|
on Computational Economics |
|
Issue of 2022‒01‒31
fifteen papers chosen by |
| By: | Garg, Karan (University of Warwick) |
| Abstract: | The rise of machine learning has revolutionised finance. Institutions across the world have increasingly turned to data science and machine learning to create trading models without the need for human intervention. This has had various implications for the financial markets that they operate in, including market efficiency. This paper simulates a financial market with agent-based modelling and Monte-Carlo style simulations, to motivate a qualitative discussion about the implications of increased algorithmic trading on financial market efficiency. It finds that algorithmic traders (ATs) can seemingly increase market efficiency through better liquidity management and more complete extraction of information from prices. However, this also comes with increased instability and potential convergence to an unstable equilibrium. The Adaptive Market Hypothesis (Lo, 2004) is suggested as an alternative framework for analysing AT behaviour. |
| Keywords: | Neural Networks ; Agent-Based Modelling ; Efficient Market Hypothesis ; Stock Market Simulation ; Financial Regulation JEL Classification: C45 ; C53 ; G14 ; G17 ; G18 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:wrk:wrkesp:11&r= |
| By: | David Anderson (University of Zurich); Urban Ulrych (University of Zurich - Department of Banking and Finance; Swiss Finance Institute) |
| Abstract: | Given the competitiveness of a market-making environment, the ability to speedily quote option prices consistent with an ever-changing market environment is essential. Thus, the smallest acceleration or improvement over traditional pricing methods is crucial to avoid arbitrage. We propose a novel method for accelerating the pricing of American options to near-instantaneous using a feed-forward neural network. This neural network is trained over the chosen (e.g., Heston) stochastic volatility specification. Such an approach facilitates parameter interpretability, as generally required by the regulators, and establishes our method in the area of eXplainable Artificial Intelligence (XAI) for finance. We show that the proposed deep explainable pricer induces a speed accuracy trade-off compared to the typical Monte Carlo or Partial Differential Equation-based pricing methods. Moreover, the proposed approach allows for pricing derivatives with path dependent and more complex payoffs and is, given the sufficient accuracy of computation and its tractable nature, applicable in a market-making environment. |
| Keywords: | American Option Pricing, Deep Neural Networks, Explainable Artificial Intelligence, Speed-Accuracy Trade-Off, Market Making, Heston Model, Computational Finance. |
| JEL: | C45 C63 G13 |
| Date: | 2022–01 |
| URL: | http://d.repec.org/n?u=RePEc:chf:rpseri:rp2203&r= |
| By: | van der Hagen, L.; Agatz, N.A.H.; Spliet, R.; Visser, T.R.; Kok, A.L. |
| Abstract: | Online grocers typically let customers choose a delivery time slot to receive their goods. To ensure a reliable service, the retailer may want to close time slots as capacity fills up. The number of cus- tomers that can be served per slot largely depends on the specific order sizes and delivery locations. Conceptually, checking whether it is possible to serve a certain customer in a certain time slot given a set of already accepted customers involves solving a vehicle routing problem with time windows. This is challenging in practice as there is little time available and not all relevant information is known in advance. We explore the use of machine learning to support time slot decisions in this context. Our results on realistic instances using a commercial route solver suggest that machine learning can be a promising way to assess the feasibility of customer insertions. On large-scale routing problems it performs better than insertion heuristics. |
| Keywords: | time slot management, vehicle routing, supervised machine learning |
| Date: | 2022–01–17 |
| URL: | http://d.repec.org/n?u=RePEc:ems:eureri:137095&r= |
| By: | Aldott, Zoltan (University of Warwick) |
| Abstract: | This paper aims to contribute to the coffee pricing literature pertaining to the Cup of Excellence (CoE) competitions by revising the feature set used and extending the modelling approach using machine learning. The specific dataset used is merged from data provided by the Alliance for Coffee Excellence and information collected through scraping public information from the Cup of Excellence website. The paper compares popular supervised learning algorithms exploring multiple interpretations of tasting notes to attain an efficient predictive model of prices. The algorithms compared include OLS, regularised linear algorithms, the decision tree, as well as, bagging and gradient-boosting ensemble methods. The best-performing models are further optimised using hyperparameter tuning and the most efficient one is selected. Based on a gradient-boosting regression, the final model is analysed to find the key relationships driving model predictions. Permutation feature importance and accumulated local effects analyses are used to provide insights into the non-linearities present in the data generating process. |
