nep-cmp New Economics Papers
on Computational Economics
Issue of 2022‒11‒28
twenty-two papers chosen by
Stan Miles
Thompson Rivers University

  1. Modeling Machine Learning By Andrew Caplin; Daniel J. Martin; Philip Marx
  2. Recovering Missing Firm Characteristics with Attention-Based Machine Learning By Beckmeyer, Heiner; Wiedemann, Timo
  3. Deep neural network expressivity for optimal stopping problems By Lukas Gonon
  4. Predicting Politicians' Misconduct: Evidence from Colombia By Gallego, Jorge; Prem, Mounu; Vargas, Juan F.
  5. Asymptotic expansion and deep neural networks overcome the curse of dimensionality in the numerical approximation of Kolmogorov partial differential equations with nonlinear coefficients By Akihiko Takahashi; Toshihiro Yamada
  6. Exploration of the Parameter Space in Macroeconomic Models By Karl Naumann-Woleske; Max Sina Knicker; Michael Benzaquen; Jean-Philippe Bouchaud
  7. Gender, Sex, and the Constraints of Machine Learning Methods By Lockhart, Jeffrey W
  8. Reservoir Computing for Macroeconomic Forecasting with Mixed Frequency Data By Giovanni Ballarin; Petros Dellaportas; Lyudmila Grigoryeva; Marcel Hirt; Sophie van Huellen; Juan-Pablo Ortega
  9. Mapping the Knowledge Space: Exploiting Unassisted Machine Learning Tools By Florenta Teodoridis; Jino Lu; Jeffrey L. Furman
  10. Incorporating Interactive Facts for Stock Selection via Neural Recursive ODEs By Qiang Gao; Xinzhu Zhou; Kunpeng Zhang; Li Huang; Siyuan Liu; Fan Zhou
  11. State-dependent asset allocation using neural networks By Bradrania, Reza; Pirayesh Neghab, Davood
  12. Predicting Politicians Misconduct: Evidence From Colombia By Gallego, J; Prem, M; Vargas, J. F.
  13. Rating Triggers for Collateral-Inclusive XVA via Machine Learning and SDEs on Lie Groups By Kevin Kamm; Michelle Muniz
  14. A parametric approach to the estimation of convex risk functionals based on Wasserstein distance By Max Nendel; Alessandro Sgarabottolo
  15. Measuring the environmental impacts of artificial intelligence compute and applications: The AI footprint By OECD
  16. Supply Chain Characteristics as Predictors of Cyber Risk: A Machine-Learning Assessment By Kevin Hu; Retsef Levi; Raphael Yahalom; El Ghali Zerhouni
  17. Newton Raphson Emulation Network for Highly Efficient Computation of Numerous Implied Volatilities By Geon Lee; Tae-Kyoung Kim; Hyun-Gyoon Kim; Jeonggyu Huh
  18. The Anatomy of Out-of-Sample Forecasting Accuracy By Daniel Borup; Philippe Goulet Coulombe; Erik Christian Montes Schütte; David E. Rapach; Sander Schwenk-Nebbe
  19. The Proof is in the Pudding. Revealing the SDGs with Artificial Intelligence By Jean-Baptiste JACOUTON; Régis MARODON; Adeline LAULANIE
  20. Twitter and Crime: The Effect of Social Movements on GenderBased Violence By Michele Battisti; Ilpo Kauppinen; Britta Rude
  21. Costs and benefits of an Individual Learning Account (ILA): A simulation analysis for the Netherlands By Henri Bussink; Bas ter Weel
  22. Contagious economic failure? Discourses around “zombie firms” in Covid-19 ridden Germany and Italy By Hilmar, Till; Paolillo, Rocco; Sachweh, Patrick

  1. By: Andrew Caplin; Daniel J. Martin; Philip Marx
    Abstract: What do machines learn, and why? To answer these questions we import models of human cognition into machine learning. We propose two ways of modeling machine learners based on this join: feasibility-based and cost-based machine learning. We evaluate and estimate our models using a deep learning convolutional neural network that predicts pneumonia from chest X-rays. We find these predictions are consistent with our model of cost-based machine learning, and we recover the algorithm's implied costs of learning.
