SEMINAR

FINANCIAL MODELLING

The life of the scientific program Financial modelling of the Centre d’Economie de la Sorbonne is organized around a monthly research seminar from October to June. The seminar takes place on the 3rd Wednesday of each month, from 17h to 19h on the 6th floor of the MSE building (Maison des Sciences Economiques, 106, boulevard de l’Hôpital, 75013 Paris). During each seminar, two to three research papers are presented.

The seminar invites French and International scholars and practitioners to present their ongoing research in all the aspects of financial modeling.

The seminar is organized in partnership with CNRS and Panthéon-Sorbonne University and benefits from a BQR (Bonus Qualité Recherche).


CONTACT: Isabelle Nagot, Christophe Chorro, Eduardo Abi Jaber

2022 ONLINE SEMINAR



Wednesday 22 June 2022 at 18h00: Ajouter au calendrier

Jean-Edouard Colliard (HEC)

Title: Algorithmic Pricing and Liquidity in Securities Markets


Abstract: We run experiments in which machine-learning algorithms play a standard market-making game under adverse selection. We study how the outcome of these experiments differs from standard equilibrium predictions. We find that a monopolist market-maker charges a price lower than the standard monopoly price. In contrast, competing market-makers charge a price at a mark-up above the competitive price. We run comparative statics exercises that deliver new empirical predictions. In particular, the mark-up decreases in the amount of adverse selection. Beyond the case of market-making, our framework is a step towards developing novel predictions on the impact of algorithmic trading in financial markets.

Joint work with Thierry Foucault, Stefano Lovo



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/92175983954?pwd=YVVyOW1UaWJwQy9qVGltWkJhZUlVZz09


ID de réunion : 921 7598 3954

Code secret : 217808


Wednesday 8 June 2022 at 18h00: Ajouter au calendrier

Thierry Foucault (HEC)

Title: Equilibrium Data Mining and Data Abundance


Abstract: We propose a new theory of information production in financial markets. In this theory, speculators search for new predictors of asset payoffs and optimally decide to trade on predictors whose signal-to-noise ratio exceeds an endogenous threshold. We use the model to derive predictions regarding the effects of progress in information technologies on quantitative asset managers' performance, the similarity of their holdings, and the informativeness of asset prices. We show that data abundance (an expansion of the search space for predictors due to greater data diversity) and greater data processing power do not have the same effects.

Joint work with Jérôme Dugast.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/96615226605?pwd=dUdMTW16RS9mVkJ2cVpVdEdjNktvUT09


ID de réunion : 966 1522 6605

Code secret : 844792



Tuesday 24 May 2022 at 18h00: Ajouter au calendrier

Christophe Geissler (Advestis)

Title: Using ESG scores in equity issuer clustering.


Abstract: We present here an original application of the non-negative matrix factorization (NMF) method, for the case of extra-financial ESG data. These data are subject to high correlations between co-variables, as well as between observations. NMF provides a more easily interpretable clustering of co-variables and observations than a simple principal component analysis (PCA). It gives rise to a natural issuer clustering based on the ESG scores.


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/97866883736?pwd=LzNlVzg4TGYvZDZaeCtDL2VOdVFaUT09


ID de réunion : 978 6688 3736

Code secret : 989561




Wednesday 11 May 2022 at 18h00: Ajouter au calendrier

Zhenyu Cui (Stevens Business School)

Title: Applications of the Dirac Delta Family Method in Implied Volatility, Risk-neutral Density, and High-dimensional Stochastic Control


Abstract: In this talk, the Dirac Delta family method is introduced, which is based on orthogonal polynomial representations of the Dirac Delta function. We show how this method can lead to new valuation results in three financial applications. First, when combined with the change of variable technique, we obtain an explicit model-free formula for the Black-Scholes implied volatility. The formula is expressed through either a limit or as an infinite series of elementary functions, and we establish that the proposed formula converges to the true implied volatility value. Numerical and empirical examples illustrate the accurateness of the formula. Second, when combined with the celebrated Carr-Madan spanning formula, we derive a novel model-free representation of the risk-neutral density in terms of market-observed options prices. Compared to the widely used method for obtaining the risk-neutral densities via the Breeden–Litzenberger device, our method yields estimate of risk-neutral densities that are model-free, automatically smooth, and in closed-form. Third, when applied to calculating the conditional expectations arising from dynamical programming, we show that it leads to a novel numerical time-stepping approach for solving corresponding HJB system. We demonstrate the accuracy and efficiency of the method through solving some one-dimensional and two-dimensional control problems.