| Keywords: | specialty coffee ; machine learning ; prediction ; Coffee Taster’s Flavor Wheel ; Cup of Excellence JEL Classification: C53 ; C81 ; D44 ; Q11 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:wrk:wrkesp:15&r= |
| By: | Songyan Hou; Thomas Krabichler; Marcus Wunsch |
| Abstract: | Using techniques from deep learning (cf. [B\"uh+19]), we show that neural networks can be trained successfully to replicate the modified payoff functions that were first derived in the context of partial hedging by [FL00]. Not only does this approach better accommodate the realistic setting of hedging in discrete time, it also allows for the inclusion of transaction costs as well as general market dynamics. |
| Date: | 2021–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2112.07335&r= |
| By: | Vladimir Pyrlik; Pavel Elizarov; Aleksandra Leonova |
| Abstract: | We assess the performance of selected machine learning algorithms (lasso, random forest, gradient boosting, and long short-term memory) in forecasting the daily realized volatility of returns of selected top stocks in the Russian stock market in comparison with a heterogeneous autoregressive realized volatility benchmark in 2018-2020. We seek to improve the predictive power of the models by including various economic indicators that carry information about future volatility. We find that lasso delivers a good combination of easy implementation and forecast precision. The other algorithms require fine-tuning and frequent re-training, otherwise they are likely to fail to outperform the benchmark often enough. Only the basic lagged log-RV values are significant explanatory variables in terms of the benchmark in-sample quality. Many economic indicators of mixed frequencies improve the predictive power of lasso though, including calendar and overnight effects, financial spillovers from local and global markets, and various macroeconomics indicators. |
| Keywords: | heterogeneous autoregressive model; machine learning; lasso; gradient boosting; random forest; long short-term memory; realized volatility; Russian stock market; mixed-frequency data; |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:cer:papers:wp713&r= |
| By: | Justin Dang (UCR); Aman Ullah (Department of Economics, University of California Riverside) |
| Abstract: | This paper proposes a new combined semiparametric estimator of the conditional variance that takes the product of a parametric estimator and a nonparametric estimator based on machine learning. A popular kernel based machine learning algorithm, known as kernel regularized least squares estimator, is used to estimate the nonparametric component. We discuss how to estimate the semiparametric estimator using real data and how to use this estimator to make forecasts for the conditional variance.Simulations are conducted to show the dominance of the proposed estimator in terms of mean squared error. An empirical application using S&P 500 daily returns is analyzed, and the semiparametric estimator effectively forecasts future volatility. |
| Keywords: | Conditional variance; Nonparametric estimator; Semiparametric models; Forecasting; Machine Learning |
| JEL: | C01 C14 C51 |
| Date: | 2021–01 |
| URL: | http://d.repec.org/n?u=RePEc:ucr:wpaper:202204&r= |
| By: | Katsafados, Apostolos; Anastasiou, Dimitris |
| Abstract: | The purpose of this study is twofold. First, to construct short-term prediction models for bank deposit flows in the Euro area peripheral countries, employing machine learning techniques. Second, to examine whether textual features enhance the predictive ability of our models. We find that Random Forest models including both textual features and macroeconomic variables outperform those that include only macro factors or textual features. Monetary policy authorities or macroprudential regulators could adopt our approach to timely predict potential excessive bank deposit outflows and assess the resilience of the whole banking sector in the Euro area peripheral countries. |
| Keywords: | Bank deposit flows; European banks; textual analysis; short-term prediction; machine learning |
| JEL: | C0 C22 C5 C51 C54 E44 E47 G10 |
| Date: | 2022–01 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:111418&r= |