    JEL: C0 D80
    Date: 2022–10
  2. By: Beckmeyer, Heiner; Wiedemann, Timo
    JEL: G10
    Date: 2022
  3. By: Lukas Gonon
    Abstract: This article studies deep neural network expression rates for optimal stopping problems of discrete-time Markov processes on high-dimensional state spaces. A general framework is established in which the value function and continuation value of an optimal stopping problem can be approximated with error at most $\varepsilon$ by a deep ReLU neural network of size at most $\kappa d^{\mathfrak{q}} \varepsilon^{-\mathfrak{r}}$. The constants $\kappa,\mathfrak{q},\mathfrak{r} \geq 0$ do not depend on the dimension $d$ of the state space or the approximation accuracy $\varepsilon$. This proves that deep neural networks do not suffer from the curse of dimensionality when employed to solve optimal stopping problems. The framework covers, for example, exponential L\'evy models, discrete diffusion processes and their running minima and maxima. These results mathematically justify the use of deep neural networks for numerically solving optimal stopping problems and pricing American options in high dimensions.
    Date: 2022–10
  4. By: Gallego, Jorge; Prem, Mounu; Vargas, Juan F.
    Abstract: Corruption has pervasive effects on economic development and the well-being of the population. Despite being crucial and necessary, fighting corruption is not an easy task because it is a difficult phenomenon to measure and detect. However, recent advances in the field of artificial intelligence may help in this quest. In this article, we propose the use of machine learning models to predict municipality-level corruption in a developing country. Using data from disciplinary prosecutions conducted by an anti-corruption agency in Colombia, we trained four canonical models (Random Forests, Gradient Boosting Machine, Lasso, and Neural Networks), and ensemble their predictions, to predict whether or not a mayor will commit acts of corruption. Our models achieve acceptable levels of performance, based on metrics such as the precision and the area under the ROC curve, demonstrating that these tools are useful in predicting where misbehavior is most likely to occur. Moreover, our feature-importance analysis shows us which groups of variables are most important upon predicting corruption.
    Date: 2022–10–18
  5. By: Akihiko Takahashi (University of Tokyo); Toshihiro Yamada (Hitotsubashi University, Japan Science and Technology Agency (JST))
    Abstract: This paper proposes a new spatial approximation method without the curse of dimensionality for solving high-dimensional partial differential equations (PDEs) by using an asymptotic expansion method with a deep learning-based algorithm. In particular, the mathematical justification on the spatial approximation is provided, and a numerical example for a 100 dimensional Kolmogorov PDE shows effectiveness of our method.
    Date: 2022–11
  6. By: Karl Naumann-Woleske (LadHyX - Laboratoire d'hydrodynamique - X - École polytechnique - CNRS - Centre National de la Recherche Scientifique); Max Sina Knicker (TUM - Technische Universität München = Technical University of Munich); Michael Benzaquen (LadHyX - Laboratoire d'hydrodynamique - X - École polytechnique - CNRS - Centre National de la Recherche Scientifique); Jean-Philippe Bouchaud (Académie des Sciences - Institut de France)
    Abstract: Agent-Based Models (ABM) are computational scenario-generators, which can be used to predict the possible future outcomes of the complex system they represent. To better understand the robustness of these predictions, it is necessary to understand the full scope of the possible phenomena the model can generate. Most often, due to high-dimensional parameter spaces, this is a computationally expensive task. Inspired by ideas coming from systems biology, we show that for multiple macroeconomic models, including an agent-based model and several Dynamic Stochastic General Equilibrium (DSGE) models, there are only a few stiff parameter combinations that have strong effects, while the other sloppy directions are irrelevant. This suggests an algorithm that efficiently explores the space of parameters by primarily moving along the stiff directions. We apply our algorithm to a medium-sized agent-based model, and show that it recovers all possible dynamics of the unemployment rate. The application of this method to Agent-based Models may lead to a more thorough and robust understanding of their features, and provide enhanced parameter sensitivity analyses. Several promising paths for future research are discussed.
