nep-cmp New Economics Papers
on Computational Economics
Issue of 2021‒10‒04
twelve papers chosen by
Stan Miles
Thompson Rivers University

  1. Application Machine Learning in Construction Management By Nguyen, Phong Thanh
  2. Multi-Transformer: A New Neural Network-Based Architecture for Forecasting S&P Volatility By Eduardo Ramos-P\'erez; Pablo J. Alonso-Gonz\'alez; Jos\'e Javier N\'u\~nez-Vel\'azquez
  3. Option return predictability with machine learning and big data By Bali, Turan G.; Beckmeyer, Heiner; Moerke, Mathis; Weigert, Florian
  4. Combining Discrete Choice Models and Neural Networks through Embeddings: Formulation, Interpretability and Performance By Ioanna Arkoudi; Carlos Lima Azevedo; Francisco C. Pereira
  5. Delta Hedging with Transaction Costs: Dynamic Multiscale Strategy using Neural Nets By G. Mazzei; F. G. Bellora; J. A. Serur
  6. Intra-Day Price Simulation with Generative Adversarial Modelling of the Order Flow By Ye-Sheen Lim; Denise Gorse
  7. Energy Pricing during the COVID-19 Pandemic: Predictive Information-Based Uncertainty Indexes with Machine Learning Algorithm By Olubusoye, Olusanya E; Akintande, Olalekan J.; Yaya, OlaOluwa S.; Ogbonna, Ahamuefula; Adenikinju, Adeola F.
  8. Estimating value-added returns to labor training programs with causal machine learning By Angell, Mintaka; Gold, Samantha; Hastings, Justine S.; Howison, Mark; Jensen, Scott; Keleher, Niall; Molitor, Daniel; Roberts, Amelia
  9. Conditional Value-at-Risk for Quantitative Trading: A Direct Reinforcement Learning Approach By Ali Al-Ameer; Khaled Alshehri
  10. Reinforcement Learning for Quantitative Trading By Shuo Sun; Rundong Wang; Bo An
  11. A network approach to consumption By Schulz, Jan; Mayerhoffer, Daniel M.
  12. Useful Results for the Simulation of Non-Optimal Economies with Heterogeneous Agents By Damián Pierri

  1. By: Nguyen, Phong Thanh
    Abstract: Machine Learning is a subset and technology developed in the field of Artificial Intelligence (AI). One of the most widely used machine learning algorithms is the K-Nearest Neighbors (KNN) approach because it is a supervised learning algorithm. This paper applied the K-Nearest Neighbors (KNN) algorithm to predict the construction price index based on Vietnam's socio-economic variables. The data to build the prediction model was from the period 2016 to 2019 based on seven socio-economic variables that impact the construction price index (i.e., industrial production, construction investment capital, Vietnam’s stock price index, consumer price index, foreign exchange rate, total exports, and imports). The research results showed that the construction price index prediction model based on the K-Nearest Neighbors (KNN) regression method has fewer errors than the traditional method.
    Keywords: Artificial Intelligence, K-Nearest Neighbors (KNN), machine learning, price index, construction management
    JEL: C53 C8 E0 L16 L74
    Date: 2020–12–29
  2. By: Eduardo Ramos-P\'erez; Pablo J. Alonso-Gonz\'alez; Jos\'e Javier N\'u\~nez-Vel\'azquez
    Abstract: Events such as the Financial Crisis of 2007-2008 or the COVID-19 pandemic caused significant losses to banks and insurance entities. They also demonstrated the importance of using accurate equity risk models and having a risk management function able to implement effective hedging strategies. Stock volatility forecasts play a key role in the estimation of equity risk and, thus, in the management actions carried out by financial institutions. Therefore, this paper has the aim of proposing more accurate stock volatility models based on novel machine and deep learning techniques. This paper introduces a neural network-based architecture, called Multi-Transformer. Multi-Transformer is a variant of Transformer models, which have already been successfully applied in the field of natural language processing. Indeed, this paper also adapts traditional Transformer layers in order to be used in volatility forecasting models. The empirical results obtained in this paper suggest that the hybrid models based on Multi-Transformer and Transformer layers are more accurate and, hence, they lead to more appropriate risk measures than other autoregressive algorithms or hybrid models based on feed forward layers or long short term memory cells.
    Date: 2021–09
  3. By: Bali, Turan G.; Beckmeyer, Heiner; Moerke, Mathis; Weigert, Florian
    Abstract: Drawing upon more than 12 million observations over the period from 1996 to 2020, we find that allowing for nonlinearities significantly increases the out-of-sample performance of option and stock characteristics in predicting future option returns. Besides statistical significance, the nonlinear machine learning models generate economically sizeable profits in the long-short portfolios of equity options even after accounting for transaction costs. Although option-based characteristics are the most important standalone predictors, stock-based measures offer substantial incremental predictive power when considered alongside option-based characteristics. Finally, we provide compelling evidence that option return predictability is driven by informational frictions, costly arbitrage, and option mispricing.