| By: | Shimon Kogan; Vitaly Meursault |
| Abstract: | A large body of literature documents the link between textual communication (e.g., news articles, earnings calls) and firm fundamentals, either through pre-defined “sentiment” dictionaries or through machine learning approaches. Surprisingly, little is known about why textual communication matters. In this paper, we take a step in that direction by developing a new methodology to automatically classify statements into objective (“facts”) and subjective (“opinions”) and apply it to transcripts of earnings calls. The large scale estimation suggests several novel results: (1) Facts and opinions are both prominent parts of corporate disclosure, taking up roughly equal parts, (2) higher prevalence of opinions is associated with investor disagreement, (3) anomaly returns are realized around the disclosure of opinions rather than facts, and (4) facts have a much stronger correlation with contemporaneous financial performance but facts and opinions have an equally strong association with financial results for the next quarter. |
| Keywords: | Subjectivity; Machine Learning; NLP; Text Analysis |
| JEL: | C00 G12 G14 |
| Date: | 2021–11–26 |
| URL: | http://d.repec.org/n?u=RePEc:fip:fedpwp:93414&r= |
| By: | Francisco Caio Lima Paiva; Leonardo Kanashiro Felizardo; Reinaldo Augusto da Costa Bianchi; Anna Helena Reali Costa |
| Abstract: | The feasibility of making profitable trades on a single asset on stock exchanges based on patterns identification has long attracted researchers. Reinforcement Learning (RL) and Natural Language Processing have gained notoriety in these single-asset trading tasks, but only a few works have explored their combination. Moreover, some issues are still not addressed, such as extracting market sentiment momentum through the explicit capture of sentiment features that reflect the market condition over time and assessing the consistency and stability of RL results in different situations. Filling this gap, we propose the Sentiment-Aware RL (SentARL) intelligent trading system that improves profit stability by leveraging market mood through an adaptive amount of past sentiment features drawn from textual news. We evaluated SentARL across twenty assets, two transaction costs, and five different periods and initializations to show its consistent effectiveness against baselines. Subsequently, this thorough assessment allowed us to identify the boundary between news coverage and market sentiment regarding the correlation of price-time series above which SentARL's effectiveness is outstanding. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2112.02095&r= |
| By: | Gabriel Borrageiro; Nick Firoozye; Paolo Barucca |
| Abstract: | We demonstrate an application of online transfer learning as a digital assets trading agent. This agent makes use of a powerful feature space representation in the form of an echo state network, the output of which is made available to a direct, recurrent reinforcement learning agent. The agent learns to trade the XBTUSD (Bitcoin versus US dollars) perpetual swap derivatives contract on BitMEX. It learns to trade intraday on five minutely sampled data, avoids excessive over-trading, captures a funding profit and is also able to predict the direction of the market. Overall, our crypto agent realises a total return of 350%, net of transaction costs, over roughly five years, 71% of which is down to funding profit. The annualised information ratio that it achieves is 1.46. |
| Date: | 2022–01 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2201.04699&r= |
| By: | Ricketts, Simon (Monash University) |
| Abstract: | We examine whether political actors engage in strategic deception on social media. We find evidence that certain groups of politicians engage in deception in response to an election. To infer deception, we construct a novel wealth inference model from text of political social media accounts. We use machine learning and natural language processing, which is accurate to within half an order of magnitude when compared to real wealth disclosures as required by law in the United States. Wealth exaggeration is not homogenous ; in an election year, the wealthiest political actors minimise their perceived wealth, while the poorest exaggerate their perceived wealth. We do not find evidence that there are differences in exaggeration due to sex, party or experience. |
| Keywords: | Strategic deception ; wealth-inference ; machine-learning ; natural language processing ; social media ; election JEL Classification: C55 ; D72 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:wrk:wrkesp:16&r= |