    Date: 2022
  7. By: Lockhart, Jeffrey W (University of Chicago)
    Abstract: Machine learning interacts with gender and sex in myriad ways, intentionally, unintentionally, and sometimes even against practitioner's concerted efforts. Some of these interactions are born out of the allure of a seemingly simple, unambiguous, binary, variable ideally aligned with the technical needs and sensibilities of ML. Most of the time, gender lurks in ML systems without any explicit invitation, simply because these systems mine data for associations, and gendered associations are ubiquitous. And in a growing body of work, scholars are using ML to actively interrogate gender and sexuality, in turn shaping what they mean and how we think about them. Machine learning brings with it new paradigms of quantitative reasoning which hold the potential to either reinscribe or revolutionize gender in not only technical systems, but scientific knowledge as well. Throughout, the key is for people in and around machine learning to pay close attention to what the technology is actually doing with gender and sex.
    Date: 2022–11–03
  8. By: Giovanni Ballarin; Petros Dellaportas; Lyudmila Grigoryeva; Marcel Hirt; Sophie van Huellen; Juan-Pablo Ortega
    Abstract: Macroeconomic forecasting has recently started embracing techniques that can deal with large-scale datasets and series with unequal release periods. The aim is to exploit the information contained in heterogeneous data sampled at different frequencies to improve forecasting exercises. Currently, MIxed-DAta Sampling (MIDAS) and Dynamic Factor Models (DFM) are the two main state-of-the-art approaches that allow modeling series with non-homogeneous frequencies. We introduce a new framework called the Multi-Frequency Echo State Network (MFESN), which originates from a relatively novel machine learning paradigm called reservoir computing (RC). Echo State Networks are recurrent neural networks with random weights and trainable readout. They are formulated as nonlinear state-space systems with random state coefficients where only the observation map is subject to estimation. This feature makes the estimation of MFESNs considerably more efficient than DFMs. In addition, the MFESN modeling framework allows to incorporate many series, as opposed to MIDAS models, which are prone to the curse of dimensionality. Our discussion encompasses hyperparameter tuning, penalization, and nonlinear multistep forecast computation. In passing, a new DFM aggregation scheme with Almon exponential structure is also presented, bridging MIDAS and dynamic factor models. All methods are compared in extensive multistep forecasting exercises targeting US GDP growth. We find that our ESN models achieve comparable or better performance than MIDAS and DFMs at a much lower computational cost.
    Date: 2022–11
  9. By: Florenta Teodoridis; Jino Lu; Jeffrey L. Furman
    Abstract: Understanding factors affecting the direction of innovation is a central aim of research in the economics of innovation. Progress on this topic has been inhibited by difficulties in measuring distance and movement in knowledge space. We describe a methodology that infers the mapping of the knowledge landscape based on text documents. The approach is based on an unassisted machine learning technique, Hierarchical Dirichlet Process (HDP), which flexibly identifies patterns in text corpora. The resulting mapping of the knowledge landscape enables calculations of distance and movement, measures that are valuable in several contexts for research in innovation. We benchmark and demonstrate the benefits of this approach in the context of 44 years of USPTO data.
