Journal of Finance and Data Science最新文献

Statistical and computational methods are being increasingly integrated into Decision Support Systems to aid management and help with strategic decisions. Researchers need to fully understand the use of such techniques in order to make predictions when using financial data. This paper therefore presents a method based literature review focused on the predictive analytics domain. The study comprehensively covers classification, regression, clustering, association and time series models. It expands existing explanatory statistical modelling into the realm of computational modelling. The methods explored enable the prediction of the future through the analysis of financial time series and cross-sectional data that is collected, stored and processed in Information Systems. The output of such models allow financial managers and risk oversight professionals to achieve better outcomes. This review brings the various predictive analytic methods in finance together under one domain.

We find economically and statistically significant gains when using machine learning for portfolio allocation between the market index and risk-free asset. Optimal portfolio rules for time-varying expected returns and volatility are implemented with two Random Forest models. One model is employed in forecasting monthly excess returns with macroeconomic factors including payout yields. The second is used to estimate the prevailing volatility. Reward-risk timing with machine learning provides substantial improvements over the buy-and-hold in utility, risk-adjusted returns, and maximum drawdowns. This paper presents a unifying framework for machine learning applied to both return- and volatility-timing.

On March 23, 2022, the SEC proposed that firms publicly disclose their cybersecurity incidents within four days of discovery. In the U.S., state-level data breach disclosure laws require firms to disclose the occurrence of a data breach, with some mandating disclosure within a deadline while others do not. Exploiting this state-level variation in disclosure deadlines, we find that, when facing a deadline, firms disclose a data breach 90 percent faster but are 58 percent less likely to disclose breach details. Investors respond negatively to delayed breach disclosures but are forgiving of a delay when it is used to gather more breach details. Our study highlights the trade-offs of mandating a disclosure deadline for cybersecurity incidents.

We analyze the performance of investable portfolios built using predicted stock returns from machine learning methods and attribute their performance to linear, marginal non-linear and interaction effects. We use a large set of features including price-based, fundamental-based, and sentiment-based descriptors and use model averaging in the validation procedure to get robust out-of-sample predictions. We find that the superiority of regression trees and neural networks comes from two points: their strong regularization mechanism and their capacity to capture interaction effects. The non-linear component of the marginal predictions on the other hand has no predictive power. Thanks to our methodology, we manage to isolate and study in detail the interaction component. We find that it has significative long term performance independent from the linear modeling and is stable through time.

Emerging literature focuses on insurers' earnings management using estimated liability for unpaid claims, known as loss reserve. An insurance company generally uses the traditional estimation methods with linear estimation to measure loss reserve error, but those methods are often criticized for several statistical shortcomings, such as estimation technique, correlated contributing variables, ignorance of the interactions, and higher-order terms. To overcome such shortcomings, this paper proposes an unsupervised-supervised machine learning approach, hierarchical clustering, and artificial neural network (ANN) by adopting a combined unsupervised-supervised method, cluster analysis (i.e., unsupervised), and various supervised machine learning algorithms such as Boostings, Support Vector Machine (SVM) and RReliefF. We show evidence that each cluster has its own foundation variables to predict and Boosting and ANN estimation provide a more efficient framework to improve insurers' reserve error. Also, the different value and order of RReliefF between Boosting and OLS show the under-or over-estimated predictor, and each year's influential variables are found to be consistent over time, which indicates that the firm's previous year's loss reserve model can predict the future loss reserve error. This paper contributes to the existing literature by suggesting a more robust, consistent, and efficient prediction method (i.e., unsupervised-supervised combination method) to improve insurers' loss reserve error prediction.

We present a simple and effective methodology for the generation of lexicons (word lists) that may be used in natural language scoring applications. In particular, in the finance industry, word lists have become ubiquitous for sentiment scoring. These have been derived from dictionaries such as the Harvard Inquirer and require manual curation. Here, we present an automated approach to the curation of lexicons, which makes automatic preparation of any word list immediate. We show that our automated word lists deliver comparable performance to traditional lexicons on machine learning classification tasks. This new approach will enable finance academics and practitioners to create and deploy new word lists in addition to the few traditional ones in a facile manner.

Using a big data set of venture capital financing and related startup firms from Crunchbase, this paper develops a machine-learning model called CapitalVX (for “Capital Venture eXchange”) to predict the outcomes for startups, i.e., whether they will exit successfully through an IPO or acquisition, fail, or remain private. Using a large feature set, the out-of-sample accuracy of predictions on startup outcomes and follow-on funding is 80–89%. This research suggests that VC/PE firms may be able to benefit from using machine learning to screen potential investments using publicly available information, diverting this time instead into mentoring and monitoring the investments they make.

We analyze the predictability of the bitcoin market across prediction horizons ranging from 1 to 60 min. In doing so, we test various machine learning models and find that, while all models outperform a random classifier, recurrent neural networks and gradient boosting classifiers are especially well-suited for the examined prediction tasks. We use a comprehensive feature set, including technical, blockchain-based, sentiment-/interest-based, and asset-based features. Our results show that technical features remain most relevant for most methods, followed by selected blockchain-based and sentiment-/interest-based features. Additionally, we find that predictability increases for longer prediction horizons. Although a quantile-based long-short trading strategy generates monthly returns of up to 39% before transaction costs, it leads to negative returns after taking transaction costs into account due to the particularly short holding periods.

The VSTOXX index tracks the expected 30-day volatility of the EURO STOXX 50 equity index. Futures on the VSTOXX index can, therefore, be used to hedge against economic uncertainty. We investigate the effect of trader inventory on the price of VSTOXX futures through a combination of stochastic processes and machine learning methods. We formulate a simple and efficient pricing methodology for VSTOXX futures, which assumes a Heston-type stochastic process for the underlying EURO STOXX 50 market. Under these dynamics, approximate analytical formulas for the implied volatility smile and the VSTOXX index have recently been derived. We use the EURO STOXX 50 option implied volatilities and the VSTOXX index value to estimate the parameters of this Heston model. Following the calibration, we calculate theoretical VSTOXX futures prices and compare them to the actual market prices. While theoretical and market prices are usually in line, we also observe time periods, during which the market price does not agree with our Heston model. We collect a variety of market features that could potentially explain the price deviations and calibrate two machine learning models to the price difference: a regularized linear model and a random forest. We find that both models indicate a strong influence of accumulated trader positions on the VSTOXX futures price.

A high-quality and granular probability of default (PD) model is on many practical dimensions far superior to any categorical credit rating system. Business adoption of a PD model, however, needs to factor in the long-established business/regulatory conventions built around letter-based credit ratings. A mapping methodology that converts granular PDs into letter ratings via referencing the historical default experience of some credit rating agency exists in the literature. This paper improves the PD implied rating (PDiR) methodology by targeting the historical credit migration matrix instead of simply default rates. This enhanced PDiR methodology makes it possible to bypass the reliance on arbitrarily extrapolated target default rates for the AAA and AA⁺ categories, a necessity due to the fact that the historical realized default rates on these two top rating grades are typically zero.