We propose a numerical recipe for risk evaluation defined by a backward stochastic differential equation. Using dual representation of the risk measure, we convert the risk valuation to a stochastic control problem where the control is a certain Radon-Nikodym derivative process. By exploring the maximum principle, we show that a piecewise-constant dual control provides a good approximation on a short interval. A dynamic programming algorithm extends the approximation to a finite time horizon. Finally, we illustrate the application of the procedure to financial risk management in conjunction with nested simulation and on an multidimensional portfolio valuation problem.

Recent advances in computing power and the potential to make more realistic assumptions due to increased flexibility have led to the increased prevalence of simulation models in economics. While models of this class, and particularly agent-based models, are able to replicate a number of empirically-observed stylised facts not easily recovered by more traditional alternatives, such models remain notoriously difficult to estimate due to their lack of tractable likelihood functions. While the estimation literature continues to grow, existing attempts have approached the problem primarily from a frequentist perspective, with the Bayesian estimation literature remaining comparatively less developed. For this reason, we introduce a Bayesian estimation protocol that makes use of deep neural networks to construct an approximation to the likelihood, which we then benchmark against a prominent alternative from the existing literature. Overall, we find that our proposed methodology consistently results in more accurate estimates in a variety of settings, including the estimation of financial heterogeneous agent models and the identification of changes in dynamics occurring in models incorporating structural breaks.

We develop a large-scale deep learning model to predict price movements from limit order book (LOB) data of cash equities. The architecture utilises convolutional filters to capture the spatial structure of the limit order books as well as LSTM modules to capture longer time dependencies. The proposed network outperforms all existing state-of-the-art algorithms on the benchmark LOB dataset [1]. In a more realistic setting, we test our model by using one year market quotes from the London Stock Exchange and the model delivers a remarkably stable out-of-sample prediction accuracy for a variety of instruments. Importantly, our model translates well to instruments which were not part of the training set, indicating the model's ability to extract universal features. In order to better understand these features and to go beyond a "black box" model, we perform a sensitivity analysis to understand the rationale behind the model predictions and reveal the components of LOBs that are most relevant. The ability to extract robust features which translate well to other instruments is an important property of our model which has many other applications.

Ergodicity describes an equivalence between the expectation value and the time average of observables. Applied to human behaviour, ergodic theory reveals how individuals should tolerate risk in different environments. To optimise wealth over time, agents should adapt their utility function according to the dynamical setting they face. Linear utility is optimal for additive dynamics, whereas logarithmic utility is optimal for multiplicative dynamics. Whether humans approximate time optimal behavior across different dynamics is unknown. Here we compare the effects of additive versus multiplicative gamble dynamics on risky choice. We show that utility functions are modulated by gamble dynamics in ways not explained by prevailing economic theory. Instead, as predicted by time optimality, risk aversion increases under multiplicative dynamics, distributing close to the values that maximise the time average growth of wealth. We suggest that our findings motivate a need for explicitly grounding theories of decision-making on ergodic considerations.

We showcase how Quantile Regression (QR) can be applied to forecast financial returns using Limit Order Books (LOBs), the canonical data source of high-frequency financial time-series. We develop a deep learning architecture that simultaneously models the return quantiles for both buy and sell positions. We test our model over millions of LOB updates across multiple different instruments on the London Stock Exchange. Our results suggest that the proposed network not only delivers excellent performance but also provides improved prediction robustness by combining quantile estimates.

In this study, we develop a deterministic nonlinear filtering algorithm based on a high-dimensional version of Kitagawa (1987) to evaluate the likelihood function of models that allow for stochastic volatility and jumps whose arrival intensity is also stochastic. We show numerically that the deterministic filtering method is precise and much faster than the particle filter, in addition to yielding a smooth function over the parameter space. We then find the maximum likelihood estimates of various models that include stochastic volatility, jumps in the returns and variance, and also stochastic jump arrival intensity with the S&P 500 daily returns. During the Great Recession, the jump arrival intensity increases significantly and contributes to the clustering of volatility and negative returns.

In a market with a rough or Markovian mean-reverting stochastic volatility there is no perfect hedge. Here it is shown how various delta-type hedging strategies perform and can be evaluated in such markets in the case of European options. A precise characterization of the hedging cost, the replication cost caused by the volatility fluctuations, is presented in an asymptotic regime of rapid mean reversion for the volatility fluctuations. The optimal dynamic asset based hedging strategy in the considered regime is identified as the so-called `practitioners' delta hedging scheme. It is moreover shown that the performances of the delta-type hedging schemes are essentially independent of the regularity of the volatility paths in the considered regime and that the hedging costs are related to a vega risk martingale whose magnitude is proportional to a new market risk parameter. It is also shown via numerical simulations that the proposed hedging schemes which derive from option price approximations in the regime of rapid mean reversion, are robust: the `practitioners' delta hedging scheme that is identified as being optimal by our asymptotic analysis when the mean reversion time is small seems to be optimal with arbitrary mean reversion times.

In this paper I re-examine the COMPAS recidivism score and criminal history data collected by ProPublica in 2016, which has fueled intense debate and research in the nascent field of `algorithmic fairness' or `fair machine learning' over the past three years. ProPublica's COMPAS data is used in an ever-increasing number of studies to test various definitions and methodologies of algorithmic fairness. This paper takes a closer look at the actual datasets put together by ProPublica. By doing so, I find that ProPublica made an important data processing mistake when it created some of the key datasets most often used by other researchers. In particular, the datasets built to study the likelihood of recidivism within two years of the original COMPAS screening date. As I show in this paper, ProPublica made a mistake implementing the two-year sample cutoff rule for recidivists in such datasets (whereas it implemented an appropriate two-year sample cutoff rule for non-recidivists). As a result, ProPublica incorrectly kept a disproportionate share of recidivists. This data processing mistake leads to biased two-year recidivism datasets, with artificially high recidivism rates. This also affects the positive and negative predictive values. On the other hand, this data processing mistake does not impact some of the key statistical measures highlighted by ProPublica and other researchers, such as the false positive and false negative rates, nor the overall accuracy.

Monte Carlo methods are core to many routines in quantitative finance such as derivatives pricing, hedging and risk metrics. Unfortunately, Monte Carlo methods are very computationally expensive when it comes to running simulations in high-dimensional state spaces where they are still a method of choice in the financial industry. Recently, Tensor Processing Units (TPUs) have provided considerable speedups and decreased the cost of running Stochastic Gradient Descent (SGD) in Deep Learning. After having highlighted computational similarities between training neural networks with SGD and stochastic process simulation, we ask in the present paper whether TPUs are accurate, fast and simple enough to use for financial Monte Carlo. Through a theoretical reminder of the key properties of such methods and thorough empirical experiments we examine the fitness of TPUs for option pricing, hedging and risk metrics computation. We show in the following that Tensor Processing Units (TPUs) in the cloud help accelerate Monte Carlo routines compared to Graphics Processing Units (GPUs) which in turn decreases the cost associated with running such simulations while leveraging the flexibility of the cloud. In particular we demonstrate that, in spite of the use of mixed precision, TPUs still provide accurate estimators which are fast to compute. We also show that the Tensorflow programming model for TPUs is elegant, expressive and simplifies automated differentiation.