    JEL: C55 C80 O3 O31 O32
    Date: 2022–10
  10. By: Qiang Gao; Xinzhu Zhou; Kunpeng Zhang; Li Huang; Siyuan Liu; Fan Zhou
    Abstract: Stock selection attempts to rank a list of stocks for optimizing investment decision making, aiming at minimizing investment risks while maximizing profit returns. Recently, researchers have developed various (recurrent) neural network-based methods to tackle this problem. Without exceptions, they primarily leverage historical market volatility to enhance the selection performance. However, these approaches greatly rely on discrete sampled market observations, which either fail to consider the uncertainty of stock fluctuations or predict continuous stock dynamics in the future. Besides, some studies have considered the explicit stock interdependence derived from multiple domains (e.g., industry and shareholder). Nevertheless, the implicit cross-dependencies among different domains are under-explored. To address such limitations, we present a novel stock selection solution -- StockODE, a latent variable model with Gaussian prior. Specifically, we devise a Movement Trend Correlation module to expose the time-varying relationships regarding stock movements. We design Neural Recursive Ordinary Differential Equation Networks (NRODEs) to capture the temporal evolution of stock volatility in a continuous dynamic manner. Moreover, we build a hierarchical hypergraph to incorporate the domain-aware dependencies among the stocks. Experiments conducted on two real-world stock market datasets demonstrate that StockODE significantly outperforms several baselines, such as up to 18.57% average improvement regarding Sharpe Ratio.
    Date: 2022–10
  11. By: Bradrania, Reza; Pirayesh Neghab, Davood
    Abstract: Changes in market conditions present challenges for investors as they cause performance to deviate from the ranges predicted by long-term averages of means and covariances. The aim of conditional asset allocation strategies is to overcome this issue by adjusting portfolio allocations to hedge changes in the investment opportunity set. This paper proposes a new approach to conditional asset allocation that is based on machine learning; it analyzes historical market states and asset returns and identifies the optimal portfolio choice in a new period when new observations become available. In this approach, we directly relate state variables to portfolio weights, rather than firstly modeling the return distribution and subsequently estimating the portfolio choice. The method captures nonlinearity among the state (predicting) variables and portfolio weights without assuming any particular distribution of returns and other data, without fitting a model with a fixed number of predicting variables to data and without estimating any parameters. The empirical results for a portfolio of stock and bond indices show the proposed approach generates a more efficient outcome compared to traditional methods and is robust in using different objective functions across different sample periods.
    Keywords: asset allocation; portfolio optimization; market state, machine learning; neural networks; performance ratio
    JEL: C1 C10 C15 C18 C53 C55 C58 G0 G1 G11 G12 G17
    Date: 2021–02–01
  12. By: Gallego, J; Prem, M; Vargas, J. F.
    Keywords: Prediction, Corruption, Machine Learning, Colombia
    Date: 2022–10–18
  13. By: Kevin Kamm; Michelle Muniz
    Abstract: In this paper, we model the rating process of an entity by using a geometrical approach. We model rating transitions as an SDE on a Lie group. Specifically, we focus on calibrating the model to both historical data (rating transition matrices) and market data (CDS quotes) and compare the most popular choices of changes of measure to switch from the historical probability to the risk-neutral one. For this, we show how the classical Girsanov theorem can be applied in the Lie group setting. Moreover, we overcome some of the imperfections of rating matrices published by rating agencies, which are computed with the cohort method, by using a novel Deep Learning approach. This leads to an improvement of the entire scheme and makes the model more robust for applications. We apply our model to compute bilateral credit and debit valuation adjustments of a netting set under a CSA with thresholds depending on ratings of the two parties.
    Date: 2022–11
  14. By: Max Nendel; Alessandro Sgarabottolo
    Abstract: In this paper, we explore a static setting for the assessment of risk in the context of mathematical finance and actuarial science that takes into account model uncertainty in the distribution of a possibly infinite-dimensional risk factor. We allow for perturbations around a baseline model, measured via Wasserstein distance, and we investigate to which extent this form of probabilistic imprecision can be parametrized. The aim is to come up with a convex risk functional that incorporates a sefety margin with respect to nonparametric uncertainty and still can be approximated through parametrized models. The particular form of the parametrization allows us to develop a numerical method, based on neural networks, which gives both the value of the risk functional and the optimal perturbation of the reference measure. Moreover, we study the problem under additional constraints on the perturbations, namely, a mean and a martingale constraint. We show that, in both cases, under suitable conditions on the loss function, it is still possible to estimate the risk functional by passing to a parametric family of perturbed models, which again allows for a numerical approximation via neural networks.