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/95067055748?pwd=UWdveVB6clcvdUNrR0pBNThYVExEZz09


ID de réunion : 950 6705 5748

Code secret : 012438



Wednesday 20 April 2022 at 18h00: Ajouter au calendrier

Serge Tabachnik (Lombard Odier) and Eric Benhamou (AI for alpha)

Title: Machine Learning Allocation for Volatility Targeting Portfolios



Abstract: In the context of proactive risk management and volatility targeting portfolio construction, finding robust predictors of future realized volatility is paramount to achieving optimal performance. Up to now, this has motivated the search for a stream of models to forecast volatility based on multiple methodologies. However, because of the complexity and non-stationary behavior of volatility, the appropriate choice of the volatility estimator, and hence the volatility targeting model relying on it, remains an open question. This presentation proposes the implementation of two Machine Learning (ML) based methodologies, to determine the optimal allocation between volatility targeting models. The first methodology is based on Deep Reinforcement Learning (DRL), while the second is based on Gradient Boosting Decision Trees (GBDT). These two innovative methodologies are intrinsically different and distinguished from one another by their learning approach. GBDT uses supervised learning contrary to DRL that falls into the unsupervised learning. Both of these dynamic allocations rely on Adaptive Machine Learning (AML) methods that aim to take the typical regime changes of volatility into account. While this adaptive aspect is inherent to reinforcement learning, in the case of the supervised learning approach it comes from the addition of a filtering step that aims to take the regime changes further into account.


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/91779346457?pwd=YllKUnkxUkNQdTdvYmVOT1N1OTJwUT09


ID de réunion : 917 7934 6457

Code secret : 424113



Wednesday 6 April 2022 at 18h00: Ajouter au calendrier

Damien Challet (Centrale Supélec)

Title: Filtering the covariance matrix of non-stationary systems with constant eigenvalues


Abstract: We propose a data-driven way to reduce the noise of covariance matrices of nonstationary systems. Our method rests on long-term averaging of the influence of the future on present eigenvalues. This zeroth-order approximation outperforms the latest optimal methods designed for stationary systems as soon as the system is not stationary as seen in large-scale empirical investigations on optimal global minimum variance portfolios and with synthetic data.


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/92265608992?pwd=MkJFWE9BTGJoM2Fkai9EcDZBaEhwdz09

ID de réunion : 922 6560 8992

Code secret : 502356



Wednesday 9 March 2022 at 18h00: Ajouter au calendrier

Guillaume Bernis (Natexis)

Title: Clustering Effects for Derivatives Pricing: Applications of Hawkes Processes in Finance



Abstract: We propose an extension of the $\Gamma$-OU Barndorff-Nielsen and Shephard model taking into account jump clustering phenomena.

We assume that the intensity process of the Hawkes driver coincides, up to a constant, with the variance process. By applying the theory of continuous-state branching processes with immigration, we prove existence and uniqueness of strong solutions of the SDE governing the asset price dynamics. We exploit a measure change of self-exciting Esscher type in order to describe the relation between the risk-neutral and the historical dynamics, showing that the $\Gamma$-OU Hawkes framework is stable under this probability change. By exploiting the affine features of the model we provide an explicit form for the Laplace transform of the asset log-return, for its quadratic variation and for the ergodic distribution of the variance process.