| By: | Carbonero, Francesco (University of Turin); Davies, Jeremy (East Village Software Consultants); Ernst, Ekkehard (ILO International Labour Organization); Fossen, Frank M. (University of Nevada, Reno); Samaan, Daniel (ILO International Labour Organization); Sorgner, Alina (John Cabot University) |
| Abstract: | AI is transforming labor markets around the world. Existing research has focused on advanced economies but has neglected developing economies. Different impacts of AI on labor markets in different countries arise not only from heterogeneous occupational structures, but also from the fact that occupations vary across countries in their composition of tasks. We propose a new methodology to translate existing measures of AI impacts that were developed for the US to countries at various levels of economic development. Our method assesses semantic similarities between textual descriptions of work activities in the US and workers' skills elicited in surveys for other countries. We implement the approach using the measure of suitability of work activities for machine learning provided by Brynjolfsson et al. (2018) for the US and the World Bank's STEP survey for Lao PDR and Viet Nam. Our approach allows characterizing the extent to which workers and occupations in a given country are subject to destructive digitalization, which puts workers at risk of being displaced, in contrast to transformative digitalization, which tends to benefit workers. We find that workers in Lao PDR are less likely than in Viet Nam to be in the "machine terrain", where workers will have to adapt to occupational transformations due to AI and are at risk of being partially displaced. Our method based on semantic textual similarities using SBERT is advantageous compared to approaches transferring AI impact scores across countries using crosswalks of occupational codes. |
| Keywords: | artificial intelligence, machine learning, digitalization, labor, skills, developing countries |
| JEL: | J22 J23 O14 O33 |
| Date: | 2021–12 |
| URL: | http://d.repec.org/n?u=RePEc:iza:izadps:dp14944&r= |
| By: | Yu, Bo (Deakin University); Tran, Trang (University of Maryland at College Park); Lee, Wang-Sheng (Monash University) |
| Abstract: | Following market liberalisation, the vehicle population in China has increased dramatically over the past few decades. This paper examines the causal impact of the opening of a heavily used high speed rail line connecting two megacities in China in 2015, Chengdu and Chongqing, on air pollution. We use high-frequency and high spatial resolution data to track pollution along major highways linking the two cities. Our approach involves the use of an augmented regression discontinuity in time approach applied on data that have been through a meteorological normalisation process. This deweathering process involves applying machine learning techniques to account for change in meteorology in air quality time series data. Our estimates show that air pollution is reduced by 7.6% along the main affected highway. We simultaneously find increased levels of ozone pollution which is likely due to the reduction in nitrogen dioxide levels that occurred. These findings are supported using a difference-in-difference approach. |
| Keywords: | air pollution, China, green infrastructure, high-speed railway, regression discontinuity, machine learning |
| JEL: | L92 O18 Q53 Q54 R41 |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:iza:izadps:dp14900&r= |
| By: | Suppawong Tuarob; Thanapon Noraset; Tanisa Tawichsri |
| Abstract: | Mental health problems are among major public health concerns during the COVID-19 pandemic, given heightened uncertainties and drastic changes in lifestyles. However, mental health problem prevention and monitoring could be greatly improved given advancements in deep-learning techniques and readily available social media messages. This research uses deep learning algorithms to extract emotion, mood, and psychological cues from social media messages and then aggregates these signals to track population-level mental health. To verify the accuracy of our proposed approaches, we compared our findings to the actual number of patients treated for depression, attempted suicides, and self-harm cases reported by Thailand’s Department of Mental Health. We discovered a strong correlation between the predicted mental signals and actual depression, suicide, and self-harm (injured) cases. Finally, we also create a database and user-friendly interface to facilitate researchers and policymakers to explore our extracted mental signals for further applications such as policy sentiment assessment. |
| Keywords: | Mental Health; Natural Language Processing; Deep Learning; Social Networks |
| JEL: | I10 |
| Date: | 2022–01 |
| URL: | http://d.repec.org/n?u=RePEc:pui:dpaper:169&r= |