    Date: 2022–10
  15. By: OECD
    Abstract: Artificial intelligence (AI) systems can use massive computational resources, raising sustainability concerns. This report aims to improve understanding of the environmental impacts of AI, and help measure and decrease AI’s negative effects while enabling it to accelerate action for the good of the planet. It distinguishes between the direct environmental impacts of developing, using and disposing of AI systems and related equipment, and the indirect costs and benefits of using AI applications. It recommends the establishment of measurement standards, expanding data collection, identifying AI-specific impacts, looking beyond operational energy use and emissions, and improving transparency and equity to help policy makers make AI part of the solution to sustainability challenges.
    Date: 2022–11–15
  16. By: Kevin Hu (Massachusetts Institute of Technology); Retsef Levi (Massachusetts Institute of Technology); Raphael Yahalom (Massachusetts Institute of Technology); El Ghali Zerhouni (Massachusetts Institute of Technology)
    Abstract: This paper provides the first large-scale data-driven analysis to evaluate the predictive power of different attributes for assessing risk of cyberattack data breaches. Furthermore, motivated by rapid increase in third party enabled cyberattacks, the paper provides the first quantitative empirical evidence that digital supply-chain attributes are significant predictors of enterprise cyber risk. The paper leverages outside-in cyber risk scores that aim to capture the quality of the enterprise internal cybersecurity management, but augment these with supply chain features that are inspired by observed third party cyberattack scenarios, as well as concepts from network science research. The main quantitative result of the paper is to show that supply chain network features add significant detection power to predicting enterprise cyber risk, relative to merely using enterprise-only attributes. Particularly, compared to a base model that relies only on internal enterprise features, the supply chain network features improve the out-of-sample AUC by 2.3\%. Given that each cyber data breach is a low probability high impact risk event, these improvements in the prediction power have significant value. Additionally, the model highlights several cybersecurity risk drivers related to third party cyberattack and breach mechanisms and provides important insights as to what interventions might be effective to mitigate these risks.
    Date: 2022–10
  17. By: Geon Lee; Tae-Kyoung Kim; Hyun-Gyoon Kim; Jeonggyu Huh
    Abstract: In finance, implied volatility is an important indicator that reflects the market situation immediately. Many practitioners estimate volatility using iteration methods, such as the Newton--Raphson (NR) method. However, if numerous implied volatilities must be computed frequently, the iteration methods easily reach the processing speed limit. Therefore, we emulate the NR method as a network using PyTorch, a well-known deep learning package, and optimize the network further using TensorRT, a package for optimizing deep learning models. Comparing the optimized emulation method with the NR function in SciPy, a popular implementation of the NR method, we demonstrate that the emulation network is up to 1,000 times faster than the benchmark function.
    Date: 2022–10
  18. By: Daniel Borup; Philippe Goulet Coulombe; Erik Christian Montes Schütte; David E. Rapach; Sander Schwenk-Nebbe
    Abstract: We develop metrics based on Shapley values for interpreting time-series forecasting models, including “black-box” models from machine learning. Our metrics are model agnostic, so that they are applicable to any model (linear or nonlinear, parametric or nonparametric). Two of the metrics, iShapley-VI and oShapley-VI, measure the importance of individual predictors in fitted models for explaining the in-sample and out-of-sample predicted target values, respectively. The third metric is the performance-based Shapley value (PBSV), our main methodological contribution. PBSV measures the contributions of individual predictors in fitted models to the out-of-sample loss and thereby anatomizes out-of-sample forecasting accuracy. In an empirical application forecasting US inflation, we find important discrepancies between individual predictor relevance according to the in-sample iShapley-VI and out-of-sample PBSV. We use simulations to analyze potential sources of the discrepancies, including overfitting, structural breaks, and evolving predictor volatilities.