We show that the proposed model exhibits a larger flexibility in comparison with the $\Gamma$-OU model, in spite of the same number of parameters required. We calibrate the model on market vanilla option prices via characteristic function inversion techniques, we study its sensitivities and propose an exact simulation scheme. The main financial result is that implied volatility of options written on VIX is upward shaped due to the self-exciting property of Hawkes processes, in contrast with the usual downward slope in the $\Gamma$-OU Barndorff-Nielsen and Shephard model.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/92648842718?pwd=ekRKWkl5R1dRN2Yxa0VoNk8xVWQ1dz09

ID de réunion : 926 4884 2718

Code secret : 586575




Wednesday 9 February 2022 at 18h00: Ajouter au calendrier

Ryan Donnelly (King's College)

Title: Optimal Execution with Exploratory Trading



Abstract: An agent wishes to liquidate a block of shares subject to price impact effects, but also desires to explore regions of the state and control space that would be avoided according to the optimal strategy of the model. This is accomplished by incorporating unpredictable randomness in the trading strategy at each point in time, and offering a reward in the form of Shannon's differential entropy of the distribution of this random component. We propose a framework in discrete time with captures this objective and solve for the optimal distribution of trades. At each point in time the optimal trade distribution is Gaussian with parameters that are given in terms of the solution to a backwards stochastic difference equation. The solution to this backwards stochastic difference equation can be approximated by a continuous time analogue, which can be solved in closed form. Using this approximation, we demonstrate the relation between this discrete time model and other pieces of literature which work in continuous time.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/98404199876?pwd=NlM0ZnRueFI0VDJCSkFuckpoVit5dz09

ID de réunion : 984 0419 9876

Code secret : 603818


Wednesday 26 January 2022 at 15h45: Ajouter au calendrier

Blanka Horvatz (TU Munich)

Title: Data - Driven Market Simulators & some applications of signature kernel methods in mathematical finance



Abstract: Techniques that address sequential data have been a central theme in machine learning research in the past years. More recently, such considerations have entered the field of finance-related ML applications in several areas where we face inherently path dependent problems: from (deep) pricing and hedging (of path-dependent options) to generative modelling of synthetic market data, which we refer to as market generation.

We revisit Deep Hedging from the perspective of the role of the data streams used for training and highlight how this perspective motivates the use of highly accurate generative models for synthetic data generation. From this, we draw conclusions regarding the implications for risk management and model governance of these applications, in contrast to risk-management in classical quantitative finance approaches.

Indeed, financial ML applications and their risk-management heavily rely on a solid means of measuring (and efficiently computing) similarity-metrics between datasets consisting of sample paths of stochastic processes. Stochastic processes are at their core random variables on path space. However a consistent notion of and efficiently computable similarity-metrics for stochastic processes remained a challenge until recently. We propose such appropriate similarity metrics and contrast them with returns-based similarity metrics. Finally, we discuss the effect of incorporating the information structure (the filtration) of the market into these similarity metrics and the implications of such metrics on options prices.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/96518678430?pwd=YjdUV3RDR2VZb3NnUzRBaWRoalVYZz09

ID de réunion : 965 1867 8430

Code secret : 223789





Wednesday 12 January 2021 at 6pm: Ajouter au calendrier

Adeline Fermanian (UPMC)

Title: Learning from time-dependent data: signatures, RNN, and neural ODE


Abstract: Time-dependent data arise in many fields of research, such as quantitative finance, medicine, or computer vision. We will be concerned with a novel approach for learning with such data, called the signature transform, and rooted in rough path theory. Its basic principle is to represent multidimensional paths by a graded feature set of their iterated integrals, called the signature. After a general overview of signatures in machine learning, we show its application on one specific problem. Building on the interpretation of a recurrent neural network (RNN) as a continuous- time neural differential equation, we show, under appropriate conditions, that the solution of a RNN can be viewed as a linear function on the signature. This connection allows us to frame a RNN as a kernel method in a suitable reproducing kernel Hilbert space. As a consequence, we obtain theoretical guarantees on generalization and stability for a large class of recurrent networks.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/97915263949?pwd=QWlnQXo2emY1dWhDandZcGNsZjZvUT09


ID de réunion : 979 1526 3949

Code secret : 683848





2021 ONLINE SEMINAR



Wednesday 9 June 2021 at 6pm:


Maciej Augustyniak (Université de Montreal)

Title: A Discrete-Time Hedging Framework with Multiple Factors and Fat Tails: On What Matters


