v2 [econ.EM]
- Onatski, A., C. Wang (2018) Alternative asymptotics for cointegration tests in large VARs, Econometrica, 86, 1465-1478.