    Keywords: Machine learning,big data,option return predictability
    JEL: G10 G12 G13 G14
    Date: 2021
  4. By: Ioanna Arkoudi; Carlos Lima Azevedo; Francisco C. Pereira
    Abstract: This study proposes a novel approach that combines theory and data-driven choice models using Artificial Neural Networks (ANNs). In particular, we use continuous vector representations, called embeddings, for encoding categorical or discrete explanatory variables with a special focus on interpretability and model transparency. Although embedding representations within the logit framework have been conceptualized by Camara (2019), their dimensions do not have an absolute definitive meaning, hence offering limited behavioral insights. The novelty of our work lies in enforcing interpretability to the embedding vectors by formally associating each of their dimensions to a choice alternative. Thus, our approach brings benefits much beyond a simple parsimonious representation improvement over dummy encoding, as it provides behaviorally meaningful outputs that can be used in travel demand analysis and policy decisions. Additionally, in contrast to previously suggested ANN-based Discrete Choice Models (DCMs) that either sacrifice interpretability for performance or are only partially interpretable, our models preserve interpretability of the utility coefficients for all the input variables despite being based on ANN principles. The proposed models were tested on two real world datasets and evaluated against benchmark and baseline models that use dummy-encoding. The results of the experiments indicate that our models deliver state-of-the-art predictive performance, outperforming existing ANN-based models while drastically reducing the number of required network parameters.
    Date: 2021–09
  5. By: G. Mazzei; F. G. Bellora; J. A. Serur
    Abstract: In most real scenarios the construction of a risk-neutral portfolio must be performed in discrete time and with transaction costs. Two human imposed constraints are the risk-aversion and the profit maximization, which together define a nonlinear optimization problem with a model-dependent solution. In this context, an optimal fixed frequency hedging strategy can be determined a posteriori by maximizing a sharpe ratio simil path dependent reward function. Sampling from Heston processes, a convolutional neural network was trained to infer which period is optimal using partial information, thus leading to a dynamic hedging strategy in which the portfolio is hedged at various frequencies, each weighted by the probability estimate of that frequency being optimal.
    Date: 2021–09
  6. By: Ye-Sheen Lim; Denise Gorse
    Abstract: Intra-day price variations in financial markets are driven by the sequence of orders, called the order flow, that is submitted at high frequency by traders. This paper introduces a novel application of the Sequence Generative Adversarial Networks framework to model the order flow, such that random sequences of the order flow can then be generated to simulate the intra-day variation of prices. As a benchmark, a well-known parametric model from the quantitative finance literature is selected. The models are fitted, and then multiple random paths of the order flow sequences are sampled from each model. Model performances are then evaluated by using the generated sequences to simulate price variations, and we compare the empirical regularities between the price variations produced by the generated and real sequences. The empirical regularities considered include the distribution of the price log-returns, the price volatility, and the heavy-tail of the log-returns distributions. The results show that the order sequences from the generative model are better able to reproduce the statistical behaviour of real price variations than the sequences from the benchmark.
    Date: 2021–09
  7. By: Olubusoye, Olusanya E; Akintande, Olalekan J.; Yaya, OlaOluwa S.; Ogbonna, Ahamuefula; Adenikinju, Adeola F.
    Abstract: The study investigates the impact of uncertainties on energy pricing during the COVID-19 pandemic using five uncertainty measures that include the COVID-Induced Uncertainty (CIU), Economic Policy Uncertainty (EPU), Global Fear Index (GFI); Volatility Index (VIX), and the Misinformation Index of Uncertainty (MIU). The data, which span between 2-January, 2020 and 19-January, 2021, corresponding to the period of the COVID-19 pandemic. The study finds energy prices to respond significantly to the examined uncertainty measures, with EPU seen to affect the prices of most energy types during the pandemic. We also find predictive potentials inherent in VIX, CIU, and MIU for global energy sources.
