The recent economic crisis has heavily affected the banking sector with consequences on the business model. Risk management and measurement has become the main goal for insurances and Banking regulatory authorities. Banks and insurance companies have to Implement and validate risk models in line with their exposures and adequately define capital regulatory requirements. New multivariate risk measures have to be developed and implemented using adequate methodologies which require the use of scenario analysis and latent variables. Systemic risk, liquidity risk, market risk and credit risk have become the major risk sources which allow to estimate the Risk Weighted Assets and the regulatory capital. Scenario analysis will be developed to estimate the probability distributions of the various exposures and the corresponding dynamics. The project aims to develop adequate methodologies for risk measurement and management in the banking and insurance sectors. The use of quantitative and qualitative variables to identify strategic choices represents key inputs. The final goal is to provide banks and insurances with the adequate risk measurement tool consistent with the new regulatory framework. Financial institutions should be able to run their usual business and play the crucial role for the economic system and growth.
In literature, several typologies of scenario generators have been proposed in financial problems. The classic approach tries to mimic the joint return behaviour by Monte Carlo generation of future scenario.
Clearly, scenario generation has to take into consideration return characteristics. In particular, we will provide real world portfolio and risks management solution in the framework of volatile markets with thousand of risk variables (factors), to incorporate the latest and more transparent advances in analytics, including comprehensive treatment of real world fat-tailed, skewed asset returns exhibiting volatility clustering and long range dependence. While the market return characteristics are widely explored, the path dependent portfolio strategies show specific challenges. To tackle the issues we will use the Markovian approach which allows:
1) Either to develop particular parametric Markov processes. We will mainly base our studies on the following papers: Staino et al. (2007) for the Lévy processes, Duan and Simonato (2001) and Duan et al. (2006) for the GARCH-type processes with Markovian innovations;
2) The computation of the statistical distribution (in an acceptable computational time) of any contingent claim, and financial products priced on the market. In particular we can easily compute the distribution of stopping times or first passage times useful for path dependent portfolio strategies (see Angelelli and Ortobelli (2009));
3) The computation of the joint Markov distributions of risk variables;
Using a more accurate methodology to generate scenarios and identifying latent variables to order the different risk preferences a more adequate model to measure credit risk and operational risk will be developed. The issue of correlation measurements among different exposures will also be addressed.
Concerning the MIFID2 application, the analysis of the latent variables will be developed through the Rasch multivariate analysis. The Rasch analysis applied to the results of the tests permits to define in a homogeneous scale the proper choice with respect to the latent variables. Starting from this point, we shall discuss this analysis for sectors of the Italian market, then will extend it to markets of the Euro zone and finally to some emerging markets.
The novelty is to build a set of latent variables (such as individual predisposition to earn/risk, capacity to product market success goods, capacity of time management) and addresses potential investors on the optimal dynamic strategy that accounts their preferences. Clearly the idea developed in this project can be used or for proper economic strategies of a country or for financial and economic strategies of firms.
Optimal choices can be identified dynamically and to capture the individual preferences among several admissible optimal choices. Historical return series can be measured with a proper factor model that accounts for most of the variability. Finally, using different scenarios concepts, we will mimic the joint behaviour of future market returns, paying attention to the modelling of all distributional aspects of the asset return series. This system will allow to compare the ex-post sample paths of the wealth obtained optimizing some performance strategies and using different dimensionality reduction techniques.
Another innovative aspect concerns the fair value of the participating life insurance contracts, usually composed by several options. The interaction among them increases the risk exposure and makes the evaluation more complex.
Focus on specific segments of mutual funds and/or portfolios to consider the overlap of the investments enhance the analysis of risk in financial markets.
The copula approach to integration and diversification adds new elements to the existing literature. First, portfolio owners are not considered as external to the market, but they are part of the market. This implies the introduction of the concepts of integration and diversification; such an approach creates a bridge between the literature on companies performances and the one on companies interactions, where the embedding of a company in a network is a key factor. Empirical analyses complete and give sound benchmark to the work.
Further references
Angelelli, E., Ortobelli S. (2009). American and European portfolio selection strategies: the Markovian approach. In: Financial Hedging, Novascience, New York.
Duan, J., Simonato, J. G. (2001). American option pricing under GARCH by a Markov chain approximation. Journal of Economic Dynamics and Control, 25(11): 1689--1718.
Duan, J., Ritchken, P., Sun, Z. (2006). Approximating GARCH-jump models, jump-diffusion processes, and option pricing. Math. Fin. 16, 21-52.
Staino, S., Ortobelli, S., Massabò, I. (2007). A Comparison among Portfolio Selection Strategies with Subordinated Lévy Processes . Int. j. comput. sci. new., 7: 224-233.