We introduce a random network model for business relationships as for instance an insurance market, overlapping portfolios, or Operational Risk. Using Pareto-tailed losses (as are observed for large risks) with a dependence structure introduced by the graph we study systemic risk measures, which are based on the Value-at-Risk and the Expected Shortfall. We show that the dependence on the network structure plays a fundamental role for the individual agent’s risk as well as for the market risk. The focus of our analysis lies in the study of the influence of the random graph on risk measures, where we consider the Bernoulli graph and a Rasch-type graph as examples. In particular, we explain the influence of the network structure on diversification in such models. This is joint work with Oliver Kley and Gesine Reinert.
This paper analyzes the reliability of standard approaches for financial risk analysis. We focus on the difference between Value-at-Risk and expected shortfall, their small sample properties, the scope for underreporting risk and how estimation can be improved. Overall, we find that risk forecasts are extremely uncertain at low sample sizes, with Value-at-Risk more accurate than expected shortfall. Value-at-Risk is easily deliberately underreported without violating regulations and control mechanisms. Finally, we discuss the implications for academic research, practitioners and regulators, along with best practice suggestions.
In the 100th birth anniversary of the mathematician and Economics Nobel winner Janos Karoly Harsanyi, we explore possible fertile connections between some of Harsanyi’s key results in game theory and open questions in Quantitative Finance. Harsanyi dividends, unknown to the large risk community public, represent a canonical way to decompose the result of a cooperative game into cohort contributions, showing interference effects at all orders. Interpreting risk or return as the game result, and risk factors, portfolio positions or investment decisions as the players of the game, we immediately obtain new paradigms to risk attribution, performance contribution and performance attribution. We compare common best practices with these new paradigms and we highlight the challenges (computational, theoretical) that these methods open, in the hope to spur future research.
The environment in which financial service providers operate has changed significantly since the financial crisis. Risk management had to address various issues over that period motivated in part by stricter regulations. However, changes in the global economy as well as changes in the focus of global regulatory agenda present new questions for risk management to answer. This talk intends to highlight some of these from a practitioner’s point of view.
Model Risk Management (MRM) within banks has been evolving from model validation towards becoming an effective and value-centric function. What are the drivers, stages, and results of this evolution? Is it possible to measure model risk? How can we use Artificial Intelligence (AI) / Machine Learning (ML) models? These are the questions we will address in the talk.
The demand for market risk measurement over long horizons increases as mutual funds and pension plan managers want to control their long-term risk/return profile. The naïve extension of the classic risk management framework fails to capture key components of long term risk. Mean reversion and regime shifting are important factors shaping long horizon scenarios, while portfolio rebalancing with state dependent transaction cost, redemptions and forced trading are examples of wrong way liquidity risk.
Investors are becoming increasingly aware of the fundamental idea of inevitable uncertainty in future estimates, and would like to reflect this in their asset allocation decisions. The tectonic shift of financial markets towards sustainable solutions underlines the importance of this mindset, given the variation in different views and beliefs, and the magnitude of its impact. We develop a systematic way to construct portfolios, residing on robust and stochastic optimisation techniques, minimising portfolio VaR/CVaR. We demonstrate how this method can be used to account for uncertainty in return estimates, and create less concentrated portfolios.
We introduce the generalized autoregressive score (GAS) models by Creal et al 2014 which provide a flexible yet computationally tractable framework to model latent variables such as volatility or default intensity. By tying the dynamics of the latent variables to the distributional assumption on the observations GAS models coherently accommodate, and appropriately moderate, the impact of outliers on parameter estimates. We showcase novel empirical adaptations of GAS models to equity risk forecasting, highlighting their intuitive appeal, computational simplicity and empirical benefits relative to current industry standard volatility forecasting models.
email: email@example.comThe main sponsor is