Abstract: This article presents a quadratic hedging framework for a general class of discrete-time affine multi-factor models and investigates the extent to which multi-component volatility factors, fat tails, and a non-monotonic pricing kernel can improve the hedging performance. A semi-explicit hedging formula is derived for our general framework which applies to a myriad of the option pricing models proposed in the discrete-time literature. We conduct an extensive empirical study of the impact of modelling features on the hedging effectiveness of S&P 500 options. Overall, we find that fat tails can be credited for half of the hedging improvement observed, while a second volatility factor and a non-monotonic pricing kernel each contribute to a quarter of this improvement. Moreover, our study indicates that the added value of these features for hedging is different than for pricing. A robustness analysis shows that a similar conclusion can be reached when considering the Dow Jones Industrial Average. Finally, the use of a hedging-based loss function in the estimation process is investigated in an additional robustness test, and this choice has a rather marginal impact on hedging performance.

A draft of the paper can be accessed here for those interested:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3728995



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/94584704249?pwd=MlFDZFRJMEVRMS9mS2FOMGdFZGhTQT09


ID de réunion : 945 8470 4249

Code secret : 369060


Wednesday 26 May 2021 at 6pm:


HAORAN WANG (Senior Research Data Scientist at Vanguard)

Title: Continuous-time mean–variance portfolio selection: A reinforcement learning framework


Abstract: We approach the continuous-time mean–variance portfolio selection with reinforcement learning (RL). The problem is to achieve the best trade-off between exploration and exploitation, and is formulated as an entropy-regularized, relaxed stochastic control problem. We prove that the optimal feedback policy for this problem must be Gaussian,with time-decaying variance. We then prove a policy improvement theorem, based on which we devise an implementable RL algorithm.We find that our algorithm and its variant outperform both traditional and deep neural network based algorithms in our simulation and empirical studies. (This is a joint work with Xun Yu Zhou at Columbia University.)


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/97779025385?pwd=U0hZdVBvSi9MVHpUNUdEY2t3Y3FPdz09


ID de réunion : 977 7902 5385

Code secret : 806013



Wednesday 12 May 2021 at 5pm:

BRUNO DUPIRE (head of quantitative research, Bloomberg )

Title: Functional Itô Calculus


Abstract:

We extend some results of the Itô calculus to functionals of the current path of a process to reflect the fact that often the impact of randomness is cumulative and depends on the history of the process, not merely on its current value. We express the differential of the functional in terms of adequately defined partial derivatives to obtain an Itô formula. We develop an extension of the Feynman-Kac formula to the functional case and an explicit expression of the integrand in the Martingale Representation Theorem. We establish that under certain conditions, even path dependent options prices satisfy a partial differential equation in a local sense. We exploit this fact to find an expression of the price difference between two models and compute variational derivatives with respect to the volatility surface.


Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/96858121629?pwd=NVhZM2k0MDREQ0FOdjFzOG4xd1ZiQT09

ID de réunion : 968 5812 1629
Code secret : 585175



Wednesday 14 April 2021 at 6pm:


MATTHIEU GARCIN (Léonard de Vinci Pôle universitaire, Paris)

Title: Fractional models: estimation, forecast, and market efficiency


Abstract:


The Hurst exponent describes the scaling properties of a time series. One also often links its value to the persistence of the series and consequently to one’s ability to forecast it: if H=1/2 there is no autocorrelation, if H>1/2 the series is persistent, and if H<1/2 the series is anti-persistent. However, the interpretation of the Hurst exponent strongly depends on the model describing the dynamic. Beyond the fractional Brownian motion (fBm), we are interested in various models: non-Gaussian extensions of the fBm, stationary transforms of a fBm, and multifractional motions, which rely on the assumption that the Hurst exponent is time-varying or even is a random process. We expose the specificities of the estimation of the Hurst exponent for all these models as well as the way one can forecast such series, using accuracy metrics that are relevant in the perspective of a portfolio manager. We also present various indicators of market efficiency based on these models.



Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/91041223916?pwd=anBPYkdyT3R5YTNCSGxOWDcxcWJ0QT09


ID de réunion : 910 4122 3916

Code secret : 096455


Wednesday 24 MARCH 2021 at 6pm


JULIEN GUYON (BLOOMBERG)

Title: The Joint S&P 500/VIX Smile Calibration Puzzle Solved and related topics


Abstract:


The very high liquidity of S&P 500 (SPX) and VIX derivatives requires that financial institutions price, hedge, and risk-manage their SPX and VIX options portfolios using models that perfectly fit market prices of both SPX and VIX futures and options, jointly. This is known to be a very difficult problem. Since VIX options started trading in 2006, many practitioners and researchers have tried to build such a model. So far the best attempts, which used parametric continuous-time jump-diffusion models on the SPX, could only produce approximate fits. In this talk we solve this long standing puzzle for the first time using a completely different approach: a nonparametric discrete-time model. Given a VIX future maturity T1, we build a joint probability measure on the SPX at T1, the VIX at T1, and the SPX at T2 = T1 + 30 days which is perfectly calibrated to the SPX smiles at T1 and T2, and the VIX future and VIX smile at T1. Our model satisfies the martingality constraint on the SPX as well as the requirement that the VIX at T1 is the implied volatility of the 30-day log-contract on the SPX.


The model is cast as the unique solution of what we call a Dispersion-Constrained Martingale Schrodinger Problem which is solved by duality using an extension of the Sinkhorn algorithm, in the spirit of (De March and Henry-Labordere, Building arbitrage-free implied volatility: Sinkhorn's algorithm and variants, 2019). We prove that the existence of such a model means that the SPX and VIX markets are jointly arbitrage-free. The algorithm identifies joint SPX/VIX arbitrages should they arise. Our numerical experiments show that the algorithm performs very well in both low and high volatility environments. Finally, we discuss how our technique extends to continuous-time stochastic volatility models, via what we dub VIX-Constrained Martingale Schrodinger Bridges, inspired by the classical Schrodinger bridge of statistical mechanics. The resulting stochastic volatility model is numerically implemented and is shown to achieve joint calibration with very high accuracy.


Time permitting, we will also briefly discuss a few related topics:

(i) a remarkable feature of the SPX and VIX markets: the inversion of convex ordering, and how classical stochastic volatility models can reproduce it;

(ii) why, due to this inversion of convex ordering, and contrary to what has often been stated, among the continuous stochastic volatility models calibrated to the market smile, the local volatility model does not maximize the price of VIX futures;

(iii) what are the optimal model-free bounds on the prices of VIX futures given SPX smiles.


ZOOM INVITATION


https://zoom.univ-paris1.fr/j/91712644694?pwd=VVkwNTBiN1NhMVArN2NlMDlnZDdCdz09


ID de réunion : 917 1264 4694

Code secret : 803722



Wednesday 17 MARCH 2021 at 6pm

SERGIO PULIDO (ENSIIE)

Title: American options in the rough Heston model

Abstract: Rough volatility models have emerged as compelling alternatives to classical semimartingale models to capture important stylized features of the implied volatility surface and the time series of realized volatility. The rough Heston model is particularly appealing because its affine structure facilitates the pricing of European options using Fourier techniques. In this work we consider the problem of pricing American options in the rough Heston model. The complexity of the problem stems from the absence of a Markovian-semimartingale structure in the model. To overcome this difficulty, we work with a Markovian multi-factor semimartingale stochastic volatility model, which approximates the rough Heston model. In this approximated model, American options can be priced using a backward approach and simulation-based methods. We prove the convergence of American options prices in the multi-factor model towards the prices in the rough Heston model. The proof relies on the explicit expression of the conditional characteristic function of the joint forward process and the spot price, which is a consequence of the affine structure of the model. We illustrate with some numerical examples the behavior of American option prices with respect to some parameters in the model. This is joint work with Etienne Chevalier and Elizabeth Zuñiga.


