What we're about
Upcoming events (4)
Abstract: Using a randomized experiment in the auto lending industry, we provide causal evidence of higher loan profitability with algorithmic machine underwriting, relative to human underwriting. Machine-underwritten loans generate 10.2% higher loan-level profit than human-underwritten loans in a sample of 140,000 randomly assigned applications. Comparing loans to otherwise identical borrowers, the loans underwritten by machines not only have higher interest rates, but also realize a 6.8% lower incidence of default, relative to loans underwritten by humans. The performance gap is more pronounced with more complex loans and at discrete cutoffs. These results are consistent with findings on the human's limited capacity for analyzing complex problems and with agency conflicts in the underwriting process. Bio: Mark Jansen is an Assistant Professor of Finance. His primary research and teaching interests are in entrepreneurial finance, household finance, and corporate finance. Prior to joining the University of Utah, Dr. Jansen worked in private equity as managing director at Holland Park Capital and was responsible for strategy and investor relations. In this capacity, Dr. Jansen was a member of the Young President’s Organization. Prior to this Dr. Jansen worked in management consulting and in the chemical Industry. He received a Ph.D. in finance from the University of Texas at Austin, an M.B.A. from London Business School and a dual B.S. in management science and mechanical engineering from t
PLEASE NOTE THAT THIS WEBINAR WILL START ON WEDNESDAY, 2 DECEMBER, 2020, AT 6:30 PM ***GMT (LONDON TIME)*** (1.30 PM ***EST (NEW YORK TIME)***) FULL TITLE: Hedging with linear regressions and neural networks ABSTRACT We study the use of neural networks as nonparametric estimation tools for the hedging of options. To this end, we design a network, named HedgeNet, that directly outputs a hedging strategy given relevant features as input. This network is trained to minimise the hedging error instead of the pricing error. Applied to end-of-day and tick prices of S&P 500 and Euro Stoxx 50 options, the network is able to reduce the mean squared hedging error of the Black-Scholes benchmark significantly. We illustrate, however, that a similar benefit arises by a simple linear regression model that incorporates the leverage effect. Finally, we argue that outperformance of neural networks previously reported in the literature is most likely due to a lack of data hygiene. In particular, data leakage is sometimes unnecessarily introduced by a faulty training/test data split, possibly along with an additional ‘tagging’ of data. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3580132 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3486363 (Joint work with Weiguan Wang) BIOGRAPHY Johannes Ruf is a full Professor at the London School of Economics (LSE) and a leading academic in mathematical finance. Prior to LSE, he was a Senior Research Fellow at the Oxford-Man Institute of Quantitative Finance and a Senior Lecturer at the University College London (UCL). Johannes was awarded his Ph.D. in Statistics at Columbia University in New York. Johannes’ research interests include machine learning and portfolio theory. His work received several industry prizes including the ‘Morgan Stanley Prize for Excellence in Financial Markets’ and a Savvy Investor recognition for the ‘Best Factor Investing Papers of 2018.’ Johannes’ research was covered by Risk Magazine. He was a Fulbright scholar and won several teaching prizes at Columbia University and LSE. He coauthored numerous published research articles with practitioners and academics from different fields including Finance, Economics, and Operations Research. Johannes is also an associated member at the UCL Centre for Blockchain Technologies and an associate editor of Applied Mathematical Finance and Stochastic Models. He served on the Expert Council for the ‘Pilot Project on Environmental Stress Testing - Testing Corporate Loan Portfolios for Drought Scenario,’ launched by the United Nations Environmental Programme. Johannes also served as the director of the MSc programme in Financial Mathematics at LSE.
PLEASE NOTE THAT THIS WEBINAR WILL START ON WEDNESDAY, 9 DECEMBER, 2020, AT 6:30 PM ***GMT (LONDON TIME)*** (1.30 PM ***EST (NEW YORK TIME)***) FULL TITLE: Impacting the Bottom Line with Bayesian Decision Making and PyMC3 ABSTRACT There are often high expectations on the business impact that data science can produce. Unfortunately, various barriers often hinder the full realization of that impact, starting with communication challenges between different parts of the organization with different backgrounds. To bridge this chasm, data scientists and stake-holders should agree on a business relevant loss function to optimize. This way, any improvements in the model can directly be measured in their impact on the bottom line. In this talk, I will show how probabilistic programming frameworks like PyMC3 can be used to solve an applied problems with an example from capital allocation. This approach allows us to accurately and flexibly map a real-world problem to a statistical model that can be quickly iterated and improved on. I will then show how the results of such a model, which are usually arcane and non-actionable posterior probability distributions, can be coupled with a loss function based on business mechanics, to (i) derive business related outcome measures, and (ii) suggest the optimal decision to make, rather than inform it. BIOGRAPHY Thomas Wiecki is the Chief Executive Officer at PyMC Labs. Prior to that Thomas was the lead data science researcher at Quantopian, where he used probabilistic programming and machine learning to help build the world’s first crowdsourced hedge fund. Among other open source projects, he is involved in the development of PyMC—a probabilistic programming framework written in Python. A recognized international speaker, Thomas has given talks at various conferences and meetups across the US, Europe, and Asia. He holds a PhD from Brown University.
A Machine Learning Approach to Analyze and Support Anti-Corruption Policy. A webinar by Elliott Ash with Sergio Galletta & Tommaso Giommoni Abstract: Can machine learning support better governance? In the context of Brazilian municipalities,[masked], we have access to detailed accounts of local budgets and audit data on the associated fiscal corruption. Using the budget variables as predictors, we train a tree-based gradient-boosted classifier to predict the presence of corruption in held-out test data. The trained model, when applied to new data, provides a prediction-based measure of corruption which can be used for new empirical analysis or to support policy responses. We validate the empirical usefulness of this measure by replicating, and extending, some previous empirical evidence on corruption issues in Brazil. We then explore how the predictions can be used to support policies toward corruption. Our policy simulations show that, relative to the status quo policy of random audits, a targeted policy guided by the machine predictions could detect more than twice as many corrupt municipalities for the same audit rate. Bio: Elliott Ash is Assistant Professor of Law, Economics, and Data Science at ETH Zurich's Center for Law & Economics, Switzerland. Elliott's research and teaching focus on empirical analysis of the law and legal system using techniques from applied micro-econometrics, natural language processing, and machine learning. Prior to joining ETH, Elliott was Assistant Professor of Economics at University of Warwick, and before that a Postdoctoral Research Associate at Princeton University’s Center for the study of Democratic Politics. He received a Ph.D. in economics and J.D. from Columbia University, a B.A. in economics, government, and philosophy from University of Texas at Austin, and an LL.M. in international criminal law from University of Amsterdam.