St.Petersburg Spring School in Risk Management, Insurance and Finance
General Information About the School
The main objective of the School is to disseminate knowledge and results of latest research in the area of Risk Management, Insurance, Finance, and related fields. The target audience is academics, students and practitioners from Russia and other countries of the former Soviet Union, although participants from the rest of the world are welcome to join the School as well.
The School is held at the European University at St.Petersburg (EUSP) in March or April each year.
The working language of the School is English.
Contact Information: Mrs Marina Kondrashova Department of Economics 

Email: mkondrashova@eu.spb.ru Phone/fax: +7 (812) 3867632 
Please refer to year tabs for further relevant information on past and future schools.
2014
The Third School is to be held at the European University at St.Petersburg (EUSP) and is organised with the generous support from Barclays Bank.
The School dates are March 2022, 2014. Participants requiring a visa to Russia must register before January 15, 2014. Other participants must register before March 1, 2014.
To register, please send your name, affiliation and contact information to the email address given below in the contact details.
As in two previous years the School is free to partake in, i.e. there is no participation fee in 2014.
The Organising Committee
 Jan Dhaene (KULeuven)
 Andrey Kudryavtsev (EUSP)
 Maxim Bouev (EUSP)
 Marina Kondrashova (EUSP)
The Programme
Click on a particular day below to see more details.
20 March, 2014: Risk Management day (6 hours)
From Risk Measures to Solvency Accounting by Professor KarlTheodor Eisele, University of Strasbourg, France 
Summary of the Course
Our starting point are convex risk measures (or their negative values: concave risk assessments) and their robust representations in the one and multiperiod case. They may be regarded as possible answers to the problem of Knightian uncertainty or model risk. Several examples will be given; among them the representation of asymptotically stable assessments. Of particular interest are timeconsistent risk assessments, the representations of which use in the coherent case mstable sets of probabilities.
Our next step is to regard risk assessments of nontradeable assets or liabilities in the presence of a financial market. Here, a central point to be discussed is the socalled supervisory arbitrage. If supervisory arbitrage is excluded, we are led to the problem of the existence of optimal replicating portfolios (ORP). Under ORPs, risk assessments are marketconsistent and thus define their marketconsistent envelope.
It turns out that this concept will become the essential tool of prudent supervision of banks and insurances. Based on a business plan e.g. of an insurance company, we introduce the concepts of supervisory provision and – margin. Next, we divide these items into their parts assigned to external capital and to own funds, respectively. This will lead us to the notions of risk margin and supervisory capital requirement. The relation between them is the costofcapital ratio. With the help of a bidprice operator, we give examples how the costofcapital ratio is calculated in the case of a finite risk horizon or in the stationary infinite case.
Finally we present the internal model of the European Solvency II project in its bottomup version, and show how it coincides with the topdown version where the concepts workedout so far will be used.
(Participants should have a good knowledge about measure and probability theory, including L^{p} spaces.)
21 March, 2014: Finance day (6 hours)
Analysis of the Monte Carlo CVA and Basel III CVA During and After the Financial Crisis of 20082009 by Dr Olga Pavlova, Managing Partner at Grizzly Bear Capital, London, UK 
Summary of the Course
The course will focus on the main driving forces behind counterparty risk for a derivatives portfolio. First, we shall run through main simulation techniques for CVA. Then we shall compare the simulated CVA with Basel III CVA using historical data from 2008 to present. We shall comment on strengths and weaknesses of the Basel approximation in different market conditions using the simulated CVA as a benchmark.
A more detailed summary will appear here in due course.