ZOOM INVITATION:

Sujet : Financial Modelling seminar : SERGIO PULIDO ENSIIE

Heure : 17 mars 2021 06:00 PM Paris

Participer à la réunion Zoom

https://zoom.univ-paris1.fr/j/91851743203?pwd=aFhGck9WZW9GM0Y4UjUvcW9ycTQ4Zz09

ID de réunion : 918 5174 3203

Code secret : 633782



2019-2020

A VENIR:


le 18/12 Matthieu Garcin (ESILV) +MIGUEL ANGEL SANCHEZ GRANERO (Alméria) [Self similarities]

le 15/01 Jentzen (Université de Munster), Alvaro Cartea (Oxford) [Machine Learning]

le 12/02 Roy cerquetti ( Sapienza Rome)+TBA

le 18/03 Alexandru Badescu (Calgary)+Sébastien Laurent (AMSE à confirmer) [Econometrics]

le 22/04 Sergio Pulido (ENSIEE) et Thibaut Mastrolia (Ecole Polytechnique) [Analyse stochastique]

le 27/05 Andrzej Nowak (Univ. Varsovie) + Dehua Shen (Univ. Tianjin) [Finance comportementale]

le 24/06 TBA [Nowcasting]


Wednesday, November 27 th 2019 (Room S17)


17h: Bernard de Meyer et Moussa Dabo (CES, Paris 1)

The CMMV Pricing Model in Practice

Abstract: Mainstream financial econometrics methods are based on models well tuned to replicate price dynamics, but with little to no economic justification. In particular, the randomness in these models is assumed to result from a combination of exogenous factors. In this paper, we present a model originating from game theory, whose corresponding price dynamics are a direct consequence of the information asymmetry between private and institutional investors. This model, namely the CMMV pricing model, is therefore rooted in market microstructure. The pricing methods derived from it also appear to fit very well historical price data. Indeed, as evidenced in the last section of the paper, the CMMV model does a very good job predicting option prices from readily available data. It also enables to recover the dynamic of the volatility surface.

https://arxiv.org/abs/1910.10005


6pm: Ramy Sukarieh (United Nations JSPF )

Stock selection and Portfolio construction using VAR models and Genetic Algorithm

We seek to construct portfolios with the following intuitive characteristics: ordered, segregated, integrated, conditional, and concentrated. To this end, we use vector auto-regressive VAR models and Genetic Algorithm GA. The VAR models were estimated using normal equations and MLE. The GA randomly permutes from the uniform discrete distribution without replacement, compute rank based fitness scaling function, and selects via stochastic universal sampling function (SUS). We use the Akaike Information Criterion (AIC) as the fitness function to the GA algorithm. This research shows that combining linear models with GA lead to well behaved investable portfolios with reasonable performance when compared to the market. In addition, it provides a simple yet rich framework from which one can bifurcate into more sophisticated model such as, Bayesian in time-varying VAR TVP-VAR with stochastics volatility, where inputs can be obtained from nonlinear models such as Poly-models.


Wednesday, October 16 th 2019

17h: Eduardo Abi Jaber (Paris 1)

Invariance and viability theory: from finite dimensional ODEs to infinite dimensional SDEs

Abstract: We will give a general overview of viability theory for arbitrary closed sets with respect to deterministic and stochastic systems. A closed set D is said to be viable for a dynamical system, if there exists a D-valued solution to this system. We will provide a new geometric characterization for finite-dimensional stochastic differential equations (with jumps), which extends the well-known inward pointing Stratonovich drift condition to the case where the diffusion matrix may fail to be differentiable. Finally, we will discuss the extension to infinite dimensional S(P)DEs.


18h: Luciano Campi (London School of economics)

N-player games and mean-field games with smooth dependence on past absorptions

Abstract: N-player games and mean-field games with smooth dependence on past absorptions Mean-field games with absorption is a class of games, that have been introduced in Campi and Fischer (2018) and that can be viewed as natural limits of symmetric stochastic differential games with a large number of players who, interacting through a mean-field, leave the game as soon as their private states hit some given boundary. In this talk, we push the study of such games further, extending their scope along two main directions. First, a direct dependence on past absorptions has been introduced in the drift of players' state dynamics. Second, the boundedness of coefficients and costs has been considerably relaxed including drift and costs with linear growth. Therefore, the mean-field interaction among the players takes place in two ways: via the empirical sub-probability measure of the surviving players and through a process representing the fraction of past absorptions over time. Moreover, relaxing the boundedness of the coefficients allows for more realistic dynamics for players' private states. We prove existence of solutions of the mean-field game in strict as well as relaxed feedback form. Finally, we show that such solutions induce approximate Nash equilibria for the N-player game with vanishing error in the mean-field limit as N goes to infinity. This talk is based on a joint work with Maddalena Ghio and Giulia Livieri (SNS Pisa).