    Keywords: variable importance; out-of-sample performance; Shapley value; loss function; machine learning; inflation
    JEL: C22 C45 C53 E37 G17
    Date: 2022–11–07
  19. By: Jean-Baptiste JACOUTON; Régis MARODON; Adeline LAULANIE
    Abstract: The use of frontier technologies in the field of sustainability is likely to accompany its visibility, and the quality of information available to decision makers. This paper explores the possibility of using artificial intelligence to analyze Public Development Banks’ annual reports.
    JEL: Q
    Date: 2022–10–05
  20. By: Michele Battisti; Ilpo Kauppinen; Britta Rude
    Abstract: This paper asks whether social movements taking place on Twitter affect genderbased violence (GBV). Using Twitter data and machine learning methods, we construct a novel data set on the prevalence of Twitter conversations about GBV. We then link this data to weekly crime reports at the federal state level from the United States. We exploit the high-frequency nature of our data and an event study design to establish a causal impact of Twitter social movements on GBV. Our results point out that Twitter tweets related to GBV lead to a decrease in reported crime rates. The evidence shows that perpetrators commit these crimes less due to increased social pressure and perceived social costs. The results indicate that social media could significantly decrease reported GBV and might facilitate the signaling of social norms.
    Keywords: Economics of gender, US, domestic abuse, public policy, criminal law, illegal behavior and the enforcement of law
    JEL: J12 J16 J78 K14 K42 O51
    Date: 2022
  21. By: Henri Bussink (SEO Amsterdam Economics); Bas ter Weel (University of Amsterdam)
    Abstract: This study analyses costs and benefits of a public-private funded individual learning account (ILA) for the labour force in the Netherlands. We consider an ILA that is funded by subsidies targeted at low- and medium-educated workers and co-funded by training levies as a share of the wage bill. We simulate two alternative steady-state scenarios about the uptake of resources and increase in training activity, using a lifecycle model of human capital investments. We derive predictions for gross earnings, income inequality and costs (training subsidies and tax deductions) and benefits (tax revenues and fewer unemployment benefits). Our results show how the balance of costs and benefits depends on the interplay between take-up rates, returns to training and the deadweight loss of subsidizing an ILA for the whole labour force. Our model and results contribute to policy trade-offs about the introduction of ILA’s to stimulate the resilience of the labour fo
    Keywords: Human capital investments, Individual learning accounts, Lifelong learning
    JEL: J24 J33
    Date: 2022–11–13
  22. By: Hilmar, Till; Paolillo, Rocco (Jacobs University); Sachweh, Patrick
    Abstract: As the spread of Covid-19 hit societies around the world, governments stepped up to contain the pandemic’s economic shock. Governments saved businesses and stabilized employment through extensive fiscal relief packages. In this paper, we analyze public debates around zombie firms – businesses that are unprofitable and/or unable to pay interests on their debt but still receive public aid. In popular culture, zombies are contagious creatures that threaten the “healthy” order of society; in economic discourse, we suggest, the zombie trope is no less rich in cultural meaning and metaphor. Specifically, we are interested in how narrative meaning around zombie firms can be employed to describe the role of the state and the role of the market in addressing economic crises and economic transformations. We comparatively explore how zombie firms are discussed in two societies, Germany and Italy, during the first two years of the Covid-19 pandemic. Combining computational and qualitative text analysis, we investigate how newspapers on the left and on the right imbue this term with meaning. Our results show that German and Italian debates are more similar than expected. Right-leaning newspapers in both countries depict zombie firms as a problem of debt, which is seen as caused and aggravated by undesired state intervention in market dynamics. Left-leaning newspapers articulate more nuanced positions: while German left-leaning newspapers tend to reject the trope of zombie firms and defend pandemic state interventions such as short-time work programs, their Italian counterparts also argue for the need for such policies but associate zombie firms with doubts about the efficiency of state support in the long term.
    Date: 2022–11–03

This nep-cmp issue is ©2022 by Stan Miles. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.