About the lecturer:
Olga Pavlova holds a PhD in Mathematics and a Master in Finance degrees. Her work experience in the City started at Deutsche Bank (London) on the Market Risk Modelling team. She then joined the Quantitative Risk Analytics team at Barclays Capital (London). There, she was a Global Head of Credit Risk Analytics, covering risk modelling for all derivative classes. Among other projects, she represented the bank in regulatory approvals of Basel II and III capital requirements. She subsequently joined UBS (London) as a Head of Quantitative Business Analysis. In this most recent role, Olga looked at various capital optimisation techniques for portfolios of derivatives. She currently works as an independent consultant giving lectures on various topics in financial modelling with the main emphasis being on counterparty risk. She also supervises masters degree theses at the Department of Financial Engineering, Imperial College (London).
22 March, 2014: Insurance day (6 hours)
Claims Reserving in NonLife Insurance and its Prediction Uncertainty by Professor Mario V. Wüthrich, ETH Zurich, Switzerland 
Summary of the Course
Lecture 1: The claims reserving problem and the chainladder method
We introduce the claims reserving problem and the claims reserving terminology. Claims reserving is one of the most important tasks in a nonlife insurance company. It aims at predicting cash flows of the outstanding loss liabilities. These predictions are needed for pricing of insurance products, for accouting of insurance business and for risk management purposes. We discuss these problems and we give first techniques that help to predict these cash flows.
Lecture 2: Stochastic chainladder model and prediction uncertainty
We present the stochastic chainladder model for claims reserving. The stochastic chainladder model is the most popular stochastic model that is used for claims reserving. On the one hand it is simple and on the other hand it provides quite reasonable predictions. We discuss this model and analyze the quantification of prediction uncertainty.
Lecture 3: The claims development result for solvency considerations
The quantification of prediction uncertainty is usually done over the entire runoff of the outstanding loss liabilities (longterm behavior). For modern solvency considerations we should also investigate the shortterm behavior, i.e. by which amount the claims reserves may change over the next accounting year. This leads to the consideration of the claims development result. We discuss this claims development result and analyze the quantification of prediction uncertainty in this shortterm view.
Lecture 4: Dependence modeling on claims reserving triangles
All the previous considerations are done under rather restrictive model assumptions about dependence, namely one makes strong assumptions about independence. In this lecture we relax these independence assumptions and we analyze the sensitivities of the results by allowing for dependence. We will observe that this increases the confidence bounds substantially.
2013
The Second School is to be held by the European University at St.Petersburg (EUSP) and is organised with the generous support from Barclays Bank.
The School dates are April 13, 2013. Participants must register before March 15, 2013.
To register, please send your name, affiliation and contact information to the email address given below in the contact details.
As in 2012 the School is free to partake in, i.e. there is no participation fee in 2013.
The Organising Committee
 Jan Dhaene (KULeuven)
 Andrey Kudryavtsev (EUSP)
 Maxim Bouev (EUSP)
 Natalia Voinova (EUSP)
The Programme
Click on a particular day below to see more details.
April 1st, 2013: Risk Management day (6 hours)
Comparison and Contrast of Risk Measures to Decision Principles, Solvency Capital and Insurance Premium Principles by Professor Marc Goovaerts, Katholieke Universiteit Leuven, Belgium 
Summary of the Course
The talk is divided into four parts:
Part 1. Risk Measures. In actuarial research, distortion, mean value and Haezendonck risk measures are concepts that are usually treated separately. In this part we indicate and characterize the relation between the different risk measures, as well as their relation to convex risk measures. While it is known that the mean value principle can be used to generate premium calculation principles, we will show how they also allow to generate solvency calculation principles. Moreover, we explain the role provided for the distortion risk measures as an extension of the Tail ValueatRisk (TVaR) and Conditional Tail Expectation (CTE).
Part 2. Decision Principles. We argue that a distinction exists between risk measures and decision principles. Though both are functions assigning a real number to a random variable, we think there is a hierarchy between the two concepts. Risk measures operate on the first “level”, quantifying the risk in the situation under consideration, while decision principles operate on the second “level”, often being derived from the risk measure. We illustrate this distinction with several canonical examples of economic situations encountered in insurance and finance. Special attention is paid to the role of axiomatic haracterizations in determining risk measures and decision principles. Some new axiomatic characterizations of families of risk measures and decision principles are also presented.