2018-2019


Tuesday 18 June 2019

12:30: Marcos Costa Santos Carreira (Ecole Polytechnique - CMAP)

Learning Interest Rate Interpolation

Abstract: The usual methods for interest rate interpolation consider only the values and time to maturity of spot rates as the inputs, and differ mainly on the continuity of the implied forward rates. We treat the interpolation problem as a replication problem, where a bond (or interest rate future/swap) is priced as a function of the minimum variance replicating portfolio of the traded bonds (or derivatives). In this view, the hedging ratios determined by the interpolation are as important (if not more) than getting the "right" interpolated rate; this is similar to the adjustments to the Black and Scholes delta as a consequence of the joint dynamics of the asset price and volatility in the different volatility models. We show how to learn the parameters of the weight functions and apply this method to the overnight rate indexed interest rates derivatives in Brazil. We then extend the concept from interpolating broken dates to the market references, in order to determine which points are key to the shape and dynamics of the curve and which points can be replicated by these real anchors.


1:30: Stéphane Crépey (Université Evry Val d’Essonne)

XVA analysis from the balance sheet

Abstract: Since the 2008-09 financial crisis, derivative dealers charge to their clients various add-ons, dubbed XVAs, meant to account for counterparty risk and its capital and funding implications. As banks cannot replicate jump-to-default related cash-flows, deals trigger wealth transfers from bank shareholders to bondholders and shareholders need to set capital at risk. On this basis, we devise a theory of XVAs, whereby the so-called contra-liabilities and cost of capital are sourced from bank clients at trade inceptions, on top of the fair valuation of counterparty risk, in order to compensate shareholders for wealth transfer and risk on their capital. The resulting all-inclusive XVA formula, meant incrementally at every new deal, reads (CVA+FVA+KVA), where C sits for credit, F for funding, and where the KVA is a cost of capital risk premium. All these XVAs are nonnegative and, even though we do crucially include the default of the bank itself in our modeling, unilateral in a certain sense. The corresponding XVA policy ensures to bank shareholders a submartingale wealth process corresponding to a target hurdle rate h on their capital at risk, consistently between and throughout deals.


Wednesday, May 22 th 2019, 6th Floor MSE


17h-18h: Maxence Soumare, Ekimetrics

Machine learning applied to Healtcare – How Ekimetrics Datascience expertise translates to the medical sector

Machine learning (ML) is a current hot topic in computer science research. This new promotion alongside an ever increasing ease to access and harness computing power has seen an explosion of ML applications to various fields. Although the medical domain is not Ekimetric’s main activity, the proposed use case demonstrates the capacity of the company to use data science to approach a wide range of questions and sectors. The presentation will introduce Ekimetric’s core activities and will then list some current Machine learning endeavors in Healthcare. We will then reintroduce some basics about neural networks, first for the biological ones followed by the same for their artificial counterparts. Lastly we will develop (methodology and technical choices) a use case that aims at answering a subject from Institut Gustave Roussy to detect renal cortex on radiologies.


18h-19h: Srinivas Raghavendra, department of economics, Galway University, Ireland

Dynamics of conflict between shareholders and managers: Revisiting the theory of firm under financialization

The period since 1980s, referred to as the “financialization” era, has witnessed profound changes in the way in which the financial markets and institutions interacted with the real economy. Although there is no one commonly agreed definition of the phenomenon “financialization”, the term is often used to refer the growing ascendency of ‘shareholder value’ as a mode of corporate governance. The shareholder power over manager is one of the channels through which financialization is argued to impact on the accumulation and growth decisions of firms. Here, we revisit the conventional theory of firm and introduce the shareholders-managers conflict to study the implications of financialization on the real economy in terms of some of the macroeconomic variables such as capacity utilization and growth.


