In this paper we use doubly stochastic processes (or Cox processes) in order to model the random evolution of mortality of an individual. These processes have been widely used in the credit risk literature in modelling default arrival, and in this context have proved to be quite flexible, especially when the intensity process is of the affine class. We investigate the applicability of affine processes in describing the individual's intensity of mortality, and provide a calibration to the Italian and UK populations. Results from the calibration seem to suggest that, in spite of their popularity in the financial context, mean reverting processes are not suitable for describing the death intensity of individuals. On the contrary, affine processes whose deterministic part increases exponentially seem to be appropriate. As for the stochastic part, negative jumps seem to do a better job than diffusive components. Stress analysis and analytical results indicate that increasing the randomness of the intensity process results in improvements in survivorship.
The paper explores the properties of a class of multivariate Lévy processes used for asset returns. We focus on describing both linear and non-linear dependence in an economic sensible and empirically appropriate way. The processes are subordinated Brownian motions. The subordinator has a common and an idiosyncratic component, to reflect the properties of trade, which it represents. A calibration to a portfolio of 10 US stock indices returns over the period 2009–2013 shows that the hyperbolic specification has a very good fit to marginal distributions, to the overall correlation matrix and to the return distribution of both long-only and long-short random portfolios, which also incorporate non-linear dependence. Their tail behaviour is also well captured by the variance gamma specification. The main message is not only the goodness of fit, but also the flexibility in capturing dependence and the ease of calibration on large sets of returns.
Artificial intelligence (AI) is a tool that financial intermediaries and insurance companies use or are willing to use in almost all their activities. AI can have a positive impact on almost all aspects of the insurance value chain: pricing, underwriting, marketing, claims management, and after-sales services. While it is very important and useful, AI is not free of risks, including those related to its robustness against so-called adversarial attacks, which are conducted by external entities to misguide and defraud the AI algorithms. The paper is designed to review adversarial AI and to discuss its implications for the insurance sector. We give a taxonomy of adversarial attacks and present an original, fully fledged example of claims falsification in health insurance, as well as some remedies which are consistent with the current regulatory framework.
This paper studies the dependence between coupled lives, i.e., the spouses’ dependence, across different generations, and its effects on prices of reversionary annuities in the presence of longevity risk. Longevity risk is represented via a stochastic mortality intensity. We find that a generation-based model is important, since spouses’ dependence decreases when passing from older generations to younger generations. The independence assumption produces quantifiable mispricing of reversionary annuities, with different effects on different generations. The research is conducted using a well-known dataset of double life contracts.
In this paper we suggest the adoption of copula functions in order to price multivariate contingent claims. Copulas enable us to imbed the marginal distributions extracted from vertical spreads in the options markets in a multivariate pricing kernel. We prove that such kernel is a copula function, and that its super-replication strategy is represented by the Frechet bounds. As applications, we provide prices for binary digital options, options on the minimum and options to exchange one asset for another. For each of these products, we provide no-arbitrage pricing bounds, as well as the values consistent with independence of the underlying assets. As a final reference value, we use a copula function calibrated on historical data.