Part 3. Solvency capital. We examine properties of risk measures that can be considered to be in line with some “best practice” rules in insurance, based on solvency margins. We give ample motivation that all economic aspects related to an insurance portfolio should be considered in the definition of a risk measure. As a consequence, conditions arise for comparison as well as for addition of risk measures. We demonstrate that imposing properties that are generally valid for risk measures, in all possible dependency structures, based on the difference of the risk and the solvency margin, though providing opportunities to derive nice mathematical results, violates best practice rules. We show that socalled coherent risk measures lead to problems. In particular we consider an exponential risk measure related to a discrete ruin model, depending on the initial surplus, the desired ruin probability
Part 4. Insurance principles. The paper derives many existing risk measures and premium principles by minimizing a Markov bound for the tail probability. Our approach involves two exogenous functions v(S) and w (S,p) and another exogenous parameter a. Minimizing a general Markov bound leads to a unifying equation: E(w(S,p))=aE(v(S)). For any random variable, the risk measure p is the solution to the unifying equation. By varying the functions w and v, we derive the mean value principle, the zeroutility premium principle, the Swiss premium principle, Tail VaR, Yaari’s dual theory of risk, mixture of Esscher principles and more.
April 2nd, 2013: Insurance day (6 hours)
Health Insurance: Actuarial Models by Professor Ermanno Pitacco, University of Trieste, Italy 
Summary of the Course
The talk presents various technical aspects of health insurance products (including sickness benefits, disability annuities, and so on), focusing on the basic actuarial structures. Special attention is paid to the following topics:
1. The need for healthrelated insurance covers
2. Products in the area of health insurance
3. Between Life and NonLife insurance: the actuarial structure of health insurance products
4. Actuarial models for sickness benefits
5. Actuarial models for disability benefits (Income Protection and Long Term Care)
6. Some problems in current scenarios
April 3rd, 2013: Finance day (6 hours)
Applications of Backward Stochastic Differential Equations with Jumps to Pricing and Hedging of Contingent Claims in Incomplete Markets by Professor Łukasz Delong, Warsaw School of Economics, Poland 
Summary of the Course
In the first part of the course we recall basic techniques of stochastic calculus for Brownian motions and random measures. We next introduce Backward Stochastic Differential Equations with jumps (BSDEs with jumps). We investigate theoretical properties of BSDEs which are crucial for financial modelling. We comment on analytical and numerical methods for solving BSDEs. In particular, we discuss methods based on Itô formula and partial integrodifferential equations, Malliavin derivatives and Least Squares Monte Carlo.
In the second part of the course we show how to use Brownian motions and random measures for modelling equity, default and insurance risks. We focus on the problem of pricing and hedging of contingent claims in incomplete financial markets. Since in an incomplete market perfect hedging is not possible, we investigate different pricing and hedging objectives, including superhedging, quadratic hedging under an equivalent martingale measure, quadratic hedging under a realworld measure, hedging under local meanvariance Markowitz risk measure, robust hedging under model risk, nogooddeal pricing. We show how to solve these optimization problems and define optimal prices and hedging strategies by applying BSDEs. Advantages of applying BSDEs are discussed.
2012
The School was held at the European University at St.Petersburg on April 24, 2012.
Participants had to register before March 15, 2012. Participation was free for all participants.
The Organising Committee
 Jan Dhaene (KULeuven)
 Andrey Kudryavtsev (EUSP)
 Alexander Surkov (EUSP)
 Natalia Voinova (EUSP)
The Programme
Click on a particular day below to see more details.