Wednesday, March 20 th 2019, 6th Floor MSE


17h-18h: Eric Jondeau, HEC Lausanne

Long-Term Financial Returns: VAR vs. DSGE Model – A Horse-Race

This paper considers an institutional investor who is implementing a long-term portfolio allocation using forecasts of financial returns. We compare the predictive performance of two competing macro-finance models —an unrestricted Vector AutoRegression (VAR) and a fully structural Dynamic Stochastic General Equilibrium (DSGE) model— for investment horizons from 2 to 15 years. We show that, although the performances are similar for short horizons, the DSGE model outperforms the unrestricted VAR at forecasting financial returns in the long term. This model also generates substantially higher Sharpe ratios for mean-variance allocations. Even if it contains fewer unknown parameters, the DSGE model benefits from economically grounded restrictions that help anchor financial returns in the long term.


18h-19h: Florian Ielpo, Unigestion SA, Paris 1 Sorbonne University, Labex ReFi, and IPAG

Factor Timing Revisited: Alternative Risk Premia Allocation Based on Nowcasting and Valuation Signals

Alternative risk premia are encountering growing interest from investors. The vast majority of the academic literature has been focusing on describing the alternative risk premia (typically, momentum, carry and value strategies) individually. In this article, we investigate the question of allocation across a diversified range of cross-asset alternative risk premia over the period 1990-2018. For this, we design an active (macro risk-based) allocation framework that notably aims to exploit alternative risk premia’s varying behaviour in different macro regimes and their valuations over time. We perform backtests of the allocation strategy in an out-of-sample setting, shedding light on the significance of both sources of information.


Wednesday, February 20 th 2019, 6th Floor MSE

17h-18h: Othmane Mounjid, Ecole Polytechnique

Optimal liquidity-based trading tactics (with Charles-Albert Lehalle and Mathieu Rosenbaum)

We consider an agent who needs to buy (or sell) a relatively small amount of asset over some fixed short time interval. We work at the highest frequency meaning that we wish to find the optimal tactic to execute our quantity using limit orders, market orders and cancellations. To solve the agent’s control problem, we build an order book model and optimize an expected utility function based on our price impact. We derive the equations satisfied by the optimal strategy and solve them numerically. Moreover, we show that our optimal tactic enables us to outperform significantly naive execution strategies.

Wednesday, January 23 th 2019, 6th Floor MSE

17h-19h: Giulia Rotundo, Sapienza University of Rome

Complex networks modeling for financial data

The seminar aims to show some usage of complex networks analysis for modeling financial data. The main focus is going to be on three applications

- A copula approach to cross-ownership of companies,

- Herding in mutual funds,

- Assessing the impact of incomplete information on the resilience of financial networks.

In details:

- The cross-ownership of companies creates no trivial links among them. Within this respect, the diversification is intended to describe the holdings; and the integration represents the number of other companies that hold the shares. A high integration allows to spread fluctuations on the other companies, but it reduces the amount of profit to be kept in the company; a high diversification allows to spread the sources of risk, but at the same time it increases the probability to be exposed to fluctuations of other companies.

A copula approach for the detection of the most instable network topologies gives the results for the interaction among integration and diversification. A case study is used to outline the method.

- Herding in mutual funds

Some previous studies were emphasizing that correlations among stocks were loose during expansion periods of the market; and that they were stronger during recessions. Does it hold for mutual funds? We are going to build the network of mutual funds through the correlation network and to insert in the regression of the herding the centrality measures as explanatory variables. The results show that the behavior of mutual funds is just the opposite of the stocks.

- Assessing the impact of incomplete information on the resilience of financial networks

The paper faces the problem of the robustness of the network of interbank exposures against the propagation of failure cascades. The available data are retrieved through the BIS database and report only incomplete information, and show a core-periphery structure. A model of financial contagion is set up to estimate the width and length of the cascades, both in real and simulated data, that insert some percentages of the missing links. Simulations show that the network is far from the worst scenario for the propagation of contagions, and that the detection of the missing links is not trivial in the overall dynamic.