    Keywords: Coronavirus pandemic; Energy market; Machine Learning; Uncertainty
    JEL: D8 D81 Q41
    Date: 2021–09–21
  8. By: Angell, Mintaka; Gold, Samantha; Hastings, Justine S.; Howison, Mark; Jensen, Scott; Keleher, Niall; Molitor, Daniel; Roberts, Amelia
    Abstract: Technology may displace tens of millions of workers in the coming decades. Part of the explanation for the projected displacement is an expanding mismatch in skills that employers seek and the skills that workers possess. Effects of labor force displacement disproportionately affect low-income workers and workers within industries where technological change replaces labor. As a result, a great deal of emphasis is placed on training and reskilling workers to ease transitions into new careers. However, utilization of training programs may be below optimal levels if workers are uncertain about the returns to their investment in training. While the U.S. spends billions of dollars annually on reskilling programs and unemployment insurance, there are few measures of program effectiveness that workers and government can use to guide training investment decisions and ensure delivery of valuable reskilling and improved outcomes. In a nationwide conjoint survey experiment, we find job seekers prefer information on the value-added returns to earnings following enrollment in training and reskilling programs. We identify a clear demand for value-added measures. For every 10% increase in expected earnings, workers are 17.4% more likely to express interest in a training program. To meet this demand for information, governments can provide return on investment measures. Fortunately, the data to estimate these returns are available in state administrative data. We demonstrate a causal machine learning method that provides these missing causal estimates of value-added that workers prefer and that can provide correct incentives in the market for labor training. Focusing on a set of workforce training programs in Rhode Island, our causal machine learning estimates suggest that training increases enrollees’ future quarterly earnings by \$605. We estimate that return on investment ranges between -\$1,570 in quarterly earnings for the lowest value-added program to \$3,470 in quarterly earnings for the highest value-added program.
    Date: 2021–09–24
  9. By: Ali Al-Ameer; Khaled Alshehri
    Abstract: We propose a convex formulation for a trading system with the Conditional Value-at-Risk as a risk-adjusted performance measure under the notion of Direct Reinforcement Learning. Due to convexity, the proposed approach can uncover a lucrative trading policy in a "pure" online manner where it can interactively learn and update the policy without multi-epoch training and validation. We assess our proposed algorithm on a real financial market where it trades one of the largest US trust funds, SPDR, for three years. Numerical experiments demonstrate the algorithm's robustness in detecting central market-regime switching. Moreover, the results show the algorithm's effectiveness in extracting profitable policy while meeting an investor's risk preference under a conservative frictional market with a transaction cost of 0.15% per trade.
    Date: 2021–09
  10. By: Shuo Sun; Rundong Wang; Bo An
    Abstract: Quantitative trading (QT), which refers to the usage of mathematical models and data-driven techniques in analyzing the financial market, has been a popular topic in both academia and financial industry since 1970s. In the last decade, reinforcement learning (RL) has garnered significant interest in many domains such as robotics and video games, owing to its outstanding ability on solving complex sequential decision making problems. RL's impact is pervasive, recently demonstrating its ability to conquer many challenging QT tasks. It is a flourishing research direction to explore RL techniques' potential on QT tasks. This paper aims at providing a comprehensive survey of research efforts on RL-based methods for QT tasks. More concretely, we devise a taxonomy of RL-based QT models, along with a comprehensive summary of the state of the art. Finally, we discuss current challenges and propose future research directions in this exciting field.
    Date: 2021–09
  11. By: Schulz, Jan; Mayerhoffer, Daniel M.
    Abstract: The nexus between debt and inequality has attracted considerable scholarly attention in the wake of the global financial crisis. One prominent candidate to explain the striking co-evolution of income inequality and private debt in this period has been the theory of upward-looking consumption externalities leading to expenditure cascades. We propose a parsimonious model of upward-looking consumption at the micro level mediated by perception networks with empirically plausible topologies. This allows us to make sense of the ambiguous empirical literature on the relevance of this channel. Up to our knowledge, our approach is the first to make the reference group to which conspicuous consumption relates explicit. Our model, based purely on current income, replicates the major stylised facts regarding micro consumption behaviour and is thus observationally equivalent to the workhorse permanent income hypothesis, without facing its dual problem of 'excess smoothness' and 'excess sensitivity'. We also demonstrate that the network topology and segregation has a significant effect on consumption patterns which has so far been neglected.
    Keywords: Agent-Based Computational Economics,Consumption,Inequality,Relative Income Hypothesis,Positional Goods,Aggregation
    Date: 2021
  12. By: Damián Pierri (Universidad Carlos III Madrid & IIEP-BAIRES (UBA-CONICET))
    Abstract: This paper deals with infinite horizon non-optimal economies with aggregate uncertainty and a finite number of heterogeneous agents. It derives sufficient conditions for the existence of a recursive structure, an ergodic, a stationary, and a non-stationary equilibria. It also gives an answer to the following question: is it possible to derive a general framework which guarantees that numerical simulations truly reflect the behavior of endogenous variables in the model? We provide sufficient conditions to give an affirmative answer to this question for endowment economies with incomplete markets and uncountable exogenous shocks. These conditions guarantee the ergodicity of the process and hold for a particular selection mechanism. For economies with finitely many shocks or for an arbitrary selection in economies with uncountable shocks, it is only possible to show that a computable, time independent and recursive representation generates a stationary Markov process. The results in this paper suggest that often a well-defined stochastic steady state in heterogenous agent models is sensitive to the initial conditions of the economy; a fact which imply that heterogeneity may have irreversible long-lasting effects.
    Date: 2021–08

This nep-cmp issue is ©2021 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.