Computational Economics最新文献

Fuzzy candlestick models have been widely used to forecast the stock market due to their capability to handle ubiquitous nonlinearities and the knowledge of investors. However, such models take only partial historical data into account and make the prediction exclusively by the selected historical data without considering the estimation errors and also lack long-term sequence information. To address these problems, a hybrid model (WEF-GRU) combines the so-called weighted error-based fuzzy candlestick (WEF) model and the improved gated recurrent unit (GRU) network is designed to reflect the influence of historical data and investor sentiment on the predicted result adequately and properly. In this study, the WEF model is established to map the fuzzy inputs to rough output to extract effective features based on the experience and knowledge of investors. Meanwhile, the GRU network is employed to maintain the long-term sequence information according to technique indicators, and then the final predicted result is derived by fusing the outputs of the WEF model and the GRU model. Finally, experimental results on eight real-world stock data which contain daily data demonstrate that the proposed hybrid model outperforms the baseline models.

Artificial intelligence has shortened edges in many areas, especially the economy, to support long-term and accurate forecasting of financial indicators. Traditional statistical methods perform poorly compared to those based on artificial intelligence, which can achieve higher rates even with high-dimensional datasets. This method still needs evolution and studies. In emerging countries, decision-makers and investors must follow the basic interest rate, such as in Brazil, with a Special System of Settlement and Custody (Selic). Prior works used deep neural networks (DNNs) for forecasting time series economic indicators such as interest rates, inflation, and the stock market. However, there is no empirical evaluation of the prediction models for the Selic interest rate, especially the impact of training time and the optimization of hyperparameters. In this paper, we shed light on these issues and evaluate, through a fair comparison, the use of DNNs models for Selic time series forecasting. Our results demonstrate the potential of DNNs with an error rate above 0.00219 and training time above 84.28 s. Our findings open up opportunities for further investigations toward real-time interest rate forecasting, facilitating more reliable and timely forecasting of interest rates for decision-makers and investors.

This paper introduces statistical models Wordscores and Wordfish to study and predict banking crises. While Wordscores is akin to supervised learning, Wordfish is analogous to unsupervised learning. Both methods estimate the position of banking distress on a tranquil-to-crisis spectrum. Findings suggest that the two statistical methods signal banking crisis up to two-years in advance, with robust results from AUROC, Granger causality and VAR impulse responses. Both methods outperform random forests in predicting crises using textual data. The Wordscores index highlights increased usage of banking sector nomenclature two years preceding a crisis, and Granger causes a crisis series with one and two lag lengths. Results from the Wordfish technique, a statistical model with Poisson distribution, show the index spikes before and during the Global Financial Crisis, when a large share of the countries in the world encountered banking crises. This paper contributes to literature on text-based models of banking crises by bolstering the preemptive policy responses available to policy makers. Given their early warning signals, both Wordscores and Wordfish can be considered a part of the toolset to monitor the stability and resilience of the banking sector.

This study employs dynamic model averaging and selection of Vector Autoregressive and Time-Varying Parameters Vector Autoregressive models to forecast out-of-sample monthly returns of US stocks, bonds, and Real Estate Investment Trusts (REITs) indexes from October 2006 to December 2021. The models were recursively estimated using 17 additional predictors chosen by a genetic algorithm applied to an initial list of 155 predictors. These forecasts were then used to dynamically choose portfolios formed by these assets and the riskless asset proxied by the 3-month US treasury bills. Although we did not find any predictability in the stock market, positive results were obtained for REITs and especially for bonds. The Bayesian-based approaches applied to just the returns of the three risky assets resulted in portfolios that remarkably outperform the portfolios based on the historical means and covariances and the equally weighted portfolio in terms of certainty equivalent return, Sharpe ratio, Sortino ratio and even Conditional Value-at-Risk at 5%. This study points out that Constant Relative Risk Averse investors should use Bayesian-based approaches to forecast and choose the investment portfolios, focusing their attention on different types of assets.

The use of machine learning models and techniques to predict economic variables has been growing lately, motivated by their better performance when compared to that of linear models. Although linear models have the advantage of considerable interpretive power, efforts have intensified in recent years to make machine learning models more interpretable. In this paper, tests are conducted to determine whether models based on machine learning algorithms have better performance relative to that of linear models for predicting the size of the informal economy. The paper also explores whether the determinants of such size detected as the most important by machine learning models are the same as those detected in the literature based on traditional linear models. For this purpose, observations were collected and processed for 122 countries from 2004 to 2014. Next, twelve models (four linear and eight based on machine learning algorithms) were used to predict the size of the informal economy in these countries. The relative importance of the predictive variables in determining the results yielded by the machine learning algorithms was calculated using Shapley values. The results suggest that (i) models based on machine learning algorithms have better predictive performance than that of linear models and (ii) the main determinants detected through the Shapley values coincide with those detected in the literature using traditional linear models.