April 2nd, 2012: Finance day (6 hours)
Superhedging Strategies for Index Options and Measuring Systemic Risk in Stock Markets by Professor Jan Dhaene, Katholieke Universiteit Leuven, Belgium

Summary of the Course
We investigate static superreplicating strategies for Europeantype call options written on a stock index, that is a weighted sum of stock prices. Both the infinite market case (where prices of the plain vanilla options are available for all strikes) and the finite market case (where only a finite number of plain vanilla option prices are observed) are considered. We show how to construct a portfolio consisting of the plain vanilla options on the different stocks, whose payoff superreplicates the payoff of the index option. As a consequence, the price of the superreplicating portfolio is an upper bound for the price of the index option. The superhedging strategy is modelfree in the sense that it is expressed in terms of the observed option prices on the individual stocks. In a second part of the course, we introduce an easy to calculate measure for systemic risk in stock markets. This measure is baptized the Herd Behavior Index (HIX). It is modelindependent and forward looking, based on observed option data. In order to determine the degree of systemic risk or herd behavior in a stock market one should compare the observed market situation with the extreme (theoretical) situation under which the whole system is driven by a single factor. The Herd Behavior Index (HIX) is defined as the ratio of an optionbased estimate of the riskneutral variance of the market index and an optionbased estimate of the corresponding variance of this extreme market situation. The HIX can be determined for any market index provided an appropriate series of vanilla options is traded on this index as well as on its components. As an illustration, we determine historical values of the 30days implied Herd Behavior Index for the Dow Jones Industrial Average.
April 3rd, 2012: Risk Management day (6 hours)
Principles and Methods of Capital Allocation for Enterprise Risk Management by Professor Emiliano Valdez, University of Connecticut, USA 
Summary of the Course
There is a growing need and interest among financial institutions not only to determine the total company capital requirement, but also to allocate this total capital across various business units or product lines. The term "capital allocation" has been used to refer to a fair and equitable subdivision of this total capital requirement; it is indeed similar in concept to a fair division of capital in a diversified portfolio of investments. In this talk, we will examine the principles behind capital allocation for enterprise risk management: what is considered fair and equitable, what are the regulatory requirements, and what methods can be used. Along these principles, we examine a unifying framework for allocating the aggregate capital of a financial firm to its various business units or product lines. This unifying approach relies on an optimization argument, requiring that the weighted sum of measures for the deviations of the business unit's losses from their respective allocated capital be minimized. In essence, this leads us to some degree of fairness and equity because this requires capital to be close to the risk that necessitates holding it. Additionally, this approach is very flexible in the sense that different forms of the objective function can reflect alternative definitions of company risk tolerance. Owing to this flexibility, the general framework reproduces several capital allocation methods that appear in the literature and allows for alternative interpretations and possible extensions. Several examples will be discussed to illustrate and to understand the implications of the results arising from these various methods.
April 4th, 2012: Insurance day (6 hours)
The chainladder method: influence analysis, robustification and diagnostic tool by Tim Verdonck, Katholieke Universiteit Leuven, Belgium 
Summary of the Course
The chainladder method is a widely used technique to forecast the reserves that have to be kept regarding claims that are known to exist, but for which the actual size is unknown at the time the reserves have to be set. Such claims are often represented in a runoff triangle and hence the goal of claims reserving is to obtain predictions for the lower part of the triangle. Several traditional actuarial methods to complete a runoff triangle, such as the chainladder method, can be described by one Generalized Linear Model (GLM). In practice it can be easily seen that even one outlier can lead to a huge over or underestimation of the overall reserve when using the chainladder method. This indicates that individual claims can be very influential when determining the chainladder estimates. The effect of contamination can be mathematically analyzed by calculating influence functions in the GLM framework corresponding to the chainladder method. It is proven that the influence functions are unbounded, confirming the sensitivity of the chainladder method to outliers. Robust alternatives are introduced to estimate the future claims reserves in a more outlier resistant way. Based on the influence functions and the robust estimators, a diagnostic tool is presented highlighting the influence of every individual claim on the classical chainladder estimates. With this tool it is possible to detect immediately which claims have an abnormally positive or negative influence on the reserve estimates. The robust methodology and the diagnostic tool will be illustrated on some artificial and real data sets.