We present a novel, straightforward, robust, and precise calibration algorithm for local volatility surfaces based on observed market call and put option prices. The proposed local volatility reconstruction method is based on the widely recognized generalized Black–Scholes partial differential equation, which is numerically solved using a finite difference scheme. In the proposed method, sample points are strategically placed in the underlying and time domains. The unknown local volatility function is represented using the scattered interpolant function. The primary contribution of this study is that our proposed algorithm not only optimizes the volatility values at the sample points but also optimizes the positions of the sample positions using a least squares method. This optimization process improves the accuracy and robustness of our calibration method. Furthermore, we do not use the Tikhonov regularization technique, which was frequently used to obtain smooth solutions. To validate the practical efficiency and superior performance of the proposed reconstruction method for local volatility functions, we conduct a series of computational experiments using real-world market option prices such as the KOSPI 200, S &P 500, Hang Seng, and Euro Stoxx 50 indices. The proposed algorithm offers financial market practitioners a reliable tool for calibrating local volatility surfaces using only market option prices, enabling more accurate pricing and risk management of financial derivatives.

Private equity (PE) represents the acquisition of stakes in non-listed companies, often long-term, with the objective of improving the performance and value of the company to obtain significant benefits at time of disinvestment. PE has gained particular importance in the global financial system for delivering superior risk-adjusted returns. Knowing the PE return drivers has been of great interest among researchers and academics, and some studies have developed statistical models to determine PE return drivers. Still, the explanatory capacity of these models has certain limitations related to their precision levels and exclusive focus on groups of countries located in Europe and the EE.UU. Therefore, in the current literature, new models of analysis of the PE return drivers are demanded to provide a better fit in worldwide scenarios. This study contributes to the accuracy of the models that identify the PE return drivers using computational methods and a sample of 1606 PE funds with a geographical focus on the world’s five regions. The results have provided a unique set of PE return drivers with a precision level above 86%. The conclusions obtained present important theoretical and practical implications, expanding knowledge about PE and financial forecasting from a global perspective.

Abstract

In this paper, we present a new approach for modelling matrix-variate time series data that accounts for smooth changes in the dynamics of matrices. Although stylized facts in several fields suggest the existence of smooth nonlinearities, the existing matrix-variate models do not account for regime switches that are not abrupt. To address this gap, we introduce the matrix smooth transition autoregressive model, a flexible regime-switching model capable of capturing abrupt, smooth and no regime changes in matrix-valued data. We provide a thorough examination of the estimation process and evaluate the finite-sample performance of the matrix-variate smooth transition autoregressive model estimators with simulated data. Finally, the model is applied to real-world data.

Abstract

Since currency price fluctuations hinder economic activity, exchange rate dynamics have an effect on national economies. To have a proper exchange rate policy in place, these dynamics are essential for nations with a trade economy. This study presents and examines a distinctive stochastic dynamics exchange rate model (ESI) in order to address the challenges associated with predicting the behavior of participants in some complex economic systems, which might lead to the system’s collapse. To address the issue of ESI stability by central bank interventions (managed currency) in a specified target value, a target value technique is also provided and tested. Last but not least, we examine the noise traders’ role as a major source of market uncertainty as we look at the market efficiency hypothesis for the foreign exchange market (FX).

This study leverages decentralized liquidity pool data sourced from UNISWAP-V2 to forecast Value-at-Risk (VaR) and Expected Shortfall (ES) employing the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model with varied error distributions and the deep learning probabilistic forecasting algorithm known as DeepAR. Performance evaluations of these distinct forecasting methodologies are conducted using an appropriate loss function. Results indicate that the GARCH model with a normal distribution consistently outperforms other models, particularly when forecasting VaR. Conversely, the DeepAR model exhibits limited effectiveness in VaR forecasting across all scenarios, except for liquidity pools involving at least one stablecoin. However, it demonstrates greater reliability in predicting most ES risk measures and associated data. Our findings underscore that in a subset of the data, providing liquidity to pairs involving at least one stablecoin entails statistically significant lower risk compared to holding an equivalent amount of crypto assets. Furthermore, this research contributes to the advancement of novel risk management tools and strategies tailored for liquidity providers.